• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

PhysX Not working

Soldato
Joined
26 Apr 2003
Posts
5,746
Location
West Midlands
I have a 470GTX and a 8800GT as my PhysX card. However when I set the 8800GT to be my PhysX card in the control panel it just just uses the CPU in game. There is no activity on that GPU what so ever according to GPUZ. I'm currently running the latest beta drivers but I've also tried several previous drivers too.

I've tested in Mafia II and FluidMark.

Windows 7 Home Premium 64bit
Gigabyte P55 USD5
4GB XMS3 DDR RAM
i5 760 @ 4.0Ghz
EVGA 470 GTX Superclocked
Standard 8800GT 512MB
Coolmaster SilentPro 700w

All devices have the latest drivers.

Here is my settings in the nVidia control panel.

physxv.png
 
Last edited:
So it appears that Batman works fine using the 8800GT. I must have picked the only two packages that don't!
 
Did you get any difference in 3dmark vantage or 3dmark11 physx scores? I have a 9800GTX that i usefor physx and the scores never changed. Batman however did seem a bit smoother to play.
 
ok, so you have a gtx470 as a render card, and a 8800 for a physX card.
but i think im correct when i say, if you crossfire / Sli, both cards will run at the speed of the slowest card.
If you had one card at 850mhz core 1gb GDDR5, and one 600mhz 512mb GDDR3.
then it would knock the faster card so that both are 600mhz 512mb GDDR3.....

so why would you want to make the gtx470 run at the same apouling speed as the 8800?
 
ok, so you have a gtx470 as a render card, and a 8800 for a physX card.
but i think im correct when i say, if you crossfire / Sli, both cards will run at the speed of the slowest card.
If you had one card at 850mhz core 1gb GDDR5, and one 600mhz 512mb GDDR3.
then it would knock the faster card so that both are 600mhz 512mb GDDR3.....

so why would you want to make the gtx470 run at the same apouling speed as the 8800?

nvidia allow people to slot in a second, slower, card and use it to do all the PhysX processing so that the main card is solely used for the graphics. It doesn't match the speeds like SLI.
 
It does if the card is in SLI but the 8800 in this case isnt. Its solely being used for the purpose of physx... You can do the same with something like a 9600 or 8200 but it wont reduce performance of a GTX470 down to the performance of a 9600gt etc...
 
yepp, and the link above is with a slower card, not in SLI, just set to be a dedicated physx card, but still made the performance suffer.

watch it and see. no sli bridge, he explains about it being dedicated physx, and you see the benchmarks, worse with a slower card as physx
 
an 8800GT should be fine for PhysX. I would look for at least 100+ CUDA Cores if you are looking for a dedicated PhysX card. The 32 cores of that 8600 GTS just isn't going to cut it for PhysX.

Also, benching Vantage with PhysX enabled to the NV GPU will be misleading if you don't understand what is going on. For example, if you have GTX 580 and GTX 550 and you run Vantage while having NV PhysX enabled, it will produce a better score with just the GTX 580 in the system. This is normal since if you are performing the same PhysX test on GTX 580 and GTX 550, of course the GTX 580 will score better. The more accurate test as some have mentioned would be in actual games using PhysX. Also mentioned already, PhysX load can vary from game to game and one configuration may be better for one game but not another.
 
Right the issue I'm having is that it keeps thinking that the 8800GT is my main card so keeps trying to display an image on that rather than from my 470GTX. As you can see my GTX is the main card and the 8800GT is set to PhysX so I'm at a loss here.
 
ok, so you have a gtx470 as a render card, and a 8800 for a physX card.
but i think im correct when i say, if you crossfire / Sli, both cards will run at the speed of the slowest card.
If you had one card at 850mhz core 1gb GDDR5, and one 600mhz 512mb GDDR3.
then it would knock the faster card so that both are 600mhz 512mb GDDR3.....

so why would you want to make the gtx470 run at the same apouling speed as the 8800?

You're right in some sense, but wrong at the same time. Since the cards are not in SLi together both cards can run at their own native speeds. It however depends on how well the cards can perform in rendering/physx. If the main card is rendering too fast, it must wait for the physx card to finish the physx calculations before it can render the result picture to send to the monitor. Hence why it can cause a loss in performance if the physx card is too slow to keep up with the main card. The 8800GT should be fast enough not to hold the 470 down.

@OP It's weird that some games don't work and some games do. I really don't know why it's acting like this :confused:.
 
Last edited:
It doesn't want to boot from my main card now, the 470GTX so I have to hook my monitor upto the 8800GT. This is what I'm seeing in the control panel. As you can see the 470GTX has vanished.

physx2.png



So to sum it, it started off as some games using the 8800GT as the Physx card and some ignoring it and passing it onto the CPU. Now the system thinks that the 8800GT is my only card and will only display an image on it. Despite my having my TV conncted to the 470GTX via HDMI.
 
You're right in some sense, but wrong at the same time. Since the cards are not in SLi together both cards can run at their own native speeds.

I don't see how I am wrong but Ok... I never claimed the clockspeeds would be the same either. Clockspeeds and this argument is irrelevant as clockspeeds can be set to different values in SLI as well.

All you need is a decent performing card. Like I said earlier, I would recommend 100+ CUDA Cores for the applications that I see out today.



Back to the OP's issue, does this also happen when you are only using one display?
 
I don't see how I am wrong but Ok... I never claimed the clockspeeds would be the same either. Clockspeeds and this argument is irrelevant as clockspeeds can be set to different values in SLI as well.

All you need is a decent performing card. Like I said earlier, I would recommend 100+ CUDA Cores for the applications that I see out today.



Back to the OP's issue, does this also happen when you are only using one display?

Oh god I just realised I quoted the wrong guy! I meant gamerdude1989 not you Nick! My bad LOL :p.

EDIT: Fixed the post now, sorry about that Nick :rolleyes:
 
Last edited:
Out of interest - is your rendering GPU (470) in the slot closest to the CPU and the 8800GT in a 2nd slot further away from the CPU?
 
BTW do you have anything connected to the 8800GT when gaming on the 470? it should be fixed now but back along the PhysX card had to have a display connected to it to work could still be some legacy issues with that if nothing is plugged into it.
 
Back
Top Bottom