should I use old card for physx

Personally, I'd sell it.

A GTX 780 can take care of graphics + physics without a problem, especially if you're only planning on gaming on one monitor.
 
If you play borderlands 2 yes use it. Game runs WAY better with 500 series as dedicated physx. Pretty much all other games will run well with physx on the 780. Also you'll probably need an 850watt or more psu to run both together.
 
the problems happen when you are using vastly higher performing cards with an extremely weak card dedicated to this task. 580 are not at all weak cards, but I will say obviously make sure you have enough juice to run them both, and well this is where testing comes in an FPS meters and such to see the impact it has or other.
 
Personally, I'd sell it.

A GTX 780 can take care of graphics + physics without a problem, especially if you're only planning on gaming on one monitor.

This, I very much agree with this. GTX 780 is actually pretty good solo without dedicated physx in my experience, I would give the GTX 780 solo a shot. I did that for many months (prior to going SLI) and was pleased with physx performance even without a dedicated card.

That said, if you still want a dedicated physx card - perhaps sell the 580 and grab a 750ti. The 750ti is like 90% of the 580s speed while sipping only a fraction of the power - making it a better option for dedicated physx if one chooses that route.
 
That said, if you still want a dedicated physx card - perhaps sell the 580 and grab a 750ti. The 750ti is like 90% of the 580s speed while sipping only a fraction of the power - making it a better option for dedicated physx if one chooses that route.

Protip.

This is exactly what I was going to say.
 
This, I very much agree with this. GTX 780 is actually pretty good solo without dedicated physx in my experience, I would give the GTX 780 solo a shot. I did that for many months (prior to going SLI) and was pleased with physx performance even without a dedicated card.

That said, if you still want a dedicated physx card - perhaps sell the 580 and grab a 750ti. The 750ti is like 90% of the 580s speed while sipping only a fraction of the power - making it a better option for dedicated physx if one chooses that route.

think i'll do just that... if I can sell the 580 for 150. Don't really want to spend more. Should be able to get 150 for a super clocked 580 with back plate and high flow bracket.
 
I tried to use my old 570gtx as dedicated PhysX card but it totally killed my temps. My top 780gtx was over 90' and my low one was around 85' I dropped the 570 and got a 750ti for my dedicated PhysX.
 
I tried to use my old 570gtx as dedicated PhysX card but it totally killed my temps. My top 780gtx was over 90' and my low one was around 85' I dropped the 570 and got a 750ti for my dedicated PhysX.
You're running GTX 780 SLI... and still decided to add a dedicated PhysX card for some reason? :eek:

Wouldn't a 750Ti be too slow to keep up with the GTX 780's, dragging down overall performance in PhysX-intensive scenes?
 
Last edited:
I tried to use my old 570gtx as dedicated PhysX card but it totally killed my temps. My top 780gtx was over 90' and my low one was around 85' I dropped the 570 and got a 750ti for my dedicated PhysX.
You're running GTX 780 SLI... and still decided to add a dedicated phys card for some reason? :eek:

Wouldn't a 750Ti be too slow to keep up with the GTX 780's, dragging down overall performance in PhysX-intensive scenes?
I'm interested in this as well. I have an old GTX 570 lying around I was thinking about adding to my GTX 780, but I was thinking about grabbing a 750 Ti for the lower power usage and better GPGPU performance. Do you (Modboxx) happen to have any performance data showing any difference between the 570 and 750 Ti in PhysX-enabled games?
 
It would certainly be an interesting comparison.

In absolute terms, the GTX 780 is a faster PhysX card. In the case of (relatively) graphically-simple scenes, a GTX 780 will be able to handle FAR more physics interactions than a GTX 750 Ti without slowdown.

In the above case, using a dedicated GTX 780 for graphics and a dedicated GTX 750 Ti for PhysX would actually result in lower framerates. The GTX 780 would be constantly waiting on data from the GTX 750 Ti. Fluidmark is a good example of this behavior.

The only way a weaker card can become beneficial to overall performance is if the primary graphics card(s) are so swamped with rendering that the remaining PhysX performance of the primary card(s) becomes slower than that of the weaker dedicated card.


Edit: And don't forget, SLI systems will already load-balance PhysX between cards. This effectively means that you have (up to) a full GTX 780's worth of raw compute power available to you for physics calculations in a GTX 780 SLI system (if that level of performance is called upon by a game at any given moment). With that in mind, I fail to see how dedicating a GTX 750 Ti to PhysX would do anything but make such an SLI system slower.
 
Last edited:
No numbers guys but i'll be happy to test? just let me know what you would like to see?

The reason I got the 750 was because when I tried COD Ghost and I was running next to Riley my fps would collapse with Phys X on high. The same thing happened in Batman Origins. The Normal game mode was fine but in those combat challenges when PhysX was set to high there was slowdown. At first I chalked it up to me sucking but then one day I turned on Fraps and noticed that frames were falling. That was when I decided to try the 570gtx for PhysX, I immediately noticed that the game felt smoother and Fraps confirmed it for me as the game was now pegged at 60fps. I only tested it on the third combat challenge but it was good enough for me.
 
Borderlands 2 is a cartoon so why would one need all that to play it. I own the game.

Because borderlands 2 was never patched to support the 600 and 700 series cuda cores properly. So game runs mint on a 580 but runs like an absolute bag of shit on 6 and 7s.
 
Because borderlands 2 was never patched to support the 600 and 700 series cuda cores properly. So game runs mint on a 580 but runs like an absolute bag of shit on 6 and 7s.
You can actually fix that yourself... Open up the Borderlands 2 game directory and find PhysxDevice.dll. Delete this DLL. Enjoy.

This will force the game to use the latest versions of this DLL (the one that lives in your System32 folder), rather than the old version included with the game.

Same fix also works for Mirrors Edge.
 
No numbers guys but i'll be happy to test? just let me know what you would like to see?

The reason I got the 750 was because when I tried COD Ghost and I was running next to Riley my fps would collapse with Phys X on high. The same thing happened in Batman Origins. The Normal game mode was fine but in those combat challenges when PhysX was set to high there was slowdown. At first I chalked it up to me sucking but then one day I turned on Fraps and noticed that frames were falling. That was when I decided to try the 570gtx for PhysX, I immediately noticed that the game felt smoother and Fraps confirmed it for me as the game was now pegged at 60fps. I only tested it on the third combat challenge but it was good enough for me.
I'd be most interested in seeing the minimum and average framerates between using the 570 and 750 as a PhysX card where it's easy to benchmark for yourself, like one of the Batman games. Would also like to see the same numbers when not using a PhysX card on your SLI system with PhysX set to Auto in the NVIDIA Control Panel.

I've heard about having a low performance card paired with a high performance one for PhysX brings overall performance down, but a lot of what I've read from personal experience overall performance and smoothness improves even with a low-end dedicated PhysX card. I have a single GTX 780 and your situation is similar to mine, so I'm interested to see what the results for you are. I also experience FPS drops and inconsistency in Arkham Origins with PhysX set to high, and I would very much like to smooth out the performance in this and other PhysX-enabled titles if possible.

My ultimate goal is to see if it is worth it investing in a GTX 750 Ti as a dedicated PhysX card and saving in power usage over plugging in my older GTX 570. Thank you, in advance.
 
well I got my 780 today and don't think a physics gpu will be necessary for quite some time... ran ac4 and arkham origins perfect with every single setting on max. will probably just wait until big price cuts and get a second 780 before i would get a different dedicated phyx card. it runs so cool and quit too...

how much power does this draw compared to the 580?
 
well I got my 780 today and don't think a physics gpu will be necessary for quite some time... ran ac4 and arkham origins perfect with every single setting on max. will probably just wait until big price cuts and get a second 780 before i would get a different dedicated phyx card. it runs so cool and quit too...

how much power does this draw compared to the 580?

Ac4 maxed out with one 780? lol ok :rolleyes:
 
Ac4 maxed out with one 780? lol ok :rolleyes:
Yeah... Arkham Origins is easy to max out on NVIDIA hardware, including the use of TXAA. AC4, however, has issues on a single GTX 780. I think I was averaging 28-35 FPS when I tried running it with everything turned up on my system @ 1920x1080 144p (i7-4770k @ stock 3.9 GHz turbo, 16GB of memory). I had to turn AA off, particle physics off, godrays to low, and environment to high to get acceptable framerate. I'm probably missing a few other setting adjustments, but you get the point.
 
yeah... i played it for a few hours last night with every setting as high as it could go and it played great. I don't know what frames i was getting, i'm sure i wasn't getting 60 fps but it was enough to be 100% playable without any sudden drops or slow downs.
 
with using the 780 and the 580? what about with just either or, I would gauge it on both cause if you can get away using 1 for gaming and save power, or you could use the 1 for mining and the other for gaming :)
 
just turned rivatuner on to see what frames i was getting.

my system:
16gb 1866 memory
i5-2500k at 4.5ghz
single gtx 780 FTW model

every setting turned up as high as possible i got:
running around in city environments 45-60fps
on the sea it stayed at a steady 30fps. i'm thinking maybe it was designed to do this because it was super steady on 30fps the whole time except aiming to shoot front cannons when it jumped up to 60.

@dragonstongue: just the 780 i haven't looked into mining at all so i don't know about that. i plan to put the 580 in our second computer my wife uses if it fits... it's in an htpc case so that's doubtful however. there's a hd6850 in there now and if i remember right it barely fit, or that could be a different case i'm thinking of... i'll probably give it to her brother if it doesn't.
 
well if you really want the 580 in there, they make dremels for a reason (if you have one of course, and don't tell the wife if you plan on doing this, many of the ladies are not fond of umm renovations :p)

For most any game that is not an FPS or high rate RTS 30FPS or so is perfectly fine, the newer cards handle dips a fair bit better then older ones did for various reasons.
 
ah, i do have a dremel and i have used it on my case as well as many other various things and my wife for some reason is in full support of my "renovation" habits... so that is probably what will happen.

so is mining worth it? say i put the 580 in the other computer and had it mining pretty much 24/7 what kind of return am i looking at? it's got an i5-750 and 8gb if those are relevant to mining.
 
the cpu and ram don't matter much at all, the gpu I do not directly know as I do not use a Nvidia card to do this, can go here and ask as a general question.
http://hardforum.com/forumdisplay.php?f=137&order=desc&page=4

I scanned through some threads I do know specifically calls for a cudaminer setup for Nvidia cards for best performance, though I do not know much about it form the Nvidia side as stated.
 
Hello again.

Well I finally found the other PCI-E power cables for my PSU and plugged in my GTX 570 to see how performance would be affected. I did some quick benchies in Arkham City and Arkham Origins. I let my GTX 780 "warm up" by running the render test in GPU-Z to put load on the card for a 30-minute period.

First, my system:

  • Motherboard: Gigabyte Z87X-UD3H
  • CPU: Intel i7-4770 @ stock (3.4 GHz, 3.9 GHz Turbo)
  • Primary GPU: EVGA NVIDIA GTX 780 SC (2783KR) @ stock (I've seen it hit ~1100 MHz Boost)
  • PhysX GPU: EVGA NVIDIA GTX 570 HD 2.5GB
  • Memory: 16GB DDR3-1600 CAS9
  • Sound: Creative Labs Sound Blaster Zx
Batman: Arkham City

  • Ambient Occlusion: Yes
  • Anti-aliasing: FXAA (High)
  • Detail Level: Extreme
  • DirectX 11 Features: MVSS and HBAO
  • Display Mode: Full-screen
  • Distortion: Yes
  • DX11 Tessellation: High
  • Dynamic Shadows: Yes
  • Hardware Accelerated PhysX: High
  • Lens Flares: Yes
  • Light Shafts: Yes
  • Reflections: Yes
  • Resolution: 1920x1080
Results (First row indicates which hardware is accelerating PhysX)

Code:
FPS|  570  780  CPU
---|---------------
Min|   46   39   20
Max|  156  142  141
Avg|   99   91   58
Minimum FPS comes from scene where Penguin is shooting Mr. Freeze's gun and ice chunks are flying everywhere.

Batman: Arkham Origins

  • Ambient Occlusion: DX11 Enhanced
  • Anti-aliasing: GeForce TXAA High
  • Depth of Field: DX11 Enhanced
  • Display Mode: Full-screen
  • Distortion: On
  • Dynamic Shadows: DX11 Enhanced
  • Geometry Detail: DX11 Enhanced
  • Hardware Accelerated PhysX: High*
  • Lens Flares: On
  • Light Shafts: On
  • Reflections: On
  • Resolution: 1920x1080
Code:
FPS|  570  780  *CPU
---|----------------
Min|   65   49    26
Max|  124   93   148
Avg|   94   64    94
Lowest FPS comes in prison corridor when two guys are walking away from the camera.
*Hardware Accelerated PhysX is changed to Normal in-game when CPU is selected as PhysX source in Control Panel. It cannot be changed to High.*

Conclusions

  • Arkham Origins saw a 47% increase in average framerate by dedicating GTX 570 to PhysX.
  • Arkham City's benchmark does not have many PhysX-intensive scenes included. Nevertheless saw an 18% increase in minimum framerate during most intensive scene described above.
  • Hitching was present during dynamic loading periods in the Arkham City benchmark across all 3 choices for PhysX source, but was least noticeable when the CPU was chosen as PhysX source.
  • Played a 20 minute story mode session with Catwoman in Arkham City with all 3 choices. Framerates not recorded, but experience was noticeably the smoothest with GTX 570 dedicated to PhysX.
  • Adding the mid-range card that is 2 generations old did not hinder non-PhysX game performance while playing Titanfall all day yesterday. GTX 570 sat idle at 50 MHz using 913 mV the entire time.
Would still like to see how my results stack up with adding a GTX 750 Ti as a PhysX card ;).
 
Conclusions

  • Arkham Origins saw a 47% increase in average framerate by dedicating GTX 570 to PhysX.
  • Arkham City's benchmark does not have many PhysX-intensive scenes included. Nevertheless saw an 18% increase in minimum framerate during most intensive scene described above.
  • Hitching was present during dynamic loading periods in the Arkham City benchmark across all 3 choices for PhysX source, but was least noticeable when the CPU was chosen as PhysX source.
  • Played a 20 minute story mode session with Catwoman in Arkham City with all 3 choices. Framerates not recorded, but experience was noticeably the smoothest with GTX 570 dedicated to PhysX.
  • Adding the mid-range card that is 2 generations old did not hinder non-PhysX game performance while playing Titanfall all day yesterday. GTX 570 sat idle at 50 MHz using 913 mV the entire time.
Would still like to see how my results stack up with adding a GTX 750 Ti as a PhysX card ;).
Great work, there! So a GTX 570 as a dedicated PhysX card is just fast enough to be a worthwhile net-benefit for a GTX 780.

I'm also curious how a GTX 750 Ti would work, especilly since Passmark DirectCompute scores between the GTX 570 and GTX 750 Ti are identical.
 
Great work, there! So a GTX 570 as a dedicated PhysX card is just fast enough to be a worthwhile net-benefit for a GTX 780.

I'm also curious how a GTX 750 Ti would work, especilly since Passmark DirectCompute scores between the GTX 570 and GTX 750 Ti are identical.

That is the reason I bought a 750Ti (and mining). Overall, it should work even better because of the lower power requirement
 
Back
Top