X670 resource

[Shamino]

ill use this thread to collect some new test bioses for the boards, maybe also to explain some less understood options

to disable cores ccd go here and choose ccd xx bit map down core.
each ones stand for an enabled core
best to disable from the back, ie:
110000
instead of 0011000
after selection press downcore apply changes or discard if made mistake

ocpak/octools

FAQ:
7950x not boosting pass 5.5G -> check that CStates is not disabled
Detailed Explanation on CState Boot Limiter


Test BIOSes:

new:
X3D OC Preset for those MB with asynch BCLK Support: (for simple slight perf boost for X3D)


DOCP/EXPO Tweaked: (for simple timings tightening)


strixe-e 1515 

strixe-f 1515 

strix e a 1515 

crosshair hero 1515 

crosshair gene 1515 

crosshair extreme 1515 

creator 670 1515

creator b650 1515

strix 650E I

strix 670 itx

for crosshair and strix e-e:

explanation of segment2 Loadline:

customize a heterogenous loadline for a dual segment workload range.

example above shows loadline=L6 when current is in range of 0~40A, and Level4 when current is above 40A.

Adds for x3d

dynamic ccd priority switch with core flex, os / driver agnostic so win10 win11 ok





Algo as follows:
If condition reached and ccd0 specified, then check current mem/cache activity > threshold and hysteresis reached, if fulfilled then switch
If condition reached and ccd1 specified, then check current mem/cache activity <=threshold and hysteresis reached,, if fulfilled then switch
Default hysteresis =4

Can combine multiple algos for ccd priority so combinations are wide

works on non x3d too but of course senseless on it. detailed explanation here.

[SkyBeam]

---

@MagicHp wrote:

Those are my main use of my computer, therefore even if everything else works I would call it unstable.

Anyone has an idea?

---

Games are a particularly bad example of a stability test. Why? Because...

• Gaming does not put the system under consistent high load; actually depending on the settings games are often limited by a single component like GPU so the load on CPU and memory is not nearly as high as it can become in scientific applications for example.
• Gaming involves stability of the graphics subsystem too where the software layer (DirectX/Vulcan/..., GPU drivers...) is much more important.
• People tend to use the latest and greatest game (just what they currently like to play) as a reference. Specifically recently released games are prone to software bugs and various issues making it crash even on a system which is entirely stable by any other definition.

So what to do:

• Use synthetic load to test limits of stability under extreme load situations. even of those situations might never appear in real-life scenarios as real-life scenarios are almost always a mix of different workloads.
• Test each component as isolated as possible to see which of the components (CPU/ALU/FPU, memory, GPU, disk) is eventually becoming unstable.
• Test multiple of the above component tests in parallel to see if any cross-subsystem effects cause instability.

The consequence of course is that you can't just fire up your current favorite game and have fun while "testing" your system. Instead you might have to keep your system running stability tests for hours to get an idea about its stability.

What tools to use?

• Memtest86+: Create a boot media and boot memtest86+. Keep it running at least until a complete pass (test 1-10) are completed. It will show a big PASS banner if your system completed a cycle. It will show a big FAIL banner immediately if there is any memory errors.
  Note that sometimes memory errors happen already 30 seconds into the first tests but do not assume it's stable if it runs for a few seconds. Later tests will perform different patterns and they might be particularly demanding for your memory subsystem. In my Ryzen 7000 system I have often seen failures in test 6-8 after about 20 minutest of run time.
  It's recommended to keep a memtest86+ running over night.
• y-cruncher: Y-cruncher is a scientific program and you might have to learn first how to use it. First download and extract the ZIP. Then run y-cruncher.exe (Windows). You will see a menu. Hit "1" on your keyboard to enter "Component Stress Tester". Here you can select the tests to run. In current Version there are 9 different tests. It's known that specifically "VSR - Vector Transform" with AVX512 Float load is putting particularly heavy load on the memory controller and is supposed to be very sensitive to memory subsystem issues.
  To select only VST do the following:
  • Type "11" and hit enter to disable test 11, BTK - Basecase + Karatsuba
  • Type "12" and hit enter to disable test 12, BBP - BBP Digit Extraction
  • Type "13" and hit enter to disable test 13, SFT - Small In-Cache FFT
  • Type "17" and hit enter to disable test 17, HNT - Hybrid NTT
  • Type "18" and hit enter to enable test 18, VST - Vector Transform
  • Then type "0" and hit enter to start stress testing
• Typically if y-cruncher VST is stable for a few minutes (2-3 test cycles) it is stable even keeping it run for a night or even days. If the test fails (you will see some red output and the test will stop) then you might face a memory controller related issue.
  You can of course also keep the tests enabled which are selected by default (see instructions above) but you might perfectly pass those tests before reaching VST which is specifically sensitive to memory subsystem stability. So I currently save a bit of time running VST only after doing any change and to get a quick indication about stability.
  
  Also note that a failed VST test might still allow you to run games perfectly fine. You might even claim your system to be stable in day-by-day use even if it dashes out during VST after a few seconds. Reason is that in real-life scenarios it is very unlikely for any workload to create nearly as much load on the memory subsystem as y-cruncher VST is. So here we can start a debate if a system failing y-cruncher VST is "stable" or not. I would personally call it "partially stable" if it is practically usable but does not sustain a continuous synthetic load.
• For GPU Stresstest there are dedicated tools like FurMark or many benchmarks - that's actually not the main topic here.

[MagicHp]

Well, that is the issue, you say "games are bad to stability test" , I say "games that have a tendancy to crash make the best tests" ... It all depends on what you are testing, "stability" is a very broad term. In my case, "stable" means "no crash whatever I do", not "passed some tests succesfully". If passing those tests doesn't automatically means you'll get no crash, then why pass those tests in the first place?!?

To be more precise, no amount of tests I did made my system crash, but 5 minutes in a precise game doing precise things ( waiting in the lobby ^^ ) crashed it immediately. That isn't stable in any way and none of the test you mention would have revealed that.

I even ran memtest for several hours back then, no issues detected. Yet a crash every 5 minutes does sound like an issue ^^

In my opinion, a game has more chances to crash especially because it has uneven load, because it uses the latest software technologies and because those indeed have software bugs that could lead to issues. So if you can run these for hours, in different load situations and get no crash when all you see on forums about these games are "crash" threads, then your system is actually stable.

You mentionned "real-life scenarios".... Well, that's it ^^ I don't do "just" gaming on my computer, but yes this is probably 80% of my use.

And as I said, if a 0.002V difference turns a crash fest into the most stable piece of software, it is pretty safe to assume that those 0.002V made your system unstable. Not as unstable as "crashing randomly", but still unstable.

Your answer looks like one of a professional IT, and that is part of the issue. "you can't just fire up your current favorite game and have fun while "testing" your system"... Well of course you can, depends on what you are testing... Benchmarks, stress tests, memtest, no issue. Launch a game, crash. You will first try to get the game running before redoing your tests and then play (or even just leave it running) for a longer period to test stability in time.

Again I tested my system, using some of the tools you mentionned, and none revealed the issue. Besides, that is pretty much "testing 101" that you did here ^^ "Test each components separately" , "Run tests for hours", etc... That isn't my first ride 😉 And I'm saying yet again, your tests wouldn't have revealed my issue, while mine revealed it in 5 minutes.

Hence my question that I will extend a bit: Does anyone has an idea on how to test properly and fully the stablity on these systems? Tests that would have revealed the 0.002V issue I had for example, knowing that it did work before, just not in every situation.

Also, I can assure you, given the amount of broken bios we had here, not many people will try each of those running a full night on tests with every settings possible, therefore not a good way to test stability in the current situation.

[84toli]

100% fact so many people mislead others by saying *My CO is stable because I can run this or these stress tests* without knowing that some games actually are way better to test CO then any stress test out there.

[DDavid33]

Don't bother with that guy, he acts like he's pro but he keeps on telling nonsenses.

BTW forget about global CO, it's plain crap, do your proper per core CO.
For example my brother's CO goes from +10 to -10 with 103 bclk and 0 to -25 with 100 bclk.
My stable per core for 1303 is: 29 22 22 28 25 32 25 36 20 12 23 18 18 14 29 21 (no async bclk on my e-a)
If we used global CO my brother would be stable at +10 or 0 and mine would be at 12, huge loss !

Our protocol (not meant to be universal or mandatory for everyone, just our protocol based on our experiences and tests that point cores on error):
- core cycler's ycruncher 13-HSW ~ Airi all test random 1.5m no suspend
- then core cycler's ycruncher 22-ZN4 ~ Kizuna all test random 1.5m no suspend
- then OCCT small extreme variable auto sse
- then OCCT large extreme variable auto avx2
- then OCCT large extreme variable auto avx512
In this order until each step throws no error, daily and overnight for core cycler and overnight only for OCCT.

Since then we had no games, video encoding or music encoding crash, no bsod at all even at idle.

[MagicHp]

Thanks I'll check those tools, OCCT seems particularly interesting ^^

I had started to do per core settings, but it was too long and tedious for me. Maybe I'll do a second pass, I should have some free time next week.

[iamgarffi]

At least MCR and PowerDown enabled works okay for “me” with EXPO off/on (1303). Cold to Windows sign in in average of 20-31s. 

2x32G G.Skill Trident Z5 Neo CL30 (QVL)

With Expo off, PBO on, MCR on, PowerDown on

4800MT/s > 20-21s till windows sign in

With Expo on, PBO on, MCR on, PowerDown on

6000MT/s > 23-31s till windows sign in

TL/DR

sticking to 1303 - no compromise here for me 🙂

[xbohdpukc]

well, the legal disclamers are gone from bios downloads...

[iamgarffi]

What does it mean then? Full support nonetheless? 

[Sib3riX]

Does 1303 bios fix soc voltages or we need to wait for proper bios fix?

[DDavid33]

It doesn't but no problem if you manually fix vsoc at 1.25V.

However vsoc fixed stable bios and hopefully correct OCP will be after 1412 which is test bios so no support.