• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

GPUZ 2.45.0 causes BSODs

Joined
Apr 21, 2011
Messages
54 (0.01/day)
Location
England
System Name \\RED
Processor Intel Core i9 Extreme 9980XE (Skylake-X)
Motherboard ASUS WS X299 Sage
Memory 128GB - 8 x 16GB - 1 x CMR64GX4M4A2666C16 + 2 x CMR32GX4M2A2666C16
Video Card(s) 1 x GeForce RTX 3080 + 1 x GeForce RTX 2080 + 1 x GeForce GT 1030
Storage 1 x Intel SSDPEKNW010T9 + 5 x Intel SSDPEKKW256G8 + 1 x Intel SSDPED1D280GA + 6 x Samsung 860 EVO
Display(s) 2 x Dell UP3017 (DP) + 1 x Dell 2407WFP (Digital) + 1 x Dell 2007FP (Digital)
Power Supply Thermaltake TPG-1250D-T
Software Windows 10 x64 Enterprise V10.00 Build 19045 22H2
When I run GPUZ 2.45.0 it causes a BSOD as it's starting. Windows minidump attached. Please can this be fixed?
 

Attachments

Last edited:
Hi,
At what point does it cause bsod
It opens just fine I use the rog themed


1649600792753.png

Ran the render test for a little while to.
 
What's the system configuration?
 
What's the system configuration?
Most of what you need should be in the .DMP + .TXT files, the GPU was an S3 Trio 3D/2X and attached should tell you the rest.

I might be able to tell more in I had the driver .SYS + .PDB files.

Update: When I changed the GPU to a GeFore FX 5200 there was not a BSOD so it seems the trigger is the S3 Trio 3D/2X.

Confirmed: with the S3 Trio 3D/2X installed GPUZ causes a BSOD on start-up every time.

Looking at the new dump things are clearer, but I need the .PDB to tell more.

Looking at https://www.felixcloutier.com/x86/movd:movq then the movq xmm0 is SSE2 which is not present on these CPUs. I suspect the sensible solution is to build a new driver that does not use SSE2

After reading https://docs.microsoft.com/en-us/wi...l/using-floating-point-or-mmx-in-a-wdm-driver then I wonder if the GPUZ driver does this, does it?

The full kernel .DMP is 15 MB when zipped and it won't let me upload it.
 

Attachments

  • S3 Trio 3D 2X.png
    S3 Trio 3D 2X.png
    67.8 KB · Views: 160
  • WINDBG.png
    WINDBG.png
    56.5 KB · Views: 145
Last edited:
Why are you trying to run a piece of software written in 2022 against an OS released in 2003, a CPU released in 2001 that does not support SSE2, and a GPU originally released in NINETEEN NINETY SIX?

Just... what?
 
Last edited by a moderator:
One could argue that S3Trio isn't a GPU, but there should not be blue-screens with this fine program. :)

Precisely, any program that triggers a BSOD on any system should be fixed. It's not just the S3Trio that triggers this, an NVidia NV17 [GeForce4 MX 440] also does. Does it count as a GPU?
 

Attachments

  • nVidia NV17 [GeForce4 MX 440].jpg
    nVidia NV17 [GeForce4 MX 440].jpg
    123.3 KB · Views: 155
  • nVidia NV17 [GeForce4 MX 440].png
    nVidia NV17 [GeForce4 MX 440].png
    46.4 KB · Views: 102
  • RAA.png
    RAA.png
    76.2 KB · Views: 119
Precisely, any program that triggers a BSOD on any system should be fixed. It's not just the S3Trio that triggers this, an NVidia NV17 [GeForce4 MX 440] also does. Does it count as a GPU?
But what if that fix would compromise stability on newer platforms like Win10/11? Its not that i like these (actually using win7 + i9 7900x setup atm), but win XP definitely is getting long in the tooth so to speak. For retro hardware i usually pick retro software, but for something like MX4 which doesnt support DirectX 9.0c at all (dx 8 and 9 is what win xp is good for, gaming wise). For dx7 win98 se would be better choice (btw it also runs dos natively so it could be a neat opportunity to combine retro dos and windows gaming in one place).
 
But what if that fix would compromise stability on newer platforms like Win10/11?

This is something that only the GPUZ developer can answer. There is no way it can compromise W11 as there is no x32 W11 release.

With my drivers I have builds for x32 NT4 + x32 W2K ->W10+ x64 XP-64 -> W11 + ia64 Itanium 2003 -> 2008R2 and even one for DEC Alpha NT4 -> W2K! My x64 Driver (SIVX64.sys) meets the latest Windows 11 HVCI Kernel Mode compatibility requirements and also happily runs on XP-64/2003 x64. For GPUZ it's only the x32 driver that needs to be fixed as all CPUs capable on running Windows x64 have SSE2.

I suspect all the GPUZ developer needs to do is set option not to use SSE2, recompile the driver and get it signed. As I recall GPUZ itself used to use SSE2 and crash, but that was fixed and I guess they overlooked the GPUZ driver.

Given what https://docs.microsoft.com/en-us/wi...l/using-floating-point-or-mmx-in-a-wdm-driver specifies I would not be surprised if the current GPUZ x32 driver needs fixing for systems that have SSE2.
 
Last edited:
But what if that fix would compromise stability on newer platforms like Win10/11? Its not that i like these (actually using win7 + i9 7900x setup atm), but win XP definitely is getting long in the tooth so to speak. For retro hardware i usually pick retro software, but for something like MX4 which doesnt support DirectX 9.0c at all (dx 8 and 9 is what win xp is good for, gaming wise). For dx7 win98 se would be better choice (btw it also runs dos natively so it could be a neat opportunity to combine retro dos and windows gaming in one place).
XP is I believe 20 yo software.
 
Precisely, any program that triggers a BSOD on any system should be fixed. It's not just the S3Trio that triggers this, an NVidia NV17 [GeForce4 MX 440] also does. Does it count as a GPU?
Of course; nVidia coined the term GPU with the original GeForce256, so GeForce4 MX440 surely is a GPU.
 
Does this build fix the bluescreen? KMD is compiled with /arch:IA32 instead of "not set"
 

Attachments

Does this build fix the bluescreen? KMD is compiled with /arch:IA32 instead of "not set"

Thank you for looking into this. I think that at some stage Microsoft changed the default from /arch:IA32 to /arch:SSE2 which would explain things.

The good news if that the BOSD has gone and the bad news is that GPUZ failed to report the clocks, but I suspect you don't care.
 

Attachments

  • GPU-Z 2.45.1.png
    GPU-Z 2.45.1.png
    217.1 KB · Views: 155
Thanks for confirming, much appreciated.

I think that at some stage Microsoft changed the default from /arch:IA32 to /arch:SSE2 which would explain things.
Exactly my thoughts

is that GPUZ failed to report the clocks, but I suspect you don't care.
GeForce MX is just too old, also AGP makes things much harder to debug and repro. Maybe something for a rainy day
 
Thank you for looking into this. I think that at some stage Microsoft changed the default from /arch:IA32 to /arch:SSE2 which would explain things.

The good news if that the BOSD has gone and the bad news is that GPUZ failed to report the clocks, but I suspect you don't care.
If you don't mind me asking, do you use this vintage system for some specialized workload that a newer machine cannot do? Or is it just hobbyist tinkering? Curiosity got the best of me :toast:

Still glad to see w1zz got your back, even if GPU-Z's usefulness can be limited on hardware going that far back. Back then graphics cards were quite rudimentary compared to the super advanced processing chips of today. Cheers
 
If you don't mind me asking, do you use this vintage system for some specialized workload that a newer machine cannot do? Or is it just hobbyist tinkering? Curiosity got the best of me :toast:
GeForce MX is just too old, also AGP makes things much harder to debug and repro.

I use them mainly to check my SIV utility works on all versions of Windows from NT 4.00 (it's half OK on NT 3.51) up to the latest Windows 11 x64 Enterprise V10.00 Build 22593 which I run on my ASUS Maximus Z690 Hero. Another consideration I can't sensibly use my main development system to test my drivers, some of my systems are setup to make testing easy for me such as my Dual P-3 with 5 GPUs!

Back in the day I never really had the time to implement all the support I wished to, but now I am retired I am retrospectively adding this. When adding support for GPUs that can't be reported using NVidia's NVAPI, AMD's ATIADL or Intel's oneAPI I like to compare what SIV reports with other utilities, hence I tried GPUZ. As W1zzard highlighted doing this without appropriate test systems is tricky.
 

Attachments

  • Dual P-3 with 5 GPUs.png
    Dual P-3 with 5 GPUs.png
    153.9 KB · Views: 83
I use them mainly to check my SIV utility works on all versions of Windows from NT 4.00 (it's half OK on NT 3.51) up to the latest Windows 11 x64 Enterprise V10.00 Build 22593 which I run on my ASUS Maximus Z690 Hero. Another consideration I can't sensibly use my main development system to test my drivers, some of my systems are setup to make testing easy for me such as my Dual P-3 with 5 GPUs!

Back in the day I never really had the time to implement all the support I wished to, but now I am retired I am retrospectively adding this. When adding support for GPUs that can't be reported using NVidia's NVAPI, AMD's ATIADL or Intel's oneAPI I like to compare what SIV reports with other utilities, hence I tried GPUZ. As W1zzard highlighted doing this without appropriate test systems is tricky.

Sweet! It all makes sense now. You develop a system info tool too! That is super cool, and that's definitely one exotic system you have there... Dual Pentium IIIs :eek:

:lovetpu:
 
One could argue that S3Trio isn't a GPU

I suspect GPUZ would argue this. For completeness I felt I should validate the S3 and GPUZ did not do as well as it might.
 

Attachments

  • At least the BSOD is no more.png
    At least the BSOD is no more.png
    78.6 KB · Views: 82
0x2f0d is due to the root certificates not being updated on your OS, so https doesn't work
 
does tpu load in ie ?
 
does tpu load in ie ?
Sort of, it whinges and when I select continue get as attached. Things work if you don't use HTTPS, as you can see SIV was able to check for updates.

In general I never use my old systems to look at websites and have not tried at all to make things work.

I suspect the sensible option would be for GPUZ to generate a file that can to used on a W10/W11 system which is what CPUZ does.

1649769779665.png
 

Attachments

  • 1649770102156.png
    1649770102156.png
    23.8 KB · Views: 69
Last edited:
and when I select continue
Problem is that I can't just silently ignore HTTPS errors, or there would be no security.

Maybe a http-based fallback could be an option
 
Problem is that I can't just silently ignore HTTPS errors, or there would be no security.

Maybe a http-based fallback could be an option
option/prompt you can click to ignore the https error?
 
Yes, that should work. I just checked a fully patched Vista x64 system and that's OK. I know unpatched Visia + W7 systems can't run my SHA2 signed x64 driver and suspect they will also have issues with HTTPS:

I just checked my unpatched W7 x64 test system and validation worked, but the GPUZ-V2 driver failed start. I don't have an unpatched Vista system.

When I looked at the screen this reminded me I meant to ask you, "When I select a GPU how do I know which one is which?" I have one system with 4 x EVGA GTX 980 and with GPUZ have no idea which one is which in the dropdown. On this system appending BFG/XFX would work, but not on my other systems.

1649771533289.png
NVIDIA GeForce GTX 980 @ bus n.png


Something on the lines of NVIDIA GeForce GTX 980 @ bus n in the dropdown would be good. You may need to make the dropdown wider.
 
Last edited:
Back
Top