9800GTX+ installed

[lopri]

Originally posted by: Janooo
I think it makes sense it's slower. A part of the GPU does physics and graphics are not done by a full GPU. Do you know a ratio between physics/graphics allocation?

And remember that there are lots more happening on-screen so there are more things to render. I have been very curious how GPUs handle physics. It's almost like load-balancing if drivers handling the amount of physics, and if game devs put codes for GPU physics there and GPU try to render everything then that'll be very hard for a GPU. I would like to see some screenshots (like before-after).

 

[SSChevy2001]

@keysplayr2003

At 2.4Ghz UT3 might be CPU limited, but I've never tried. Your score looks better, but it's still to low even overclocked. My mins are your averages, so there's got to be something wrong. Also your running faster memory speeds and shader clock, which should more than make up for the difference in resolution. I got rivatuner set to autolevel the fan up to 70c. The 64c you quoted is under the stock fan speeds?

Anyway nice to see the 9800gtx+ has room for overclocking. Physx still looks like it needs room for improvement, but I can't complain.

My last tests were just outdoor not indoors, in doors I get much higher FPS. I just did a short 5min pass running back and forth getting flag and capture. Here's the result of that run.

Min Max Avg
34 82 51.187

@lopri There's no need to compare the difference it's night and day. Frame rates drop under 10fps real quick without physx. The game is beyond playable and lowering the resolution doesn't change the fps at all.

Also currently the heatray physx map is slightly different from the regular heatray map, so I doubt you can compare the speeds. If I had to guess though the same map without physx would at least be 2 times faster, but if you can spare the speed it's worth it.

 

[Hauk]

Originally posted by: nRollo

Originally posted by: Jhonka
@nRollo

Hope to see more impressions of the card's performance.
Do you know when this card is going to be available?

All I can say about that is "If I have it, you can VERY soon. Very, very soon".

I was told there's no NDA on these cards, but let me check if that includes launch date.

Also nRollo, even a vague answer as to 55nm GTX 280's arrival would be appreciated. I know someone who's considering an EVGA and hopes to step-up...

 

[taltamir]

Originally posted by: keysplayr2003

Originally posted by: SSChevy2001
Any advice you can give with UT3 Physx would be cool. I got a 8800GTS 512 with the modified drivers 177.41 and physx. It works fine for Vantage, but it only worked once for UT3. After I tried reloading UT3 and the drivers many times, but no luck. Hopefully when Nvidia releases the physx system software version 8.06.18 it will work again.

As of today, physx is only enabled for 9800GTX(+), GTX260 & GTX280. Shortly, Nvidia will release support for 8800 series (G80 & G92).

I'll be running some Physx benches tonight as well on the 9800GTX+.

I had it working last month on a 8800GTS 512 using hacked drivers...

I wonder how it affected my Mass Effect playthrough? I showed that mass effect is CPU bound on a 4850 with an E8400 @ 3.6... back then I was at 3.0ghz only... And mass effect DOES use physX.

I can't test it myself since the 8800GTS 512 is in the PC I built my brother and is being shipped to gunnison colorado right now. Anyone with physX enabled card and an E8400 minds testing mass effect?

 

[Keysplayr]

Originally posted by: SSChevy2001
@keysplayr2003

At 2.4Ghz UT3 might be CPU limited, but I've never tried. Your score looks better, but it's still to low even overclocked. My mins are your averages, so there's got to be something wrong. Also your running faster memory speeds and shader clock, which should more than make up for the difference in resolution. I got rivatuner set to autolevel the fan up to 70c. The 64c you quoted is under the stock fan speeds?

Anyway nice to see the 9800gtx+ has room for overclocking. Physx still looks like it needs room for improvement, but I can't complain.

My last tests were just outdoor not indoors, in doors I get much higher FPS. I just did a short 5min pass running back and forth getting flag and capture. Here's the result of that run.

Min Max Avg
34 82 51.187

@lopri There's no need to compare the difference it's night and day. Frame rates drop under 10fps real quick without physx. The game is beyond playable and lowering the resolution doesn't change the fps at all.

Also currently the heatray physx map is slightly different from the regular heatray map, so I doubt you can compare the speeds. If I had to guess though the same map without physx would at least be 2 times faster, but if you can spare the speed it's worth it.

Have you recorded a demo? If so, would you be able to send it to me if I hooked you up with an FTP account? It sounds like your just doing a manual walkthrough. I'm looking for something repeatable. The timedemo I am running has literally thousands of objects being tossed around in front of me by twisters, and by the destruction of buildings.

Let me know.

FYI, the demo shouldn't be more than 2MB's, and that is about 150 seconds worth.

Thanks.

 

[Keysplayr]

Originally posted by: taltamir

Originally posted by: keysplayr2003

Originally posted by: SSChevy2001
Any advice you can give with UT3 Physx would be cool. I got a 8800GTS 512 with the modified drivers 177.41 and physx. It works fine for Vantage, but it only worked once for UT3. After I tried reloading UT3 and the drivers many times, but no luck. Hopefully when Nvidia releases the physx system software version 8.06.18 it will work again.

As of today, physx is only enabled for 9800GTX(+), GTX260 & GTX280. Shortly, Nvidia will release support for 8800 series (G80 & G92).

I'll be running some Physx benches tonight as well on the 9800GTX+.

I had it working last month on a 8800GTS 512 using hacked drivers...

I wonder how it affected my Mass Effect playthrough? I showed that mass effect is CPU bound on a 4850 with an E8400 @ 3.6... back then I was at 3.0ghz only... And mass effect DOES use physX.

I can't test it myself since the 8800GTS 512 is in the PC I built my brother and is being shipped to gunnison colorado right now. Anyone with physX enabled card and an E8400 minds testing mass effect?

Does Mass Effect support Physx using and actual PPU? Ageia card? Or is it for GPU driven Physx? I almost bought this game the other day, but it got mixed reviews and wasn't sure I'd like it.

 

[n7]

Interesting stuff.

I'm a decently avid UT3 player (as you'd know keys ), so this is somewhat of interest to me.

I assume you cannot play these maps online?

Offline play only?

 

[bryanW1995]

Originally posted by: nRollo

Originally posted by: Jhonka
@nRollo

Hope to see more impressions of the card's performance.
Do you know when this card is going to be available?

All I can say about that is "If I have it, you can VERY soon. Very, very soon".

I was told there's no NDA on these cards, but let me check if that includes launch date.

they originally said july 16, I see no reason to think otherwise if rollo already has his.

 

[SSChevy2001]

@keysplayr2003

Never made a timedemos before, but I recorded one anyway. I ran a fraps benchmark while recording. Here's the fraps files and demo, which is about 6.5 min long.

http://www.sendspace.com/file/h4rzop

Still ended up with a 45fps average @ 1080p max settings.

Also mass effect doesn't support hardware physx.

 

[Keysplayr]

Downloaded it. Thanks Chevy, I'll check it out.

 

[SSChevy2001]

Here are 2 more benchmarks with demo files and fraps if you want.

Heatray Physx with 31 bots 1080p maxed settings avg 43.479
Tornado again with default bots 1080p maxed settings avg 44.224

http://www.sendspace.com/file/rtzeyr

 

[nRollo]

Originally posted by: Janooo
I think it makes sense it's slower. A part of the GPU does physics and graphics are not done by a full GPU. Do you know a ratio between physics/graphics allocation?

I just did a FRAPS 60 second session of constant battling and got 26 minimum, 62 maximum, and 38 fps average for the Heatray level.

So it is slower than non PhysX, but I'd choose to play with this as much as possible. This is a difference you don't have to look hard for, changes the whole game.

(which kind of surprises me, I never even considered an Ageia card when they launched)

Not bad for a ~$200 card- no stand alone card needed.

 

[nRollo]

Originally posted by: SSChevy2001
Here are 2 more benchmarks with demo files and fraps if you want.

Heatray Physx with 31 bots 1080p maxed settings avg 43.479
Tornado again with default bots 1080p maxed settings avg 44.224

http://www.sendspace.com/file/rtzeyr

I bet your Intel @ 3.5 is lifting you above my Phenom at 2.5, would be my hunch anyway.

(as UT3 is multithreaded for quad)

 

[zod96]

Is the 9800GTX+ any quieter than the original 9800GTX? Also how's the heat on the card as well?

 

[SSChevy2001]

@nRollo

CPU might be helping a little, also clock for clock the 8800GTS 512 is actually faster than 9800GTX.

Here's a website that confirms it also.
http://en.expreview.com/2008/0...rce-9800gtx-review/12/

 

[taltamir]

could it be that they loosened timing on the ram?

 

[Keysplayr]

Originally posted by: SSChevy2001
@nRollo

CPU might be helping a little, also clock for clock the 8800GTS 512 is actually faster than 9800GTX.

Here's a website that confirms it also.
http://en.expreview.com/2008/0...rce-9800gtx-review/12/

Lots of contraversy in the conversation appended to that review. I wouldn't call anything "confirmed" at this point.

 

[Keysplayr]

Chevy, I tried your demo (The first one you provided a link for). Between your demo, and the one I am running, I think there has to be about 3 times more debris in my demo. You have only a few small portions in your demo where buildings are being destroyed and the twisters are picking up and hurling debris. In mine, it's almost a constant throughout the demo. Add to this the fact I am running a higher res than you and rendering an additional 230,400 pixels at all times, I think our results are pretty accurate respectively.
I'll check out your other two demos tonight. Thanks for sending them.

At stock speeds, I am still getting around 27-35 fps on average in my demo, and about 35-42 avg. fps in your first demo.
I think it's pretty amazing how much faster the GPU runs Physx over the CPU. CPU driven Physx is hammered down into the single digits.

I'm going to try the GTX280 tonight as well.

 

[Kuzi]

Originally posted by: nRollo
...
This stuff does come at a cost with a single 9800GTX+ at 19X12 0X16X Very High Quality I could definitely tell the game was running slower than without PhysX. It was still playable and fun, but noticeably slower than without PhysX.
...

So it improves physics and slows down frame rate. Kind of like when 3D cards first released on the PC. You get improved graphics but lower fps, sometimes a slideshow. At least it's a good start.

Maybe we can expect GT200 based cards to accelerate physics without much loss in fps?

 

[Keysplayr]

Originally posted by: Kuzi

Originally posted by: nRollo
...
This stuff does come at a cost with a single 9800GTX+ at 19X12 0X16X Very High Quality I could definitely tell the game was running slower than without PhysX. It was still playable and fun, but noticeably slower than without PhysX.
...

So it improves physics and slows down frame rate. Kind of like when 3D cards first released on the PC. You get improved graphics but lower fps, sometimes a slideshow. At least it's a good start.

Maybe we can expect GT200 based cards to accelerate physics without much loss in fps?

It does slow down framerate with Physx. But remains in the very playable framerate area. At 1920x1200 with an enormous amount of action on the screen, (it's actually overdone IMHO, focusing on as much Physx content as possible in this demo). I'm averaging 27-35 with this initial driver. I'll be trying out the GTX280 later tonight. I'll update accordingly.
But first, I'll be trying out 9800GTX+ SLI with Physx.

 

[Kuzi]

Originally posted by: keysplayr2003

Originally posted by: Kuzi

Originally posted by: nRollo
...
This stuff does come at a cost with a single 9800GTX+ at 19X12 0X16X Very High Quality I could definitely tell the game was running slower than without PhysX. It was still playable and fun, but noticeably slower than without PhysX.
...

So it improves physics and slows down frame rate. Kind of like when 3D cards first released on the PC. You get improved graphics but lower fps, sometimes a slideshow. At least it's a good start.

Maybe we can expect GT200 based cards to accelerate physics without much loss in fps?

It does slow down framerate with Physx. But remains in the very playable framerate area. At 1920x1200 with an enormous amount of action on the screen, (it's actually overdone IMHO, focusing on as much Physx content as possible in this demo). I'm averaging 27-35 with this initial driver. I'll be trying out the GTX280 later tonight. I'll update accordingly.
But first, I'll be trying out 9800GTX+ SLI with Physx.

Nice. I'm very interested in how GTX280 would do with Physics, I would assume the hit in fps would be lower than on the GTX+. You know the GTX280 is way more powerfull and has much higher memory bandwidth (not sure if that matters in this case).

 

[SSChevy2001]

@keysplayr2003

Yeah Tornado level is a pain when your trying to focus on just physx. Problem is if you focus on breaking walls, is the bots kill you. If you get to close to the tornado you die. Worst FPS I get is right in the tornado 25-27fps. Also I hate CTF.

Heatray Physx though is another story, as the hail is what causes the slowdown in that level. I'm more interested in how your GTX280 and 9800GTX+ does in that level.

Edit - About the review do you know of any other reviews that counter it? I thought I seen another review will similar results, but I can't remember where. Really though there small anyway.

 

[Keysplayr]

Originally posted by: SSChevy2001
@keysplayr2003

Yeah Tornado level is a pain when your trying to focus on just physx. Problem is if you focus on breaking walls, is the bots kill you. If you get to close to the tornado you die. Worst FPS I get in that level is about 27fps if I'm right in the torando. Also I hate CTF.

Heatray Physx though is another story, as the hail is what causes the slowdown in that level. I'm more interested in how your GTX280 and 9800GTX+ does in that level.

type "god" without quotes in the console. Also, type "allammo" for unlimited ammo.
This is great for recording demos without the "inconvenience" of being fragged.

In HeatRay, I get an average of 27-30 fps with a 9800GTX+ (Again, this demo is constantly focusing on the physics, more so than you would run into in a regular game. Player just constantly shoots buildings and objects to destroy them and create debris). I have ran it at least 5 times so far and that number remains pretty consistent.

I can post a couple few screenshots later of the console result, along with fraps result.

 

[nRollo]

Originally posted by: Kuzi

Originally posted by: nRollo
...
This stuff does come at a cost with a single 9800GTX+ at 19X12 0X16X Very High Quality I could definitely tell the game was running slower than without PhysX. It was still playable and fun, but noticeably slower than without PhysX.
...

So it improves physics and slows down frame rate. Kind of like when 3D cards first released on the PC. You get improved graphics but lower fps, sometimes a slideshow. At least it's a good start.

Maybe we can expect GT200 based cards to accelerate physics without much loss in fps?

Whoa- I did not say it was a "slideshow" anywhere. With minimum fps in the mid to high 20s and averages in the high 30s, it's pretty smooth. Not like without PhysX, but nowhere near the early days of 3d, or even the early days of SM3 or HDR with AA.

 

[taltamir]

one of the unique things about first order physics... it has to be calculated even when not visible on the screen.... only second order physX (a pure graphic effect with no effect on the game) can be dropped when not visible...