Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated:
0
00:00:00,599 --> 00:00:03,780
hey guys Bill Zoid here and today we're
1
00:00:02,760 --> 00:00:06,120
going to be taking a look at another
2
00:00:03,780 --> 00:00:09,019
ddr5 overclock this time with Horizon
3
00:00:06,120 --> 00:00:12,660
970 900x on an ASRock
4
00:00:09,019 --> 00:00:14,639
b650m--hdv m.2 motherboard the CPU was
5
00:00:12,660 --> 00:00:16,859
provided by AMD so big thank you to them
6
00:00:14,639 --> 00:00:18,960
for sending it over and the motherboard
7
00:00:16,859 --> 00:00:20,880
and the memory kit were purchased by me
8
00:00:18,960 --> 00:00:23,640
so big thank you to my supporters for
9
00:00:20,880 --> 00:00:25,019
making those purchases possible uh now
10
00:00:23,640 --> 00:00:27,900
then the memory kit that I'm using here
11
00:00:25,019 --> 00:00:30,660
is a Kingston Fury 2 by 16 gigabyte uh
12
00:00:27,900 --> 00:00:32,099
memory kit it uses heynix 16 gigabit MDI
13
00:00:30,660 --> 00:00:35,040
memory chips at least the one that I
14
00:00:32,099 --> 00:00:38,160
have here theoretically this uh spec of
15
00:00:35,040 --> 00:00:40,579
like 6000 cl36 at 1.35 volts you could
16
00:00:38,160 --> 00:00:43,200
end up with high Nix a die memory chips
17
00:00:40,579 --> 00:00:44,700
uh which wouldn't really change much
18
00:00:43,200 --> 00:00:46,620
because the memory timings that I'm
19
00:00:44,700 --> 00:00:49,379
using here aren't actually like super
20
00:00:46,620 --> 00:00:51,239
like they're not minimized they're just
21
00:00:49,379 --> 00:00:54,600
kind of generic hynix timings that
22
00:00:51,239 --> 00:00:57,539
should work on both a die and MDI memory
23
00:00:54,600 --> 00:00:59,460
kits so this video will apply to both
24
00:00:57,539 --> 00:01:01,199
even though there are some very slight
25
00:00:59,460 --> 00:01:03,420
differences in terms of what kind of
26
00:01:01,199 --> 00:01:05,900
timings MDI chips will do compared to 8i
27
00:01:03,420 --> 00:01:09,540
chips but anyway that doesn't really
28
00:01:05,900 --> 00:01:11,640
matter that much so yeah that's the spec
29
00:01:09,540 --> 00:01:14,100
of the system that I'm that I'm you know
30
00:01:11,640 --> 00:01:16,799
working with here for cooling there's an
31
00:01:14,100 --> 00:01:18,720
air cooler on the CPU so the CPU did end
32
00:01:16,799 --> 00:01:20,939
up running quite hot especially for lin
33
00:01:18,720 --> 00:01:22,979
pack and there's no extra cooling for
34
00:01:20,939 --> 00:01:25,200
the memory so it the memory is basically
35
00:01:22,979 --> 00:01:28,619
just cooled by whatever air washes off
36
00:01:25,200 --> 00:01:29,280
of the CPU air cooler so anyway
37
00:01:28,619 --> 00:01:31,320
um
38
00:01:29,280 --> 00:01:33,840
let's take a look at the the settings
39
00:01:31,320 --> 00:01:36,360
here so the memory I've pushed it to
40
00:01:33,840 --> 00:01:38,340
6200 megabits per second so that's 3.1
41
00:01:36,360 --> 00:01:41,159
gigahertz actual like memory clock
42
00:01:38,340 --> 00:01:42,720
because DDR Right double data rate the
43
00:01:41,159 --> 00:01:45,000
clock is half the data rate the data
44
00:01:42,720 --> 00:01:46,500
rate is 6200 megabits per second and
45
00:01:45,000 --> 00:01:48,420
then you also map to multiply that by
46
00:01:46,500 --> 00:01:51,420
the bus width which for two sticks of
47
00:01:48,420 --> 00:01:52,500
ddr5 is 128 Bits And if you multiply all
48
00:01:51,420 --> 00:01:54,420
of that together you can actually
49
00:01:52,500 --> 00:01:57,720
calculate your theoretical maximum
50
00:01:54,420 --> 00:02:00,299
memory bandwidth uh anyway
51
00:01:57,720 --> 00:02:02,460
um yeah so 3.1 gigahertz on the memory
52
00:02:00,299 --> 00:02:04,200
3.1 gigahertz on the memory controller
53
00:02:02,460 --> 00:02:05,759
uh you want to have your memory
54
00:02:04,200 --> 00:02:07,140
controller running at the same clock as
55
00:02:05,759 --> 00:02:08,880
your memory because that gives you a
56
00:02:07,140 --> 00:02:10,200
latency Advantage as well as just a
57
00:02:08,880 --> 00:02:12,900
general performance Advantage
58
00:02:10,200 --> 00:02:15,540
technically there is a two to one mode
59
00:02:12,900 --> 00:02:17,760
for the memory controller I see
60
00:02:15,540 --> 00:02:20,580
absolutely no reason to ever use that on
61
00:02:17,760 --> 00:02:23,700
ryzen 7000 in my experience it doesn't
62
00:02:20,580 --> 00:02:27,120
really clock any higher and it does hurt
63
00:02:23,700 --> 00:02:28,800
performance so yeah at least an ambient
64
00:02:27,120 --> 00:02:30,959
it really doesn't clock any higher maybe
65
00:02:28,800 --> 00:02:32,340
with like sub ambient temperatures it
66
00:02:30,959 --> 00:02:34,500
actually helps but if you're on like air
67
00:02:32,340 --> 00:02:35,940
cooling or water cooling there is zero
68
00:02:34,500 --> 00:02:37,500
reason to use the two to one mode
69
00:02:35,940 --> 00:02:39,959
because you're just giving up
70
00:02:37,500 --> 00:02:41,099
performance and yeah there's there's no
71
00:02:39,959 --> 00:02:43,860
benefits like it doesn't clock
72
00:02:41,099 --> 00:02:46,440
significantly higher enough to to like
73
00:02:43,860 --> 00:02:48,420
justify the trade-off
74
00:02:46,440 --> 00:02:50,580
anyway the infinity fabric I have it
75
00:02:48,420 --> 00:02:53,239
clocked at 2100 megahertz the infinity
76
00:02:50,580 --> 00:02:55,260
fabric on ryzen 7000 is completely
77
00:02:53,239 --> 00:02:57,840
desynchronized from the memory
78
00:02:55,260 --> 00:02:59,519
controller because the memory controller
79
00:02:57,840 --> 00:03:01,800
like typically your memory controller
80
00:02:59,519 --> 00:03:03,660
might run at like say uh at like three
81
00:03:01,800 --> 00:03:07,200
gigahertz and your Infinity fabric at
82
00:03:03,660 --> 00:03:08,280
like two gigahertz there is no way to
83
00:03:07,200 --> 00:03:11,220
like
84
00:03:08,280 --> 00:03:14,340
cleanly transition data off of a three
85
00:03:11,220 --> 00:03:16,260
gigahertz like clock region into a two
86
00:03:14,340 --> 00:03:18,000
gigahertz clock region that just doesn't
87
00:03:16,260 --> 00:03:19,680
work right so that whole thing with
88
00:03:18,000 --> 00:03:21,300
ryzen 5000 where you want your Infinity
89
00:03:19,680 --> 00:03:23,280
fabric synchronized to the memory
90
00:03:21,300 --> 00:03:24,540
controller that works because your
91
00:03:23,280 --> 00:03:26,280
Infinity Fabric and the memory
92
00:03:24,540 --> 00:03:29,879
controller actually run at the same
93
00:03:26,280 --> 00:03:32,819
clock if the clock is like you know if
94
00:03:29,879 --> 00:03:35,580
what like three to two requires buffers
95
00:03:32,819 --> 00:03:37,379
in between the two clock regions and
96
00:03:35,580 --> 00:03:39,300
well once you're buffering the data the
97
00:03:37,379 --> 00:03:40,739
you you just have a latency penalty that
98
00:03:39,300 --> 00:03:43,379
that's just permanent that's like the
99
00:03:40,739 --> 00:03:45,120
whole reason that unlike ryzen 5000 the
100
00:03:43,379 --> 00:03:47,819
whole keeping your memory in Infinity
101
00:03:45,120 --> 00:03:49,200
fabric synchronized was a thing was if
102
00:03:47,819 --> 00:03:51,360
you did keep them synchronized you
103
00:03:49,200 --> 00:03:54,480
didn't have to buffer the data on ryzen
104
00:03:51,360 --> 00:03:56,280
7000 it always has to be buffered um
105
00:03:54,480 --> 00:03:58,739
even if you set your memory speed to
106
00:03:56,280 --> 00:04:01,440
like two gigahertz uh it still ends up
107
00:03:58,739 --> 00:04:03,420
getting buffered there's there's no uh
108
00:04:01,440 --> 00:04:05,640
like Infinity fabric to memory
109
00:04:03,420 --> 00:04:08,640
controller synchronization on ryzen 7000
110
00:04:05,640 --> 00:04:10,980
CPUs based on my testing so the good
111
00:04:08,640 --> 00:04:12,480
news about this is that if you just want
112
00:04:10,980 --> 00:04:14,220
to max out your performance you just
113
00:04:12,480 --> 00:04:16,739
push the infinity fabric clock as high
114
00:04:14,220 --> 00:04:19,680
as it goes the bad news about that is
115
00:04:16,739 --> 00:04:22,079
that the infinity fabric causes some fun
116
00:04:19,680 --> 00:04:24,240
instability like yes fun stability
117
00:04:22,079 --> 00:04:26,460
issues depending on how you look at it
118
00:04:24,240 --> 00:04:28,500
mainly in that the infinity fabric will
119
00:04:26,460 --> 00:04:29,880
absolutely not produce at least not in
120
00:04:28,500 --> 00:04:33,540
my experience it will not produce
121
00:04:29,880 --> 00:04:35,280
Hardware errors at all that thing though
122
00:04:33,540 --> 00:04:37,440
on ryzen 5000 if you push the infinity
123
00:04:35,280 --> 00:04:39,320
fabric too far on a lot of CPUs you'll
124
00:04:37,440 --> 00:04:41,940
just get a flood of errors in Windows
125
00:04:39,320 --> 00:04:43,979
you do not get that on ryzen 7000
126
00:04:41,940 --> 00:04:46,139
instead you get random performance
127
00:04:43,979 --> 00:04:48,180
regressions which are proportional to
128
00:04:46,139 --> 00:04:50,460
how heavy the test is and if you push
129
00:04:48,180 --> 00:04:53,160
the infinity fabric way too far it will
130
00:04:50,460 --> 00:04:55,740
actually out like right crash but
131
00:04:53,160 --> 00:04:57,840
basically the infinity fabric on on the
132
00:04:55,740 --> 00:04:59,340
7000 series does have some kind of error
133
00:04:57,840 --> 00:05:02,340
correction with re-transmission
134
00:04:59,340 --> 00:05:03,720
capabilities uh and so if you push it a
135
00:05:02,340 --> 00:05:05,820
little bit too far you start seeing
136
00:05:03,720 --> 00:05:07,259
performance regressions uh and if you
137
00:05:05,820 --> 00:05:09,120
push it way too far it does start
138
00:05:07,259 --> 00:05:11,460
causing outright crashing but the
139
00:05:09,120 --> 00:05:13,680
basically what that means is that you're
140
00:05:11,460 --> 00:05:15,540
going to want to run stress tests which
141
00:05:13,680 --> 00:05:17,340
actually have a performance metric which
142
00:05:15,540 --> 00:05:19,979
is why I use Lin pack for stress testing
143
00:05:17,340 --> 00:05:22,919
this because 2100 for this 7900x is
144
00:05:19,979 --> 00:05:24,720
actually really pushing it most CPUs or
145
00:05:22,919 --> 00:05:28,080
actually I'd go as far as saying that
146
00:05:24,720 --> 00:05:31,259
all CPUs should do at least 2066 on the
147
00:05:28,080 --> 00:05:33,060
infinity fabric but yeah once you start
148
00:05:31,259 --> 00:05:34,919
trying to go past that you'll find that
149
00:05:33,060 --> 00:05:38,340
some CPUs really don't want to run even
150
00:05:34,919 --> 00:05:40,380
like 2100 and this 7900x is one of those
151
00:05:38,340 --> 00:05:43,919
a lot of chips might be able to actually
152
00:05:40,380 --> 00:05:45,539
even do like 2133 but also it depends
153
00:05:43,919 --> 00:05:46,979
like what's really annoying with the
154
00:05:45,539 --> 00:05:49,500
infinity fabric is that it actually
155
00:05:46,979 --> 00:05:52,860
clocks better if your memory clock is
156
00:05:49,500 --> 00:05:54,720
worse so at like six ddr5 6000 you can
157
00:05:52,860 --> 00:05:58,440
generally bulk the infinity fabric a bit
158
00:05:54,720 --> 00:05:59,699
higher however the performance uh the
159
00:05:58,440 --> 00:06:01,440
overall performance is actually better
160
00:05:59,699 --> 00:06:03,180
if you push your memory clock first so
161
00:06:01,440 --> 00:06:04,919
you should first max out your memory
162
00:06:03,180 --> 00:06:06,720
while keeping the infinity fabric at
163
00:06:04,919 --> 00:06:07,860
like two gigahertz and then once that
164
00:06:06,720 --> 00:06:09,419
you've maxed out your memory clock
165
00:06:07,860 --> 00:06:11,160
that's when you start adjusting the
166
00:06:09,419 --> 00:06:13,740
infinity fabric doing it the other way
167
00:06:11,160 --> 00:06:16,440
around doesn't work you end up with less
168
00:06:13,740 --> 00:06:20,820
performance in the end so
169
00:06:16,440 --> 00:06:23,580
anyway uh yeah so 2100 uh I got that
170
00:06:20,820 --> 00:06:24,840
stable here linpak is running at the way
171
00:06:23,580 --> 00:06:26,220
it should be basically what you want to
172
00:06:24,840 --> 00:06:27,600
be doing is as you're raising your
173
00:06:26,220 --> 00:06:29,639
Infinity fabric clock you want to run
174
00:06:27,600 --> 00:06:31,259
Lin pack and check that the gigaflops
175
00:06:29,639 --> 00:06:33,240
aren't going down so if like two
176
00:06:31,259 --> 00:06:36,240
gigahertz you get
177
00:06:33,240 --> 00:06:39,479
um I don't know like 620 gigaflops and
178
00:06:36,240 --> 00:06:40,680
then at like 2033 you get 620 you know a
179
00:06:39,479 --> 00:06:43,020
little bit more than that it doesn't
180
00:06:40,680 --> 00:06:46,620
scale that much with the infinity fabric
181
00:06:43,020 --> 00:06:49,919
unfortunately linpak on a on ddr5 is
182
00:06:46,620 --> 00:06:52,680
actually uh like kind of latency limited
183
00:06:49,919 --> 00:06:54,960
because ddr5 just has so much memory
184
00:06:52,680 --> 00:06:56,759
bandwidth which like on ddr4 systems
185
00:06:54,960 --> 00:06:59,340
linpak was like a really memory
186
00:06:56,759 --> 00:07:00,780
bandwidth Limited Test and now it's
187
00:06:59,340 --> 00:07:02,100
really more limited by like memory
188
00:07:00,780 --> 00:07:04,500
timings which is kind of interesting
189
00:07:02,100 --> 00:07:06,180
that like ddr5 just has so much
190
00:07:04,500 --> 00:07:07,500
bandwidth available for for something
191
00:07:06,180 --> 00:07:09,539
like linpak
192
00:07:07,500 --> 00:07:11,759
that it like completely changes how the
193
00:07:09,539 --> 00:07:14,880
like performance scaling behaves but
194
00:07:11,759 --> 00:07:16,860
anyway so you'll you know see like small
195
00:07:14,880 --> 00:07:18,720
performance increases as you push the
196
00:07:16,860 --> 00:07:20,160
infinity Fabric and at some point you're
197
00:07:18,720 --> 00:07:21,780
either going to start seeing calculation
198
00:07:20,160 --> 00:07:23,039
errors or an actual reduction in
199
00:07:21,780 --> 00:07:24,960
performance and that basically means
200
00:07:23,039 --> 00:07:26,460
that the infinity fabric that like
201
00:07:24,960 --> 00:07:27,840
you've pushed the infinity fabric too
202
00:07:26,460 --> 00:07:29,759
far there are certain voltage
203
00:07:27,840 --> 00:07:30,780
adjustments you can make to try uh fix
204
00:07:29,759 --> 00:07:32,460
that
205
00:07:30,780 --> 00:07:33,900
um which is what I've done here and
206
00:07:32,460 --> 00:07:36,180
that's sort of one of the main things
207
00:07:33,900 --> 00:07:36,780
that I wanted to show in this video
208
00:07:36,180 --> 00:07:39,300
um
209
00:07:36,780 --> 00:07:40,560
but uh yeah if you're not gonna like if
210
00:07:39,300 --> 00:07:42,780
you don't want to fiddle with any of the
211
00:07:40,560 --> 00:07:45,120
voltages manually then uh or you don't
212
00:07:42,780 --> 00:07:48,060
want to push any of the voltages uh
213
00:07:45,120 --> 00:07:50,759
further than the sort of lazy option is
214
00:07:48,060 --> 00:07:52,860
to just stop at like 2066 because that
215
00:07:50,759 --> 00:07:55,680
generally doesn't run into any weird
216
00:07:52,860 --> 00:07:57,419
issues and also 2066 and 2100 perform
217
00:07:55,680 --> 00:07:58,800
kind of they're the same because they're
218
00:07:57,419 --> 00:08:01,380
not really that different from each
219
00:07:58,800 --> 00:08:03,060
other right like going from 2000 to 2100
220
00:08:01,380 --> 00:08:06,840
is a five percent clock speed increase
221
00:08:03,060 --> 00:08:09,060
going from 2066 to 2100 is like well
222
00:08:06,840 --> 00:08:11,759
less than five percent so
223
00:08:09,060 --> 00:08:13,259
uh you know you'll you'll have to weigh
224
00:08:11,759 --> 00:08:14,400
up how much effort you want to put into
225
00:08:13,259 --> 00:08:15,060
that yourself
226
00:08:14,400 --> 00:08:18,000
um
227
00:08:15,060 --> 00:08:19,379
anyway uh we're gonna take a look at the
228
00:08:18,000 --> 00:08:21,599
memory timings once we actually get into
229
00:08:19,379 --> 00:08:23,280
the BIOS I only have ryzen Master here
230
00:08:21,599 --> 00:08:25,020
for the clock readouts
231
00:08:23,280 --> 00:08:26,879
so let's take a look at all of the
232
00:08:25,020 --> 00:08:29,819
different stress tests I Ran So I did
233
00:08:26,879 --> 00:08:32,399
run linpack uh this is 77 Loops of Lin
234
00:08:29,819 --> 00:08:35,520
pack which 45 seconds per Loop works out
235
00:08:32,399 --> 00:08:38,279
to I think just under an hour I also ran
236
00:08:35,520 --> 00:08:40,200
over you know 5 000 seconds of Y
237
00:08:38,279 --> 00:08:41,940
cruncher so that's like over an hour
238
00:08:40,200 --> 00:08:43,620
like yeah that's well over an hour of Y
239
00:08:41,940 --> 00:08:45,420
cruncher why cruncher is actually really
240
00:08:43,620 --> 00:08:47,040
good at picking up on Infinity fabric
241
00:08:45,420 --> 00:08:49,380
instability
242
00:08:47,040 --> 00:08:52,200
um it will just outright crash which is
243
00:08:49,380 --> 00:08:54,120
kind of neat though I'm not 100 certain
244
00:08:52,200 --> 00:08:55,860
that it's completely consistent about
245
00:08:54,120 --> 00:08:58,620
the whole like outright crashing thing
246
00:08:55,860 --> 00:09:00,720
because some CPUs might be more tolerant
247
00:08:58,620 --> 00:09:03,360
to Infinity fabric instability than
248
00:09:00,720 --> 00:09:06,000
others so
249
00:09:03,360 --> 00:09:08,339
yeah anyway but why cruncher passed
250
00:09:06,000 --> 00:09:10,740
completely stable no problems there Lin
251
00:09:08,339 --> 00:09:12,779
pack is running at like the gigaflops it
252
00:09:10,740 --> 00:09:14,700
should be running at also no calculation
253
00:09:12,779 --> 00:09:16,140
errors whatsoever we don't have any
254
00:09:14,700 --> 00:09:17,940
errors down here though I really don't
255
00:09:16,140 --> 00:09:20,940
expect to see any errors even if Lin
256
00:09:17,940 --> 00:09:22,560
pack and white crencher were crashing uh
257
00:09:20,940 --> 00:09:24,779
just based on my experience with pushing
258
00:09:22,560 --> 00:09:26,940
the infinity fabric so far and I've also
259
00:09:24,779 --> 00:09:28,920
run three thousand percent of HCI mem
260
00:09:26,940 --> 00:09:30,779
test though the there's one thread that
261
00:09:28,920 --> 00:09:32,700
only ran to two thousand percent and
262
00:09:30,779 --> 00:09:35,880
this is a general issue with the Dual
263
00:09:32,700 --> 00:09:38,580
CCD 7000 series chips in my experience
264
00:09:35,880 --> 00:09:41,399
uh basically there is not enough memory
265
00:09:38,580 --> 00:09:43,980
bandwidth to go around and windows isn't
266
00:09:41,399 --> 00:09:45,720
good at balancing the threads so
267
00:09:43,980 --> 00:09:49,140
basically what's going on here is that
268
00:09:45,720 --> 00:09:51,000
like the first 22 uh 23 threads because
269
00:09:49,140 --> 00:09:52,920
we have thread zero over here right so
270
00:09:51,000 --> 00:09:55,080
they're started from zero so that's 23
271
00:09:52,920 --> 00:09:57,779
threads over here so the first 23
272
00:09:55,080 --> 00:09:59,279
threads basically have access to pretty
273
00:09:57,779 --> 00:10:01,440
much all of the memory bandwidth they
274
00:09:59,279 --> 00:10:03,839
need and this one's getting kind of
275
00:10:01,440 --> 00:10:05,700
neglected and that's why that one that
276
00:10:03,839 --> 00:10:07,260
one you know is only two thousand
277
00:10:05,700 --> 00:10:08,519
percent while a lot of the other ones
278
00:10:07,260 --> 00:10:09,899
are well over three thousand percent
279
00:10:08,519 --> 00:10:12,420
right like this one was only getting
280
00:10:09,899 --> 00:10:14,279
like 800 megabyte like this one was
281
00:10:12,420 --> 00:10:16,860
getting around 900 megabytes per second
282
00:10:14,279 --> 00:10:18,779
of memory bandwidth uh sort of provided
283
00:10:16,860 --> 00:10:20,640
to it while a lot of the other ones were
284
00:10:18,779 --> 00:10:21,899
running over a gigabyte of memory
285
00:10:20,640 --> 00:10:22,860
bandwidth
286
00:10:21,899 --> 00:10:24,959
um
287
00:10:22,860 --> 00:10:26,760
well over a gigabytes per second of
288
00:10:24,959 --> 00:10:28,620
memory bandwidth and the issue here is
289
00:10:26,760 --> 00:10:30,480
that there's just not enough memory
290
00:10:28,620 --> 00:10:32,339
bandwidth across the entire memory
291
00:10:30,480 --> 00:10:34,080
controller to feed all of the threads
292
00:10:32,339 --> 00:10:36,660
evenly at the same time and because
293
00:10:34,080 --> 00:10:38,160
Windows is a bit dumb you end up with
294
00:10:36,660 --> 00:10:42,060
one thread that's just lagging way
295
00:10:38,160 --> 00:10:44,519
behind this is even worse on a 7950x on
296
00:10:42,060 --> 00:10:46,680
a 7950x you'll have like eight threads
297
00:10:44,519 --> 00:10:49,680
that are just miles behind all of the
298
00:10:46,680 --> 00:10:50,940
others uh the 7900x is actually not the
299
00:10:49,680 --> 00:10:53,820
worst example because the thing with the
300
00:10:50,940 --> 00:10:56,399
7950x is like it has you have even more
301
00:10:53,820 --> 00:10:58,500
threads that pull the same amount of
302
00:10:56,399 --> 00:11:01,019
bandwidth and the basic issue with with
303
00:10:58,500 --> 00:11:02,940
ryzen 7000 is like the memory system
304
00:11:01,019 --> 00:11:05,279
here really doesn't actually have that
305
00:11:02,940 --> 00:11:06,839
much memory bandwidth you can really see
306
00:11:05,279 --> 00:11:08,880
that reflected in the like overall
307
00:11:06,839 --> 00:11:11,160
bandwidth readout over here it's only 30
308
00:11:08,880 --> 00:11:14,760
gigabytes per second like this is
309
00:11:11,160 --> 00:11:16,680
high-end ddr4 territory
310
00:11:14,760 --> 00:11:18,660
um now admittedly this is a memory you
311
00:11:16,680 --> 00:11:20,279
know like relatively sequential workload
312
00:11:18,660 --> 00:11:22,500
I would assume whereas something like
313
00:11:20,279 --> 00:11:25,380
linpak is very random which is why
314
00:11:22,500 --> 00:11:28,800
linpak like loves ddr5
315
00:11:25,380 --> 00:11:31,860
um but yeah uh and like Lin pack even on
316
00:11:28,800 --> 00:11:33,779
ryzen 7000 just loves ddr5 also it might
317
00:11:31,860 --> 00:11:36,839
be that ryzen 7000 just isn't that
318
00:11:33,779 --> 00:11:38,579
bandwidth heavy when running linpak but
319
00:11:36,839 --> 00:11:40,620
either way in mem test you can actually
320
00:11:38,579 --> 00:11:44,160
really see that like amd's memory system
321
00:11:40,620 --> 00:11:47,220
is is really not maximizing the memory
322
00:11:44,160 --> 00:11:49,620
bandwidth that's available from from uh
323
00:11:47,220 --> 00:11:51,300
ddr5 I would suspect that it's mainly
324
00:11:49,620 --> 00:11:53,700
because the infinity fabric is at like
325
00:11:51,300 --> 00:11:56,459
2.1 gigahertz while the memory is at
326
00:11:53,700 --> 00:11:58,740
like 3.1 gigahertz so you know you have
327
00:11:56,459 --> 00:12:02,640
this really fast memory hooked up to a
328
00:11:58,740 --> 00:12:05,640
not that fast CPU interconnect
329
00:12:02,640 --> 00:12:07,800
um so yeah there's like with linpack
330
00:12:05,640 --> 00:12:09,899
that's not a not really an issue it
331
00:12:07,800 --> 00:12:11,459
doesn't use as much bandwidth as memtest
332
00:12:09,899 --> 00:12:14,640
does but mem test uses a ton of
333
00:12:11,459 --> 00:12:17,100
bandwidth and so it just kind of gets
334
00:12:14,640 --> 00:12:19,380
bottlenecked on the infinity fabric from
335
00:12:17,100 --> 00:12:22,140
but based on how it behaves
336
00:12:19,380 --> 00:12:24,720
anyway here's the uh
337
00:12:22,140 --> 00:12:26,579
Ida uh performance results 55
338
00:12:24,720 --> 00:12:29,220
nanoseconds of Licensing which is pretty
339
00:12:26,579 --> 00:12:31,019
solid and then I don't think these
340
00:12:29,220 --> 00:12:32,820
bandwidth readouts are at all accurate
341
00:12:31,019 --> 00:12:35,100
but they're there
342
00:12:32,820 --> 00:12:36,660
um and uh then for a more useful
343
00:12:35,100 --> 00:12:38,100
Benchmark or at least in my opinion a
344
00:12:36,660 --> 00:12:42,300
more useful Benchmark we have Pi Prime
345
00:12:38,100 --> 00:12:45,720
2.0 over here uh 9.64 uh
346
00:12:42,300 --> 00:12:47,459
yeah 9.649 seconds now the thing with pi
347
00:12:45,720 --> 00:12:50,160
Prime is it does scale a little bit with
348
00:12:47,459 --> 00:12:52,019
CPU clock and I'm obviously on a 7900x
349
00:12:50,160 --> 00:12:54,600
so this result would not necessarily be
350
00:12:52,019 --> 00:12:57,420
comparable for say lower clocked CPUs
351
00:12:54,600 --> 00:13:00,060
like the 7600x but if you have a 7950x
352
00:12:57,420 --> 00:13:02,459
or something uh you should be like sub
353
00:13:00,060 --> 00:13:03,899
10 seconds is like the goal as far as
354
00:13:02,459 --> 00:13:06,240
I'm concerned if you're getting over 10
355
00:13:03,899 --> 00:13:08,820
seconds in pi Prime on a stock 7000
356
00:13:06,240 --> 00:13:10,620
series with overclocked memory you're
357
00:13:08,820 --> 00:13:12,360
over you could you could overclock the
358
00:13:10,620 --> 00:13:13,200
memory harder
359
00:13:12,360 --> 00:13:15,779
um
360
00:13:13,200 --> 00:13:17,940
also even lower results are possible and
361
00:13:15,779 --> 00:13:19,860
x3d CPUs actually I think run Pi Prime
362
00:13:17,940 --> 00:13:22,860
even faster because this this Benchmark
363
00:13:19,860 --> 00:13:26,160
is like crazy memory latency limited and
364
00:13:22,860 --> 00:13:27,060
so it also benefits a lot from L3 cache
365
00:13:26,160 --> 00:13:28,500
um
366
00:13:27,060 --> 00:13:31,019
so
367
00:13:28,500 --> 00:13:33,060
anyway uh oh no I did the I did the
368
00:13:31,019 --> 00:13:36,480
windows thing uh that's what I get for
369
00:13:33,060 --> 00:13:38,700
using a new OS install but anyway uh so
370
00:13:36,480 --> 00:13:40,440
yeah all of the stress tests passed I
371
00:13:38,700 --> 00:13:43,740
didn't bother running test mem 5 just
372
00:13:40,440 --> 00:13:46,200
because it's so Jank and it's so clunky
373
00:13:43,740 --> 00:13:50,160
it's a good stress test it's a good free
374
00:13:46,200 --> 00:13:52,860
stress test it's just so clunky that I
375
00:13:50,160 --> 00:13:53,700
don't like using it
376
00:13:52,860 --> 00:13:56,639
um
377
00:13:53,700 --> 00:13:57,779
anyway so yeah that's that's all of the
378
00:13:56,639 --> 00:13:58,980
stress has passed I mean at this point
379
00:13:57,779 --> 00:14:00,300
I'm going to be restarting the system
380
00:13:58,980 --> 00:14:02,040
anyway so
381
00:14:00,300 --> 00:14:05,519
um also that means I have now committed
382
00:14:02,040 --> 00:14:07,260
to this this take of this video so
383
00:14:05,519 --> 00:14:08,760
yeah actually I don't know why I'm
384
00:14:07,260 --> 00:14:10,980
manually closing everything down I
385
00:14:08,760 --> 00:14:13,440
should just hit restart
386
00:14:10,980 --> 00:14:15,899
now I do have the memory uh context
387
00:14:13,440 --> 00:14:18,300
restore disabled so this is going to
388
00:14:15,899 --> 00:14:20,519
take a while to post
389
00:14:18,300 --> 00:14:23,100
um because the memory context uh restore
390
00:14:20,519 --> 00:14:24,959
with over like with yeah if you're
391
00:14:23,100 --> 00:14:27,120
manually overclocking it can cause some
392
00:14:24,959 --> 00:14:28,860
stability issues
393
00:14:27,120 --> 00:14:30,060
um and it depends it kind of depends on
394
00:14:28,860 --> 00:14:32,160
the motherboard as well like some
395
00:14:30,060 --> 00:14:34,320
motherboards are less bad with the
396
00:14:32,160 --> 00:14:35,579
memory context restore enabled but on
397
00:14:34,320 --> 00:14:37,860
some other boards it's really bad where
398
00:14:35,579 --> 00:14:39,540
if you have it enabled for like if you
399
00:14:37,860 --> 00:14:41,220
have it enabled after a couple restarts
400
00:14:39,540 --> 00:14:42,720
you might find that your memory is no
401
00:14:41,220 --> 00:14:44,639
longer stable
402
00:14:42,720 --> 00:14:47,720
um because during every boot like if
403
00:14:44,639 --> 00:14:49,680
something slightly changes on the system
404
00:14:47,720 --> 00:14:52,920
uh you want the memory control
405
00:14:49,680 --> 00:14:56,519
controller to basically retrain the uh
406
00:14:52,920 --> 00:14:59,100
memory also please oh I missed the BIOS
407
00:14:56,519 --> 00:15:01,019
it's actually not that bad a boot up
408
00:14:59,100 --> 00:15:02,760
like considering I have memory context
409
00:15:01,019 --> 00:15:06,380
restore disabled that really didn't take
410
00:15:02,760 --> 00:15:06,380
that long but uh
411
00:15:06,480 --> 00:15:10,440
yeah which is how how also like the
412
00:15:09,060 --> 00:15:11,880
capture card doesn't help because the
413
00:15:10,440 --> 00:15:13,980
the system's actually like the capture
414
00:15:11,880 --> 00:15:18,139
card actually like lags behind the
415
00:15:13,980 --> 00:15:18,139
system like initializing the GPU
416
00:15:19,560 --> 00:15:23,720
so I'm just gonna have to mash the
417
00:15:21,120 --> 00:15:23,720
delete button
418
00:15:45,480 --> 00:15:49,339
now there we go and it's up
419
00:15:50,820 --> 00:15:55,260
yeah but still I wouldn't consider this
420
00:15:53,160 --> 00:15:57,060
an egregiously long post time
421
00:15:55,260 --> 00:15:59,160
considering that I have memory context
422
00:15:57,060 --> 00:16:00,899
restore uh disabled over here I do
423
00:15:59,160 --> 00:16:03,300
appreciate the fact that ASRock put this
424
00:16:00,899 --> 00:16:05,040
into the OC tweaker menu now then here
425
00:16:03,300 --> 00:16:07,500
you can see my settings I am pushing the
426
00:16:05,040 --> 00:16:09,720
memory voltage relatively like this is
427
00:16:07,500 --> 00:16:13,680
about as high as I would go for daily uh
428
00:16:09,720 --> 00:16:16,260
for the vdd voltage now vddq and vddio I
429
00:16:13,680 --> 00:16:19,320
have these at 1.25 volts the reason for
430
00:16:16,260 --> 00:16:20,699
this is the vddq voltage and the vddio
431
00:16:19,320 --> 00:16:24,000
these are basically the signaling
432
00:16:20,699 --> 00:16:26,220
voltages so this you can sort of so dram
433
00:16:24,000 --> 00:16:29,040
vdd voltage is like the voltage that
434
00:16:26,220 --> 00:16:32,699
powers the internals of the memory chips
435
00:16:29,040 --> 00:16:34,560
uh vddq voltage Powers the input output
436
00:16:32,699 --> 00:16:35,940
portion of the memory chip so this this
437
00:16:34,560 --> 00:16:38,399
is the voltage that's like actually used
438
00:16:35,940 --> 00:16:42,360
for talking to the cpu's memory
439
00:16:38,399 --> 00:16:44,339
controller and the vddio mem S3 over
440
00:16:42,360 --> 00:16:48,180
here that's the same deal but from the
441
00:16:44,339 --> 00:16:50,459
CPU side so the CPU uses vddio mem to
442
00:16:48,180 --> 00:16:53,600
talk to the memory and the memory uses
443
00:16:50,459 --> 00:16:56,880
video ddq to talk to the CPU
444
00:16:53,600 --> 00:16:58,500
and in a lot of cases what you will find
445
00:16:56,880 --> 00:17:01,380
is that
446
00:16:58,500 --> 00:17:02,820
um running these low cons in well
447
00:17:01,380 --> 00:17:04,439
depends on the motherboard and the BIOS
448
00:17:02,820 --> 00:17:06,480
version but running these lower can
449
00:17:04,439 --> 00:17:08,459
sometimes help with stability
450
00:17:06,480 --> 00:17:10,620
just because driving the signals with
451
00:17:08,459 --> 00:17:12,959
higher voltage doesn't necessarily mean
452
00:17:10,620 --> 00:17:16,980
that the signals are going to be easier
453
00:17:12,959 --> 00:17:19,319
to read on both ends so
454
00:17:16,980 --> 00:17:22,439
um yeah now you can also just use the
455
00:17:19,319 --> 00:17:24,419
the VD dvdq equals like the all of them
456
00:17:22,439 --> 00:17:26,280
being equal mode
457
00:17:24,419 --> 00:17:28,199
um I'm not entirely certain that what
458
00:17:26,280 --> 00:17:31,080
I've done here is actually necessary but
459
00:17:28,199 --> 00:17:34,020
it works and I'm like the the thing is
460
00:17:31,080 --> 00:17:35,100
uh depending on the motherboard and the
461
00:17:34,020 --> 00:17:36,179
settings you're running you might find
462
00:17:35,100 --> 00:17:37,980
that there's actually a whole bunch of
463
00:17:36,179 --> 00:17:39,299
different voltage combinations that do
464
00:17:37,980 --> 00:17:40,919
work
465
00:17:39,299 --> 00:17:42,600
um because with say this like on some
466
00:17:40,919 --> 00:17:45,059
other boards having high signaling
467
00:17:42,600 --> 00:17:46,980
voltages is just fine right so like
468
00:17:45,059 --> 00:17:49,500
running everything at 1.42 volts might
469
00:17:46,980 --> 00:17:52,860
have actually worked just fine
470
00:17:49,500 --> 00:17:54,960
um but uh on say I think they're like
471
00:17:52,860 --> 00:17:56,340
there's a specific set of bioses for
472
00:17:54,960 --> 00:17:58,799
like gigabyte boards where you actually
473
00:17:56,340 --> 00:18:00,840
want your signaling voltages at 1.25
474
00:17:58,799 --> 00:18:02,880
volts because above that it just causes
475
00:18:00,840 --> 00:18:04,500
issues with memory training atheroc
476
00:18:02,880 --> 00:18:05,880
doesn't like this the BIOS that I have
477
00:18:04,500 --> 00:18:07,380
on this board doesn't seem to suffer
478
00:18:05,880 --> 00:18:08,220
from that
479
00:18:07,380 --> 00:18:11,160
um
480
00:18:08,220 --> 00:18:13,320
but uh wait where does it say well bile
481
00:18:11,160 --> 00:18:15,600
swirl oh I guess on Main yeah so we're
482
00:18:13,320 --> 00:18:17,520
on the 1.21 bios also this bios does
483
00:18:15,600 --> 00:18:20,039
have the 1.3 volt SOC voltage
484
00:18:17,520 --> 00:18:22,020
restriction which uh uh well you can see
485
00:18:20,039 --> 00:18:24,120
that sort of sort of down here it's just
486
00:18:22,020 --> 00:18:26,160
like yeah uh also yeah I have that
487
00:18:24,120 --> 00:18:29,220
completely maxed out because the cpu's
488
00:18:26,160 --> 00:18:30,059
memory controller is not great
489
00:18:29,220 --> 00:18:31,500
um
490
00:18:30,059 --> 00:18:33,660
but anyway
491
00:18:31,500 --> 00:18:35,880
um yeah so the like
492
00:18:33,660 --> 00:18:37,860
so the vddio end of this this is
493
00:18:35,880 --> 00:18:40,980
completely safe for the CPU
494
00:18:37,860 --> 00:18:42,780
um right the memory vdd voltage that's
495
00:18:40,980 --> 00:18:45,240
not a problem for the memory sticks
496
00:18:42,780 --> 00:18:46,799
themselves the vddq voltage also in and
497
00:18:45,240 --> 00:18:49,200
of itself wouldn't be a problem for the
498
00:18:46,799 --> 00:18:51,059
memory stick I do have a slight concern
499
00:18:49,200 --> 00:18:52,860
because according to the ddr5
500
00:18:51,059 --> 00:18:55,320
documentation these two voltages should
501
00:18:52,860 --> 00:18:57,840
be kept relatively close together
502
00:18:55,320 --> 00:18:59,760
but on the other hand I've also seen a
503
00:18:57,840 --> 00:19:03,179
lot of people you know running at least
504
00:18:59,760 --> 00:19:04,860
High Nix chips with very big gaps in
505
00:19:03,179 --> 00:19:06,539
between these two voltages because the
506
00:19:04,860 --> 00:19:08,280
vdd voltage is sort of what really
507
00:19:06,539 --> 00:19:10,980
allows you to push your memory timings
508
00:19:08,280 --> 00:19:12,600
down whereas the vddq voltages of
509
00:19:10,980 --> 00:19:14,940
voltage you basically want to set up
510
00:19:12,600 --> 00:19:17,039
like set optimally for the like
511
00:19:14,940 --> 00:19:20,400
signaling behavior of the CPU and the
512
00:19:17,039 --> 00:19:22,380
and the motherboard that you're using so
513
00:19:20,400 --> 00:19:24,360
yeah this one sort of affects your
514
00:19:22,380 --> 00:19:26,160
timings and this one really affects your
515
00:19:24,360 --> 00:19:27,900
your clocks basically is the way you can
516
00:19:26,160 --> 00:19:30,059
think about it
517
00:19:27,900 --> 00:19:31,860
um and they don't necessarily like and
518
00:19:30,059 --> 00:19:33,539
you won't necessarily get the best
519
00:19:31,860 --> 00:19:37,679
results by having them both at the same
520
00:19:33,539 --> 00:19:39,780
value so uh yeah I wouldn't be too
521
00:19:37,679 --> 00:19:42,720
concerned about this as it's set up
522
00:19:39,780 --> 00:19:45,299
right here like yeah I I would I would
523
00:19:42,720 --> 00:19:47,640
I'd be willing to daily this if you have
524
00:19:45,299 --> 00:19:48,900
doubts about doing this then ASRock does
525
00:19:47,640 --> 00:19:51,000
have the option to just have all of the
526
00:19:48,900 --> 00:19:54,299
voltages set equal to each other in
527
00:19:51,000 --> 00:19:56,580
which case uh I mean I don't think 1.4
528
00:19:54,299 --> 00:20:00,179
volts vddio would be dangerous for the
529
00:19:56,580 --> 00:20:02,520
CPU as there are uh there are you know
530
00:20:00,179 --> 00:20:05,340
Expo memory kits with the 1.4 volts as
531
00:20:02,520 --> 00:20:08,220
their Expo voltage I would assume AMD
532
00:20:05,340 --> 00:20:11,880
wouldn't certify those if 1.4 volts of
533
00:20:08,220 --> 00:20:13,679
eddio was actually a problem so
534
00:20:11,880 --> 00:20:16,140
though there are some motherboards that
535
00:20:13,679 --> 00:20:18,600
actually default your vddio and vddq to
536
00:20:16,140 --> 00:20:21,720
1.25 volts even if you're like vdd
537
00:20:18,600 --> 00:20:24,840
voltage is at 1.4 so
538
00:20:21,720 --> 00:20:26,760
yeah this this is the thing is just
539
00:20:24,840 --> 00:20:29,160
like
540
00:20:26,760 --> 00:20:31,320
I don't have like there's not enough
541
00:20:29,160 --> 00:20:33,419
data on this like I don't work at SK
542
00:20:31,320 --> 00:20:36,480
High Nix I don't know how the memory
543
00:20:33,419 --> 00:20:38,039
chips behave internally at this from all
544
00:20:36,480 --> 00:20:39,660
everything I've seen and my own
545
00:20:38,039 --> 00:20:42,660
experience this shouldn't cause any
546
00:20:39,660 --> 00:20:44,100
issues but I can't say like with a
547
00:20:42,660 --> 00:20:45,900
hundred percent certainty that this
548
00:20:44,100 --> 00:20:47,160
won't cause any issues because I just
549
00:20:45,900 --> 00:20:48,000
don't know
550
00:20:47,160 --> 00:20:51,059
um
551
00:20:48,000 --> 00:20:52,740
anyway dram VPP voltage is at 1.8 volts
552
00:20:51,059 --> 00:20:54,720
which is the stock voltage for the VPP
553
00:20:52,740 --> 00:20:56,820
voltage it should be at 1.8 volts that's
554
00:20:54,720 --> 00:20:58,620
that's what it defaults to
555
00:20:56,820 --> 00:21:01,919
um anyway then I have the infinity
556
00:20:58,620 --> 00:21:04,799
fabric set to 2100 uclk equals mem CLK
557
00:21:01,919 --> 00:21:06,299
uh also speaking of UCL k equals mclk
558
00:21:04,799 --> 00:21:09,179
this is a really annoying thing about
559
00:21:06,299 --> 00:21:11,940
the azroc file so if we set ddr56000 you
560
00:21:09,179 --> 00:21:14,760
can see that uclk mclk is is the same if
561
00:21:11,940 --> 00:21:17,220
we set 6200 it automatically divides
562
00:21:14,760 --> 00:21:17,760
your mem CLK by two
563
00:21:17,220 --> 00:21:20,580
um
564
00:21:17,760 --> 00:21:22,140
this I think is mainly a mainly like
565
00:21:20,580 --> 00:21:24,240
ASRock is mainly doing this because if
566
00:21:22,140 --> 00:21:28,140
you have an XMP memory kit that's like
567
00:21:24,240 --> 00:21:30,840
6200 or 6400 it's more likely to work if
568
00:21:28,140 --> 00:21:33,720
they down clock the memory controller
569
00:21:30,840 --> 00:21:36,179
the obvious issue with this is like well
570
00:21:33,720 --> 00:21:37,679
if I set my memory clock myself I have
571
00:21:36,179 --> 00:21:39,360
every basically every time you adjust
572
00:21:37,679 --> 00:21:41,760
your memory clock upwards you need to
573
00:21:39,360 --> 00:21:44,280
manually set your memory controller to
574
00:21:41,760 --> 00:21:45,419
memory clock ratio which is kind of
575
00:21:44,280 --> 00:21:47,539
annoying
576
00:21:45,419 --> 00:21:47,539
um
577
00:21:48,299 --> 00:21:52,320
yeah anyway below that we have SOC
578
00:21:50,340 --> 00:21:53,820
voltage which I have at 1.3 volts which
579
00:21:52,320 --> 00:21:55,620
is as high as it goes on this BIOS
580
00:21:53,820 --> 00:21:58,080
version right I can lower it
581
00:21:55,620 --> 00:22:00,600
I cannot raise it I do appreciate how
582
00:21:58,080 --> 00:22:02,340
quick asrock's bios refreshes like there
583
00:22:00,600 --> 00:22:04,860
are some bioses that are really laggy
584
00:22:02,340 --> 00:22:05,700
convert to this so this is quite nice
585
00:22:04,860 --> 00:22:07,799
um
586
00:22:05,700 --> 00:22:09,360
anyway then we have vdd misc voltage
587
00:22:07,799 --> 00:22:10,799
which is at 1.1 volts and that's the
588
00:22:09,360 --> 00:22:12,900
actual default voltage for that I have
589
00:22:10,799 --> 00:22:15,120
not raised that at all
590
00:22:12,900 --> 00:22:17,400
um or I do think I've manually like set
591
00:22:15,120 --> 00:22:19,500
it to 1.1 volts but this is what it like
592
00:22:17,400 --> 00:22:22,380
defaults to on every single motherboard
593
00:22:19,500 --> 00:22:25,020
even on Full Auto settings this voltage
594
00:22:22,380 --> 00:22:28,080
is used to basically power the voltage
595
00:22:25,020 --> 00:22:30,600
Regulators that generate the vddg CCD
596
00:22:28,080 --> 00:22:32,700
voltage as well as the vddg IOD voltage
597
00:22:30,600 --> 00:22:36,419
now these two voltages these affect your
598
00:22:32,700 --> 00:22:39,000
Infinity fabric clock stability
599
00:22:36,419 --> 00:22:41,400
um most boards once you enable an expo
600
00:22:39,000 --> 00:22:44,100
profile will default these to 0.95 volts
601
00:22:41,400 --> 00:22:47,520
I think at all like bone stock I think
602
00:22:44,100 --> 00:22:48,840
is like 0.85 at like ddr5 5200 or
603
00:22:47,520 --> 00:22:49,620
whatever
604
00:22:48,840 --> 00:22:52,200
um
605
00:22:49,620 --> 00:22:54,120
but what I found with some CPUs if
606
00:22:52,200 --> 00:22:55,500
you're pushing the infinity fabric like
607
00:22:54,120 --> 00:22:58,500
basically like if you're trying to max
608
00:22:55,500 --> 00:23:00,480
out the infinity fabric raising the CCD
609
00:22:58,500 --> 00:23:02,460
voltage can help with this uh can help
610
00:23:00,480 --> 00:23:04,679
with stability so that's why I have the
611
00:23:02,460 --> 00:23:07,380
CCD voltage at 1.05 volts instead of
612
00:23:04,679 --> 00:23:09,240
just 0.95 because usually you would just
613
00:23:07,380 --> 00:23:11,760
use 0.95 because that's just sort of
614
00:23:09,240 --> 00:23:13,620
what it tends to default to
615
00:23:11,760 --> 00:23:16,860
um but yeah I've ended up bumping this
616
00:23:13,620 --> 00:23:19,140
up to 1.05 one thing I did run into with
617
00:23:16,860 --> 00:23:20,700
my 7600x and this was on a different
618
00:23:19,140 --> 00:23:23,460
motherboard so it might be tied to the
619
00:23:20,700 --> 00:23:26,700
motherboard to some extent is raising
620
00:23:23,460 --> 00:23:29,340
the vddg voltages past one volt can
621
00:23:26,700 --> 00:23:31,200
cause weird stability issues at idle now
622
00:23:29,340 --> 00:23:33,360
the system that we're looking at here
623
00:23:31,200 --> 00:23:35,520
was literally idling for over a hundred
624
00:23:33,360 --> 00:23:38,580
hours at this point without any crashes
625
00:23:35,520 --> 00:23:40,740
or anything so this seems to be working
626
00:23:38,580 --> 00:23:41,940
at least on whatever BIOS version this
627
00:23:40,740 --> 00:23:44,580
is
628
00:23:41,940 --> 00:23:46,440
um but uh yeah so that that is something
629
00:23:44,580 --> 00:23:48,240
to watch out for if you're messing with
630
00:23:46,440 --> 00:23:50,480
the vddg voltages and especially if
631
00:23:48,240 --> 00:23:52,980
you're pushing them like past one volt
632
00:23:50,480 --> 00:23:54,780
uh you might find that like your
633
00:23:52,980 --> 00:23:56,760
stability in like Lin pack and why
634
00:23:54,780 --> 00:23:59,580
cruncher VST gets better but then the
635
00:23:56,760 --> 00:24:02,760
system like just resets itself at idle
636
00:23:59,580 --> 00:24:04,740
which is super weird I have no idea why
637
00:24:02,760 --> 00:24:07,620
that happens but it is something that
638
00:24:04,740 --> 00:24:09,480
I've run into with a 7600x when pushing
639
00:24:07,620 --> 00:24:11,220
these voltages for the same reason in
640
00:24:09,480 --> 00:24:13,799
order to try get like uh I think I was
641
00:24:11,220 --> 00:24:15,539
trying to get 2133 or 2166 Infinity
642
00:24:13,799 --> 00:24:16,919
fabric to work
643
00:24:15,539 --> 00:24:18,360
um and that's actually the CPU where I
644
00:24:16,919 --> 00:24:20,039
first discovered that hey raising the
645
00:24:18,360 --> 00:24:22,679
CCD voltage actually kind of helps
646
00:24:20,039 --> 00:24:24,960
stability at least under load it didn't
647
00:24:22,679 --> 00:24:26,820
help but there was the the idle issues
648
00:24:24,960 --> 00:24:29,940
anyway with with this chip on this board
649
00:24:26,820 --> 00:24:31,500
this actually works for getting 2100 to
650
00:24:29,940 --> 00:24:33,659
to be stable
651
00:24:31,500 --> 00:24:35,880
so that's pretty cool then we have the
652
00:24:33,659 --> 00:24:39,059
vddp voltage this voltage is used for
653
00:24:35,880 --> 00:24:42,600
actually uh powering the Phi of the
654
00:24:39,059 --> 00:24:46,340
memory controller uh
655
00:24:42,600 --> 00:24:46,340
ddr5 bus signaling
656
00:24:48,900 --> 00:24:54,380
item in it
657
00:24:51,179 --> 00:24:54,380
that doesn't make any sense
658
00:24:55,200 --> 00:25:00,840
oh right okay well the just yeah that's
659
00:24:58,380 --> 00:25:02,640
what I figured
660
00:25:00,840 --> 00:25:05,220
anyway so yeah if you read the
661
00:25:02,640 --> 00:25:07,320
descriptions for these you have vddp is
662
00:25:05,220 --> 00:25:08,820
a voltage for the ddr5 bus signaling the
663
00:25:07,320 --> 00:25:10,620
the Phi so that's the actual physical
664
00:25:08,820 --> 00:25:12,840
interface to the memory chips but if you
665
00:25:10,620 --> 00:25:15,059
read the vddio voltage it says the same
666
00:25:12,840 --> 00:25:18,299
basically the same ddr5 bus signaling
667
00:25:15,059 --> 00:25:21,140
five uh yeah
668
00:25:18,299 --> 00:25:21,140
um so that's fun
669
00:25:25,980 --> 00:25:30,679
all right why would vddi have to be
670
00:25:27,960 --> 00:25:34,159
greater than
671
00:25:30,679 --> 00:25:34,159
SOC voltage
672
00:25:38,400 --> 00:25:43,580
like I just noticed that they have that
673
00:25:40,380 --> 00:25:43,580
rule there but it's like
674
00:25:48,120 --> 00:25:52,400
is it a reverse biasing issue
675
00:25:52,500 --> 00:25:55,980
because if that was a reverse biasing
676
00:25:54,120 --> 00:25:57,659
issue I'd expect the CPU to be dead at
677
00:25:55,980 --> 00:25:59,720
this point
678
00:25:57,659 --> 00:25:59,720
um
679
00:26:06,419 --> 00:26:10,860
also that just doesn't make a ton like
680
00:26:08,400 --> 00:26:14,720
there's like some relatively low voltage
681
00:26:10,860 --> 00:26:14,720
high speed memory kits out there
682
00:26:17,700 --> 00:26:25,340
not sure how like I'm pretty sure you
683
00:26:20,460 --> 00:26:25,340
can find like 1.2 volt ddr5 6000 kits
684
00:26:26,400 --> 00:26:30,779
I I can't imagine that the board would
685
00:26:28,380 --> 00:26:33,260
default that to like 1.35 volts on that
686
00:26:30,779 --> 00:26:33,260
one of those
687
00:26:33,480 --> 00:26:35,600
um
688
00:26:35,940 --> 00:26:39,779
man it would be really cool if AMD like
689
00:26:37,919 --> 00:26:42,120
provided actual public documentation
690
00:26:39,779 --> 00:26:47,480
like Intel does
691
00:26:42,120 --> 00:26:47,480
um except they don't so uh yeah
692
00:27:02,400 --> 00:27:09,000
man I don't want to reshoot this video
693
00:27:06,779 --> 00:27:12,020
because like this works
694
00:27:09,000 --> 00:27:12,020
this is stable
695
00:27:12,059 --> 00:27:14,179
um
696
00:27:14,580 --> 00:27:18,240
I really wonder what that like SOC
697
00:27:16,679 --> 00:27:20,600
voltage rule is supposed to be about
698
00:27:18,240 --> 00:27:20,600
though
699
00:27:21,240 --> 00:27:26,480
like that vddio has to be above SOC
700
00:27:24,000 --> 00:27:26,480
voltage
701
00:27:30,659 --> 00:27:36,059
wait I can't be right because stock so
702
00:27:33,779 --> 00:27:39,360
oh no because stock SOC voltage is one
703
00:27:36,059 --> 00:27:42,380
volt and if you use a jdeck memory kit
704
00:27:39,360 --> 00:27:42,380
you'll be at 1.1
705
00:27:49,200 --> 00:27:52,760
I really wonder about that
706
00:27:55,260 --> 00:28:01,679
that is such a weird way to Def like
707
00:27:59,159 --> 00:28:03,779
I I really wish I had a like and you
708
00:28:01,679 --> 00:28:05,640
know what's really annoying AMD back in
709
00:28:03,779 --> 00:28:07,559
the am3 plus days you know when they
710
00:28:05,640 --> 00:28:10,700
made CPUs that weren't competitive with
711
00:28:07,559 --> 00:28:10,700
apps like anything
712
00:28:10,740 --> 00:28:17,760
they published like full-on overclocking
713
00:28:13,500 --> 00:28:19,380
like guides for their CPUs I guess to
714
00:28:17,760 --> 00:28:21,659
try compensate for the fact that they
715
00:28:19,380 --> 00:28:23,460
were really slow out of the box and
716
00:28:21,659 --> 00:28:25,980
those guides would include a bunch of
717
00:28:23,460 --> 00:28:29,159
voltage recommendations
718
00:28:25,980 --> 00:28:31,860
AMD hasn't done that since ryzen 7 uh
719
00:28:29,159 --> 00:28:34,260
since ryzen 1000.
720
00:28:31,860 --> 00:28:36,600
um yeah they kind of did that for ryzen
721
00:28:34,260 --> 00:28:38,580
1000 but since like ryzen 1000 I haven't
722
00:28:36,600 --> 00:28:41,640
seen them like do anything similar to
723
00:28:38,580 --> 00:28:44,159
the the old like FX processor tuning
724
00:28:41,640 --> 00:28:45,600
guides that they had which is really
725
00:28:44,159 --> 00:28:49,039
quite annoying because it did include a
726
00:28:45,600 --> 00:28:51,720
bunch of like uh voltage setting uh
727
00:28:49,039 --> 00:28:54,179
recommendations and I'm I'm looking at
728
00:28:51,720 --> 00:28:56,279
this wondering like wait like is this a
729
00:28:54,179 --> 00:28:58,679
stability concern or is this like a
730
00:28:56,279 --> 00:29:00,299
longevity concern because like if it's a
731
00:28:58,679 --> 00:29:02,880
longevity concern at this point I'd
732
00:29:00,299 --> 00:29:04,620
expect the CPU to be dead
733
00:29:02,880 --> 00:29:06,480
um
734
00:29:04,620 --> 00:29:08,640
if it's a stability concern well
735
00:29:06,480 --> 00:29:11,340
evidently that's not true
736
00:29:08,640 --> 00:29:13,460
because this is stable
737
00:29:11,340 --> 00:29:13,460
um
738
00:29:14,100 --> 00:29:18,000
so anyway that's fun
739
00:29:16,140 --> 00:29:19,740
um
740
00:29:18,000 --> 00:29:21,659
yeah honestly I think I'm just going to
741
00:29:19,740 --> 00:29:23,659
leave this video up as is
742
00:29:21,659 --> 00:29:23,659
um
743
00:29:25,799 --> 00:29:31,440
like I'm not like yeah no screw it like
744
00:29:29,399 --> 00:29:33,659
if you if you you know if you want to
745
00:29:31,440 --> 00:29:34,440
set this voltage higher like that's up
746
00:29:33,659 --> 00:29:37,860
to you
747
00:29:34,440 --> 00:29:39,179
um I'm gonna leave it at 1.25 volts
748
00:29:37,860 --> 00:29:41,220
um
749
00:29:39,179 --> 00:29:43,320
honest and I would do that even with my
750
00:29:41,220 --> 00:29:45,419
7800 X 3D
751
00:29:43,320 --> 00:29:47,700
um there uh like I I know because like
752
00:29:45,419 --> 00:29:48,960
this is a CPU provided by AMD so it's
753
00:29:47,700 --> 00:29:50,460
like oh builds what you're just saying
754
00:29:48,960 --> 00:29:53,580
that because the CP was in Texas the
755
00:29:50,460 --> 00:29:55,860
risk no if like when I set up my well
756
00:29:53,580 --> 00:29:57,840
the thing is with my 7800x3d I probably
757
00:29:55,860 --> 00:30:00,120
wouldn't want a daily it at 1.3 volts
758
00:29:57,840 --> 00:30:03,000
SOC
759
00:30:00,120 --> 00:30:05,279
um but I'd probably daily it at 1.25 and
760
00:30:03,000 --> 00:30:07,320
I would probably still run my vddio at
761
00:30:05,279 --> 00:30:08,700
1.25 because there's a pretty decent
762
00:30:07,320 --> 00:30:11,220
chance I'm going to end up dailying a
763
00:30:08,700 --> 00:30:14,640
gigabyte motherboard at this right
764
00:30:11,220 --> 00:30:16,140
um just because well yeah it's like yeah
765
00:30:14,640 --> 00:30:17,940
there's a decent chance I'll be dealing
766
00:30:16,140 --> 00:30:20,399
a gigabyte motherboard like I haven't
767
00:30:17,940 --> 00:30:23,039
settled on what motherboard I eventually
768
00:30:20,399 --> 00:30:24,120
want to move into my daily system but
769
00:30:23,039 --> 00:30:26,279
right now it really looks like it's
770
00:30:24,120 --> 00:30:28,679
going to be a gigabyte board even though
771
00:30:26,279 --> 00:30:30,120
they're pcie like well
772
00:30:28,679 --> 00:30:32,220
we'll see
773
00:30:30,120 --> 00:30:34,440
because I I kind of want to have an
774
00:30:32,220 --> 00:30:35,039
external clock Jen anyway
775
00:30:34,440 --> 00:30:36,600
um
776
00:30:35,039 --> 00:30:38,760
yeah so these are my voltage settings
777
00:30:36,600 --> 00:30:40,980
here if you don't like them you don't
778
00:30:38,760 --> 00:30:41,940
have to use them
779
00:30:40,980 --> 00:30:43,500
um
780
00:30:41,940 --> 00:30:44,700
I haven't adjusted anything in the
781
00:30:43,500 --> 00:30:46,140
external voltage settings because
782
00:30:44,700 --> 00:30:48,240
there's just no reason to do anything
783
00:30:46,140 --> 00:30:50,820
here there's not even any LLC settings
784
00:30:48,240 --> 00:30:53,340
which is a funny thing with this board
785
00:30:50,820 --> 00:30:55,080
I'm not sure if that's necessarily a bad
786
00:30:53,340 --> 00:30:57,360
thing or a good thing because on some
787
00:30:55,080 --> 00:30:58,799
motherboards uh having LLC settings
788
00:30:57,360 --> 00:31:01,080
doesn't mean that they're actually
789
00:30:58,799 --> 00:31:04,140
useful so
790
00:31:01,080 --> 00:31:06,899
um yeah like well like I'll decide
791
00:31:04,140 --> 00:31:09,480
whether or not I'm happy with the lack
792
00:31:06,899 --> 00:31:10,799
of LLC one LLC settings once I measure
793
00:31:09,480 --> 00:31:13,380
this board with the oscilloscope I
794
00:31:10,799 --> 00:31:14,880
haven't done that yet anyway now you
795
00:31:13,380 --> 00:31:16,380
might be looking at this over here and
796
00:31:14,880 --> 00:31:20,039
thinking oh this motherboard has an
797
00:31:16,380 --> 00:31:22,679
external clock gen uh no it doesn't yeah
798
00:31:20,039 --> 00:31:23,640
it unfortunately doesn't
799
00:31:22,679 --> 00:31:25,919
um
800
00:31:23,640 --> 00:31:28,320
ASRock basically this is just a thing
801
00:31:25,919 --> 00:31:30,179
that they do they will regularly forget
802
00:31:28,320 --> 00:31:32,520
to remove settings from their lower end
803
00:31:30,179 --> 00:31:34,919
boards for functionality that they have
804
00:31:32,520 --> 00:31:36,419
on their higher end boards I am pretty
805
00:31:34,919 --> 00:31:38,220
sure this is not the first time I've run
806
00:31:36,419 --> 00:31:39,600
into an ASRock board that had a bclk
807
00:31:38,220 --> 00:31:40,799
setting that doesn't actually do
808
00:31:39,600 --> 00:31:43,860
anything like the first one actually
809
00:31:40,799 --> 00:31:46,200
works that's your like regular vclk that
810
00:31:43,860 --> 00:31:47,640
overclocks absolutely everything but the
811
00:31:46,200 --> 00:31:48,960
external clock gen that you can get on
812
00:31:47,640 --> 00:31:51,419
some high-end boards that avoids
813
00:31:48,960 --> 00:31:53,580
overclocking the PCI and stuff yeah this
814
00:31:51,419 --> 00:31:55,200
board doesn't have that so the BCL K1
815
00:31:53,580 --> 00:31:57,000
setting actually just causes the board
816
00:31:55,200 --> 00:32:02,039
to like not post if you try to change
817
00:31:57,000 --> 00:32:05,419
that well no if you do like this
818
00:32:02,039 --> 00:32:08,820
nothing happens and if you do
819
00:32:05,419 --> 00:32:11,760
this the board fails to post and if you
820
00:32:08,820 --> 00:32:14,220
do this you get a two percent bclk
821
00:32:11,760 --> 00:32:16,500
overclock so yeah
822
00:32:14,220 --> 00:32:18,179
um welcome to ASRock motherboards anyway
823
00:32:16,500 --> 00:32:20,159
then we have the spread Spectrum setting
824
00:32:18,179 --> 00:32:23,820
which I have disabled because that just
825
00:32:20,159 --> 00:32:25,740
basically uh spreads the frequency of
826
00:32:23,820 --> 00:32:27,360
the bclk around a little bit to reduce
827
00:32:25,740 --> 00:32:29,940
Emi
828
00:32:27,360 --> 00:32:31,919
um if you disable it your bclk ends up
829
00:32:29,940 --> 00:32:33,600
being more consistent which if you're
830
00:32:31,919 --> 00:32:35,520
you know pushing your frequencies to the
831
00:32:33,600 --> 00:32:37,799
limit is kind of nice that the bclk
832
00:32:35,520 --> 00:32:39,240
isn't floating around by a by a small
833
00:32:37,799 --> 00:32:39,840
amount
834
00:32:39,240 --> 00:32:42,899
um
835
00:32:39,840 --> 00:32:46,740
but anyway this can potentially cause
836
00:32:42,899 --> 00:32:49,140
Emi issues which is why this is a thing
837
00:32:46,740 --> 00:32:51,179
um but I've never run into those myself
838
00:32:49,140 --> 00:32:53,880
and I've been disabling spread Spectrum
839
00:32:51,179 --> 00:32:56,100
on motherboards for basically as long as
840
00:32:53,880 --> 00:32:59,460
I've been overclocking because that is a
841
00:32:56,100 --> 00:33:01,740
very yeah that is a and like you have
842
00:32:59,460 --> 00:33:04,020
spread Spectrum for vrms and for cl like
843
00:33:01,740 --> 00:33:07,440
for bclk and I've always been disabling
844
00:33:04,020 --> 00:33:09,960
it and never ran into any issues so
845
00:33:07,440 --> 00:33:12,299
anyway but your experience uh may vary
846
00:33:09,960 --> 00:33:13,620
with that so yeah that's basically all
847
00:33:12,299 --> 00:33:15,600
of the adjustments
848
00:33:13,620 --> 00:33:17,100
um that I've made on on this board
849
00:33:15,600 --> 00:33:18,840
there's really
850
00:33:17,100 --> 00:33:20,340
oh yeah so that's like the voltage
851
00:33:18,840 --> 00:33:22,500
settings and now we need to go over the
852
00:33:20,340 --> 00:33:24,840
timing settings I have gear down Mode
853
00:33:22,500 --> 00:33:27,600
still enabled because disabling it is
854
00:33:24,840 --> 00:33:29,700
not worth the effort in my experience
855
00:33:27,600 --> 00:33:32,100
um you just call like you get free
856
00:33:29,700 --> 00:33:34,260
stability at the slight like a
857
00:33:32,100 --> 00:33:36,299
negligible reduction in performance like
858
00:33:34,260 --> 00:33:38,880
I I'll take the free stability thank you
859
00:33:36,299 --> 00:33:40,799
very much uh if I'm pushing benchmarks
860
00:33:38,880 --> 00:33:45,360
then yeah I'll go and disable it but for
861
00:33:40,799 --> 00:33:47,880
for a like stable daily ish setup uh I
862
00:33:45,360 --> 00:33:50,820
can't be bothered anyway then we have uh
863
00:33:47,880 --> 00:33:52,860
Castle agency 30 trcd 38 this could
864
00:33:50,820 --> 00:33:54,840
probably go a little bit lower like 37
865
00:33:52,860 --> 00:33:56,700
or maybe even 36 if you're lucky
866
00:33:54,840 --> 00:33:59,159
actually I think with this member well
867
00:33:56,700 --> 00:34:01,500
this memory kit will probably not do 36
868
00:33:59,159 --> 00:34:03,120
but 37 might work if you're lucky with
869
00:34:01,500 --> 00:34:04,620
your memory kit you might be able to do
870
00:34:03,120 --> 00:34:07,440
36
871
00:34:04,620 --> 00:34:10,080
um at 1.42 volts but this kit I like I
872
00:34:07,440 --> 00:34:11,879
tried that it didn't work uh trp's at 38
873
00:34:10,080 --> 00:34:13,740
this could also potentially go lower
874
00:34:11,879 --> 00:34:16,800
depending on how lucky you get with your
875
00:34:13,740 --> 00:34:19,260
memory kit uh t-ras is at 30. now ASRock
876
00:34:16,800 --> 00:34:22,139
doesn't enforce the
877
00:34:19,260 --> 00:34:24,480
so ASRock does enforce a timing limit
878
00:34:22,139 --> 00:34:26,339
but like I'm not sure if this one is
879
00:34:24,480 --> 00:34:29,220
real because if you compare it to say
880
00:34:26,339 --> 00:34:32,520
gigabyte boards gigabyte boards enforce
881
00:34:29,220 --> 00:34:35,220
a t-ras limit of 30.
882
00:34:32,520 --> 00:34:37,859
again if AMD had public documentation
883
00:34:35,220 --> 00:34:39,780
for their CPUs like Intel does I
884
00:34:37,859 --> 00:34:42,540
wouldn't have to wonder about this but
885
00:34:39,780 --> 00:34:44,940
yeah so I'm going with the higher like
886
00:34:42,540 --> 00:34:47,339
the gigabyte timing limit
887
00:34:44,940 --> 00:34:51,300
um because uh
888
00:34:47,339 --> 00:34:53,460
it like I'm not sure if um
889
00:34:51,300 --> 00:34:55,500
if you can actually set tras below 30
890
00:34:53,460 --> 00:34:57,119
also it's worth noting that the TRS
891
00:34:55,500 --> 00:34:58,619
timing doesn't affect performance that
892
00:34:57,119 --> 00:35:00,119
much so like pushing it below 30
893
00:34:58,619 --> 00:35:03,060
probably won't make much of a difference
894
00:35:00,119 --> 00:35:05,220
now TRC does affect performance quite a
895
00:35:03,060 --> 00:35:08,400
bit however the way TRC works is if you
896
00:35:05,220 --> 00:35:10,380
set it below your trp plus trass it
897
00:35:08,400 --> 00:35:13,440
doesn't do anything
898
00:35:10,380 --> 00:35:16,140
um so this is just set to 68
899
00:35:13,440 --> 00:35:21,599
um because my trp and traps are 30 and
900
00:35:16,140 --> 00:35:23,400
38 so t-ros 68. if I set it to 67 uh it
901
00:35:21,599 --> 00:35:24,839
shouldn't really improve performance in
902
00:35:23,400 --> 00:35:27,599
any way shape or form it also shouldn't
903
00:35:24,839 --> 00:35:31,320
affect stability I haven't retested that
904
00:35:27,599 --> 00:35:34,200
since the ryzen 5000 chips but unlike
905
00:35:31,320 --> 00:35:37,260
ryzen 5000 chips you could set your like
906
00:35:34,200 --> 00:35:38,760
TRC impossibly low and then your t-ras
907
00:35:37,260 --> 00:35:42,300
would sort of actually be what would
908
00:35:38,760 --> 00:35:44,160
govern your actual like real TRC or you
909
00:35:42,300 --> 00:35:48,119
could set your t-ras impossibly low and
910
00:35:44,160 --> 00:35:48,720
then your TRC would govern your TRC
911
00:35:48,119 --> 00:35:50,760
um
912
00:35:48,720 --> 00:35:53,220
so yeah that's that's what's going on
913
00:35:50,760 --> 00:35:55,560
with that anyway like this is not
914
00:35:53,220 --> 00:35:58,980
necessarily a super tight TRC setting
915
00:35:55,560 --> 00:36:01,740
for for heineck stims at 6200 but also
916
00:35:58,980 --> 00:36:04,020
there's it's not really going to go much
917
00:36:01,740 --> 00:36:07,560
lower than this I don't think uh then we
918
00:36:04,020 --> 00:36:09,300
have TWR which is at 48 which oh as rock
919
00:36:07,560 --> 00:36:12,240
apparently and for okay I think they
920
00:36:09,300 --> 00:36:15,720
just like have arbitrary timing lists
921
00:36:12,240 --> 00:36:18,839
because 42 is just the most random TWR
922
00:36:15,720 --> 00:36:20,940
no as a 48 like I can I can understand
923
00:36:18,839 --> 00:36:23,400
why it would be 48 because that at least
924
00:36:20,940 --> 00:36:24,720
lines up with some of the ddr5 jdec
925
00:36:23,400 --> 00:36:27,599
documentation
926
00:36:24,720 --> 00:36:30,180
42 is just like a completely random
927
00:36:27,599 --> 00:36:32,640
number as far as I'm concerned so anyway
928
00:36:30,180 --> 00:36:34,920
48 at least this should be the register
929
00:36:32,640 --> 00:36:35,820
limit on the CPU the thing is if you
930
00:36:34,920 --> 00:36:37,859
push
931
00:36:35,820 --> 00:36:39,839
um if a motherboard lets you set timings
932
00:36:37,859 --> 00:36:43,079
below the register limit of the CPU the
933
00:36:39,839 --> 00:36:45,480
CPU will just round up to the actual
934
00:36:43,079 --> 00:36:47,460
register limit usually so setting the
935
00:36:45,480 --> 00:36:49,980
timings even lower will just kind of not
936
00:36:47,460 --> 00:36:52,440
do anything a lot of the time
937
00:36:49,980 --> 00:36:57,060
um anyway I have the refresh interval
938
00:36:52,440 --> 00:37:00,960
completely maxed out at 65 535 the trfc
939
00:36:57,060 --> 00:37:03,119
one is at 477. uh this could potentially
940
00:37:00,960 --> 00:37:05,820
go a little bit lower but um yeah
941
00:37:03,119 --> 00:37:08,700
slightly below 500 tends to be you know
942
00:37:05,820 --> 00:37:11,820
a good stopping point for heinek's based
943
00:37:08,700 --> 00:37:14,099
memory sticks uh then trfc2 and trfc
944
00:37:11,820 --> 00:37:16,079
same bank are not currently used by the
945
00:37:14,099 --> 00:37:19,260
AMD memory controller so I have them at
946
00:37:16,079 --> 00:37:22,680
1337 because that's just kind of funny
947
00:37:19,260 --> 00:37:24,359
um I would have them at 69.69 but they
948
00:37:22,680 --> 00:37:25,740
won't let me do that and if I do
949
00:37:24,359 --> 00:37:28,200
actually well actually it would let me
950
00:37:25,740 --> 00:37:29,880
have them at 69. so I guess I should
951
00:37:28,200 --> 00:37:31,560
have set one of them to that but yeah
952
00:37:29,880 --> 00:37:33,300
neither of these timings is actually
953
00:37:31,560 --> 00:37:35,880
used by the memory controller I have
954
00:37:33,300 --> 00:37:37,740
another video sort of testing this
955
00:37:35,880 --> 00:37:40,020
um these are an alternative refresh mode
956
00:37:37,740 --> 00:37:41,460
AMD just hasn't implemented it or if
957
00:37:40,020 --> 00:37:43,560
they have implemented it it's not
958
00:37:41,460 --> 00:37:44,160
enabled by default
959
00:37:43,560 --> 00:37:46,680
um
960
00:37:44,160 --> 00:37:49,800
on like any motherboard in my experience
961
00:37:46,680 --> 00:37:51,180
so yeah anyway that's why I just like
962
00:37:49,800 --> 00:37:53,160
you could also just leave these on auto
963
00:37:51,180 --> 00:37:54,900
like the point is these two timings
964
00:37:53,160 --> 00:37:57,480
don't actually do anything right now
965
00:37:54,900 --> 00:38:00,060
because they're just not being used
966
00:37:57,480 --> 00:38:01,920
um and we have RTP which I have set to
967
00:38:00,060 --> 00:38:04,020
12. this could probably go a little bit
968
00:38:01,920 --> 00:38:05,940
lower especially on MDI memory kits
969
00:38:04,020 --> 00:38:09,359
it'll usually go as low as like eight
970
00:38:05,940 --> 00:38:12,240
maybe even below that but actually yeah
971
00:38:09,359 --> 00:38:13,619
maybe even below that uh but uh on ADI
972
00:38:12,240 --> 00:38:16,200
memory kits you will find that generally
973
00:38:13,619 --> 00:38:18,119
it doesn't like going below 12.
974
00:38:16,200 --> 00:38:19,200
um at least not without tons and tons of
975
00:38:18,119 --> 00:38:22,140
voltage
976
00:38:19,200 --> 00:38:25,079
anyway up next we have trdl and trrds
977
00:38:22,140 --> 00:38:27,599
which I have an eight and four TR RDL at
978
00:38:25,079 --> 00:38:30,000
6200 on an MDI kit could probably do
979
00:38:27,599 --> 00:38:31,440
four instead of just eight but the
980
00:38:30,000 --> 00:38:33,900
performance difference from going like
981
00:38:31,440 --> 00:38:36,119
going from eight to four on TR RDL with
982
00:38:33,900 --> 00:38:37,920
ddr5 is relatively negligible because
983
00:38:36,119 --> 00:38:39,720
there's a lot of different bank groups
984
00:38:37,920 --> 00:38:43,020
so the memory controller shouldn't be
985
00:38:39,720 --> 00:38:45,119
using this timing very often uh and so
986
00:38:43,020 --> 00:38:47,880
yeah that's like this timing isn't like
987
00:38:45,119 --> 00:38:50,280
super performance critical
988
00:38:47,880 --> 00:38:53,280
um and trds is literally at the register
989
00:38:50,280 --> 00:38:54,900
limit it cannot be one okay I assure you
990
00:38:53,280 --> 00:38:58,980
it can't be one
991
00:38:54,900 --> 00:39:01,380
uh T Falls at 20. uh ASRock for some
992
00:38:58,980 --> 00:39:03,000
reason allows you to set it to 12 as far
993
00:39:01,380 --> 00:39:04,920
as I know the AMD memory controller
994
00:39:03,000 --> 00:39:06,839
doesn't doesn't allow it to be less than
995
00:39:04,920 --> 00:39:08,760
20 but again
996
00:39:06,839 --> 00:39:11,220
um that's mostly based off of the timing
997
00:39:08,760 --> 00:39:12,119
limits of gigabyte boards
998
00:39:11,220 --> 00:39:15,300
um
999
00:39:12,119 --> 00:39:17,040
not based off of like you know register
1000
00:39:15,300 --> 00:39:19,380
documentation
1001
00:39:17,040 --> 00:39:23,040
um which for Intel CPUs is accessible
1002
00:39:19,380 --> 00:39:23,640
but isn't for AMD CPUs because reasons
1003
00:39:23,040 --> 00:39:25,320
um
1004
00:39:23,640 --> 00:39:29,460
anyway now we have the right to read
1005
00:39:25,320 --> 00:39:31,200
timings uh TW uh wtrl is at 16.
1006
00:39:29,460 --> 00:39:33,000
I don't think you're you'll be able to
1007
00:39:31,200 --> 00:39:34,859
push this any lower than this not at
1008
00:39:33,000 --> 00:39:37,560
least without like tons of extra voltage
1009
00:39:34,859 --> 00:39:39,119
and wtrs is at four this might actually
1010
00:39:37,560 --> 00:39:41,820
go a little bit lower
1011
00:39:39,119 --> 00:39:43,200
um but it won't really do much to
1012
00:39:41,820 --> 00:39:45,000
performance at this point because it's
1013
00:39:43,200 --> 00:39:45,780
already very low
1014
00:39:45,000 --> 00:39:48,060
um
1015
00:39:45,780 --> 00:39:51,420
anyway then we have read to read scl
1016
00:39:48,060 --> 00:39:53,520
which is F4 uh this won't yeah this
1017
00:39:51,420 --> 00:39:56,760
probably won't go any lower than this uh
1018
00:39:53,520 --> 00:39:58,920
same uh SC is at one this is as low as
1019
00:39:56,760 --> 00:40:00,420
this timing goes uh then we have the
1020
00:39:58,920 --> 00:40:03,000
right to rights these are high Nix
1021
00:40:00,420 --> 00:40:06,420
memory sticks so they do very low right
1022
00:40:03,000 --> 00:40:08,160
to right timings so scl is a two and SC
1023
00:40:06,420 --> 00:40:10,079
is also at one because it doesn't go any
1024
00:40:08,160 --> 00:40:12,900
lower than that and then we have the
1025
00:40:10,079 --> 00:40:14,520
right to read uh the other set of right
1026
00:40:12,900 --> 00:40:18,119
to read timings and read to write
1027
00:40:14,520 --> 00:40:20,339
timings which are at 4 and 16. uh this
1028
00:40:18,119 --> 00:40:21,720
might go a bit lower than this but again
1029
00:40:20,339 --> 00:40:23,640
it won't really drastically change
1030
00:40:21,720 --> 00:40:26,099
performance at this point and read to
1031
00:40:23,640 --> 00:40:28,140
write is not really going to get much
1032
00:40:26,099 --> 00:40:30,780
lower than 16. uh actually I don't think
1033
00:40:28,140 --> 00:40:33,780
it'll be stable at all below 16. so yeah
1034
00:40:30,780 --> 00:40:35,220
these are sort of like these are
1035
00:40:33,780 --> 00:40:38,400
basically the memory timings I will
1036
00:40:35,220 --> 00:40:40,980
punch in on any AMD system with high Nix
1037
00:40:38,400 --> 00:40:43,200
memory sticks at around 1.4 volts on the
1038
00:40:40,980 --> 00:40:46,740
memory if you're lucky it might even
1039
00:40:43,200 --> 00:40:48,480
work at like 1.35 volts and by 1 like by
1040
00:40:46,740 --> 00:40:51,000
on the memory I mean the like vdd
1041
00:40:48,480 --> 00:40:53,460
voltage as as I've mentioned earlier the
1042
00:40:51,000 --> 00:40:55,020
vddq and the vddio voltage those are
1043
00:40:53,460 --> 00:40:57,420
more about the memory clock than they
1044
00:40:55,020 --> 00:40:58,079
are about the actual timings
1045
00:40:57,420 --> 00:40:59,820
um
1046
00:40:58,079 --> 00:41:02,460
your timing scale basically with just
1047
00:40:59,820 --> 00:41:04,619
vdd voltage because like the vdd voltage
1048
00:41:02,460 --> 00:41:06,839
Powers the circuitry that's affected by
1049
00:41:04,619 --> 00:41:09,420
the various uh memory timings that we
1050
00:41:06,839 --> 00:41:11,280
have over here whereas like vddq just
1051
00:41:09,420 --> 00:41:13,500
Powers the part of the memory chip that
1052
00:41:11,280 --> 00:41:16,680
actually outputs the data
1053
00:41:13,500 --> 00:41:18,960
um and the the data or like receive the
1054
00:41:16,680 --> 00:41:21,060
data actually it doesn't really receive
1055
00:41:18,960 --> 00:41:21,660
the data but anyway
1056
00:41:21,060 --> 00:41:24,540
um
1057
00:41:21,660 --> 00:41:26,820
yeah so
1058
00:41:24,540 --> 00:41:29,040
that's it for this um so this video was
1059
00:41:26,820 --> 00:41:31,200
a bit of a mess but uh I'm gonna I'm
1060
00:41:29,040 --> 00:41:33,660
gonna leave it this way because I mostly
1061
00:41:31,200 --> 00:41:36,119
wanted to show that this ASRock board is
1062
00:41:33,660 --> 00:41:37,859
actually like capable as far as like
1063
00:41:36,119 --> 00:41:39,660
memory overclocking and infinity fabric
1064
00:41:37,859 --> 00:41:41,220
overclocking goes like this is as high
1065
00:41:39,660 --> 00:41:45,060
as I've ever had the infinity fabric
1066
00:41:41,220 --> 00:41:46,740
clocked on this 7900x and it's also as
1067
00:41:45,060 --> 00:41:49,740
high as I've ever had the memory clocked
1068
00:41:46,740 --> 00:41:51,440
on this 7900x so uh
1069
00:41:49,740 --> 00:41:54,780
yeah the Amazon
1070
00:41:51,440 --> 00:41:56,880
b650m-hdv is you know perfectly capable
1071
00:41:54,780 --> 00:41:58,440
of like memory overclocking also it ran
1072
00:41:56,880 --> 00:42:00,900
Lin pack for like
1073
00:41:58,440 --> 00:42:03,119
uh it actually ran an impact for much
1074
00:42:00,900 --> 00:42:04,560
more than just the 77 Loops that you saw
1075
00:42:03,119 --> 00:42:06,900
in the initial stress test here because
1076
00:42:04,560 --> 00:42:08,220
I ran will impact multiple times in the
1077
00:42:06,900 --> 00:42:10,560
process of like getting to these
1078
00:42:08,220 --> 00:42:12,780
settings so yeah this board has
1079
00:42:10,560 --> 00:42:15,900
absolutely no problem handling a you
1080
00:42:12,780 --> 00:42:19,140
know 7900x even a 7950x really shouldn't
1081
00:42:15,900 --> 00:42:22,020
be any any uh like you shouldn't give
1082
00:42:19,140 --> 00:42:24,119
this motherboard any any difficulty
1083
00:42:22,020 --> 00:42:26,099
um especially because the 7900x and the
1084
00:42:24,119 --> 00:42:27,480
7950x actually have the same at least at
1085
00:42:26,099 --> 00:42:28,500
stock settings they have the same power
1086
00:42:27,480 --> 00:42:31,680
limit
1087
00:42:28,500 --> 00:42:34,320
um and current limit as well so
1088
00:42:31,680 --> 00:42:35,099
yeah anyway
1089
00:42:34,320 --> 00:42:36,960
um
1090
00:42:35,099 --> 00:42:39,119
that's it for the video so hopefully
1091
00:42:36,960 --> 00:42:42,180
this is somewhat helpful if you have an
1092
00:42:39,119 --> 00:42:43,980
ASRock uh B6 really any b650 motherboard
1093
00:42:42,180 --> 00:42:45,420
because yeah like the the thing is a lot
1094
00:42:43,980 --> 00:42:48,420
of these settings you can just kind of
1095
00:42:45,420 --> 00:42:50,220
like like on one hand you're technically
1096
00:42:48,420 --> 00:42:52,200
not supposed to copy other people's
1097
00:42:50,220 --> 00:42:54,359
settings blindly that doesn't tend to
1098
00:42:52,200 --> 00:42:55,740
well the issue with copying other
1099
00:42:54,359 --> 00:42:56,880
people's settings is if you don't know
1100
00:42:55,740 --> 00:42:58,380
what you're doing and you copy
1101
00:42:56,880 --> 00:43:00,300
somebody's settings and it doesn't work
1102
00:42:58,380 --> 00:43:02,760
you're not going to be able to fix it on
1103
00:43:00,300 --> 00:43:04,380
the other hand I do actually just kind
1104
00:43:02,760 --> 00:43:05,880
of use the same settings on basically
1105
00:43:04,380 --> 00:43:09,000
every motherboard
1106
00:43:05,880 --> 00:43:11,460
so I like copy paste my own settings
1107
00:43:09,000 --> 00:43:12,060
um so make of that what you will
1108
00:43:11,460 --> 00:43:14,339
um
1109
00:43:12,060 --> 00:43:16,140
but uh oh actually I should mention that
1110
00:43:14,339 --> 00:43:17,940
there are some CPUs out there and you
1111
00:43:16,140 --> 00:43:20,880
have to be incredibly unlucky for this
1112
00:43:17,940 --> 00:43:22,500
but if you get incredibly unlucky 6200
1113
00:43:20,880 --> 00:43:25,319
might not work
1114
00:43:22,500 --> 00:43:29,940
um yeah but that is like very very rare
1115
00:43:25,319 --> 00:43:34,380
every single CPU I have can do over 6200
1116
00:43:29,940 --> 00:43:38,819
um not 6400 6400 is very rare but 6200
1117
00:43:34,380 --> 00:43:39,839
is very doable in my experience so
1118
00:43:38,819 --> 00:43:43,020
um
1119
00:43:39,839 --> 00:43:45,900
yeah now with the fclk I mentioned like
1120
00:43:43,020 --> 00:43:47,940
2066 is very doable more than that tends
1121
00:43:45,900 --> 00:43:51,060
to depend on how lucky you get and then
1122
00:43:47,940 --> 00:43:53,160
maybe if depend you know also can depend
1123
00:43:51,060 --> 00:43:55,440
on how much you decide to push your like
1124
00:43:53,160 --> 00:43:55,980
vddg voltages
1125
00:43:55,440 --> 00:43:58,380
um
1126
00:43:55,980 --> 00:44:00,240
anyway so yeah that'll be it for this
1127
00:43:58,380 --> 00:44:01,980
video it went way longer than I was
1128
00:44:00,240 --> 00:44:05,280
hoping it would but oh well here we are
1129
00:44:01,980 --> 00:44:06,780
so thanks for watching uh like share
1130
00:44:05,280 --> 00:44:08,339
subscribe leave any comments questions
1131
00:44:06,780 --> 00:44:09,960
suggestions down in the comment section
1132
00:44:08,339 --> 00:44:11,760
below if you'd like to support what I do
1133
00:44:09,960 --> 00:44:13,680
here with actually hardcore overclocking
1134
00:44:11,760 --> 00:44:15,180
I have a patreon there's a link to that
1135
00:44:13,680 --> 00:44:16,920
down in the description below there's
1136
00:44:15,180 --> 00:44:19,319
also the hoc Teespring store where you
1137
00:44:16,920 --> 00:44:21,900
can pick up shirts uh hoodies posters
1138
00:44:19,319 --> 00:44:24,599
you know the usual YouTuber merch and
1139
00:44:21,900 --> 00:44:25,800
I've also got a band camp uh there's a
1140
00:44:24,599 --> 00:44:27,660
link to that down in the description
1141
00:44:25,800 --> 00:44:29,940
below as well
1142
00:44:27,660 --> 00:44:31,020
um and uh yeah if you'd like to support
1143
00:44:29,940 --> 00:44:33,180
the channel That would be much
1144
00:44:31,020 --> 00:44:37,579
appreciated and that's it for the video
1145
00:44:33,180 --> 00:44:37,579
so thank you for watching and goodbye
84520
Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.