Titan X + 970 GTX Physx Benchmarks

[gerardfraser]

I have 1920 x 1080 144Hz screen ,hence the 1920 x 1080 Physx benchmarks.

Just to clarify it is extremely hard to get every kill/blowing up stuff ETC with 5 minute frap runs in game exactly the same each time for Titan X and Titan X + 970 GTX .YMMV.

Neither setup was able to run AC4Blackflag/Metro 2033 Redux/Metro Last Light Redux without 100% stutter free.

May be of interest to some.




Data table for easier reading


For anyone interested Frametimes for all benches with fraps viewer.
http://www.mediafire.com/download/g9xhyq7x9ladt24/Titan+X+Frametimes+With+Viewer.7z

 

[BiH115]

Some of those games look like they REALLY benefited from the 970, pretty impressive on the minimums.

How's it compare to like, a 460 for physx, or something of a lower class? Any plans on those comparisons?

 

[Dayaks]

Thanks for sharing.

I think you'd want at least a 750ti for Physx... there are reviews on the older cards and a 460 is too old IIRC.

 

[BiH115]

I think the 460 should still be fine (though not super ideal). I was just curious on a full on comparison from high to low (970 being a VERY high physx card).

 

[Simmonz]

So I guess my wanting to have a constant 144 FPS is still a pipe dream ?

 

[gerardfraser]

So I guess my wanting to have a constant 144 FPS is still a pipe dream ?

I tested a lot of games on Titan X at 1920x1080.Still lots of games can not be maxed out due to lots of reasons.Hell I been chasing 60 FPS in games Maxed out,still not there yet.Come on AMD and Windows 10 to help out some.

 

[Nicholars]

Those scores look a bit low.

 

[Michaelius]

You might be CPU bottlenecked in some of those cases.

 

[Nicholars]

Removed

 

[gerardfraser]

Those scores look a bit low.

Well lets put it this way there not low there,they are where they should be from testing 780TISLI/970SLI/980SLI/R9 290/X Crossfire.Same machine.

There is a combination of things that affect scores and it is what it is.

You might be CPU bottlenecked in some of those cases

Yes this can be true and even with a 5960 not going to make much difference in some games.The Titan X is a fast card .Maybe Windows 10 can help somewhat in the CPU department in future games.

 

[Dayaks]

Well when you're doing 8x MSAA it's going to impact your performance a bit. Might be better off doing 1.5x DSR and 2x MSAA or something like that. In my testing doing a little bit of DSR and a little bit of MSAA looked better and was less of a performance hit.

 

[gerardfraser]

Well when you're doing 8x MSAA it's going to impact your performance a bit. Might be better off doing 1.5x DSR and 2x MSAA or something like that. In my testing doing a little bit of DSR and a little bit of MSAA looked better and was less of a performance hit.

Well of course you can turn down AA use DSR.

EG:Metro 2033 Redux 144Hz v-sync
https://www.youtube.com/watch?v=HpyyMBd6kr4

 

[Tgrove]

This is good stuff. I always wanted a reason to get a dedicated physx card. Planetside 2 would have been a good excuse but they never put the physx back in

 

[Dayaks]

Sk3tch did this review a long time ago. http://1pcent.com/?p=135

I've read people with SLI comment it adds to microstutter a lot with SLI. You have to wonder if it just bogs down one card or if Physx is split across the cards. I'd guess it's just one.

 

[Nicholars]

Well lets put it this way there not low there,they are where they should be from testing 780TISLI/970SLI/980SLI/R9 290/X Crossfire.Same machine.

There is a combination of things that affect scores and it is what it is.



Yes this can be true and even with a 5960 not going to make much difference in some games.The Titan X is a fast card .Maybe Windows 10 can help somewhat in the CPU department in future games.

Don't know but I don't remember arkham city going under 60fps at all at 1080p with everything including physx on high with a single 970 at 1425mhz. Maybe my memory is not very good

 

[Daroller]

I will have a 770 classified to use as a dedicated physx card in the next couple weeks. I'll add those results to this thread when I get the card (if I can pull myself away from GTA5 long enough to test.. ).

Your Metro results are disappointing... those are the games I'm getting the dedicated physx card for, and it looks like it won't help much at all.

 

[gerardfraser]

Yes it does not help much in the Metro games,but I can tell you it really does not matter with the metro games if you turn off 4xSSAA and use 2xSSAA or less.
Also the Metro redux games can be run in 4k really good without SSAA and no slowdowns.

 

[Dayaks]

That's interesting that Metro wasn't affected much. Did you make sure Physx effects were present in the run through? Grenades/smoke/ect? I remember a benchmark where a 290x was decimated by metro ll with Physx on (50% hit) so I thought there would if been some kind if hit.

I like having dedicated Physx to reduce stutter (increase minimums). Physx effects generally happen in short bursts and have less an effect on averages.

 

[gerardfraser]

That's interesting that Metro wasn't affected much. Did you make sure Physx effects were present in the run through? Grenades/smoke/ect? I remember a benchmark where a 290x was decimated by metro ll with Physx on (50% hit) so I thought there would if been some kind if hit.

I like having dedicated Physx to reduce stutter (increase minimums). Physx effects generally happen in short bursts and have less an effect on averages.

.

Just did another quick test with 970 SLI. CPU Physx was faster.HMMMMM interesting.All I can say this is what I get when benching.
Well i would like to say I had it setup correctly but anything is possible.

Heavy Squad with all Physx stuff going on.


2015-05-08 18:48:11 - Metro Last Light-970 SLI CPU Physx
Frames: 2630 - Time: 60000ms - Avg: 43.833 - Min: 36 - Max: 49

2015-05-08 18:55:09 - Metro Last Light-970 SLI Auto Physx(970 card 2)
Frames: 2354 - Time: 60000ms - Avg: 39.233 - Min: 29 - Max: 48

2015-05-08 19:02:47 - Metro Last Light-970 + 970 Physx
Frames: 1334 - Time: 60000ms - Avg: 22.233 - Min: 16 - Max: 51

 

[Dayaks]

.

Just did another quick test with 970 SLI. CPU Physx was faster.HMMMMM interesting.All I can say this is what I get when benching.
Well i would like to say I had it setup correctly but anything is possible.

Heavy Squad with all Physx stuff going on.


2015-05-08 18:48:11 - Metro Last Light-970 SLI CPU Physx
Frames: 2630 - Time: 60000ms - Avg: 43.833 - Min: 36 - Max: 49

2015-05-08 18:55:09 - Metro Last Light-970 SLI Auto Physx(970 card 2)
Frames: 2354 - Time: 60000ms - Avg: 39.233 - Min: 29 - Max: 48

2015-05-08 19:02:47 - Metro Last Light-970 + 970 Physx
Frames: 1334 - Time: 60000ms - Avg: 22.233 - Min: 16 - Max: 51

Your first and second test are most probably the same settings. GPU Physx can't be run on a CPU so it defaulted to the SLI setup.

 

[deff]

I'm curious if The Witcher 3 will see any benefit from a dedicated Physx card. It seems that, depending on the game, dedicating a card to Physx gives mixed results.