680 + physx bench

gerardfraser

gerardfraser

[H]ard|Gawd
Joined
Feb 23, 2009
Messages
1,366
EDIT:post #12 680 GTX SLI + 560ti Physx bench


A couple 680 + Physx benchmarks(450 Nvidia card)

2600k
5014 Mhz

680GTX
1285 Core
6552 MEM

Z68 motherboard, so lanes are running at 8x8

Mafia 2 1920x1080 Maxed all settings

680 Mafia 2 physx
60.5 FPS


Full Size
http://imageshack.us/photo/my-images/407/mafia2680.jpg/

680 Mafia 2 + 450 for physx
81.8 FPS


Full size
http://imageshack.us/photo/my-images/600/mafia2680physxcard.jpg/

Batman Arkham City Maxed all settings with FXAA DX11

680 Batman Arkham City physx
MIn=7
Max=118
AVG=64




Full Size
http://imageshack.us/photo/my-images/585/batman680.jpg/

680 Batman Arkham City 450 physx
MIn=12
Max=124
AVG=74


Full Size
http://imageshack.us/photo/my-images/213/batman680physxcard.jpg/
 
Last edited:
Thanks a lot for doing this. These certainly show that there's still gains from dedicated PhysX cards.
 
Yeah, for Mafia 2 that's a nice jump ;) Minimums in Batman increased a good bit when considered by percent also.

Doesn't Metro 2033 have a benchmark built in for Physix? Can you do that in the future with your 680GTX 450 combo?
 
Metro Physx runs off CPU but I did some benchmarks for fun. 5 passes showing 3 & 4 Runs.

DOF Disabled


AAA


4xMSAA
 
I also appreciate you doing these benches for all the Physx card naysayers out there. Also, what software is that that gives you the red text overlay?
 
Well its RRTS server with HWiNFO64"RTSS is part of all unwinder stuff like MSI afterburner,EVGA precision ETC.
You can configure it any way you want

Also lots of people are saying 680 does not need a physx card but I proved to myself a physx card is better with 680GTX.
 
gerardfraser said:
Metro Physx runs off CPU but I did some benchmarks for fun.
Click to expand...

This might explain why I wasn't seeing any real GPU load on my dedicated GTX460 768MB PhysX card in this title as discussed in the thread I started on the subject. It peaked around 2-3% GPU use, probably because it was being done on CPU... (Though that makes me wonder what the 2-3% was being caused by, as the "dedicate to physX" box is checked in drivers for the GTX460.)

It does not explain why I'm not seeing any load in RO2, but then again, that game can be buggy, so maybe it's not properly initializing the PhysX.
 
The metro Physx thing is weird.

Disclamier these runs were done with cores parked giving higher FPS

After running alot of test 680+450 is close to 680+680 Physx,due to me changing settings in between cards (parking cores)

Parking cores gave me more FPS on some games and higher mins in some games like Metro. Other games seem lower mins.

680SLI is great also.

Mafia 2 with 680+680physx


Batman arkham city 680+680physx



680 SLI Batman Arkham city


680 SLI Mafia 2
 
Yeah, I am thinking that the assertion that Metro 2033 only uses CPU PhysX is incorrect.

I have registered a load on my GTX460 PhysX card in Metro 2033, but at the settings I normally play it was very very low (almost not visible on the charts).

When I lowered the video settings to let me get a higher framerate, I could get the PhysX GPU load into the mid to high teens.
 
I think Metro may use some CPU back when I had my 580 the CPU load did seem higher in previous areas.

Maybe that was just the card not pushing as hard too?
 
CPU 3770k 4700MHzMB
MSI z77 gd65
SLI Runs---PCI-E 3.0 (8/8)
Physx Runs---PCI-E 3.0 (8/4/4) Pyhsx running PCI-E 2.0

680 SLI Batman Arkham City 2560x1440 DX11 Max + 32x CSAA
MIN=25
MAX=90
AVG=60




680 SLI + 560 Physx card Batman Arkham City 2560x1440 DX11 Max + 32x CSAA
MIN=33
MAX=105
AVG=71




680 SLI Mafia 2 2560x1440 DX9 Max + AA on + Custom FXAA
AVG=63.3 Rank B Driver 301.25




680 SLI + 560 Physx card Mafia 2 2560x1440 DX9 Max + AA on + Custom FXAA
AVG=88.2 Rank A HAd to use Driver 301.42, 301.25 crashed with Physx


Clocks
 
Metro Physx runs off CPU

that is not true. any game that offers optional hardware physx will use additional effects that run primarily on the gpu when enabled.
 
gerardfraser: are you using a single 680 for your benchies??? I just got a 690 and have a GTX 550 Ti for physX and am getting numbers like yours..... I'm really thrown off by this since yours is a single 680, I'm hoping I read it wrong!

System:
MSI P67A-GD80
i7 2600k @ 4.5GHz
16GB DDR3
EVGA GTX 690
MSI GTX 550 Ti set to dedicated for physx
Win 7 x64

vid cards @ default clock speeds

Mafia II

Settings: Bult in benchmark, 1920x1080, AA On, AF 16x, Shadows High, AO On, Geom High, Apex PhysX High

1920x1080

GTX 690 PhysX OFF: 149 fps A

GTX 690 w/ PhysX on: 64.6 fps B

GTX 690 w/ 550Ti PhysX: 80 fps A



2560x1440

GTX 690 PhysX OFF: 98.2 fps A

GTX 690 w/ 550Ti PhysX: 73.2 fps B




Metro 2033
Settings: Built in becnhmark tool, Frontline. Direct X 11, Resolution: See bench, Qual: Very High, AA AAA, Texture Filt AF 4X, Enable Advance PhysX, DX 11 Tesselation and DOF enabled

1920x1080

GTX 690 PhysX OFF
Avg: 69.97
Max: 337
Min: 10.14

GTX 690 Advance PhysX On
Avg: 53
Max: 295
min: 2.05

GTX 690 w/ 550ti PhysX
avg: 69
Max 305
min 11

2560x1440

GTX 690 PhysX off
Avg: 42
Max: 211.28
Min: 13.44

GTX 690 w/ 550Ti PhysX
Avg: 41.67
Max: 219.93
Min: 9.94




Batman Arkham City

Settings: Built in Benchmark, Reolution: See Below , FXAA High, DX 11 Feat MVSS and HBAO, DX 11 Tess High, Detail level Very High, Dynamic Shadows Yes, Motion Blur Yes, Distortion Yes, Lens Flares Yes, Light Shafts Yes, Reflections Yes, AO Yes, HW Accel PhysX High / Off


1920x1080

GTX 690 PhysX OFF
min: 20
Max: 280
Avg: 157


GTX 690 PhysX High
Min 5
Max: 114
Avg: 66


GTX 690 w/ 550ti for physX High

Min: 17
Max: 103
Avg: 63

2560x1440

GTX 690 PhysX Off
Min: 6
Max: 171
Avg: 129

GTX 690 w/ 550Ti PhysX high
Min: 5
Max: 160
Avg: 63



Why the numbers went down with a dedicated physx card and didn't seem to make a difference on Batman AC staggers me.....


Fluid Mark 1.4.0:

GTX 690 CPU PhysX
1920x1080 benchmark 1727 points 28fps

GTX 690 only.... wouldn't run right, none of the emitters were spraying the red liquid


GTX 690 w/ 550 Ti for PhysX
1920x1080 benchmark 3153 points 52fps





Unigine Benchmark

Settings: Direct X 11, shaders: high, Tesselation: extreme, Aniso: 16, AA: 8x Full Screen, 1920x1080

FPS: 80.9
Score: 2038
Min 29.9
Max: 197.1
 
Last edited:
corrected a mistake I made above and added benchmark results with PhysX off
 
nintari said:
Why the numbers went down with a dedicated physx card and didn't seem to make a difference on Batman AC staggers me.....
Click to expand...

I don't want to sound like a prig, but did you by chance have the PhysX option in the NVCP set to "Recommended" or did you manually set the 560 as the PhysX card.

Also, what driver set are you using?

I've been fooling around with the idea of getting a dedicated card (especially since UE4 and Borderlands 2 are going to make extensive use of it), so these tests are really interesting.
 
Interesting....

In my original setup I had the GTX 550 Ti in the very bottom slot (on the MSI P67A-GD80). I thought that when I used to cards in the top two slots like before when I had CFX 6970s that the last slot got bumped down to PCIe x4. I decided to try the second slot down since I didn't have any card in that slot now and had some interesting results.

Mafia II same settings as above

1920x1080: 91.4 fps A


Batman Arkham City same settings as above:

1920x1080

Min: 13
Max: 122
Avg: 72


Fluidmark:
1920x1080 benchmark Score: 4423 72 FPS



Metro 2033 stayed about the same



So I had increases across the board moving to a higher bandwidth slot! Guess that answers that question :D
 
the 550Ti was set to dedicated in the Nvidia display control panel.

oh and 301.42 drivers
 
Good to see that there was some improvement!

Also, I noticed on nVidia's site there's an updated PhysX driver...think I might be able to bribe you with e-cred for a few retests to see if it made any improvements? One of the major changes listed was fixes to their 2.8.4 PhysX engine to improve performance, but I think that was in the notes for the previous one.

I'm trying to talk my friend into letting me borrow his 560 to test as a PhysX card while he gets moved into his new place...
 
What would be awesome to see, is numbers for Physx when the GPU has gold-fingers taped off to make it run as x1/x4 and x8. :)
 
DarkAngel_ZERO said:
Good to see that there was some improvement!

Also, I noticed on nVidia's site there's an updated PhysX driver...think I might be able to bribe you with e-cred for a few retests to see if it made any improvements? One of the major changes listed was fixes to their 2.8.4 PhysX engine to improve performance, but I think that was in the notes for the previous one.

I'm trying to talk my friend into letting me borrow his 560 to test as a PhysX card while he gets moved into his new place...
Click to expand...


I may test that out a bit tonight :) thanks for pointing that out.

People may knock PhysX and say it is a gimmick....and to them I say... well I can't say it prob without getting banned. ANYTHING and I mean ANYTHING that makes Graphics in a game more detailed, realistic and immersive is desperately needed in games today. If you don't think that is true, then you sir (or madam) are an Idiot. Note though I said Graphics :D Yes there are a ton of games out there that could use better storylines, dialouge etc etc.


Do I agree with having a proprietary API that only one companies video cards can use? No, absolutely not. Do I like the way the API run like crap and brings frame rates to their knees... No, not at all, there is no reason it should cut frame rates down that low especially when you have a higher end card dedicated to processing these things. I know it adds more items to be rendered on screen, but it shouldn't cut your frame rate in half or more than half. That is complete and total BS in my opinion.

But do I like what is added to games when it is used? Abso-freakin-lutely.

When I say I want to turn up ALL the eye candy in a game to get the most out of it, I want just that ALL of the eye candy with good frame rates.
 
Mafia II same settings as above

9.12.0214 1920x1080: 91.4 fps A
9.12.0613 1920x1080: 93.4 fps A

Batman Arkham City same settings as above:

9.12.0214 9.12.0613
1920x1080 1920x1080
Min: 13 Min: 9
Max: 122 Max 138
Avg: 72 avg: 78


Fluidmark:
9.12.0214 1920x1080 benchmark Score: 4423 72 FPS
9.12.0613 1920x1080 benchmark Score: 4411 72 FPS

Metro... no difference.... it actually scored a bit lower by 3fps
 
Ok all :) gonna do this the [H]ard way :)

I have a GTX 550 Ti in my rig now... 192 cuda cores

My sons machine has a GTX 480 in his rig.... 480 Cuda cores

My HTPC at the fiancees house has a GTX 470 ....448 cuda cores :)

guess what I'll be doing soon ;)


tryign to convince someone to let me borrow his 680 for a night as well..... we'll see how that goes :D
 
Just a recap:

System:
MSI P67A-GD80
i7 2600k @ 4.5GHz
16GB DDR3
EVGA GTX 690
EVGA GTX 480 / or / GTX 550 Ti set to dedicated for physx
Win 7 x64

PhysX version: 9.12.0613

Driver Version:

vid cards @ default clock speeds

Mafia II

Settings: Bult in benchmark, 1920x1080, AA On, AF 16x, Shadows High, AO On, Geom High, Apex PhysX High

1920x1080

GTX 690 w/ 550Ti PhysX: 93.4 fps A

GTX 690 w/ GTX 480 PhysX: 96.8 fps A



Metro 2033
Settings: Built in becnhmark tool, Frontline. Direct X 11, Resolution: See bench, Qual: Very High, AA AAA, Texture Filt AF 4X, Enable Advance PhysX, DX 11 Tesselation and DOF enabled

1920x1080

GTX 690 w/ 550ti PhysX
avg: 69
Max 305
min 11

GTX 690 w/ GTX 480 PhysX
avg: 66.67
Max 271.96
min 9.77




Batman Arkham City

Settings: Built in Benchmark, Reolution: See Below , FXAA High, DX 11 Feat MVSS and HBAO, DX 11 Tess High, Detail level Very High, Dynamic Shadows Yes, Motion Blur Yes, Distortion Yes, Lens Flares Yes, Light Shafts Yes, Reflections Yes, AO Yes, HW Accel PhysX High / Off


1920x1080

GTX 690 w/ 550ti for physX High

Min: 9
Max: 138
Avg: 78

GTX 690 w/ GTX 480 for physX High

Min: 10
Max: 142
Avg: 85


Fluidmark

1920x1080 benchmark

GTX 690 w/ GTX 550Ti PhysX
Score: 4411 72 FPS


GTX 690 w/ GTX 480 PhysX
Score: 4264 70 FPS




hmmm.... Mafia went up 3.4 fps, BAC went up 7fps on average....

For the added heat, power consumption and etc that the GTX 480 brings I would say it isn't worth it for such minimal gains

I was hoping to see at least 10fps + gains on average at least.... but I didn't even get that :(

Pointless in my opinion to test the 470, the 680 though I;d still be interested in just for the heck of it...I have to know now :D
 
Just some added image features. If you like eye candy, PhysX can provide a bit more in some games.
 
You must log in or register to reply here.
Back
Top