VOGONS


Pascal GPU support in WinXP

Topic actions

Reply 160 of 169, by Trashbytes

User metadata
Rank l33t
Rank
l33t
SupraGSX wrote on 2026-11-15, 08:18:
Dimos wrote on 2026-11-15, 07:56:

It is mentioned a few posts above that Heaven 4.0 needs an installer from 2013 to run on Xp. Also in supragsx' s case the 3dm06 benchmark produces again some perplexing results. The win 7 score is much higher than the Xp one, although the cpu score is higher on Xp. This again seems pretty strange. Another thing i noticed is the mention of the 1569 mhz boost clock. I really hope that this in fact is the base clock, not the boost one, cause every 1080 ti boosts at least up to the mid 1800s - low 1900s.

No they don't I don't know where you're getting that information? Base clock of 1080ti is 1480mhz. Boost is about 100mhz higher on most cards. I can boost my evga ftw3 card up to about 1950mhz but founders cards are no where near that.

Yeah, the blower cards quite frankly suck at boost clocks, its the 3rd party 1080Ti's with huge beefy coolers that are able to boost higher.

AI is inevitable, there is nothing you can do to stop it, no Ironman snap to save you from the Basilisk.

Reply 161 of 169, by Shagittarius

User metadata
Rank Oldbie
Rank
Oldbie
SupraGSX wrote on 2026-11-15, 08:14:
Shagittarius wrote on 2026-11-15, 07:22:

On the other hand my cpu should be running circles around his core2 even with his overclock, so why do I get the same performance as a 780? Also the 3.0 benchmark reports the full 12gb of ram on xp 32bit. I shall have to try to get 4.0 installed I suppose, how does one go about that?

As I already said above, its ran in qemu the core 2 duo is emulated. It's actually an am5 9950x3d. Also it's a multi GPU system so 1080ti isn't even in primary slot. I did both benches in qemu to isolate every variable but os. Do the numbers mean anything by themselves as a score nope. Can you compare the two scores to one another yep. Did we learn anything? Probably not. Heaven is basically identical, 3dmark06 is 20 percent slower. Depending on which you believe it's either at parity or 20 percent slower. Regardless it's orders of magnitude faster than a 780ti if that's your original concern 😂

Again 4.0 is native extract and install no magic to it.

I personally don't trust emulated environments for benchmarking. I'm also not saying there is a problem, I still believe I am bottlenecked by my CPU.

Running 3DMARK06 in Win10 with the latest compatible official drivers on the same machine produces 22,467. Running in XP 32bit with your modified drivers but at 1920x180, 8x Anti, 16x Aniso produces 24,311. Clearly limited by my CPU.

Any gains I get from this are going to be from resolution and eye candy. I don't have a history for the 780ti at higher settings to compare to.

EDIT: I Overclocked my X5690 to 4.3 and here it is added into the comparison:

The attachment Overclock.jpg is no longer available

Reply 162 of 169, by Dimos

User metadata
Rank Newbie
Rank
Newbie
SupraGSX wrote on 2026-11-15, 08:18:
Dimos wrote on 2026-11-15, 07:56:

It is mentioned a few posts above that Heaven 4.0 needs an installer from 2013 to run on Xp. Also in supragsx' s case the 3dm06 benchmark produces again some perplexing results. The win 7 score is much higher than the Xp one, although the cpu score is higher on Xp. This again seems pretty strange. Another thing i noticed is the mention of the 1569 mhz boost clock. I really hope that this in fact is the base clock, not the boost one, cause every 1080 ti boosts at least up to the mid 1800s - low 1900s.

No they don't I don't know where you're getting that information? Base clock of 1080ti is 1480mhz. Boost is about 100mhz higher on most cards. I can boost my evga ftw3 card up to about 1950mhz but founders cards are no where near that.

Of course, a FE card drops its clocks after a while due to thermal throttling, but until that happens (which happens really quick), even for a couple of minutes, it boosts pretty higher.

Cpu: Intel i7 4790k
Gpu: Gigabyte Xtreme Gaming 980 ti
Ram: G-Skill Trident X F3-2400C10Q-16GTD
Mobo: Gigabyte Z97x Gaming 5
Hdd: T-Force Vulcan Z 512 gb Ssd
Psu: Corsair CX650
Soundcard: Creative SB Audigy RX
Os: Windows XP Sp3 x86

Reply 163 of 169, by Dimos

User metadata
Rank Newbie
Rank
Newbie
Shagittarius wrote on 2026-11-15, 15:30:
I personally don't trust emulated environments for benchmarking. I'm also not saying there is a problem, I still believe I am b […]
Show full quote
SupraGSX wrote on 2026-11-15, 08:14:
Shagittarius wrote on 2026-11-15, 07:22:

On the other hand my cpu should be running circles around his core2 even with his overclock, so why do I get the same performance as a 780? Also the 3.0 benchmark reports the full 12gb of ram on xp 32bit. I shall have to try to get 4.0 installed I suppose, how does one go about that?

As I already said above, its ran in qemu the core 2 duo is emulated. It's actually an am5 9950x3d. Also it's a multi GPU system so 1080ti isn't even in primary slot. I did both benches in qemu to isolate every variable but os. Do the numbers mean anything by themselves as a score nope. Can you compare the two scores to one another yep. Did we learn anything? Probably not. Heaven is basically identical, 3dmark06 is 20 percent slower. Depending on which you believe it's either at parity or 20 percent slower. Regardless it's orders of magnitude faster than a 780ti if that's your original concern 😂

Again 4.0 is native extract and install no magic to it.

I personally don't trust emulated environments for benchmarking. I'm also not saying there is a problem, I still believe I am bottlenecked by my CPU.

Running 3DMARK06 in Win10 with the latest compatible official drivers on the same machine produces 22,467. Running in XP 32bit with your modified drivers but at 1920x180, 8x Anti, 16x Aniso produces 24,311. Clearly limited by my CPU.

Any gains I get from this are going to be from resolution and eye candy. I don't have a history for the 780ti at higher settings to compare to.

EDIT: I Overclocked my X5690 to 4.3 and here it is added into the comparison:

The attachment Overclock.jpg is no longer available

You ran the benchmark in Win 10 with the same settings as in Win Xp?

Cpu: Intel i7 4790k
Gpu: Gigabyte Xtreme Gaming 980 ti
Ram: G-Skill Trident X F3-2400C10Q-16GTD
Mobo: Gigabyte Z97x Gaming 5
Hdd: T-Force Vulcan Z 512 gb Ssd
Psu: Corsair CX650
Soundcard: Creative SB Audigy RX
Os: Windows XP Sp3 x86

Reply 164 of 169, by Shagittarius

User metadata
Rank Oldbie
Rank
Oldbie
Dimos wrote on 2026-11-15, 18:07:
Shagittarius wrote on 2026-11-15, 15:30:
I personally don't trust emulated environments for benchmarking. I'm also not saying there is a problem, I still believe I am b […]
Show full quote
SupraGSX wrote on 2026-11-15, 08:14:

As I already said above, its ran in qemu the core 2 duo is emulated. It's actually an am5 9950x3d. Also it's a multi GPU system so 1080ti isn't even in primary slot. I did both benches in qemu to isolate every variable but os. Do the numbers mean anything by themselves as a score nope. Can you compare the two scores to one another yep. Did we learn anything? Probably not. Heaven is basically identical, 3dmark06 is 20 percent slower. Depending on which you believe it's either at parity or 20 percent slower. Regardless it's orders of magnitude faster than a 780ti if that's your original concern 😂

Again 4.0 is native extract and install no magic to it.

I personally don't trust emulated environments for benchmarking. I'm also not saying there is a problem, I still believe I am bottlenecked by my CPU.

Running 3DMARK06 in Win10 with the latest compatible official drivers on the same machine produces 22,467. Running in XP 32bit with your modified drivers but at 1920x180, 8x Anti, 16x Aniso produces 24,311. Clearly limited by my CPU.

Any gains I get from this are going to be from resolution and eye candy. I don't have a history for the 780ti at higher settings to compare to.

EDIT: I Overclocked my X5690 to 4.3 and here it is added into the comparison:

The attachment Overclock.jpg is no longer available

You ran the benchmark in Win 10 with the same settings as in Win Xp?

Yes default settings for the registered version. I ran it in Win10 with the overclock and I get 29,047...so still a little behind the performance with your drivers in XP 32bit. Makes sense beacause of bigger cpu overhead with win10 I think, while dx 9 holds back the 1080ti.

Reply 165 of 169, by agent_x007

User metadata
Rank Oldbie
Rank
Oldbie
Shagittarius wrote on 2026-11-15, 07:22:

I shall have to try to get 4.0 installed I suppose, how does one go about that?

The attachment Unigine.webp is no longer available

^Left one works under XP, right one will not.

Google drive link for 2013 version : LINK

Note : Heaven shows 4GB as VRAM on 1080 Ti, because that is max it can show.
Modern PC/Win10 :

The attachment Heaven 4.0 DX11 1080p.png is no longer available

Reply 166 of 169, by Shagittarius

User metadata
Rank Oldbie
Rank
Oldbie

Dx10 on xp would be a cool project.

Reply 167 of 169, by RayeR

User metadata
Rank Oldbie
Rank
Oldbie

Isn't already there via WineD3D? (Ok, it's not native D3D but goes via OpenGL, AFAIK)...

Gigabyte GA-P67-DS3-B3, Core i7-2600K @4,5GHz, 8GB DDR3, 128GB SSD, GTX970(GF7900GT), SB Audigy + YMF724F + DreamBlaster combo + LPC2ISA

Reply 168 of 169, by Shagittarius

User metadata
Rank Oldbie
Rank
Oldbie

Actually a better project would just be bringing native EAX to Win7 32bit, though I guess both are pretty impossible.

Reply 169 of 169, by Andalu

User metadata
Rank Newbie
Rank
Newbie
RedCharles wrote on 2026-11-07, 16:08:
https://github.com/RedCharles9001/P2200-XP-Driver […]
Show full quote

https://github.com/RedCharles9001/P2200-XP-Driver

I started with 382.69, and it didn't work at 4k 120hz or 5120x1440 at 120hz. These were the errors:

ASUS 4K120: The old driver’s parser skipped CTA video codes above 34, including the ASUS monitor’s 4K120 code, 118. The driver had the code but skipped it.

Samsung 5120 × 1440 at 120 Hz: The monitor declared three EDID blocks, but the XP driver’s internal capture contained only the first two.

Codex fixed both of them after a few tries, but if you had it set up to do these resolutions in general, there's some miscommunication going on between the driver inf and the monitor edid on newer monitors.

I'm having codex investigate further.

Thank you so much for sharing the modified Forceware driver that allows the Quadro P2200 to work in XP.

Unfortunately, I don't have a 4K monitor with a refresh rate higher than 60Hz:
https://ibb.co/tMkDkZ7R

I also tried it on a 34" monitor (3440 x 1440), and it allows the maximum possible refresh rate (165Hz):
https://ibb.co/DDyC4VFv

Not tested with games.