Crosshair VIII 2501's for testing

[Redline9k]

Hi Shamino,

Just wanted to post and let you know that after updating to 2701 onwards (including 2702) I'm now getting the BIOS reset issues that others have. Gigabyte looks to be doing well with this issue as of BIOS F31L.
I'm a member of a rather large Australian forum where we have a Zen-specific thread and you can read about others issues here:
https://forums.whirlpool.net.au/thre...q9?p=-1#bottom
In every case after the soft-reset issue happens, Event Viewer shows "Error in processor core" as the fault. Given that myself and so many others are also having the issue, I'm doubtful it's a faulty CPU.

I've confirmed it's not RAM or anything otherwise as I'm a fairly advanced user myself. Additionally, after about the 5th time it did it within a couple hours the board (C8H WiFi) refused to POST and gave debug codes AD, then A6 & then 0D.

Hope this is of some help in getting this issue resolved. I'm back on 2502 for the moment.

[Silent Scone@ROG]

I would highly recommend anyone experiencing instability problems to shift to Karhu RAM Test version 1.1.0.0 or greater. This is by far the most effective memory stress test for Zen-based platforms and should help diagnose instability faster.


For instance, Google Stress App test (GSAT) may take over 1 hour to throw an error whereas Karhu may find an error on the same system in under 5 minutes.

https://www.karhusoftware.com/ramtest/

[Redline9k]

I would highly recommend anyone experiencing instability problems to shift to Karhu RAM Test version 1.1.0.0 or greater. This is by far the most effective memory stress test for Zen-based platforms and should help diagnose instability faster.


For instance, Google Stress App test (GSAT) may take over 1 hour to throw an error whereas Karhu may find an error on the same system in under 5 minutes.

https://www.karhusoftware.com/ramtest/

It's not a memory issue. I've been around the bend with this kit and have stress tested it accordingly. I'm now back on 2502 and not experiencing the random crashes or restarts. I know it's definitely not a RAM issue because I can reproduce it with or without XMP, and with or without my tuned timings which were, stable on my 3900x and stable until the last 2 BIOS releases.
Something is definitely off with the 2 latest revisions of BIOS' and, IMO, shouldn't be up for download on the ASUS website. I'm happy to help however I can, but this setup was completely stable with my 3900x, and completely stable with my new 5950x prior to upgrading to 2701 and 2702.
The system only ever reboots (and others in the thread I linked prior have verified it) under low-load scenarios. We can't get it to happen whenever there's a decent amount of load on the CPU. This to me seems like maybe idle voltage is dropping too low or something else is causing the system to reset when the CPU is largely unused.

[Silent Scone@ROG]

It's not a memory issue. I've been around the bend with this kit and have stress tested it accordingly. I'm now back on 2502 and not experiencing the random crashes or restarts. I know it's definitely not a RAM issue because I can reproduce it with or without XMP, and with or without my tuned timings which were, stable on my 3900x and stable until the last 2 BIOS releases.
Something is definitely off with the 2 latest revisions of BIOS' and, IMO, shouldn't be up for download on the ASUS website. I'm happy to help however I can, but this setup was completely stable with my 3900x, and completely stable with my new 5950x prior to upgrading to 2701 and 2702.
The system only ever reboots (and others in the thread I linked prior have verified it) under low-load scenarios. We can't get it to happen whenever there's a decent amount of load on the CPU. This to me seems like maybe idle voltage is dropping too low or something else is causing the system to reset when the CPU is largely unused.

Not everyone is experiencing the same issues as you are, so the route cause is unclear for individual cases. I have no instability issues on my CH8 / 5950X setup whatsoever on any of these builds even when pushing FCLK past 1800. The post is directed at people who need guidance when setting up memory, whether that includes you or it doesn't.

May be best to wait for the next AGESA before flashing again if the newer builds are causing instability.

[darkmessias94]

It's not a memory issue. I've been around the bend with this kit and have stress tested it accordingly. I'm now back on 2502 and not experiencing the random crashes or restarts. I know it's definitely not a RAM issue because I can reproduce it with or without XMP, and with or without my tuned timings which were, stable on my 3900x and stable until the last 2 BIOS releases.
Something is definitely off with the 2 latest revisions of BIOS' and, IMO, shouldn't be up for download on the ASUS website. I'm happy to help however I can, but this setup was completely stable with my 3900x, and completely stable with my new 5950x prior to upgrading to 2701 and 2702.
The system only ever reboots (and others in the thread I linked prior have verified it) under low-load scenarios. We can't get it to happen whenever there's a decent amount of load on the CPU. This to me seems like maybe idle voltage is dropping too low or something else is causing the system to reset when the CPU is largely unused.

I had the same issues with the random restarts in each and every BIOS version. For me it seems to be the combination of FMAX Enhancer (in PBO settings) and the curve optimizer. I saw that the FMAX Enhancer uses less voltage on single core workloads (e.g. Cinebench R20/R23 it showed 5.25 GHZ clock with 1.41V only). I was always able to finish Cinebench etc. without an issue, but light workloads (e.g. windows defender scan or starting a memory benchmark in AIDA) lead to a crash.
Also it crashed with each and every number deposited in the curve optimizer, does not matter if it is -2 or -15 if FMAX enhancer is on. Also FMAX enhancer leads to a lower single core score in cinebench (Fmax Enhancer on: 613 points in C20, Fmax Enhancer off: 648) due to stretching clocks like crazy. HWinfo reported 5.25 ghz on each core after some gaming, but the effective clock was about 4.7ghz (even when cinebenching single core). Without FMAX enhancer it is around 5ghz effective clock with -15 all core curve optimizer. Without FMAX Enhancer the core voltages goes up to 1.49v now in single threaded workloads and no more crashes.

Long story short: Try to disable Fmax Enhancer while using the curve optimizer.

Kind regards

[Baio73]

Hey there buddy,

I'm running 4 single sticks of GTZR or whatever the mem is, 3800 CL14. Two kits.

So try following:

Push there 2701 BiOS, clear CMOS, set leave everything as it is, try this:
1. pick D.O.C.P. for RAM.
2. downclock RAM to 3200 MHz
3. pick 1600 MHz FCLK
4. check out whether the timings autodetected by BiOS's DB fit the description, at least first five values or whatever.

Boot it.

Hi Uplink,
Thanks for your help.
I went through all the steps you listed, RAM gets set to 18-17-17-17-37 instead of D.O.C.P. 17-17-17-17-37, according to both CPU-Z and DRAM Calculator.
Obviously GearDownMode is on Auto, like all the rest... I only disabled SATA, Wi-Fi an Bluetooth controllers as I don't use them.

Baio

[uplink]

Hi Uplink,
Thanks for your help.
I went through all the steps you listed, RAM gets set to 18-17-17-17-37 instead of D.O.C.P. 17-17-17-17-37, according to both CPU-Z and DRAM Calculator.
Obviously GearDownMode is on Auto, like all the rest... I only disabled SATA, Wi-Fi an Bluetooth controllers as I don't use them.

Baio

You're welcome And is it now more stable for You?

[Stephpd]

Purchased the 5600x and only issue i have is with memory on the Crosshair viii Hero wifi.
Bios 2702 lets me run 2 sticks of single rank Patriot viper steel at 4000mhz, but not 4 sticks.
Had all 4 sticks able to run at 3600mhz on the 2502 bios but couldn't get 3800mhz.

Other then that no crashes.
Able to overclock the 5600x up to 3700mhz.

[Redline9k]

Not everyone is experiencing the same issues as you are, so the route cause is unclear for individual cases. I have no instability issues on my CH8 / 5950X setup whatsoever on any of these builds even when pushing FCLK past 1800. The post is directed at people who need guidance when setting up memory, whether that includes you or it doesn't.

May be best to wait for the next AGESA before flashing again if the newer builds are causing instability.

I'm thinking I'll have to wait. For reference I wasn't saying you were wrong or specifically taking your post personally; I know it was general advice :P
People are reporting the same soft reset issues on Gigabyte boards with the latest AGESA as well.
To me this seems like a firmware level issue.
https://www.tweaktownforum.com/forum...s-xtreme-5950x

I've also heard it could be a while:
https://www.tweaktownforum.com/forum...589#post975589

"AMD for Agesa 1.1.8.0 won't allow any official BIOS on websites etc. They will have a beta, but for now 1.1.8.0 breaks the Vermeer functionality, so is only work with Cezanne CPUs. It's unknown the time for now.
As for PBO2 it will come with Agesa 1.1.9.0. Beta BIOS should be out some time in January and official bios mid-February.
Agesa 1.1.0.0 D is coming next with a beta BIOS around mid-December. It's basically an official update for 400-series with Vermeer CPUs."

[Redline9k]

I had the same issues with the random restarts in each and every BIOS version. For me it seems to be the combination of FMAX Enhancer (in PBO settings) and the curve optimizer. I saw that the FMAX Enhancer uses less voltage on single core workloads (e.g. Cinebench R20/R23 it showed 5.25 GHZ clock with 1.41V only). I was always able to finish Cinebench etc. without an issue, but light workloads (e.g. windows defender scan or starting a memory benchmark in AIDA) lead to a crash.
Also it crashed with each and every number deposited in the curve optimizer, does not matter if it is -2 or -15 if FMAX enhancer is on. Also FMAX enhancer leads to a lower single core score in cinebench (Fmax Enhancer on: 613 points in C20, Fmax Enhancer off: 648) due to stretching clocks like crazy. HWinfo reported 5.25 ghz on each core after some gaming, but the effective clock was about 4.7ghz (even when cinebenching single core). Without FMAX enhancer it is around 5ghz effective clock with -15 all core curve optimizer. Without FMAX Enhancer the core voltages goes up to 1.49v now in single threaded workloads and no more crashes.

Long story short: Try to disable Fmax Enhancer while using the curve optimizer.

Kind regards

That's indeed some interesting behaviour. I'm not using the curve optimiser myself, but I have explicitly enabled a +200mhz boost limit with PBO. I don't remember enabling Fmax Enhancer either, but checking the bios settings .txt I have saved on my BIOS USB shows that it is indeed enabled. It's the same behaviour for me; light workloads crashing but anytime I properly load the CPU it works fine.
I've now been on the previous 2502 for hours now without issue at all, whereas previously it'd crash 5+ times in an hour while idle. So it's definitely some recent change that has been made.
I'll try to disable Fmax Enhancer later on and see how I go. Thanks for your post though -- looks like Fmax Enhancer might be a little broken. If all looks well after playing around I may re-update to the test version and keep Fmax Enhancer disabled to see how we go. Seems possible the clock-stretching behaviour might be causing issues with voltages?