• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Best PHYSX carD?

S

soulesschild

Supreme [H]ardness
Joined
Feb 18, 2007
Messages
6,176
Now with the addition of PhysX+ATI, what's the best cheap NVIDIA card to get for physX?

I've heard 9600GT is a good bet, any others?
 
Any 8-series or above with at least 32 cores (shaders) and 256MB memory should get the job done.
 
GotNoRice said:
Any 8-series or above with at least 32 cores (shaders) and 256MB memory should get the job done.
Click to expand...

"get the job done" does not equal Best.

Shure a low end card will work.. but the PhysX performance will suck.
 
I might have missed an announcement to the contrary, but as far as I know, using an ATI card with a NVidia for PhysX isnt something you should be spending money on.
Maybe the patch to re-enable PhysX hardware support will stay ahead of the game, but thats a risk.
I look forward to being enlightened :)
 
Nenu said:
I might have missed an announcement to the contrary, but as far as I know, using an ATI card with a NVidia for PhysX isnt something you should be spending money on.
Maybe the patch to re-enable PhysX hardware support will stay ahead of the game, but thats a risk.
I look forward to being enlightened :)
Click to expand...

The patch work with the latest Nvidia drivers....

The thing about patching the drivers is that the only thing they are changing is probably a couple jumps so it goes to the same place whether or not it sees a non-Nvidia card...

Either that, or it just jumps past that non-NVidia test code and doesn't even bother with it... most likely a lot simpler solution.

It can't be just a byte-replacement patch as it wouldn't stay the same for different driver releases. It almost has to look for specific code or specific calls and changes it to either do something else.. like nothing when it tries to check for a non NVidia card.
 
cyclone3d said:
"get the job done" does not equal Best.

Shure a low end card will work.. but the PhysX performance will suck.
Click to expand...

The author specifically mentioned "cheap". No one besides the OP can quantify exactly what that means. I was merely posting a minimum recommendation.
 
my recommendations:

9600gt - minimum suggested
9800gt - good price/ performance
gts250 - solid pick for somewhat better performance but only if cost over above is marginal
gtx260/275 - money is no object
gtx285 - only if you find a ridiculously good deal

so anything in terms of most "cost-effective" would be in the 9600gt-9800gt-gts250 category. so look for the best deal you can find. ebay works.
 
kumquatsrus said:
my recommendations:

9600gt - minimum suggested
9800gt - good price/ performance
gts250 - solid pick for somewhat better performance but only if cost over above is marginal
gtx260/275 - money is no object
gtx285 - only if you find a ridiculously good deal

so anything in terms of most "cost-effective" would be in the 9600gt-9800gt-gts250 category. so look for the best deal you can find. ebay works.
Click to expand...

Thanks! That's what I was looking for :)
 
If you run a dedicated PhysX GPU, buy big...

If you intend to really use PhysX, get the most powerful single-GPU (currently PhysX only uses one GPU) card you can. PhysX scales well, so more GPU simply results in more performance. A lot of games will do great with a 9800GT or higher, but newer games are using PhysX more, so a more powerful GPU will only help in the future.
 
I went the 9800gt rout. Best Buy had the PNY low-power version (no supplemental PCI-e power needed) for $111 -60 in points.
 
To give an idea of the performance using a GTX260 with and without an 8800GT for PhysX, here are the numbers clocked and unclocked.
(these have been posted before)
They demonstrate that at least for Batman running at 1080p a stock clocked 8800GT is enough.
Cryostasis is much tougher to get high enough framerates but I have nothing to report as I havent tested it.
I agree with kumquatsrus recommendations.


Results for E8400 @ 4.1GHz, GTX260 clocked/unclocked, 8800GT clocked/unclocked for PhysX.
These are to test the effect of adding an 8800GT and what clocking each card gives when paired together.
all tests are at 1080p res

All game settings to highest, vsync off, PhysX set to highest.
(both cards are mild overclock versions so slightly higher default clocks)

GTX260 clocked (720/1475, 1208), no PhysX card, PhysX set to highest.
AA Min, Max, Average
0xAA 34, 76, 56
2xAA 30, 69, 51
4xAA 29, 68, 50
8xAA 29, 63, 47
8QAA 24, 60, 38
16xAA 27, 64, 47
16QAA ran very very slow in places and crashed at the end.

GTX260 stock (626/1350, 1053), 8800GT stock (625, 1566, 900)
AA Min, Max, Average
0xAA 41, 108, 73
4xAA 41, 92, 67
8xAA 38, 93, 64
8QAA 32, 82, 54
16xAA 36, 90, 63
16QAA 30, 75, 50

GTX260 clocked (720/1475, 1208), 8800GT stock (625, 1566, 900)
AA Min, Max, Average
0xAA 46, 109, 81
16xAA 41, 96, 70
16QAA 35, 87, 58

GTX260 clocked (720/1475, 1208), 8800GT clocked a bit (723, 1753, 916)
AA Min, Max, Average
0xAA 46, 111, 81
16xAA 43, 100, 71
16QAA 36, 86, 58
Click to expand...


At 0xAA, the 8800GT jumps framerate from 34 min to 46 min, 35% extra.
At 16xAA, the 8800GT jumps framerate from 27 min to 41 min, 52% extra.
The GTX260 doesnt have enough memory to run 16QAA and highest PhysX without the 8800GT.

Clocking the main gfx card gives a good performance boost.
There is no value in clocking the 8800GT (PhysX).
Clearly you cant test PhysX on highest without an extra PhysX card so these results wont apply directly.
It does demonstrate that for this game running at 1080p, an unclocked 8800GT is enough.
 
Will a Geforce 8 GTS 640mb with 96 shader cores be good enough, or not worth it?
 
my dedicated 9800gt rocked batman: aa with physx on high. it was a great gaming experience too.
 
Lamina said:
Will a Geforce 8 GTS 640mb with 96 shader cores be good enough, or not worth it?
Click to expand...

It should raise framerate, I wouldnt expect it to lower the framerate.
Performance shouldnt be too far off my 8800GT results, maybe even the same...
Clocking the 8800GT while running PhysX didnt help the average framerate at all, so its already fast enough at 1080p alongside a GTX260.
Maybe yours GTS is just as good?

Then again it might also depend what you pair it with, what is your main gfx card?
 
The main card is a Geforce 280 GTX. I'm waiting for a new motherboard in the mail, will try it out when I get that. If its not good enough, I can always sell it and get a 9800GT.
 
I just got a 9800GT (112 stream processors) for a PhysX card. Should apir up nicely with my current GTX260.

I'll post up my results when it gets here tomorrow.
 
Unknown-One said:
I just got a 9800GT (112 stream processors) for a PhysX card. Should apir up nicely with my current GTX260.

I'll post up my results when it gets here tomorrow.
Click to expand...

That's what I am running on a 22" monitor. So far, it's worked very well.
 
I always recommend a 9800gt as a minimum. :)

BTW, I tested Batman's benchmark running at full detail with the GTX 295 doing physx (max setting) and then with the gtx 9800 (had to try twice as the card is going back for RMA) doing it.
The 9800gtx+ beat the GTX 295 only configuration easily. I would get the numbers I wrote on a piece of paper later tonight if interested.
I wanted to do a test with a GTX280 when I got the 295 but my crappy USPS delivery girl forgot about the card so I had to hunt her down to get it and when I get home I only had time to pick up the GTX 280 and send it to the new owner as promised. :(
 
Last edited:
Modred189 said:
That's what I am running on a 22" monitor. So far, it's worked very well.
Click to expand...

Well, I'm running triple-22" monitors (5040x1050 total resolution), so as you can imagine, my poor little GTX260 doesn't have much power left over for PhysX. :p

I expect a seriously massive speed boost using the 9800GT.
 
They didn't test a GTX260, nor did they test a 9800GT...they also didn't test resolutions above 2560x1600.
 
Unknown-One said:
They didn't test a GTX260, nor did they test a 9800GT...they also didn't test resolutions above 2560x1600.
Click to expand...

true, but you can kind of extrapolate what the results might be since a 9800gt would be between a gts250 and 9600gt and a gtx260 would be between a gtx275 and gts250. still, i would imagine they would be very close to each other in terms of performance gains.

lol @ testing res above 2560x1600.
 
kumquatsrus said:
lol @ testing res above 2560x1600.
Click to expand...

SoftTH, TripleHead2Go, and Eyefinity all allow resolutions GREATLY in excess of 2560x1600. It would be nice to know.

Even the low-end of triple-head, 5040x1050, is more total pixels than 2560x1600.
 
Unknown-One said:
SoftTH, TripleHead2Go, and Eyefinity all allow resolutions GREATLY in excess of 2560x1600. It would be nice to know.

Even the low-end of triple-head, 5040x1050, is more total pixels than 2560x1600.
Click to expand...

yeah i understand, but those are extremely niche resolutions. even 4mp isn't that widely used. maybe later on it will become more prevalent, but right now that isn't happening. all the major tech sites only benchmark performance up to 2560x1600 as of now. i can only imagine how much more difficult and time consuming it will be to test with even more of the various configurations that multimon brings ( number of mon: 3/5/etc., 1680x1050/1920x1200/1920x1080/2560x1600/etc.) aside from the standard single mon resolutions. and triplehead has been around for several years and websites still don't use the higher resolutions afforded by it to test as part of their benchmarking.
 
Unknown-One said:
Oh really? Lots of 5040x1050 benchmarks here, including Batman Arkham Asylum...
http://www.widescreengamingforum.com/wiki/index.php/WSGF_Benchmarking

There are quite a lot of triple-head users, more than you might think.
Click to expand...

that's a site dedicated to multimon dude (i wouldn't consider it a major tech site either). and yes, i know that there are plenty of triplehead users. but that is a drop in the bucket compared to the vast majority of single mon users. which is why toms, anand, techspot, hard, firingsquad, driverheaven, guru3d, and others test at up to 2560x1600 since that is the highest res. on a single monitor. therefore, only a site like wsgamingforum will have multimon benchmarks.
 
Last edited:
I believe that a GTX 285 would be the best single GPU PhysX card and a GTX 295 would be the best dual GPU PhysX card. However you'd be spending waaaaay too much wattage for that particularly narrow application.
 
A low power 9800gt (the EE editions), would suffice, however, at about ~120USD (20usd over normal 9800gt price). it doesn't take that extra PCIe 6 pin, and runs off of the PCIe slot power.
 
jeremyshaw said:
A low power 9800gt (the EE editions), would suffice, however, at about ~120USD (20usd over normal 9800gt price). it doesn't take that extra PCIe 6 pin, and runs off of the PCIe slot power.
Click to expand...

The "green" 9800GT's are way cheaper than that. You can usually get them for around $95

Example:
http://www.amazon.com/gp/product/B0021HQCKM/ref=ox_ya_oh_product

As an added bonus, there's no sales tax when buying from Amazon.com in every US state (except Washington), and there's free shipping available. So that's $95 said-and-done, out the door.
 
Last edited:
As an Amazon Associate, HardForum may earn from qualifying purchases.
I use a BFG gtx275 OC with a BFG 8800gt OC2 for physx. I tried out Batman, and it was supremely better with the 8800gt. I reckon that you don't need a higher end card than an 8800gt to do physx well, but obviously bigger is better. That being said, I will be relegating my 275 to physx when the next gen nvidias come out, just so that I am future proofed.
 
Unknown-One said:
The "green" 9800GT's are way cheaper than that. You can usually get them for around $95

Example:
http://www.amazon.com/gp/product/B0021HQCKM/ref=ox_ya_oh_product

As an added bonus, there's no sales tax when buying from Amazon.com in every US state (except Washington), and there's free shipping available. So that's $95 said-and-done, out the door.
Click to expand...

+1.
I got mine for $51 at Best Buy when it was on clearance and didn't think twice. I almost feel bad about using it as a PhysX card, with my HTPC using a 4650...
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
Is the performance of a green edition card fine compared to the regular 9800gt? cos i'm thinking of getting a Green Edition over a regular edition and hoping that a 520w psu will suffice.
 
I think the GTS 250 is the best bang for the buck. I have a GTX275, GTS250 combo and very happy with what I get. really can go wrong with ~70$ on a GTS 250
 
Zardoz said:
I think the GTS 250 is the best bang for the buck. I have a GTX275, GTS250 combo and very happy with what I get. really can go wrong with ~70$ on a GTS 250
Click to expand...

I can't find it for that cheap anywhere...
 
One of our Polish tech sites have done Physx benchmarking with cards using various numbers of sp:

http://www.egielda.com.pl/?str=art&id=5447-4

If you look at them it shows that using something like 9400 will hurt your physx performance and that physx sometimes introduces another bottleneck in gaming.
 
You must log in or register to reply here.
Back
Top