Hack gets PhysX and Radeon working again

[SamHughe]

That only confirms my findings. PhysX calculations alone just do not require a huge amount of GPU power to be processed efficiently. In my own tests between a 8800GS and my GTS250 used as a dedicated PhysX card the difference 1 - 2 fps at most in games that use PhysX. The most significant thing the GTS250 did over the 8800GS was boost my Vantage score by 500 points.

Yep. That's what I'm saying. It totally confirms your findings. As far as the 500 vantage score boost goes, it is probably either the clocks (8800gs 550/1375 vs. GTS250 738/1836) or the memory bandwidth 8800gs 30.8 GB/s vs GTS250 70.4 GB/s). So even though some people claim they are based on the same chip, gts250 is a quite bit of improvement over 8800 series. You wouldn't wanna spend money on a 8800 for PhysX right know unless there is one laying around.

[CraptacularOne]

Yep. That's what I'm saying. It totally confirms your findings. As far as the 500 vantage score boost goes, it is probably either the clocks (8800gs 550/1375 vs. GTS250 738/1836) or the memory bandwidth 8800gs 30.8 GB/s vs GTS250 70.4 GB/s). So even though some people claim they are based on the same chip, gts250 is a quite bit of improvement over 8800 series. You wouldn't wanna spend money on a 8800 for PhysX right know unless there is one laying around.

I'm pretty sure the increase in Vantage came from the extra stream processors on the GTS250 as it has 128 of them where the 8800GS has 96. The 8800GS I had was run at 750/1800/2000. It was a little overclocking demon for sure. In any case, I'm happy to confirm my findings with a review that mirrors my results. The bottom line is that you just don't need a very powerful card if it's just going to be a dedicated PhysX card

[STEvil]

Anything 8600GT or up is good enough for PhysX really.

[KoHaN69]

It seems to me that the 240 GT would be the best physX card due to its low power consumption
http://www.newegg.com/Product/Produc...-138-_-Product

Yes?

also, do you all overclock your cards beyond artifacting? what i mean is, if a video card is artifacting due to high clocks, cant it still be used a a physx card without negative results?

[SamHughe]

It seems to me that the 240 GT would be the best physX card due to its low power consumption
http://www.newegg.com/Product/Produc...-138-_-Product

Yes?

also, do you all overclock your cards beyond artifacting? what i mean is, if a video card is artifacting due to high clocks, cant it still be used a a physx card without negative results?

I'd get this one.
http://www.newegg.com/Product/Produc...82E16814500131

It's cheaper one slot solution and comes with 3D glasses. Noise has never been a problem with NVida cards.

[viccan]

So let me ask then... Having started out with an agia ppu w/ 128megs that wouldn't even work w/ batman, to a 8800gt w/ 640 megs and being somewhat sluggish at times, not for all scenes but for some ( like when papers would fly as you'd run over them while fighting the deformed 1st experimental black gate prisoner) then a 9800 gt w/ 512 megs same stuttering to the 285 gtx with no slowdown anywhere to be seen??
I mean I know the review said so but I have all three cards in my possession and these where my results... Now possibly its due to the 2560 resolution I run at or ?????? I dunno Im just stating my experience and not some reviews. The 8800 and 9800 had slight stuttering (again not every frame but in certain high physX scenes) while the same box with the 285 had none.. Doesnt make much sense..

[SamHughe]

So let me ask then... Having started out with an agia ppu w/ 128megs that wouldn't even work w/ batman, to a 8800gt w/ 640 megs and being somewhat sluggish at times, not for all scenes but for some ( like when papers would fly as you'd run over them while fighting the deformed 1st experimental black gate prisoner) then a 9800 gt w/ 512 megs same stuttering to the 285 gtx with no slowdown anywhere to be seen??
I mean I know the review said so but I have all three cards in my possession and these where my results... Now possibly its due to the 2560 resolution I run at or ?????? I dunno Im just stating my experience and not some reviews. The 8800 and 9800 had slight stuttering (again not every frame but in certain high physX scenes) while the same box with the 285 had none.. Doesnt make much sense..

You're right it doesn't make any sense. Maybe it was your 285 running the whole thing and not 4870X2 + 285. There is only one way to solve this puzzle. You do the same Batman:AA benchmarks with all four cards and post the results. Stuttering means you are getting less than 20 FPS maybe less that means what your 4870x2 can do by itself. Maybe with the 8800 and 9800 the physX wasn't enabled properly. I don't think Ageia PPU should even work with the new NVdia drivers.

[STEvil]

I'm going to vote that PhysX was broken on the 8800/9800.

Make sure to do taskmanager/fraps graphs.

[gamervivek]

So let me ask then... Having started out with an agia ppu w/ 128megs that wouldn't even work w/ batman, to a 8800gt w/ 640 megs and being somewhat sluggish at times, not for all scenes but for some ( like when papers would fly as you'd run over them while fighting the deformed 1st experimental black gate prisoner) then a 9800 gt w/ 512 megs same stuttering to the 285 gtx with no slowdown anywhere to be seen??
I mean I know the review said so but I have all three cards in my possession and these where my results... Now possibly its due to the 2560 resolution I run at or ?????? I dunno Im just stating my experience and not some reviews. The 8800 and 9800 had slight stuttering (again not every frame but in certain high physX scenes) while the same box with the 285 had none.. Doesnt make much sense..

If all those three cards were only being utilised for PhysX then the memory size is reduntant as a performance criteria.

If the first card has 640 megs of memory then you had an older G80 8800gts which has 96 shader processors, G92 9800gt has 112 of them clocked higher than the G80 8800gts and gtx285 has 240 of them.

From my limited understanding the resolution should not have an effect on the physics calculation and that the physics calculations are primarily affected by the computational power of the gpu, i.e. the shaders.

But the shaders processors and their organisation aren't the same in all the three GPUs, the 112 shaders arranged in 7 groups of 16 shaders in 9800gt and 240 shaders of the gt200 being arranged in 10 groups of 24 each.

The recommended settings call for a 9800gtx for physX at high, meaning an extra shader cluster compared to the 9800gt and is in a way similar to running gta4 on a dual core versus a tri core or higher.

[viccan]

If all those three cards were only being utilised for PhysX then the memory size is reduntant as a performance criteria.

If the first card has 640 megs of memory then you had an older G80 8800gts which has 96 shader processors, G92 9800gt has 112 of them clocked higher than the G80 8800gts and gtx285 has 240 of them.

From my limited understanding the resolution should not have an effect on the physics calculation and that the physics calculations are primarily affected by the computational power of the gpu, i.e. the shaders.

But the shaders processors and their organisation aren't the same in all the three GPUs, the 112 shaders arranged in 7 groups of 16 shaders in 9800gt and 240 shaders of the gt200 being arranged in 10 groups of 24 each.

The recommended settings call for a 9800gtx for physX at high, meaning an extra shader cluster compared to the 9800gt and is in a way similar to running gta4 on a dual core versus a tri core or higher.

Well, this makes total sense. Thank you for relieving me the headache of running benches w/ all 4 cards again just to show Im not crazy... I do in fact have the 9800 gt and not gtx and the 8800 is about 3-4 years maybe?
But given this scenario to askimos & purecains question of what cards to match up,. it does make a difference for high settings (at least in batman)...

Thanks for the post gamervivek tremendous help

[adpr_02]

Hey, how does the Nvidia GPU's performance affect the overall performance with physx enabled?

Is it the same above a certain gpu speed, or would a gtx260 give much better fpx than a 9800?

I'm thinking of getting a 50$ 8600gt. Should I spend more and get a 9800/8800 and oc them?

I have a HD5850 @ 1020/1270

Edit: Did a bunch of reading would anything above a gt 220 make any difference?

[CraptacularOne]

Hey, how does the Nvidia GPU's performance affect the overall performance with physx enabled?

Is it the same above a certain gpu speed, or would a gtx260 give much better fpx than a 9800?

I'm thinking of getting a 50$ 8600gt. Should I spend more and get a 9800/8800 and oc them?

I have a HD5850 @ 1020/1270

Edit: Did a bunch of reading would anything above a gt 220 make any difference?

I wouldn't bother with an 8600GT it's a horribly slow card to begin with and only just meets the requirements to process PhysX with it's 32 stream processors. My GTS250 runs everything that uses PhysX (all 5 games...lol) just fine. You don't even remotely need something like a GTX260 just for PhysX.

[Metroid]

Keeping my 8800GT at bay was a good decision.