PhysX in Borderlands 2

thought I replied but dont see it ....oh well

I finally settled on a GTX 550 Ti for dedicated physx with a 480 I only gained 3FPS max on most of the benchmarks I tried out

for me the added heat, and power consumption wasn't worth 3fps
 
nintari said:
thought I replied but dont see it ....oh well

I finally settled on a GTX 550 Ti for dedicated physx with a 480 I only gained 3FPS max on most of the benchmarks I tried out

for me the added heat, and power consumption wasn't worth 3fps
Click to expand...

Did the game play feel smoother? Did it let you max things out more? Or was it just a flat-out bare minimum gain of around 3 FPS?
 
nintari said:
thought I replied but dont see it ....oh well

I finally settled on a GTX 550 Ti for dedicated physx with a 480 I only gained 3FPS max on most of the benchmarks I tried out

for me the added heat, and power consumption wasn't worth 3fps
Click to expand...

Unfortunately, benchmarks tend to be a bit deceiving when evaluating physX performance. A dedicated card only helps during game play that is stressful to the main card by offloading the physX calculations. A 3fps average increase in a benchmark will probably indicate a substantial performance increase during those most stressful times, but then shows no value during the times that are very easy.

In my anecdotal experience, I'd estimate that 90%+ of my gameplay in borderlands2 is in areas doing activities that have very low physX demands. During these periods, my dedicated card is a waste. It is only during the most intense areas and fights that it is helping. For me this is not a problem. I like looking over at my gpu-z graph after a big fight and seeing that my physX card was being pushed much harder than in the easy areas. A dedicated physX card is a luxury.

Another thing that I think can be confusing for many people is that besides the phyx calculations, there are a ton of physX related objects that are generated during the high physX fights. These objects all increase the load on the main rendering gpu. So if a person goes from no physX to full physX with a dedicated gpu, they may see a drop in performance as the main gpu has to deal with all of the extra junk being generated. More work is being done, but the end user experience is lower.

~daPhoosa
 
sk3tch said:
Hi all. I am looking to settle on a dedicated PhysX card for my HTPC rig (running a GTX 680 reference as primary). Mostly for Borderlands 2. Specs are in my sig.

I have purchased the following today to determine which will be best for my dedicated PhysX needs:
- EVGA GTX 580 SC 1.5GB DDR5 ($210, used)
- EVGA GT 640 2GB DDR3 ($95 AR)
- EVGA GTX 650 SC 1GB DDR5 ($110 AR)

I will post results when I have them. This should help us understand exactly how much juice is required. So far, it looks like the GTX 650 SC is equivalent to a GTX 460 as a dedicated PhysX processor (info here: http://www.evga.com/forums/tm.aspx?m=1743982) but I'd like to see if throwing more at it helps (GTX 580 SC). Also, perhaps it's just all about the CUDA cores and you don't need to bother with speed? In that case the GT 640 would be ideal for the cost. We'll soon find out... :)
Click to expand...

Something to keep in mind: It's been a common occurrence that having a dedicated PhysX card (SOMETIMES), doesn't net you much of any performance increases with PhysX based games, if the primary adapter is already powerful enough, like in your case, it is more than plenty for a single monitor setup. I think you'd be safe with a 192 cuda card as a primary PhysX card.
 
pj-schmidt said:
Unfortunately, benchmarks tend to be a bit deceiving when evaluating physX performance. A dedicated card only helps during game play that is stressful to the main card by offloading the physX calculations. A 3fps average increase in a benchmark will probably indicate a substantial performance increase during those most stressful times, but then shows no value during the times that are very easy.

In my anecdotal experience, I'd estimate that 90%+ of my gameplay in borderlands2 is in areas doing activities that have very low physX demands. During these periods, my dedicated card is a waste. It is only during the most intense areas and fights that it is helping. For me this is not a problem. I like looking over at my gpu-z graph after a big fight and seeing that my physX card was being pushed much harder than in the easy areas. A dedicated physX card is a luxury.

Another thing that I think can be confusing for many people is that besides the phyx calculations, there are a ton of physX related objects that are generated during the high physX fights. These objects all increase the load on the main rendering gpu. So if a person goes from no physX to full physX with a dedicated gpu, they may see a drop in performance as the main gpu has to deal with all of the extra junk being generated. More work is being done, but the end user experience is lower.

~daPhoosa
Click to expand...

This was when I was testing out Hybrid PhysX with 6970s in crossfire

Benchmarks I mean built in benchmarks with physX enabled in Mafia II, Batman Arkham Asylum, Batman Arkham City and Metro 2033. Then generic physX benchmarking I used Fluidmark.

I'm running a GTX 690 now with the 550 and did some basic testing again to see what difference it would make and there was a gain with the dedicated physx so I kept it. Didn't jot those numbers down I don't think
 
FWIW I have played about 8 hours of the game, about 4 hours with no phyx and the other 4 with CPU phyx on. I'm running the rig in my sig, it's a 5850 running at 5870 speeds and I have a 2600K running somewhere around 4.3-4.5Ghz, can't 100% remember right now.

The 4 hours with the phyx on I've not really had any super jaw droping changes in gameplay for good or bad. The overal experience hasn't changed a whole lot for me. Sometimes something cool happens then I think wow that was neat but then I forget about phyx again.

For whatever reason it's giving me a mild headache and I didn't have that happen without it on. I'm guessing it's slowing it down just enough at times to mess with my brain. I will say that my next GPU might be an Nvdia unit.
 
BiH115 said:
Something to keep in mind: It's been a common occurrence that having a dedicated PhysX card (SOMETIMES), doesn't net you much of any performance increases with PhysX based games, if the primary adapter is already powerful enough, like in your case, it is more than plenty for a single monitor setup. I think you'd be safe with a 192 cuda card as a primary PhysX card.
Click to expand...

Understood. This is one reason I've run the gamut between the $90-$250 range as far as cards go (you could easily go for the GTX 660 versus the GTX 680 if you wanted a 6xx series high-end dedicated PhysX card). My research led me to believe that I'd need a $200+ 6XX class card to help accentuate my 680s PhysX performance. I want to test this hypothesis. For my setup, I want to see what the best fit is. The performance I experience at 1080P is quite good. But it does get a bit dicey when big battles occur (I have everything maxed). I'm hoping that even if I don't see a huge FPS increase, I will get a huge fluidity and gaming "feel" increase. If that's the case - I'll keep whatever gives me that and costs the lowest.
 
sk3tch said:
Understood. This is one reason I've run the gamut between the $90-$250 range as far as cards go (you could easily go for the GTX 660 versus the GTX 680 if you wanted a 6xx series high-end dedicated PhysX card). My research led me to believe that I'd need a $200+ 6XX class card to help accentuate my 680s PhysX performance. I want to test this hypothesis. For my setup, I want to see what the best fit is. The performance I experience at 1080P is quite good. But it does get a bit dicey when big battles occur (I have everything maxed). I'm hoping that even if I don't see a huge FPS increase, I will get a huge fluidity and gaming "feel" increase. If that's the case - I'll keep whatever gives me that and costs the lowest.
Click to expand...

I'm really looking forward to your results. I have been playing some coop and noticed that my GTX 460 has actually hit 75% usage while my GTX 680 has dropped to 50% usage (and the game drops to 20 fps or so) at the same time. Either something is buggy, or my PhysX card is a bottleneck in that case. :(

I'm contemplating selling my spare 6950 to fund part of a GTX 660 purchase, then just move the GTX 460 to the backup computer since no one games on it.

Did you see that they are going to release a benchmark for Borderlands 2? It is talked about here: http://physxinfo.com/news/9425/borderlands-2-is-cpu-capable-of-handling-the-physx-effects/
 
jolli said:
Did you see that they are going to release a benchmark for Borderlands 2? It is talked about here: http://physxinfo.com/news/9425/borderlands-2-is-cpu-capable-of-handling-the-physx-effects/
Click to expand...
that review certainly matches up my experience so far. I haven't implemented a hybrid setup yet and with PhysX set to med. on my CPU, Caustic Caverns is an absolute nightmare, with framerates dropping down to 15-20 FPS. Its pretty solid on the CPU the vast majority of the time in other maps
 
I had a GTX 460 that I tossed on with my 670 just to see what it would do. It seems to have helped a bit. A noticed a few times in heavy firefights that I'd drop down to 30-40 fps with my card. Since I added the 460, the lowest dip I saw was 45 fps. Its not a huge jump, but it makes a noticeable difference in smoothness at a point in the game when it matters most.

I'll keep the card in till I'm done with my first playthrough. By the time I'm onto round 2, i'll likely have a second 670 (christmas + anniversary coming up :D)
 
I went ahead and ordered a GTX 660 for kicks. When I finish playing BL2, I'm going to stick it in my backup computer.
 
jolli said:
I'm really looking forward to your results. I have been playing some coop and noticed that my GTX 460 has actually hit 75% usage while my GTX 680 has dropped to 50% usage (and the game drops to 20 fps or so) at the same time. Either something is buggy, or my PhysX card is a bottleneck in that case. :(

I'm contemplating selling my spare 6950 to fund part of a GTX 660 purchase, then just move the GTX 460 to the backup computer since no one games on it.

Did you see that they are going to release a benchmark for Borderlands 2? It is talked about here: http://physxinfo.com/news/9425/borderlands-2-is-cpu-capable-of-handling-the-physx-effects/
Click to expand...

my 680 is constantly hovering around 50% with maxed settings + physx @ 1920x1200. never goes higher than 60%. only running one card right now too btw. is this normal? :confused:
 
tacos4me said:
my 680 is constantly hovering around 50% with maxed settings + physx @ 1920x1200. never goes higher than 60%. only running one card right now too btw. is this normal? :confused:
Click to expand...

Is your game capped at 60FPS or are you using v-sync?
 
Thanks for the review. I am surprised that Grady had so many issues with PhysX and AMD cards, since my experience has been completely flawless other than the inevitable performance hits.
 
jolli said:
I went ahead and ordered a GTX 660 for kicks. When I finish playing BL2, I'm going to stick it in my backup computer.
Click to expand...

You may want to hold off on that if you're going to use it just for PhysX. I've been testing different combinations since yesterday: GTX 680 @ 1250 MHz alone, and with a GT 640 or a GTX 580. Turns out the GT 640 is around the same performance as the GTX 580 as a dedicated PhysX processor! I thought for sure the $100 card would do it. I'll be putting my numbers together and writing this up soon. Doing a few more tests. Didn't even bother to open the EVGA GTX 650 SC that I have here because the GT 640 has the same amount of CUDA cores.

I tested within the same area during about a 30 second fight (small bandit camp right outside sanctuary). I did the test 3 times with each configuration.

This was all on my i5 2500k box, in my sig below.
 
Interesting to hear! Unfortunately, Amazon has shipped the card already. I will have to find the spot that brought my GTX 460 to its knees and see if the GTX 660 helps.

Did you happen to note the load on the dedicated card? Mine seemed fine under 60%, but the gameplay experience was terrible once it went over 70%.
 
jolli said:
Interesting to hear! Unfortunately, Amazon has shipped the card already. I will have to find the spot that brought my GTX 460 to its knees and see if the GTX 660 helps.

Did you happen to note the load on the dedicated card? Mine seemed fine under 60%, but the gameplay experience was terrible once it went over 70%.
Click to expand...

I haven't been tracking that. Do you just use Precision X or Afterburner to display on the OSD? I noticed that my 680 info no longer displays (except FPS) when I have either the 580 or 640 in (using Precision X).
 
I use Precision X and display it on a G13. I haven't tried displaying it on my monitor.
 
sk3tch said:
Didn't even bother to open the EVGA GTX 650 SC that I have here because the GT 640 has the same amount of CUDA cores.
Click to expand...

I suspect the MUCH faster GDDR5 in the 650 is going to make it outperform the 640 by quite a bit.
 
Parja said:
I suspect the MUCH faster GDDR5 in the 650 is going to make it outperform the 640 by quite a bit.
Click to expand...

Well, except PhysX doesn't really use memory that much. And as I said in my post - the GT 640 is performing very similarly to a GTX 580 when being used as a dedicated PhysX processor. So I do not know how you come to the conclusion that the GTX 650 would be much faster. :) I am going to be running more tests tonight. I'll post up some numbers. So far, in a very similar 30 second scenario the 580 and 640 perform very similarly. Over the course of a longer play session (not as similar, just a long session with benchmarking on) - they are also similar. Extremely similar. So I just don't see it being worthwhile to invest in anything more than a GT 640 for dedicated PhysX in Borderlands 2 if you currently have a GTX 680.
 
Last edited:
I really wish they would release the patch with the built in benchmark. At least we could all compare results that way.
 
sk3tch said:
Well, except PhysX doesn't really use memory that much. And as I said in my post - the GT 640 is performing very similarly to a GTX 580 when being used as a dedicated PhysX processor. So I do not know how you come to the conclusion that the GTX 650 would be much faster. :) I am going to be running more tests tonight. I'll post up some numbers. So far, in a very similar 30 second scenario the 580 and 640 perform very similarly. Over the course of a longer play session (not as similar, just a long session with benchmarking on) - they are also similar. Extremely similar. So I just don't see it being worthwhile to invest in anything more than a GT 640 for dedicated PhysX in Borderlands 2 if you currently have a GTX 680.
Click to expand...

Well, could just be BL2 isn't that demanding for PhysX. From what I've read, the 640 is pretty crappy for PhysX in games like Batman: AA.
 
Parja said:
Well, could just be BL2 isn't that demanding for PhysX. From what I've read, the 640 is pretty crappy for PhysX in games like Batman: AA.
Click to expand...

Well, the 640 and the 650 only have like a $10 price difference. So I guess I don't know why it would even matter nor why the 650 would be better (other than faster clock speed) as the CUDA cores are equal. I'd rather have the 640 due to it not requiring external power. You're tempting me to open this 650 I have here. But I saw so little performance difference between the 640 and 580 that I figured it would be pointless. :)

Also, I don't play Batman anymore. I'm only interested in the newest implementations of PhysX. Per the HardOCP overview of Borderlands, GTX 680s take a significant hit when you enable PhysX high. So I'm guessing it is quite a robust implementation of PhysX.
 
sk3tch: physX has always been demanding, take a look at the frame rates for Alice Maddness Returns, Batman Arkham City or asylum or Mafia II with PhysX off, and then Again with PhysX set to high. Even with a Dedicated PhysX card the frame rate drops like a brick. Same now goes for Borderlands 2.

And yes I have done this myself :) many times, over and over again trying to determine what was the best dedicated PhysX card for my setup. Back then it was CFX 6970s, now it is a GTX 690. I haven't tested recently with Borderlands 2 though :) so we'll leave that out for now lol I still have to test and see (but I no longer have my other cards besides the GTX 550 Ti and a GTX 480 now so really cant do comprehensive testing like I did before)
 
I was pleasantly surprised this evening. I tried Hybrid PhysX with a Radeon HD 6870 as my primary GPU, and a GeForce 9600 GSO (the 92-bit crippled one) as the PhysX GPU.

With PhysX on high, I would only get about 6-8 FPS during intense firefights when utilizing my CPU (i5-2500k, stock speed). With hybrid PhysX, I get above 60 consistently, never dipping below 20 in intense firefights when utilizing the Geforce 9600 GSO.

I didn't think this old card had enough juice to make that much of a noticeable difference. I think I'll set PhysX to Medium and enjoy a smooth experience the whole way through.

EDIT: I followed this guide. Very handy. http://www.overclock.net/t/1307142/borderlands-2-with-hybrid-physx
 
there is no such thing as a 92 bit 9600 GSO. in fact there is no such thing as 92 bit all. technically it would have to be 96 bit but even that has never been used on a gpu. the 9600 GSO was either 256 or 192 and even a 128 bit version was used later.
 
Quite the reply for missing the "1"...

Its one of the 192-bit versions. I know for a fact that its the one with 1.5 GB VRAM, and I'm 90% certain its this one specifically:

http://www.newegg.com/Product/Product.aspx?Item=N82E16814130543

Its crippled because it only has 48 CUDA cores instead of 96, although, admittedly, I don't know enough about graphics cards to tell if this would make a big difference or not.

After looking it up again, I was pretty amazed at how many versions there were of this card... different memory sizes, bandwidth, shader cores, power reqs. Some of these were known to be similar to one of the 8800 series cards while other variants fell short, like the one I have.

For the $35 I paid for this card a year and a half ago, I'm very happy with it.
 
yeah the 9600 GSO probably sets the record for having the most versions. its really quite silly as the performance can vary quite a bit from the slowest to fastest version.
 
nintari said:
sk3tch: physX has always been demanding, take a look at the frame rates for Alice Maddness Returns, Batman Arkham City or asylum or Mafia II with PhysX off, and then Again with PhysX set to high. Even with a Dedicated PhysX card the frame rate drops like a brick. Same now goes for Borderlands 2.

And yes I have done this myself :) many times, over and over again trying to determine what was the best dedicated PhysX card for my setup. Back then it was CFX 6970s, now it is a GTX 690. I haven't tested recently with Borderlands 2 though :) so we'll leave that out for now lol I still have to test and see (but I no longer have my other cards besides the GTX 550 Ti and a GTX 480 now so really cant do comprehensive testing like I did before)
Click to expand...

Not sure why you're addressing me - I'm not the one stating that perhaps Borderlands 2 PhysX is not as demanding as prior games. That's Parja. :)

Well, I'm the last post from yesterday so I may as well edit this one. :)

Here's the results of my tests...EVGA GTX 680 @ 1250 MHz boost clock as the primary GPU and I tested both the EVGA GT 640 and the EVGA GTX 580 SC as dedicated PhysX GPUs. Bottom line: for $100 the GT 640 is a great deal and will improve your Borderlands 2 FPS around 20%! Very similar performance to the GTX 580 SC. Details are here: http://1pcent.com/?p=135.
 
Last edited:
Is it just me or is physx over hyped? i watched the youtube videos and thought it looked really cool, but now that im playing the game in a lot of scenarios i just wish physx was turned off.

I like the explosions, i hate the stupid curtains flapping around, blocking my view in the shoot outs...
 
buttons said:
Is it just me or is physx over hyped? i watched the youtube videos and thought it looked really cool, but now that im playing the game in a lot of scenarios i just wish physx was turned off.

I like the explosions, i hate the stupid curtains flapping around, blocking my view in the shoot outs...
Click to expand...

You can shoot them and they'll go away. :D

I don't know, I personally really love the PhysX effects in Borderlands 2. It really adds to the immersion to me. The towns and stuff (when empty - before you approach) seem to have more life because of all the flags and such whipping in the air.
 
sketch did you happen to grab numbers for PhysX disabled on your setup? Just curious to see the performance hit with it on high VS disabled

I haven't had the time to look myself, or do the benchies :( was going to do some testing last night but had an incident I needed to take care of first
 
nintari said:
sketch did you happen to grab numbers for PhysX disabled on your setup? Just curious to see the performance hit with it on high VS disabled

I haven't had the time to look myself, or do the benchies :( was going to do some testing last night but had an incident I needed to take care of first
Click to expand...

I did not. Sorry. That's a good idea. Actually though - HardOCP's article (Kyle linked to it earlier in this thread) has some numbers around this area that may help you get an idea of the hit.
 
nintari said:
physX has always been demanding...Same now goes for Borderlands 2.
Click to expand...

Well, the numbers are showing it's really not. If a 48-shader 9600 GSO can handle the job fairly well and a GT 640 can handle it as well as a GTX 580 SC, the PhysX in BL2 clearly isn't THAT demanding.
 
You must log in or register to reply here.
Back
Top