WinXP Drivers for GTX 9xx and 10xx video cards

[TrevMUN]

19 minutes ago, asdf2345 said:

The 980 TI, Titan, Quadro, and lower end cards are newer than the driver, so there's no way it could support them without driver modification

Oh, that explains a lot, then. It's not a matter of .inf modding, is it? Either way, it would be a better bet to follow the guide from Matt's Repository, since that opens up the newer drivers (up until nVidia discontinued XP/XP64 support) and provides more opportunities to find a driver that works best for our machines.

[asdf2345]

54 minutes ago, TrevMUN said:

Oh, that explains a lot, then. It's not a matter of .inf modding, is it? Either way, it would be a better bet to follow the guide from Matt's Repository, since that opens up the newer drivers (up until nVidia discontinued XP/XP64 support) and provides more opportunities to find a driver that works best for our machines.

They're all Maxwell cards, so it'll probably work with just inf modification, but not stock

Edited January 29, 2021 by asdf2345

[UCyborg]

I never said anything else about 368.81 besides the problem regarding monitor detection, which I brought up because that's something that could've been looked into if they were to be modded.

I haven't actually used any other version with GTX 750 Ti on XP x64. I encountered 2 other problems that could be related to graphics driver, but can't say for certain if they are (Serpent browser freezing after resuming from standby/sleep and DOOM 3 consuming a lot more memory on XP than on newer OS).

There's another known issue with said 64-bit version of the driver and multi-monitor setups, known to occur with GTX 9xx series cards, which someone patched.

 

Edited January 31, 2021 by UCyborg
Elaborated 2 problems

[Dixel]

14 hours ago, UCyborg said:

I never said anything else about 368.81 besides the problem regarding monitor detection, which I brought up because that's something that could've been looked into if they were to be modded.

I haven't actually used any other version with GTX 750 Ti on XP x64. I encountered 2 other problems that could be related to graphics driver, but can't say for certain if they are (Serpent browser freezing after resuming from standby/sleep and DOOM 3 consuming a lot more memory on XP than on newer OS).

There's another known issue with said 64-bit version of the driver and multi-monitor setups, known to occur with GTX 9xx series cards, which someone patched.

 

Well , that's at least 3 by you and a couple by me . I can confirm all mentioned by you and add these to the list.

1 - Significantly decreased performance starting with 35.x

2 - Problems with idle modes on GTX980/970 , GTX780 (new revision, reissue with another chip) , Titan Black.

Though, the first GTX 780 (from May 2013) and the original Titan did not behave oddly during idle.

Mind you , this happened on both , Vista and XP , so XP has nothing to do with it , it's the drivers themselves.

My opinion : 344 series are the best for XP. If someone wants newer , 364 drivers seem to be OK , but not as fast 344.11 and they need to be modded

to work with 9xx series.

For some raging publishers : I didn't test with Titan X , so no comment.

EDIT : I did not test with lower end cards .

 

Edited January 31, 2021 by Dixel

[UCyborg]

I didn't detect any perceptible performance difference between 368.81 and 344.11 on my GTX 750 Ti. That problem with Serpent browser hanging after PC out of stand by...isn't actually specific to it...seems can effect any D3D app that was running at the time...and it happens with 344.11 as well.

And the DOOM 3 memory problem, which might not necessarily be a problem, it has something to do with reading through 1 GB worth of archives containing textures. When handles to those files are closed, shown memory usage goes down.

Running Unigine Heaven 4.0 with 344.11 driver has shown visual artifacts, at least in D3D9 mode, I haven't tried OpenGL mode.

Seems I'm not missing anything crucial for my use cases by keeping 368.81 on XP.

[Dixel]

16 hours ago, UCyborg said:

I didn't detect any perceptible performance difference between 368.81 and 344.11 on my GTX 750 Ti. That problem with Serpent browser hanging after PC out of stand by...isn't actually specific to it...seems can effect any D3D app that was running at the time...and it happens with 344.11 as well.

And the DOOM 3 memory problem, which might not necessarily be a problem, it has something to do with reading through 1 GB worth of archives containing textures. When handles to those files are closed, shown memory usage goes down.

Running Unigine Heaven 4.0 with 344.11 driver has shown visual artifacts, at least in D3D9 mode, I haven't tried OpenGL mode.

Seems I'm not missing anything crucial for my use cases by keeping 368.81 on XP.

I'm not sure how GTX750ti is relevant to this discussion , being lower end and from another series. Yeah , I know it's the first gen. of Maxwell , but still another series . Btw, artifacts in Unique Heaven oftenly tell that somethig's wrong with the videocard itself . No such issue for me with high end 900 series and 344 drivers. What about the other issue with displays ? Besides , what version of XP you were using ? 32 or 64 bit ?

[UCyborg]

No glitches with 368.81 driver in Unigine Heaven on XP x64 here, the GPU is fine. Though this particular case isn't the main point, just that the older drivers (and software in general) can have bugs that newer don't (and vice versa, though the goal is to eliminate issues with newer releases).

[Dixel]

@UCyborg, I wanted to ask you , does Nvidia's HDMI work fine for you with XP64 and these drivers ? 

EDIT . My mobo has PCI express gen.2 . Windows Vista shows gen.2 , but with XP64 Nvidia CPL shows that it's gen.1 only.

Any idea why is that ? Also , tried some DX9 games and the performance is much worse on the same hardware and with same driver versions.

I'm not going to use it , of course , it's just out of curiosity .

Thanks. 

Edited February 22, 2021 by Dixel

[UCyborg]

No idea, I only use DVI ports.

[Dixel]

13 hours ago, UCyborg said:

No idea, I only use DVI ports.

I have found the source of the problem with HDMI . In case you will ever use Nvidia's HDMI (Azalia) or any other HD drivers you have to install KB901105 x64 before the drivers. V3 worked for me . It's very important to install this KB before. Still , I do not know why the Nvidia CPL shows gen.1 PCI express .

Could you look at yours ?

[Rod Steel]

14 hours ago, Dixel said:

I do not know why the Nvidia CPL shows gen.1 PCI express .

Maybe it is power saving under low load? Check latest GPU-Z, it newest version still working with XP. In GPU-Z it is question mark button "PCI-E render test" near Bus Interface line.

Also it is hard to diagnose problem when you don't tell what exact your hardware is.

Also it can be that Nvidia CPL just show incorrect information - that happened sometimes with different programs - they are not flawless.

P.S. Here is link to EVGA forum with program to benchmark PCI-E bandwitch on Nvidia cards:https://forums.evga.com/PCIE-bandwidth-test-cuda-m1972266.aspx

Edited February 23, 2021 by Rod Steel

[Dixel]

16 hours ago, Rod Steel said:

Maybe it is power saving under low load? Check latest GPU-Z,

Thanks for the suggestion , actually GPUZ was the first thing I've tried .

But any version of GPU-z starting from 0-8-5 hangs up the whole windows XP64 system to the point of hard reset . Please keep in mind that I'm talking about 64 bit , not x86. Older versions of GPU-z work fine , but they do not properly detect this GPU. By the way , Nvidia CPL must show the highest version of PCI express available, it shouldn't depend on the current load. And it shows gen. 2.0 on this exact PC with Vista x64. My card is not EVGA , I try to avoid EVGA as much as I can , I've had quite a few problems with their quality. It's just Nvidia (no other branding, from their website) , reference design. 

I do notice poor 3d performance with games on XP64, the same game is like 2x times faster with Vista.

Hardware: 

GPU: Geforce GTX Titan 6GB mfg date February 2013, 

Motherboard: Fujitsu Siemens D3012-A11-GS3 DDR3 RAM 16GB with Intel chipset Q45 mfg date April 2011.

CPU: Intel® Xeon® Processor X3370 12M Cache, 3.00 GHZ (Quad).

PSU: FSP (Fujitsu Siemens Power) 600W ,rare design and size ,supplied with this PC, fits only Fujitsu/Siemens housings.

[Dixel]

On 2/23/2021 at 9:38 PM, Rod Steel said:

P.S. Here is link to EVGA forum with program to benchmark PCI-E bandwitch on Nvidia cards:https://forums.evga.com/PCIE-bandwidth-test-cuda-m1972266.aspx

Well , I've finally found some time to test and compare , thanks for the programme , it's quite interesting . The results are : with XP the bandwith is approx. 25-30% lower than with Vista. It's strange because it actually tells that gen. 2.0 is used , but not to the maximum (with XP64). 

[Rod Steel]

On 2/24/2021 at 1:08 PM, Dixel said:

I do notice poor 3d performance with games on XP64, the same game is like 2x times faster with Vista.

Yea this is not how it suppose to be. XP64 is faster then Vista. Hm. It looks like something works weird in your system with XP64 install. Maybe drivers on motherboard, or maybe in BIOS of this rare motherboard something not set right, like not forced PCI-E 2.0 mode, or processor in power-saving mode constantly while running XP? Maybe you need reinstall XP-64 from different ISO, i sent u an email about that.

I also see infamous Fujitsu Siemens motherboard. It have incredibly bad reputation in Overclockers community many years ago. It look like it is a server motherboard and BIOS in F-S servers stuff is usually archaic and obsolete. I really low opinion about they motherboards. In reviews that i read like 15 years ago it was the worst in test from like 15 motherboards. Maybe you should consider upgrade on I7-3770K or FX-8350 - this is fast systems that officially support XP, DDR3 and still good.

[Rod Steel]

On 2/24/2021 at 1:08 PM, Dixel said:

But any version of GPU-z starting from 0-8-5 hangs up the whole windows XP64 system to the point of hard reset .

Hm. Maybe it cause to old GPU driver version. Or more fresh Visual Studio needed to be installed.

Here they state that for XP GPU-Z ver 2.22.0 is recommended.

Or maybe your XP-64 version screwed up...