Overclock.net banner
Cancel
Create thread New Forums

ASUS ROG X570 Crosshair VIII Overclocking & Discussion Thread

3 reading
Jump to Latest
12,641 - 12,660 of 13,920 Posts
#12,642 ·
Kelutrel said:
Ok, hopefully this weekend I will have the time to clean it up a bit and add some instructions and release it
Click to expand...
Quick question for you re. the deep bowls of Asus X570 bios - would the X570 (Dark Hero) have an asynchronous clock option somewhere (ie. use 105 MHz for CPU, RAM but keep PCIe at 100) ? I know the 670E has it, but I haven't found anything similar in X570. Tx :)
 
#12,643 ·
J7SC said:
Quick question for you re. the deep bowls of Asus X570 bios - would the X570 (Dark Hero) have an asynchronous clock option somewhere (ie. use 105 MHz for CPU, RAM but keep PCIe at 100) ? I know the 670E has it, but I haven't found anything similar in X570. Tx :)
Click to expand...
Not that I know of. In the past I thought that the "Extreme Tweaker\Tweaker's Paradise\Force OC Mode Disable" setting was somehow related to unlinking the BCLK from the other clocks, because of its description, but I have never been able to make it work and I am not sure if it has any meaning in that regard.
 
  • Rep+
Reactions: J7SC
Save
Reply Quote Rep+ Rep+
#12,644 ·
Kelutrel said:
Not that I know of. In the past I thought that the "Extreme Tweaker\Tweaker's Paradise\Force OC Mode Disable" setting was somehow related to unlinking the BCLK from the other clocks, because of its description, but I have never been able to make it work and I am not sure if it has any meaning in that regard.
Click to expand...
Thanks ! I checked 'Force OC Mode Disable' before a while back; couldn't make it work then but might give it another shot when I have more time.
 
#12,645 · (Edited)
stimpy88 said:
A release of the tool you made would be beneficial to the community, as ScatterBencher seems to want to keep his to himself...
Click to expand...
Here I release the windows executable for the tool to show the peak frequency of the cores when progressively loaded: ropbench_v1.51.zip
I named it RopBench, please read the README for instructions.
It should provide very similar data to the NopBench that SkatterBencher used.
I grant that the archive and the files are virus free, but feel free to take any additional precaution you may see fit.
The SHA-256 of the original archive is: 80c0bb285b8d0a342e9c32c0cd574de736c6c764579b041db483cfe9e5a3d577
 
#12,646 ·
Kelutrel said:
Here I release the windows executable for the tool to show the peak frequency of the cores when progressively loaded: ropbench.zip
I named it RopBench, please read the README for instructions and required tools.
It should provide very similar data to the NopBench that SkatterBencher used.
I grant that the archive and the files are virus free, but feel free to take any additional precaution you may see fit.
The SHA-256 of the original archive is: 381ff1ebfc7164b356ebd34a69092330e68b55960e51ccc7d070bc0977eb1a8d
Click to expand...
Now it will be very intersting to check bios 3801 vs 4201 behavior :D
 
Save
Reply Quote Rep+ Rep+
#12,648 · (Edited)
CyrIng said:
BIOS 4201
EDC Tweak
  • EDC @ 10A
View attachment 2580459
View attachment 2580460

Limit of 3950X is found @ 4775 MHz with absolute frequency peak @ 4800 MHz

View attachment 2580461

Vcore offset is being tuned for the same frequency results but keeping low voltage and temperature while idle
Click to expand...
Uhmmmm... I nearly halved my usual performance with this, I have a 5950X and I just changed EDC to 10 and undervolted the cpu by 0.03125v .

I got this in BenchMaestro with EDC 10:
Font Line Screenshot Technology Parallel



While my usual configuration with EDC 140 and no undervolt gets this:
Font Line Screenshot Technology Parallel


Is your experience any different ?
 
Save
Reply Quote Rep+ Rep+
#12,649 ·
Kelutrel said:
Is your experience any different ?
Click to expand...
Our Processors and OS differ but my experience using the integrated benchmark of 7-zip is giving the following:
  1. EDC Tweak is apparently still possible with BIOS version 4201
  2. Tweak is resulting with ~ 2000 total MIPS
A complete Linux Kernel build and also CoreFreq' Conic Compute are stable with EDC Tweak; measured temperature (71 C) and consumed Package / Cores Power (145 W / 139 W) remain slightly above of the Processor specs.
Various idle cases remain stable so far also.

Keeping a BIOS change log here

Code: 
***
BIOS 3801 - PBO[Auto]
***

7-Zip [64] 17.04 : Copyright (c) 1999-2021 Igor Pavlov : 2017-08-28
p7zip Version 17.04 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,64 bits,32 CPUs x64)

x64
CPU Freq: - 64000000 - - - - - - -

RAM size:   32008 MB,  # CPU hardware threads:  32
RAM usage:   7060 MB,  # Benchmark threads:     32

                       Compressing  |                  Decompressing
Dict     Speed Usage    R/U Rating  |      Speed Usage    R/U Rating
         KiB/s     %   MIPS   MIPS  |      KiB/s     %   MIPS   MIPS

22:     118924  2981   3881 115690  |    1435181  3166   3867 122391
23:     110640  2959   3810 112729  |    1417588  3170   3870 122666
24:     107608  2962   3906 115701  |    1398690  3178   3863 122767
25:     105546  2999   4019 120508  |    1378238  3184   3852 122654
----------------------------------  | ------------------------------
Avr:            2975   3904 116157  |             3174   3863 122619
Tot:            3075   3883 119388

***
BIOS 4201 - PBO[Manual] - EDC Tweak 10A
***

7-Zip [64] 17.04 : Copyright (c) 1999-2021 Igor Pavlov : 2017-08-28
p7zip Version 17.04 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,64 bits,32 CPUs x64)

x64
CPU Freq: - - - - - - - - -

RAM size:   32012 MB,  # CPU hardware threads:  32
RAM usage:   7060 MB,  # Benchmark threads:     32

                       Compressing  |                  Decompressing
Dict     Speed Usage    R/U Rating  |      Speed Usage    R/U Rating
         KiB/s     %   MIPS   MIPS  |      KiB/s     %   MIPS   MIPS

22:     120504  3005   3901 117227  |    1467766  3160   3961 125170
23:     111952  2933   3889 114066  |    1438227  3139   3965 124452
24:     108923  2962   3954 117114  |    1428968  3174   3952 125425
25:     106890  2986   4088 122043  |    1412278  3186   3945 125683
----------------------------------  | ------------------------------
Avr:            2972   3958 117613  |             3165   3956 125182
Tot:            3068   3957 121397
 
#12,650 · (Edited)
CyrIng said:
Our Processors and OS differ but my experience using the integrated benchmark of 7-zip is giving the following:
  1. EDC Tweak is apparently still possible with BIOS version 4201
  2. Tweak is resulting with ~ 2000 total MIPS
A complete Linux Kernel build and also CoreFreq' Conic Compute are stable with EDC Tweak; measured temperature (71 C) and consumed Package / Cores Power (145 W / 139 W) remain slightly above of the Processor specs.
Various idle cases remain stable so far also.

Keeping a BIOS change log here

Code: 
***
BIOS 3801 - PBO[Auto]
***

7-Zip [64] 17.04 : Copyright (c) 1999-2021 Igor Pavlov : 2017-08-28
p7zip Version 17.04 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,64 bits,32 CPUs x64)

x64
CPU Freq: - 64000000 - - - - - - -

RAM size:   32008 MB,  # CPU hardware threads:  32
RAM usage:   7060 MB,  # Benchmark threads:     32

                       Compressing  |                  Decompressing
Dict     Speed Usage    R/U Rating  |      Speed Usage    R/U Rating
         KiB/s     %   MIPS   MIPS  |      KiB/s     %   MIPS   MIPS

22:     118924  2981   3881 115690  |    1435181  3166   3867 122391
23:     110640  2959   3810 112729  |    1417588  3170   3870 122666
24:     107608  2962   3906 115701  |    1398690  3178   3863 122767
25:     105546  2999   4019 120508  |    1378238  3184   3852 122654
----------------------------------  | ------------------------------
Avr:            2975   3904 116157  |             3174   3863 122619
Tot:            3075   3883 119388

***
BIOS 4201 - PBO[Manual] - EDC Tweak 10A
***

7-Zip [64] 17.04 : Copyright (c) 1999-2021 Igor Pavlov : 2017-08-28
p7zip Version 17.04 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,64 bits,32 CPUs x64)

x64
CPU Freq: - - - - - - - - -

RAM size:   32012 MB,  # CPU hardware threads:  32
RAM usage:   7060 MB,  # Benchmark threads:     32

                       Compressing  |                  Decompressing
Dict     Speed Usage    R/U Rating  |      Speed Usage    R/U Rating
         KiB/s     %   MIPS   MIPS  |      KiB/s     %   MIPS   MIPS

22:     120504  3005   3901 117227  |    1467766  3160   3961 125170
23:     111952  2933   3889 114066  |    1438227  3139   3965 124452
24:     108923  2962   3954 117114  |    1428968  3174   3952 125425
25:     106890  2986   4088 122043  |    1412278  3186   3945 125683
----------------------------------  | ------------------------------
Avr:            2972   3958 117613  |             3165   3956 125182
Tot:            3068   3957 121397
Click to expand...
I am also on BIOS 4201. On my 5950X under Windows 11 the performances nearly halved when using EDC@10 and undervolting the CPU by 0.03125v.
The peak single-core frequency of the single thread benchmark peaked slightly higher, but the average effective frequency was much lower and so the final performance was quite low.

On your specific hardware and OS and configuration, the result of a multithreaded benchmark which uses all the cores is higher or lower when you set EDC@10+Undervolt compared to the normal default settings without undervolt ?

What I am saying is that on a 5950X, if you set EDC to 10 and undervolt the cpu, you will get lower temperatures, less used watts, and only half the performances. And I am curious to understand if on a 3950X this is different instead.
 
Save
Reply Quote Rep+ Rep+
#12,651 ·
Kelutrel said:
And I am curious to understand if on a 3950X this is different instead.
Click to expand...
It seems doing the reverse
  • PBO/EDC [Auto]
    Colorfulness Screenshot Font Parallel Pattern


  • PBO [Manual] EDC [10A] w/ Undervolt
    Colorfulness Rectangle Font Screenshot Parallel
Case EDC of 10A:
  • About same temperature of 70 C
  • +7 W total cores power
  • -0.02 V peak core voltage
  • +150 MHz instant absolute frequency
I still have to relax other limits such as thermal but throttling may happen for very short periods that I have not monitor yet.
 
#12,652 · (Edited)
Just because I was curious, I sampled the first 30ms of frequency change of a 5950X core when placed under sudden load, with a sampling rate of a 20th of a millisecond.
I post the picture here because it may be informative. It doesn't clearly appear in the chart, but after the first 30ms you get 5080MHz when at EDC140, 5090MHz when at EDC120, and 5100MHz when at EDC0+FMAX (this last one is slightly better for burst single-core loads but looses in the long run on every MT benchmark).

Rectangle Slope Parallel Font Paper product



Zoom on the interesting bit:
Rectangle Slope Font Parallel Circle
 
Save
Reply Quote Rep+ Rep+
#12,653 · (Edited)
CyrIng said:
It seems doing the reverse
Case EDC of 10A:
  • About same temperature of 70 C
  • +7 W total cores power
  • -0.02 V peak core voltage
  • +150 MHz instant absolute frequency
I still have to relax other limits such as thermal but throttling may happen for very short periods that I have not monitor yet.
Click to expand...
I'm ok with the slightly higher peak frequency of a single core after a bunch of ms, although on the 5950X I have not observed such a large variation as you.

But on a 5950X, using a single core, when at EDC10+Undervolt, the final average frequency is much lower. It's like if it peaks for a very short burst to a higher frequency, like 5070MHz, and then drops to 4000MHz and stays there so the performances in the end suffer compared to the normal EDC. And when placing the load on all the cores it is even worse, it literally burst to 3700MHz only for a very short duration and then drops to around 2000MHz. You can see this from the pictures of the BenchMaestro benchmark I posted above.

This is why I asked you, if you run a benchmark that uses all the cores and runs for at least 10 seconds, first using the EDC10+Undervolt, and then using the same BIOS version and configuration but without undervolt and with EDC set at default, everything else staying the same, what scores you would get on the 3950X.
 
Save
Reply Quote Rep+ Rep+
#12,654 ·
Kelutrel said:
This is why I asked you, if you run a benchmark that uses all the cores and runs for at least 10 seconds, first using the EDC10+Undervolt, and then using the same BIOS version and configuration but without undervolt and with EDC set at default, everything else staying the same, what scores you would get on the 3950X.
Click to expand...
Somehow I'm close to your testing protocol but without duration of 10 secs.
Conic (home made) is stressing forever. What I check after-while are the Min, Max
I think I have set all BIOS settings combinations you mentioned.
I will add what I have observed is that without Undervolt the Max absolute frequency is 4800 MHz whereas frequency peak does not go above 4775 MHz when a negative offset is applied to Vcore.

I remember reading in this thread that choosing BIOS 4201 is making you loose the ability of the EDC Tweak.
With my 3950X, versions 3801 and 4201 are both providing the EDC tweak.
Is this only true for a Zen2 processor ;)
 
#12,655 · (Edited)
CyrIng said:
Somehow I'm close to your testing protocol but without duration of 10 secs.
Conic (home made) is stressing forever. What I check after-while are the Min, Max
I think I have set all BIOS settings combinations you mentioned.
I will add what I have observed is that without Undervolt the Max absolute frequency is 4800 MHz whereas frequency peak does not go above 4775 MHz when a negative offset is applied to Vcore.

I remember reading in this thread that choosing BIOS 4201 is making you loose the ability of the EDC Tweak.
With my 3950X, versions 3801 and 4201 are both providing the EDC tweak.
Is this only true for a Zen2 processor ;)
Click to expand...
The EDC Tweak was originally discovered by a guy named TheStilt and consisted to set EDC to 1, and then it would have been ignored by the PBO limits logic, thus making the cores peak faster to the max freq of the silicon without being limited by the EDC budget. This tweak was later changed to be a proper feature on Asus motherboards, and was called FMax Enhancer. This tweak on Zen 3 still makes the cores peak faster to their max frequency but due to the other PBO limits and, probably, somewhat more complex logics related to those limits, the actual effective clock rate is slower in my tests when it is enabled, and the benchmark scores are also lower. Additionally, it requires to review the PBO curves because the frequency change is more sudden and introduce some additional instability to a previously stable configuration. Afaik all the Asus BIOSs, including 4201, support the FMax enhancer.

Setting the EDC to 10 is not the original EDC tweak as it does not disable the PBO logics for the EDC. Additionally, on Asus motherboards, when you set EDC to 1 it simply resets it to the default 140 on Zen3, and to actually enable the EDC Tweak you have to directly use the FMax Enhancer option.

On BIOS 4201 (and on any BIOS using Agesa 1.2.0.6 or later, if I remember correctly) the EDC Tweak still works (using the FMax Enhancer option), but they also introduced a new behavior on the Zen3 and unrelated to the EDC Tweak. This new behavior is that if you raise your EDC above the default 140 then the max voltage provided to the CPU is not 1.5v anymore but 1.475v (maybe this 1.475v new limit is on BIOS 4201, and may have been lower on the previous Agesa versions). And this has been called EDC Bug, so you may have been confused by the name.

This change was added (afaik) to prevent electromigration at higher EDC and to prevent the CPU going over its max voltage of 1.5v at times. This unexpected overvoltage at higher EDC, that then imho caused the max voltage reduction at EDC>140 in the later Agesa versions, was discovered by Pier in this same forum ( the original thread ) and discussed with TheStilt, but it is not related to the EDC Tweak.

Or at least this is what I know.

Going back to your suggestion, what OS are you using ? Maybe GeekBench runs on it, and you can use that benchmark to show any difference that you get on the 3950X when setting EDC10+Undervolt compared to the normal EDC and no undervolt. Because the min/max frequency is not really telling the whole story, and a benchmark score would be better instead.
 
  • Rep+
Reactions: stimpy88
Save
Reply Quote Rep+ Rep+
#12,656 ·
flyinion said:
Maybe I misread the articles but I thought this was only an issue with the new 7000 series? I know 5000 had it originally at Win11 launch and it was eventually fixed. I interpreted the articles I read as saying the new issue was specific to dual CCD 7000 series CPU's.
Click to expand...
CapFrameX op Twitter: "Windows 11 22H2 can cause performance issues on PCs with Ryzen CPUs. This is a comparison of feature update 22H2 vs reinstallation OS (including 22H2). â–¶ R9 5900 on MSI B550 + RX 6800 XT @PaulyAlcorn @TechPowerUp @hms1193 @areejs12 #Win11 #Gaming #Ryzen https://t.co/nYzzh7fXlT" / Twitter

not yet tested myself ( I am at sea away of pc) but I will reinstall, and see what happens..
 
Save
Reply Quote Rep+ Rep+
#12,657 ·
Kelutrel said:
Setting the EDC to 10 is not the original EDC tweak as it does not disable the PBO logics for the EDC
Click to expand...
Thanks for those explanations.
EDC @ 0A is same as Auto with my Zen2
EDC from 1A to 9A is permanently throttling: about the P2 P-state frequency.
EDC > specs is same as 140 A

Meanwhile, I have decrease the monitoring interval to 100 ms (rather than default 1 sec) and getting peak of 4800 MHz. Thus not linked to Vcore or offset.
 
#12,658 ·
CyrIng said:
Thanks for those explanations.
EDC @ 0A is same as Auto with my Zen2
EDC from 1A to 9A is permanently throttling: about the P2 P-state frequency.
EDC > specs is same as 140 A

Meanwhile, I have decrease the monitoring interval to 100 ms (rather than default 1 sec) and getting peak of 4800 MHz. Thus not linked to Vcore or offset.
Click to expand...
Yes, sorry, EDC set to 1 is the EDC Tweak, not zero. I updated the post above. Anyway the behavior you described seems to not really be the same on Zen3, so it may be specific for the Zen2 only.
 
Save
Reply Quote Rep+ Rep+
#12,659 ·
Kelutrel said:
Yes, sorry, EDC set to 1 is the EDC Tweak, not zero. I updated the post above. Anyway the behavior you described seems to not really be the same on Zen3, so it may be specific for the Zen2 only.
Click to expand...
Fmax enabled, without Undervolt, is providing less peak and all Cores frequencies.
EDC 10A provides best MIPS
Log is updated with Fmax screenshots.

At first, I have followed " EDC = 1, PBO TURBO BOOST " thread; reason I set 10A
 
#12,660 ·
CyrIng said:
EDC 10A provides best MIPS
Click to expand...
Not on a Zen 3, unless there is some other BIOS option or setting that enables this improvement.
I have looked at the thread you linked, and it looks like on Zen 2 they talk of EDC from 10A to 20A as optimal. This doesn't correspond to my testing on Zen 3, and at least on my hardware it looks like there is no way that that little energy can sustain my cpu.
I may be interested in understanding what changed in the role of EDC between Zen 2 and Zen 3, but I reiterate that I am not seeing the same behavior occurring on Zen 3.
 
  • Rep+
Reactions: stimpy88
Save
Reply Quote Rep+ Rep+
12,641 - 12,660 of 13,920 Posts
About this Discussion
14K Replies
820 Participants
Last post:
DirtyCasual
Overclock.net
posts
28M
members
562K
Since
2004
The premier forum for overclocking experts and enthusiasts. Discuss hardware optimization, custom builds, benchmarking, cooling solutions, and pushing the boundaries of computing performance. From beginner guides to extreme overclocking, join our technical community to master system tuning.
Full Forum Listing
Explore Our Forums
Intel CPUs Hardware News NVIDIA PC Gaming Water Cooling