1usmus
patreon
Zen 4 diagnostic - your thoughts?
🕑 Added 2023-08-03 07:56:33 +0000 UTCI think many of you have encountered the problem when the values after diagnostics for CO are positive, not even 0.
Here is my example, the core has already a value of +2. So the factory values are unstable?
Let's get to the bottom of this.
Prime95 is currently being used to diagnose the processor. I am neither its developer nor co-author. I and other well-known enthusiasts in the process of research have set up settings that effectively find errors. As it happens, the current load level created by Prime95 in AVX mode demonstrates even the instability of the stock processor, and temperatures are at the threshold of 95 degrees.
AVX in Prime95 turns out to be even heavier than AVX512. SSE = AVX2. There are 3 outcomes of this event:
1) Wait for Prime95 author to make adjustments to the code of his product or AMD to make throttling adjustments for such scenarios. I don't believe any of them will do anything.
2) Continue to use this approach and get information that some cores are unstable even in stock, but in this case even a super mega heavy stress test will be passed.
3) Refuse to use AVX for Zen 4 in Prime95. Test with SSE and get results that will satisfy stability in 3D Mark, Games, Cinebench R23 and so on. The problem with this approach is that users who are used to testing in mega-heavy super-stress tests will get instability, and as a consequence a flurry of indignation in all forums and chats.
What to do in this case?
Note 1. Fortunately Zen 3 is past such a problem, Prime 95 generates the appropriate load. The problem only exists for Zen 4.
Note 2. OCCT I do not consider it because the problem is similar there + error detection as such is absent (except of course for scenarios of super mega load activation which will prove that CO 0 is unstable).
__________________________________________________________________
Alternative. At the moment I'm doing research that might be a solution to the Zen 4 diagnostic problem. No details yet, but keep your fingers crossed for me.
Comments
Fuyu
AMD 7900X3D asus strix b650e-e bios 1414 asus rtx 3070 ti Win 11 V.22H2 Hi, I recently switched back to AMD and bought the X3D series because I use the pc more for gaming. I learned about Hydra from a friend who bought it and used it in his pc and laptop with noticeable improvements. So I wanted to try it to optimize my 7900X3D. HYDRA 1.5B PRO: With this version I was able to finish the test. With the first version it was stuck (maybe because of HWINFO) on CO04. With this update it almost ran smoothly except for the following problems encountered: -If the IGPU is disable from Windows, Hydra won't recognize both GPU, if the IGPU is disable from the bios, Hydra recognize the GPU correctly. -I had to copy the dll again, because it was reporting the following error: System.FormatException: Input string was not in a correct format. at System.Number.StringToNumber(String str, NumberStyles options, NumberBuffer& number, NumberFormatInfo info, Boolean parseDecimal) at System.Number.ParseInt32(String s, NumberStyles style, NumberFormatInfo info) at A. .(Int32 ) at A. .(Boolean ) at A. ..ctor() Once I restarted in safe mode and recopied the dlls, I did not get any kind of error. Diagnostic: AMD-CO Preset Normal C00 through C03 I had no problems at all. C04 C05 I had in both the exact same problems : Blue screen and inability of the system to reboot automatically. After manually rebooting the diagnostics restarted but when it tried to open cinebench it failed to do so and led to another manual reboot. Once rebooted it continued its progress managing to open cinebench. From C06 to C11 I did not experience any kind of problem. The test ended without error. C00 -24 C01 -22 C02 -18 C03 -18 C04 -24 C05 -24 C06 -12 C07 -12 C08 -6 C09 -14 C10 -14 C11 -10 I could not get Hydra to start with the PBO settings. So I copied the settings into the bios. After some tests done, I didn't notice any improvement, neither in wattage and consequently in temperature. The only difference is that in game, the system becomes unstable and leads to several reboots. So for me there is no improvement. Maybe I did something wrong? It's strange to me that with the voltages set in minus, the temperature and wattages haven't changed. So for now I will not use Hydra, and wait for a version or updates. I hope my tests will help with future updates.
Have you tried it stock, with PBO disabled and no CO? In my experience (tested two cpus, a 7950x and a 7950x3d), they are fine stock, but once I mess with CO, some cores seems to need a little positive nudge. It also seems to be pr CCD, ie if I don't tweak CCD0 cores, they are fine stock, and I'm allowed to tweak CCD1. Also, AMD might have tweaked boost over the different AGESAs, as this was more of an issue with early versions. One BIOS upgrade gave me around-10CO on avg. Of course it's the end boost speed that counts, so my guess is that they just lowered it a bit? And if you guys use CoreCycler, it has a line in the script: Add-Content $configFile2 ('EnableSetAffinity=0') # Don't let Prime automatically assign the CPU affinty, we're doing this on our own That for some reason makes P95 even more cranky. Took me a while to figure out, as my OC passed P95 if I cycled cores manually, but failed CoreCycler.
Dmitry Ghitu
In case if it's not complicated to implement, maybe better approach would be to ask user to decide ? Something like "Do you like stability or performance" and possible options: 1. Stability. Test with AVX. 2. Performance. Tests with SSE.
