PSA: 535 system stability concerns.

These new drivers are more aggressive in some way that is provoking memory related bsods and Device Loss crashes in D3D12 and Vulkan,

This is not a Driver Bug pər-ˈsē, the prevailing crash dumps are either revealing off by one bit errors in the UMD (Device Loss source), or in critical kernel drivers (BSOD Source).

Off by one bit errors can be antagonised by software operating more aggressively, but they are not software caused, the error can be caused in 2-3 different ways,

• one is the CPU tells IMC to write Value X but it Writes value Y because of signal noise on the trace between imc and memory
  
• another is the Value in memory corrupts on row refresh
  
• and another concerns the Command Rate, where a latch times out on the first bit and its never written in the first place.

At this point, I suspect the onslaught of users playing DSR and RDR2 that are encountering crashes have hit this stability concern and need to tweak their system timings from the start,

and it also means that some people are going to be pissy because their expensive XMP ram is not stable, and likely never ever was.


Let me leave off with a reminder that when you deviate from the specifications the CPU vendor lists on the CPU's support page, the system is in the stability grey area, no board vendor or cpu vendor can guarantee the stability of "Performance Memory" or a minor cpu overclock across Kernel driver updates, the OS itself or when taken into context against the CPU's own Eratta.

For Context, my own home theatre build which I have overclocked and was using 1N command rate for the last year stable, became unstable with the 535.98 update.

It is stable again with CR2, please confirm your system configurations before reporting bugs, a driver can and often does reveal system weaknesses in the wake of significant alterations deep in the driver, and it is incredibly unlikely that a memory management bsod with the parameter 0x401 - 0x403 is software caused.

 

Good & stable OC is a must IMhO.
I have UV for CPU - mostly 4.2GHz at mere 0.94v *
DRAM is at peak of Zen3+ capabilities = 3800MHz 1900IF CL16 1T but i managed to get it very stable.

Note:
Additional Fan for DRAM helped me a lot if not of this i can go max 3733MHz CL16 in summer (Hot days).

* I have also 4.4GHz & 4.6Ghz manual per CCX with low V. for max FPS if needed.

 

Accented letters cause the word's tense to change in Latin. It's per (e sound, not ö sound) se (e sound, not ay sound), not per-say.

 

At this point, I am content that you only suspect the RDR2 issue is related to these stability concerns you've mentioned.
I can play DSR and RDR2 just fine. I just have to use DX12 instead of Vulkan in RDR2.

Good read.

 

Have Rockstar updated RDR2 DLSS to the latest builds yet? I wanted to replay it after I got my RTX 3090 (which I did only for testing early on), but I am waiting for them to update their DLSS version before actually playing the whole game again.

 

I think they haven't touched the dll since the original release. However, you can always swap the .dll for the newest one yourself, but it may introduce some bugs like haloing foliage, blocky shadows, a broken sharpening slider and an overcontrasted image on camera movement.

You might also want to use the Lenny Mod Loader, as there are countless mods that improve the DLSS/TAA/FSR implementation.

 

I'm experiencing something weird in Diablo 4, randomly the game may "lose the window focus" for a brief second and then it's fine. Dunno if it's the divers or the game, my system has no OC except RAM XMP.

 

I have been having a potentially related issue for almost a year. Sometimes, seemingly out of nowhere, a background window (usually a file explorer window) pops into focus without any user interaction whatsoever. Although I never lost focus of a game or media player this way but it often (one or twice a week but that's enough if it goes on for a year or so) disturbs my web browsing with Chrome.

 

Yeah that's exactly what i'm seeing while playing Diablo, a popup window coming in focus for a brief second and then dissappears, what is it?

 

I don't know. I would have provided an explanation or workaround had I known one. Sorry.
But it's good to know I am not alone with this issue. My subconscious mind often wondered if this could be some kind of "virus" (malicious software generating fake user input to execute things so fast after a focus change that I can't possibly see what it did). But I shrugged that feeling off (it's not an elegant way for a "virus" sophisticated enough to survive the anti-virus protections for long).
I guess this is a generic Win11 bug (I suspect it happens when the background window refreshes based on some trigger event, like a change in the contents of the directory opened in the file explorer [even just the modification time of any item] -> it might think I am interested to see that things changed...).

 

Pretty sure that's a Windows 11 Explorer bug.

 

I think the window is a command prompt..

 

Tried DLSSTweaks with it after initially testing when I got the 3090, got a warning from Rockstar Launcher that tampering with game files may result in a banned account as soon as I ran the game. Decided to uninstall the game and the launcher altogether after that. So, I will wait until they officially update it themselves.

 

Perhaps get the Steam version when it's on sale, there are no such problems there.

 

Can confirm that.

 

I don't know if it's related but I got two internal parity errors in the event viewer when playing Hogwarts Legacy. Yesterday I could play for a while with no issues whatsoever, now it just crashes to desktop. The second time it crashed I got these errors. They had disappeared for a while after I slightly increased the vcore of my CPU, but now they are back.

 

I am not sure if it's related but today I installed the recently released game DESORDRE.
It's a DX12 UE5 game and it keeps crashing with "device hung and recovered" errors.
This seems a common thing with UE5 on my system, I had similar crashes during The Finals Beta test.
What's worrying is that it doesn't seem to be a GPU stability issue, even after removing the overlock the driver would still crash and the same happened with The Finals a few months ago.
Thre is also a whole thread on the UE forums filled with people reporting the same issue: https://forums.unrealengine.com/t/ue5-gpu-crashed-or-d3d-device-removed/524297/138

 

Auto-correct did this, i knew it was pər-ˈsē but just let the corrector do its thing lol.

 

Vulkans a weird one, you can get Device Loss errors using perfectly legal implementation if memory is a bit iffy, or if you lack it in some way, you can also get it if you have HAGS on.

HAGS can also mask timing related glitches that you'd otherwise usually use Barriers

This is could be windows terminal opening with the games own exception handler service, try setting the system back to Conhost, which implements hidden operation (Windows terminal doesn't)

This is an artifact of the Skylake++++++++++ architecture, you can stop these by raising the ringbus voltage.

 

I'm going to increase the voltage a bit on my memory and see if that helps to prevent another crash. If it doesn't, I will have to relax the timings.