Borderlands 2 GPU/CPU benchmarks [TechSpot/HardOCP/Others]

[railven]

At this point you should try reading the Tech Spot article because again I have to quote you what the author said EXACTLY.

This game looks really bad without FXAA or injected aliasing. Jaggies are everywhere and are especially bad since it's a cell-shaded game. Again, you seem to confuse 2 things here:

1) TechSpot author said he had limited time to test review so his preformance was not 100% indicative of all sections of the game; however, the game is playable via unloading PhysX to the CPU;

With PhysX set to high, the GTX 680 became 19% slower at 1920x1200, averaging just 60fps instead of 74fps. Surprisingly, the HD 7970 did slightly better dropping 15% from 72fps to 61fps, and as far as we could tell, the PhysX effects looked identical on both brand of cards.

This statement/conclusion is WRONG. If you haven't noticed, he claims PhysX High with the Radeon + CPU averages 61 FPS, that should be the first red flag. I already demonstrated this wrong. This should lead you to the second flag - what was he testing? If you read the comment section, he admits to not fully testing the implementation of PhysX, so one concludes:

a) he didn't properly test PhysX thus his findings are invalid.
b) his compliments that the Radeon had a smaller performance dip than the GeForce, even though he didn't properly test the system [he'd have seen the Radeon tank and the GeForce flying high.]
c) his conclusion IS WRONG

His own conclusion:

In fact, based on our testing, Radeon cards seem to handle PhysX slightly better than their GeForce counterparts. When running Borderlands 2 at 1920x1200, the HD 7970 only took a 15% performance hit after enabling PhysX (dropping from 72fps to 61fps), whereas the GTX 680 fell 19% from 74fps to 60fps.

The bold part is WRONG.

2) At least 2 other review sites tested PhysX High on the CPU (GameGPU and another one during an intensive boos fight) and both showed frames dipping to 40s on the CPU. Unless you consider dips to 40 fps unplayable, then offloading PhysX to the CPU is still doable.

So again, the biggest problem you seem to have here is that the game dips to 30-45 fps in certain sections. So it's not that TechSpot's author or me or GameGPU were wrong. You can easily play this game on a CPU with PhysX High but it just means you won't be gaming at 60 fps minimums. If you to maintaining 60 fps at all times is a requirement, then yes, you need an NV GPU, but then if you want 60 fps minimums everywhere with PhysX High, even GTX680 will probably dip below that in places.

You're 100% right and their findings don't even mirror Tech Spots, not even close. They mirror my findings, which is what I've been saying since this thread popped up.

I don't have a problem, I'm playing this game most likely further than these reviewers, and along the way I decide to test PhysX without offloading and guess what I find - HUGE performance drops. I report those drops. You're the only who seems to have an issue with my findings. You quote some reviewer, who mirrors me position, as if you've proven a point. The game dips, it doesn't dip to 30-45 FPS all the time, sometimes it dips to 22. And, no I never said this was an issue, I said if people can put up with it, you have the option else, you can use a mod to offload.

You keep referring to the Tech Spot article like they did something right, they didn't, and you should not be using them as reference. Thus me saying "if you're going to push GPUs [ie translation: recommend], you stop misinformation whenever it arises."

In another thread you told someone to buy Radeon's because, ready for this, they ran PhysX BETTER than GeForce and you linked Tech Spot. Now do you see why I even responded to you?

Not even close. It appears that a modern Intel CPU can handle the entire PhysX of BL2 with a Radeon HD7970 dedicated to graphics. You get the same performance as having a GTX680 doing graphics + PhysX.

*note: User had GTX 580 SLI.


<gif of mind blown>

EDIT: Since you added to your post:

Your blanket statement that you "cannot play this game with PhysX High" being offloaded to the CPU is not accurate. Yes, you can play this game with PhysX High on the CPU but it just won't be 60 fps minimums. TechSpot author, nor myself, nor GameGPU, nor the 3rd review site never said you can get 50-60 fps minimums but it doesn't mean the game is unplayable as you continue to imply.

This game is actually pretty forgiving with respect to dipping framerates, far below 60 fps on a 7970 + PhysX High, but even with these dips it still feels smooth, contrary to what you keep implying.

Borderlands 2 PhysX test without Nvidia GPU.
AMD HD 7970 stock
i5-2500k @ 4.5GHz
http://www.youtube.com/watch?v=WYhyfKVdN0g

^ Notice how even at 27-30 fps this game still feels remarkable smooth, despite those frame-rate killing fluids all over the screen? You just said in this thread over and over that PhysX High off the CPU + Radeon 7970 is unplayable supposedly or "cannot be done" with fluids/blood and so on. Yet in the video the game is running fairly well in terms of motion. Looks like the authors of TechSpot/GameGPU were right in that you can play this game without much trouble by offloading the PhysX High to an overclocked Core i CPU. If you want 60 fps minimums for a mental peace of mind, sure, but this particular game doesn't need it like Quake 3 Arena or Tekken Tag Tournament 2 or Unreal Tournament 99 where superb aiming and combo precision requires 60 fps+.

Really, that's how we play games? We just sit in an area and shoot stuff? What isn't being calculated here? Enemy AI? Oh sure, that affects CPU performance too, correct? I can post a video of 130+ FPS with PhysX High - and I'd still say, what I've been saying. The game bogs down, in intense fights - A SHIT TON. And again, I said, if you have no issues with <30 FPS sure, the game is "playable."

Things I learned today: <30 FPS is what enthusiast with $400+ GPUs and $200+ CPUs consider satisfactory/smooth, and when we play games - we just shoot the ground.

Roger roger, checking out this time for reals.

 

[RussianSensation]

When you start claiming 30-45 fps is as good as 60fps, you need to stop trying to claim any sort of objectivity.

Wow. That's just a new low for this forum.

No amount of walls of text or mostly unrelated pictures will support that, sorry.

I never once said 30-45 fps is as good as 60 fps in general but in this particular game it's a lot more forgiving similar to Crysis for example. You can view the video made on a 7970 I just linked to you. At the very least it gives people an idea of how the game actually performs in motion with PhysX High on the CPU without only relying on charts of fps.

All I did is post a video showing you BL2 using PhysX High off the CPU and how smooth it actually looks. Other shooters (or racing games) are not like this and are far more punishing at lower frames per second. Please don't generalize the statement about 30-45 vs. 60 fps to all games since I never made such as statement.

You get labeled as a sales man because non stop your pushing what you think the world should do. Pushing what people should or shouldnt buy. I hope you at least think about it, i am not trying to put you down or anything. Please dont think of it like that.

Thanks for the feedback. If you think buying a better price/performance GPU or a faster GPU and overclocking it or seeking price/performance are all bad advise, I will stop giving it as such. That's the only advice I keep pushing here. You should have seen the pattern since you've been here since at least 2008. Generally speaking when people ask for advice on what GPU they want, these factors come up quite often, or otherwise a person would just go and buy a $500-1000 GPU and not even care.

I already linked a video of an HD7970 GPU + Core i5 2500K @ 4.5ghz that gives us an even better sense of the performance hit from forcing PhysX to the CPU. At the very least people now get to see how this game feels in motion with fluid all over the screen being calculated on the CPU.

 

[ocre]

I picked up the game on the 18th and loving every min of it. Performance is good on my rig at 1200p maxed. I do notice the odd drops in fps when a tons of stuff is on screen but nothing game changing.

I've been considering adding a second gpu to my rig to enable PhysX for awhile now.

So what I need to know is what model geforce would be a good match up for what I have?

budget $150-$200

I intended on running this mod and can pickup Batman AC for additional testing.

you could pick up a used 460 for cheap these days. it would make for a great hybrid system. There is plenty of help out there for getting the AMD + nvidia physx GPU to work fine. This combo works out with great success. Overall i think it would be pretty cheap to drop in a used nvidia card.

But the other option is great to. sell your GPU and buy a gtx670. The performance upgrade will be there all the time and this would benefit all the time, not just a physx improvement. This is up to you.

I cant wait to pick this game up when i get some time off work. Sounds fantastic.

 

[SPBHM]

^ Notice how even at 27-30 fps this game still feels remarkable smooth,

really? going from 60 to 30 feels smooth to you?! for me going from 45 to 35 is pretty noticeable already...

anyway, look the video I made with the i3 2100
http://www.youtube.com/watch?v=WWjOgWnt5W8#t=3m15s

right image is physx high, left low, under 20fps for most of the time with physx high (GPU usage is pretty low)

now compare to your graphic showing
minimum at 30 for the same CPU!?

 

[RussianSensation]

This statement/conclusion is WRONG.

Yes, if he claimed that Radeons handle Physics better than GeForce, I don't agree with that now after looking at 2 other reviews. So yes, I was wrong for relying only on TechSpot's review in general. You are right. To me I took his review to mean that you can run the game with PhysX High and not need an NV GPU.

I don't have a problem, I'm playing this game most likely further than these reviewers, and along the way I decide to test PhysX without offloading and guess what I find - HUGE performance drops. I report those drops. You're the only who seems to have an issue with my findings.

I told you repeatedly, I believe your findings over and over and you don't seem to read what I am saying. I not once questioned your performance findings, I questioned your implication that you cannot play this game with PhysX High on the CPU since it's basically unplayable according to you.

In another thread you told someone to buy Radeon's because, ready for this, they ran PhysX BETTER than GeForce and you linked Tech Spot. Now do you see why I even responded to you?

Now you are just twisting my own words. I never said Radeons perform PhysX better....that's nonsense.

 

[RussianSensation]

And again, I said, if you have no issues with <30 FPS sure, the game is "playable."

Exactly my point. This particular game appears to be way more forgiving with frame rate drops than other FPS titles. Now, if a gamer thinks it's still unplayable, he/she can go out and buy a $150 dedicated NV GPU and set up a Hybrid PhysX as you have done. That's not the same as implying you can't offload PhysX High to the CPU in this title.

Things I learned today: <30 FPS is what enthusiast with $400+ GPUs and $200+ CPUs consider satisfactory/smooth, and when we play games - we just shoot the ground.

I didn't link the video from YouTube to disprove your testing or to imply that someone who buys $400 GPUs should settle for 30 fps in all games in the world. It's there to show others how it feels at these rates with CPU doing PhysX High. That's all. Again, stop sensationalizing things I never stated explicitly.

really? going from 60 to 30 feels smooth to you?! for me going from 45 to 35 is pretty noticeable already...

Not in all games, but in some games frame rate drops are way less forgiving. For example, F1 2010-2012 games run good below 60 fps but other racing games stutter with drops from 60 to 52 fps. You can't just make these generalizations to all games. You guys keep implying over and over in this thread a user practically NEEDS an NV card to experience PhysX High in this game because otherwise the game is "unplayable"? Right?

Did you actually watch this recording?
http://www.youtube.com/watch?v=WYhyfKVdN0g

Does that look like a slideshow to you? Would you say that's unplayable and that you have to go out and buy a $150 GPU to offload PhysX to the GPU in this game?

If that person removed the frame rate counter in the top left hand corner in that video, somehow I doubt many people would be able to tell that this part of the game is not running smoothly. That's my point. It's not about 35 vs. 60 fps in general, but in this particular game.

 

[Ferzerp]

Hey guys, look at my 30 fps youtube video with no enemies on the screen to show how great my AMD card is with physx in this game!

 

[AtenRa]

The game with the FX8150@4.6GHz and MSI HD6950 2GB at 1080p, highest settings, FXAA and PhysX at High is unplayable in heavy fighting, fps dropping to sub 10 like flies

The most annoying thing is that debris and blood stays for a long time making fps stay at 10 or less for quite some time.

 

[SPBHM]

You guys keep implying over and over in this thread a user NEEDS an NV card to experience PhysX High in this game because otherwise the game is "unplayable"? Right?

Did you actually watch this recording?
http://www.youtube.com/watch?v=WYhyfKVdN0g

Does that look like a slideshow to you? Would you say that's unplayable and that you have to go out and buy a $150 GPU to offload PhysX to the GPU in this game?

If that person removed the frame rate counter in the top left hand corner in that video, somehow I doubt many people would be able to tell that this part of the game is not running smoothly. That's my point. It's not about 35 vs. 60 fps in general, but in this particular game.

did you watch my video?

the video you posted doesn't show enough action going on to judge how it performs,

YES you need a Geforce to play this game with PhysX on High properly.

The game with the FX8150@4.6GHz and MSI HD6950 2GB at 1080p, highest settings, FXAA and PhysX at High is unplayable in heavy fighting, fps dropping to sub 10 like flies

The most annoying thing is that debris and blood stays for a long time making fps stay at 10 or less for quite some time.

exactly, that's basically what I see here, and what my video is showing...

 

[Ferzerp]

I predict another wall of text with more unrelated pictures to try to shift the focus again.

 

[RussianSensation]

Hey guys, look at my 30 fps youtube video with no enemies on the screen to show how great my AMD card is with physx in this game!

It has nothing to do with enemies on screen. It's just there to show the actual impact on motion with frame rate dips in an area with a certain amount of PhysX fluid/particle effects that together result in a significant frame rate drop. This video is not encompassing of the entire game by any means but it shows a real time performance drop from 60 fps to 26 fps and thus shows to other people in this thread what they would experience by offloading PhysX High to their CPU.

BTW, you are quick to attack while all I did was offer more information (without once turning it into HD7970 vs. 680) but what exactly did you personally contribute to this thread to help gamers see what happens with PhysX High on CPU vs. GPU? Did you link reviews, charts, performance links, videos? Grow up, if you can't respond in a mature way to another point of view, don't even bother responding at all.

 

[Ferzerp]

It has nothing to do with enemies on screen. It's just there to show the actual impact on motion with frame rate dips. This performance drops at the very least shows real world scenario of what people would experience by offloading PhysX High to their CPU vs. anything that you contributed to this thread. Grow up, if you can't respond in a mature way, don't even bother responding at all.

You continually try to mislead, push your views, and won't admit when you're so very, very wrong. That isn't contributing. That's just making more walls of text, and pictures that don't really help anything.

Just admit it. You're wrong.

edit: new prediction, you'll never admit that you are wrong, but feel that you win the argument when everyone else just gets tired of pointing out over and over how wrong you are and eventually abandon the thread because walloftext.

 

[AtenRa]

OK, same behavior with Core i7 920@4GHz and ASUS HD6950 1GB, fps drops to below 10

GPU and CPU usage is very low, i will say that no more than 3-4 CPU threads are used.

 

[wand3r3r]

Here's a few screenshots maxed out on an i7 980 @4.2 GHz, with a gtx 690 @1900x1200. Every setting is @max. Note* Physx is at high.

I tried to catch some action but typically when the fireworks were fully in session it took everything to stay alive and fight so they aren't at the ideal moment.

Names removed as they are random ppl.

 

[RussianSensation]

wand3r3r, sorry, your links don't seem to work.

 

[sontin]

OK, same behavior with Core i7 920@4GHz and ASUS HD6950 1GB, fps drops to below 10

GPU and CPU usage is very low, i will say that no more than 3-4 CPU threads are used.

I guess that the PhysX resolver runs in 1-2 Threads and that will hurt AMD user much more than people with an Intel cpu because of the better single-thread performance.

 

[wand3r3r]

Fixed now ^

 

[RussianSensation]

edit: new prediction, you'll never admit that you are wrong, but feel that you win the argument when everyone else just gets tired of pointing out over and over how wrong you are and eventually abandon the thread because walloftext.

I'll admit that I am wrong on which part? I already told railven that I admit that TechSpot's review didn't accurately test PhysX High since it didn't consider the most demanding aspects and I was wrong to use that review as an accurate representation of the more demanding areas of PhysX early on. That was addressed already.

However, after I went further and showed the performance hit in an area that was demanding on the CPU since it had plenty of PhysX effects and indeed the system incurred a huge performance hit, but the game was still playable, despite drops to 27-30 fps. Is that optimal performance? OF course not, not against a GTX680 or GTX680 SLI for example. Is it playable without absolutely needing a $150 slave NV GPU to do PhysX? I think it is. At least that's my view.

So, a $200-500 NV GPU still provides a superior gaming experience with PhysX in terms of performance than offloading PhysX to the CPU.

What I don't agree with you guys on is that you NEED an NV GPU for PhysX High in this game. If you have a fast enough CPU, it's manageable.

the video you posted doesn't show enough action going on to judge how it performs,
YES you need a Geforce to play this game with PhysX on High properly.

I saw your video. You have a Core i3 2100 stock + HD5750 vs. Core i5 2500k @ 4.5ghz with HD7970? Comparing apples and oranges, don't you think?

You realize you have 2 cores with HT and a 4x weaker GPU running this game? You are massively CPU and GPU limited, nevermind PhysX High. How else did you expect it to perform? Core i3 + 5750 can't even maintain 60 fps minimums without PhysX throughout this game I bet.

Fixed now ^

wand3r3r, from your screenshots, is the GTX690 performing PhysX High and yet you are still getting drops to 15 - 26 fps? Just trying to understand if even GTX690 can't manage > 30 fps in some places with PhysX on high?

 

[ocre]

Dont forget that with videos 24fps is all that is needed to look fluid. With very consistent timing in between frames, videos dont have to be 60fps to be smooth.

Things are different for games and actual processing the frames. The time between frames vary and 24fps is way to low to look smooth while gaming. This is common knowledge i assume?

What you see in the youtube videos isnt the frames per second on the counters. Each frame is consistently spaced and as a result can appear less erratic. The video is not displaying in 60fps nor is it in 15fps. It will make a difference. The video captures a frame at preset times and these samples are then played back in perfect timing. It is not exactly what you experience while playing it.

I am not saying videos arent useful, but it will miss a lot of the hitches and because of the evenly paced outcome the results must not be considered what you will actually experience while rendering it first hand.

 

[AtenRa]

Im sorry but if GTX690 gets 15 fps then i dont know what to say other than dont use PhysX at high.

 

[AtenRa]

I guess that the PhysX resolver runs in 1-2 Threads and that will hurt AMD user much more than people with an Intel cpu because of the better single-thread performance.

Nope, same crappy fps with Core i7 920 @ 4GHz

 

[wand3r3r]

Im sorry but if GTX690 gets 15 fps then i dont know what to say other than dont use PhysX at high.

Generally it's from 60-200 fps.. When the fireworks fly it drops like a rock.

 

[AnandThenMan]

Wow some people are desperate to prove that you HAVE to have an Nvidia GPU to enjoy the game with PhysX enabled.

 

[SPBHM]

I saw your video. You have a Core i3 + 5750 vs. Core i5 2500k @ 4.5ghz with HD7970? Comparing apples and oranges.

You realize you have 2 cores with HT and a 3-4x weaker GPU running this game? How else did you expect it to perform?

you realize that the graphics card have nothing to do with physx high if it's an AMD card don't you?

look at the left video, is the same system with physx low and compare the framerate
obviously the GPU is not causing the low framerate with physx high,
GPU usage goes down to 30% or something, with a simple HD5750 when trying to play with physx at high.


you posted this saying it was showing physx on the CPU
http://gamegpu.ru/images/stories/Te...0;&#1077;&#1089;&#1090;/proz/b2 proz ph h.png

the i3 is not to far from the faster CPUs....

and we have another user with an i7 (8 threads) claiming 10fps with physx at high...
so?

Nope, same crappy fps with Core i7 920 @ 4GHz

 

[RussianSensation]

So hold on, do we have a situation where an honest person posted i7 980 @4.2 GHz + GTX690 and the game still tanks to 15, 26, 36 fps in places on a $1000 GPU setup with PhysX High and yet some people keep claiming Radeon users need to go out now and buy a slave NV card, but even that might not fix it? So now we just went from PhysX is 'unplayable' via CPU to PhysX High is "unplayable" overall if you require > 60 fps minimums?