Batman Arkham City, no physics at all if you don't use physx ?

[blackened23]

Yes I sure do.

Benchmarks

DX11 high preset, DOF off, physx enabled: 55.60 fps avg, 13.05 minimum
DX11 high preset, DOF off, physx disabled: 58.53fps avg, 14.32 minimum

DX9 very high preset, physx on: 36.57fps avg, 10.67fps minimum
DX9 very high preset, physx off: 38.87fps avg, 8.89fps minimum

As I said, minimal performance hit. Do you believe my results or are you going to pull a toyota on me?

Averages. In actual gameplay the performance hit is severe. In fact, I can run at DX11 / medium / DOF off / physx disabled / tesselation on and it NEVER dips below 60 fps during that benchmark.

But when I enable phsyx, it dips to 38 in the middle part of the benchmark.

Of course your numbers are skewed because of the averages. Because 95% of the game takes place indoors, your framerate will go incredibly high in certain areas and will throw the numbers off - what toyota said is 100% correct. What you see with this benchmark of yours is not what takes place in actual gameplay. You definitely notice it when the framerate dips suddenly to 30 fps, which happens with physx enabled.

In actual gameplay with physx on I never dip below 60 fps, but some areas go down to 35 fps with physx on. Of course the metro2033benchmark.exe won't show you this because its all based on averages.

Hold on and i'll post some screenshots of the benchmark in action

 

[tviceman]

and that's the benchmark which is not reflective of actual game in spots.

you're right, the benchmark is more demanding than actually playing the game.

I tested this crap too a while back. my minimum framerate was 20-25 fps higher with physx off then with it on. there are plenty of places where I would drop into the upper 20s with physx on in that game. with it off I rarely dropped out of the mid 50s for a minimum. so again you can keep defending physx all you want but many people like me think its a joke for the actual performance hit that you get at times.

Ok toyota. You keep telling me how it's running on my system. I guess Mafia II is the ONLY GAME where I don't notice significant dips in performance (i.e. below ~40 fps).

Do you have a batman AA save game file right before one of the scarecrow levels? I do not, I did a google search and could not find one, but if you do and would like to once and for all settle this (even though it's wtih a different game) I'd be more than happy to put my money where my mouth is.

 

[tviceman]

Averages. In actual gameplay the performance hit is severe. In fact, I can run at DX11 / medium / DOF off / physx disabled / tesselation on and it NEVER dips below 60 fps during that benchmark.

But when I enable phsyx, it dips to 38 in the middle part of the benchmark.

Of course your numbers are skewed because of the averages. Because 95% of the game takes place indoors, your framerate will go incredibly high in certain areas and will throw the numbers off.

In actual gameplay with physx on I never dip below 60 fps, but some areas go down to 35 fps with physx on. Of course the metro2033benchmark.exe won't show you this because its all based on averages.

Hold on and i'll post some screenshots of the benchmark in action

I gave you the MINIMUMS from each benchmark as well as the averages - the differences in minimums are negligible.

 

[toyota]

Yes I sure do.

Benchmarks

DX11 high preset, DOF off, physx enabled: 55.60 fps avg, 13.05 minimum
DX11 high preset, DOF off, physx disabled: 58.53fps avg, 14.32 minimum

DX9 very high preset, physx on: 36.57fps avg, 10.67fps minimum
DX9 very high preset, physx off: 38.87fps avg, 8.89fps minimum

As I said, minimal performance hit. Do you believe my results or are you going to pull a toyota on me?

the minimum in the Metro benchmark means nothing and you know it. in fact if you watch the screen it will never dip that low. its only a single instant dip in the benchmark recording itself that is not applicable to the real performance.

in the REAL game your minimum framerete will most certainly suffer in spots with physx on. at least it did in the areas I checked it. that's why I tuned it off in the game.

 

[tviceman]

you're right, the benchmark is more demanding than actually playing the game.



Ok toyota. You keep telling me how it's running on my system. I guess Mafia II is the ONLY GAME where I don't notice significant dips in performance (i.e. below ~40 fps).

Do you have a batman AA save game file right before one of the scarecrow levels? I do not, I did a google search and could not find one, but if you do and would like to once and for all settle this (even though it's wtih a different game) I'd be more than happy to put my money where my mouth is.

I gave you the MINIMUMS from each benchmark as well as the averages - the differences in minimums are negligible.

the minimum in the Metro benchmark means nothing and you know it. in fact if you watch the screen it will never dip that low. its only a single instant dip in the benchmark recording itself that is not applicable to the real performance.

in thr REAL game your minimum framerte will most certainly suffer with physx on. at least it did in the areas I checked it. that's why i tuned it off in the game.

The top graph has physx on. The bottom graph has physx off. The graphs look pretty similar to me, but since toyota proved that I can't tell when Mafia II drops to super low frame rates on my setup, please point out the massive performance drop. PLEASE. PLEASE.

 

[blackened23]

Post pictures of the same scene with physx on and physx off. I just explained averages, as did Toyota, and you COMPLETELY ignored it.

 

[toyota]

http://i83.photobucket.com/albums/j303/tviceman/metrodx11physx.jpg
http://i83.photobucket.com/albums/j303/tviceman/metrodx11.jpg

The top graph has physx on. The bottom graph has physx off. The graphs look pretty similar to me, but since toyota proved that I can't tell when Mafia II drops to super low frame rates on my setup, please point out the massive performance drop. PLEASE. PLEASE.

I am just saying the very minimum framerate in the benchmark is not indicative of game performance. jeez if i knew somebody was going to claim there was no difference in the game I would have recorded a video at that time. maybe it was just the spot I checked or maybe physx just put me over the edge right there for what my system could handle.

you keep enjoying that system of yours though since you seem unaffected by the clothing in Mafia 2 or the Scarecrow levels in Batman either. the rest of us are just liars I guess.

 

[Carfax83]

and that's the benchmark which is not reflective of actual game in spots. I tested this crap too a while back. my minimum framerate was 20-25 fps higher with physx off then with it on. there are plenty of places where I would drop into the upper 20s with physx on in that game. with it off I rarely dropped out of the mid 50s for a minimum. so again you can keep defending physx all you want but many people like me think its a joke for the actual performance hit that you get at times.

Dude, the performance impact of PhysX is entwined with your graphical settings, if you're using a single GPU.

So for example, someone that plays Mafia 2 at a low resolution like 1280x720 with a powerful video card like the GTX 560 Ti, should have no problem maxing out every setting in the game a long with PhysX, because the GPU isn't being taxed at that resolution from a graphical stand point; even after throwing AA and the other extras into the mix.

At 1920x1200 though, things may not be so rosey, because the higher resolution in combination with the other graphical effects now means the GPU is being taxed, so enabling PhysX will have a much larger performance impact.

So what I recommend to people that want to enable PhysX but love to play at high rez with all the bells and whistles turned on, is to get a cheap dedicated physx card.

That takes the PhysX calculations off of your primary GPU completely, letting it do 3D rendering only.

 

[SirPauly]

well that goes to show how crappy the implementation can be which is one of the things I have been complaining about. you have to mod the freaking game for it to run smoothly. so again I still say physx is nothing but a joke.

You can say what ever you want. But I welcome innovation with CPU and GPU Physics.

 

[SirPauly]

Dude, the performance impact of PhysX is entwined with your graphical settings, if you're using a single GPU.

So for example, someone that plays Mafia 2 at a low resolution like 1280x720 with a powerful video card like the GTX 560 Ti, should have no problem maxing out every setting in the game a long with PhysX, because the GPU isn't being taxed at that resolution from a graphical stand point; even after throwing AA and the other extras into the mix.

At 1920x1200 though, things may not be so rosey, because the higher resolution in combination with the other graphical effects now means the GPU is being taxed, so enabling PhysX will have a much larger performance impact.

So what I recommend to people that want to enable PhysX but love to play at high rez with all the bells and whistles turned on, is to get a cheap dedicated physx card.

That takes the PhysX calculations off of your primary GPU completely, letting it do 3D rendering only.

It's what I do and my older, dated GPU garners value as a PhysX card instead of a paper weight or collecting dust. Granted, would desire more adoption to put it to more use though!

 

[toyota]

Dude, the performance impact of PhysX is entwined with your graphical settings, if you're using a single GPU.

So for example, someone that plays Mafia 2 at a low resolution like 1280x720 with a powerful video card like the GTX 560 Ti, should have no problem maxing out every setting in the game a long with PhysX, because the GPU isn't being taxed at that resolution from a graphical stand point; even after throwing AA and the other extras into the mix.

At 1920x1200 though, things may not be so rosey, because the higher resolution in combination with the other graphical effects now means the GPU is being taxed, so enabling PhysX will have a much larger performance impact.

So what I recommend to people that want to enable PhysX but love to play at high rez with all the bells and whistles turned on, is to get a cheap dedicated physx card.

That takes the PhysX calculations off of your primary GPU completely, letting it do 3D rendering only.

yes I agree that you need a dedicated physx card for best results. that is sort of the point here you know. I am saying that with a single card, performance looks good from an average fps standpoint in a benchmark but during the actual game its a different story in physx heavy parts.

 

[tviceman]

I am just saying the very minimum framerate in the benchmark is not indicative of game performance.

And I am just saying the graphs I posted show very little variation in fps performance across the entire benchmark with and without physx disabled, backing up my original statement that physx in Metro2033 has a minimal impact on overall performance of that game.

I feel like I'm having an argument with a religious zealot. "THE EARTH IS ONLY 18,000 YEARS OLD I KNOW IT IS BECAUSE IT SAYS SO AND IT DOESN"T MATTER HOW MUCH IRREFUTABLE EVIDENCE YOU SHOW ME I WON'T BELIEVE IT ALL THE EVIDENCE YOU SHOW ME IS FALSE BECAUSE YOUR EQUIPMENT DOESN"T KNOW(*W#RJ#@(&$#

 

[OCGuy]

All the people who scream that Physx is a dead technology, scream when they don't get to use it. Kind of funny, actually.

 

[blackened23]

metro2033benchmark.exe

DX11 / tesselation on / advanced DOF off / advanced physx off / medium detail

77.78 fps

DX11 / tesselation off / advanced DOF off / advanced physx on / medium detail

48.29 fps

If you're saying physx has no impact on performance, you're either intentionally misrepresenting the truth or haven't played the game. In actual gameplay you get all kinds of
performance dips with physx, but its not noticeable in benchmarks because the entire game is indoors and you get huge framerate spikes (upward and downward).

 

[tviceman]

metro2033benchmark.exe

DX11 / tesselation on / advanced DOF off / advanced physx off / medium detail

80.70 fps

http://steamcommunity.com/id/xoleras/screenshot/650994986876383389?tab=public

DX11 / tesselation off / advanced DOF off / advanced physx on / medium detail

48.29 fps

on an hd6970?

 

[blackened23]

My GPU settings for both:

MSI lightning GTX 580 clocked at 963

 

[tviceman]

Post pictures of the same scene with physx on and physx off. I just explained averages, as did Toyota, and you COMPLETELY ignored it.

I went one step better than you and posted a second by second graph of my frame rates. Examine my graphs and point out to me where my physx performance drop off is huge.

 

[blackened23]

Like I stated earlier, I've played through the game multiple times and my sweet spot was 1080p, medium detail, 4x MSAA with dx11 - tessellation turned on with physx turned off. With these settings, my framerate never drops below 60.

But with physx on at those same settings, some areas become near unplayable due to choppiness, with huge spikes to 35-40 fps. It is very noticeable in actual gameplay but the benchmark won't pick this up. I can post more screenshots in areas where this happens, but its VERY bad in lighted areas. The performance dip is extremely noticeable with physx turned on - whereas its completely 100% smooth and faster than 60 fps with physx turned off.

I think this is what myself and toyota are getting at. Averaged benchmarks don't really pick up on this but when you're actually playing the game, such a huge nosedive in performance is very noticeable. It happens in certain areas in metro 2033, but because nearly all of the game is indoors the framerate goes REALLY high at times and of course will throw a benchmark average off.

I don't know how I got caught up in this, but i'll close this by saying -- once again -- I think physx is a good thing. But saying that physx doesn't incur a performance penalty, I don't agree with that at all because i've seen big performance hits with it.

 

[SirPauly]

That's fine, and there are gamers that don't like to go under 60. But, for me, go under 60 all the time and will sacrifice some performance for immersion and image quality --and 40's are fine. I tend to look at a 60FPS average for enjoyable frame-rate for me but others may differ. Thankfully, there is flexibility and choice for gamers subjective tastes and tolerances.

 

[toyota]

heck I can "average" 50 fps in the Metro 2033 benchmark with advanced physx on my cpu. have fun with the mid to low 20s though when actual physx effects get rendered.

 

[Keysplayr]

Mebbe you should just agree to disagree. It's obvious neither will budge.

 

[blackened23]

That's fine, and there are gamers that don't like to go under 60. But, for me, go under 60 all the time and will sacrifice some performance for immersion and image quality --and 40's are fine. I tend to look at a 60FPS average for enjoyable frame-rate for me but others may differ. Thankfully, there is flexibility and choice for gamers subjective tastes and tolerances.

I'll say this, some of the physx lighting effects are pretty nice in metro 2033. Its something good to have, I guess....hopefully the next installment of metro will be slightly more optimized to run better.

 

[tviceman]

My GPU settings for both:

MSI lightning GTX 580 clocked at 963

Is it in the same rig as your hd6970's? And could you post the graphs of the two benchmarks (one with physx on, the other off) - your method isn't exactly scientifically repeatable.

 

[blackened23]

Is it in the same rig as your hd6970's? And could you post the graphs of the two benchmarks (one with physx on, the other off) - your method isn't exactly scientifically repeatable.

How exactly can you have an hd6970 and a GTX 580 in the same rig? Is that even possible? Sorry, I don't keep my sig up to date. I have 1 computer with crossfire 6970's and another that i'm typing on now, GTX 580. Its easily repeatable. If you've played through the entire game you can pick which chapter/scene to play, play it once with fraps running and keep a tab on your framerate in heavily lighted areas. Now do the same with physx enabled.

The benchmark doesn't pick up on the huge discrepancy because of various reasons myself and toyota have mentioned. Its based on averages. When a physx scene is rendered, your framerate dips extremely low but quickly returns to the "average" after the physx lighting is gone. Of course with physx disabled your framerate will not dip at all. Thats why I took screenshots of the same settings in the same scene, one with physx enabled and one without.

 

[Grooveriding]

I'll say this, some of the physx lighting effects are pretty nice in metro 2033. Its something good to have, I guess....hopefully the next installment of metro will be slightly more optimized to run better.

Basically the whole issue with it. Stupid performance hits, not as good as what we've seen done on the CPU and CPU physics cost nothing performance wise.

You play BF3 and watch destruction of buildings having a real impact on gameplay, terrain suffering destruction, impressive destruction of everything around you in the game and so on, all with no performance cost. Mafia 2; stupid amounts of shards of rock and glass flying all over the place and a trenchcoat flapping around, big performance cost. Could of been done on the CPU without it.

It's a failure in its current iteration. It certainly has potential but in its current state is just irrelevant and is a detriment to gameplay with the superior and more immersive CPU physics options available.