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

[Ferzerp]

And?


Run any other modern game at 1680x1050 with FXAA on a top-end GPU, and many will show the same thing (i.e. primarily CPU limited).

People are acting like this is something new, but it isn't. That's the point.

I get you now.

 

[OVerLoRDI]

I have a feeling a driver update will get things more back in line. Latest update from AMD was 12.8, middle of August. A BL2 driver is likely coming down the pipe shortly.

 

[Silverforce11]

Don't put too much value in that, there are things amiss. They don't state if they used a second card for offloading (which I'm confident they did) and they don't state what PhysX effects they tested.

Debris and flags run fine off the CPU with just an HD 7970, but once you hit some blood/goo, good bye reliable FPS.

From their statement, it seems theres no other cards as they claim a drop of 19% perf for the 680 running physx vs a drop of 15% for a 7970 running physx.

Thats why im shocked.. if its all running off the CPU, causing a bottleneck around 60 fps, it makes sense. That means any reasonably fast CPU can run this game at 60 fps with physx maxed and its a huge difference from how much Physx crushes CPU in past games.

 

[exar333]

1920x1200 + max quality:

Nvidia Geforce 680 = 74 fps avg.
Amd Radeon 7970 = 72 fps avg.

Thats a ~2,7% differnce in performance.
Hell a 7970 1ghz edition might even beat out that 680's score.



Yeap looks very demanding on the CPU.




Looks like its one of those, better overclock your CPU before you try it.
Doesnt look like GPU matters nearly as much as what CPU you have.



Amd FX 4170 = 120$ = 47 fps. ( 0.39 fps pr $ ) (higher is better)
Intel i5-2500k = 220$ = 61 fps. ( 0,27 fps pr $ ) (higher is better)

Looks like you get more "value" out of a Smaller FX-4170 than you do a i5-2500k.
That said... the 29% differnce in performance (fps-wise) is kinda noticeable.




Yep looks like AMD hasnt had a chance to optimise drivers for this game yet.
Probably performance updates to come for this game with newer drivers.

So by your logic Option A is better?

CPU A $100 10fps (.1fps/$)
CPU B $320 30fps (.094fps/$)

I would rather choose the playable one...I know that's a little exaggerated but I would certainly not reserve judgement for anyone that takes the FX-4170 as the best option based on this review.

 

[Lonbjerg]

PhysX run via CUDA.
Good luck getting CUDA to run on a AMD GPU...They are DirectCompute/OpenCL only...no CUDA.

An NVIDIA dosn't see any benefit in making GPU PhysX run on anything but CUDA...people should loook more into things before posting.

 

[Ferzerp]

Well, if they are running it off the CPU, realize that they have the consumer cpu with the most raw processing power out there right now, a 3960X.

 

[Silverforce11]

Well, if they are running it off the CPU, realize that they have the consumer cpu with the most raw processing power out there right now, a 3960X.

Stock.

OC sandy will match or beat it. Which brings back to the important point, Physx running off the CPU without crippling frame rate.. thats a win for gamers.

 

[Termie]

1920x1200 + max quality:

Nvidia Geforce 680 = 74 fps avg.
Amd Radeon 7970 = 72 fps avg.

Thats a ~2,7% differnce in performance.
Hell a 7970 1ghz edition might even beat out that 680's score.

...

Let's be serious here:

I'm not trying to bash on AMD - obviously nVidia has done everything in its power to make this game play better on nVidia cards, and yes, an AMD driver update will definitely bring these cards more in line with each other.

What I'm more curious about, however, is that the newer AMD cards don't perform as well as older AMD cards in the same or lower performance class. 7850 below a 6870? 7870 just matching a 6970? I'd like AMD to remedy that for sure.

 

[Ferzerp]

1920x1200 max quality with no physx appears to be right at that tipping point where the processor is still holding back the 680 (75 fps seems the most their test cpu could produce in the game), but the 7970 is the source of the 72 fps number. It's really obvious once you bump up to 2560x1600.

 

[Silverforce11]

Let's be serious here:

I'm not trying to bash on AMD - obviously nVidia has done everything in its power to make this game play better on nVidia cards, and yes, an AMD driver update will definitely bring these cards more in line with each other.

What I'm more curious about, however, is that the newer AMD cards don't perform as well as older AMD cards in the same or lower performance class. 7850 below a 6870? 7870 just matching a 6970? I'd like AMD to remedy that for sure.

GCN isnt running very well yet, but different bench, different results.

It's not a very demanding game at all for the GPU, a 5850 stock will run it great maxed 1080p. OC and its still 60 fps+

 

[Ferzerp]

Those benches have removed the cpu bottleneck with a pretty big CPU OC. The 7970 number is similar to the first set we saw, but with a faster proc, the 680 pulls way ahead. This agrees with what I put forth earlier regarding the 7970 being the bottleneck at 1920x1200, but the CPU still being the bottleneck for the 680.

Though, one is 1080p, the other is 1920x1200....

 

[Termie]

Thanks for those benches, Silverforce. Very interesting - one of the big differences is the use of a heavily overclocked 3930 and PhysX - again apparently running on the CPU.

Anyway, all this makes me very sorry I sold both of my 5850s - I should have kept one instead of my 460, which is getting clobbered here. An OC'd 5850 would easily beat a stock 7850.

 

[boxleitnerb]

In my experience, Borderlands 2 needs 60fps to be really playable. Don't know if that is due to SLI and PhysX on my system, but once I go below 60, and if it's only 50, it feels sluggish.

 

[railven]

From their statement, it seems theres no other cards as they claim a drop of 19% perf for the 680 running physx vs a drop of 15% for a 7970 running physx.

Thats why im shocked.. if its all running off the CPU, causing a bottleneck around 60 fps, it makes sense. That means any reasonably fast CPU can run this game at 60 fps with physx maxed and its a huge difference from how much Physx crushes CPU in past games.

That's because they aren't disclosing everything OR are confused. I knew I had uploaded some pics, rig used in Sig [this are before I swapped my 9800 GTX+ with my GTX 460, and even before I got the mod to work).

These two screen shots were captured in the same area, difference is one has blood (notice my GPU1 [Radeon] Load and FPS):
Settings: Max everything, No FXAA, PhysX High

Present: Debris, Blood, Flags - PhysX

Now, the blood cleared up, notice my GPU1 load and FPS:

Present: Debris, Flags - PhysX

They either didn't properly test the PhysX features OR they are not disclosing a second GPU. Let me see if I can find an image of once I got my GPU2 to load properly.

 

[railven]

Found it, same settings as above, this time fixed the PhysX Mod issue I had:

Notice the load, improved FPS, and once I swapped out the 9800 GTX+ with the GTX 460 I was getting 60 FPS with Blood (minus one little hitch, still investigating that though.)

 

[Termie]

I have to say, while the PhysX effects look cool (and are probably even better seen in motion), the relatively primitive nature of the graphics engine means that PhysX kind of highlights its limitations. Tremendous aliasing, particles that seem to float on top of the ground, etc.

And in the interest of full disclosure, I got the game on PS3, mostly because that's where my friends are going to play it, but also because I had a feeling that even maxed out on a PC, the graphics were not going to be what sold this game. I do miss K/M controls though, as well as adjustable FOV. It's pretty bad on the console. Also, I can't believe how challenging it is single-player. I died a bunch of times even early on.

 

[railven]

This was the author of that article's response to someone asking if he used a second card for PhysX on the Radeon side:

You don't need a Nvidia graphics card to use PhysX in Borderlands 2.
First open the Borderlands 2 folder -> WillowGame -> Config folders.
(Path should be something like: "C:\Users\\Documents\My Games\Borderlands 2\WillowGame\Config")
Then find and open the WillowEngine.ini file.
Then find the line that reads "PhysXLevel="
(It probably says PhysXLevel=0)
And finally set the PhysX level to either 1 or 2 if you want to enable it on AMD cards, PhysXLevel=1 is Medium and PhysXLevel=2 is High.
As far as I can tell the Radeon HD 7970 and GeForce GTX 680 look the same with PhysX set to high but I have not played a huge amount of the game and I have only use a few guns. However the cloth effects and rock debris effects from shooting stuff with the Gearbox Rifle look the same.

He totally fudged that test. I already showed in my two screen shots that PhysX (proper effects) cripples any non-nVidia setup. If he just sat there and shot the walls to create debris and looked at some waving tarps, he did NOT test PhysX Hybrid.

His review should be pulled or marked with a huge disclaimer. Outside of his comment section, is there any way to directly contact the author? (Lots of options are blocked at my job.) Shoddy article, at best, in terms of Hybrid PhysX.

 

[Final8ty]

http://www.youtube.com/watch?feature=player_embedded&v=J3CYBz_PMMM
http://www.youtube.com/watch?v=KrYxaiuG74w
http://www.youtube.com/watch?v=XazRggzur28

 

[Gunslinger08]

Looks like I might get a 680 super clocked to play this at 1440p at 60 fps on high. Need to find the best one Amazon sells.

 

[Yuriman]

Run any other modern game at 1680x1050 with FXAA on a top-end GPU, and many will show the same thing (i.e. primarily CPU limited).

People are acting like this is something new, but it isn't. That's the point.

While this is true, you're missing an important detail: The CPU bottleneck in Borderlands 2 is not at 500fps. An overclocked FX-8150 can only achieve 51fps (average?) with the settings they used. Bottlenecks only matter if they hurt playability, and in this case the CPU bottleneck does.


You can achieve over 60fps with a stock 3770K so I'd find little point in overclocking an Ivy Bridge CPU for this game except maybe to raise minimum frames (not covered by the review), but an 8150, even overclocked, will not net you 60fps regardless of graphical settings.


What settings in Borderlands 2 are there to reduce CPU load? I know in many games you simply don't have the option. If you have a chip that can't net 60fps+ (and you want 60fps) you're SOL, but if your GPU isn't up to the task of maxing the game out graphically you can drop visual detail until you get your desired framerate.

 

[Gonad the Barbarian]

Well this is disappointing. This was one of the games I was hoping an upgrade to a 7850 would hold me over on my current rig until next summer and Haswell. Even if there is room for improvement in a driver update I'm still looking at a mostly CPU bottleneck. Damn.

 

[AtenRa]

CPU: AMD FX 8150 (overclocked at 4.6GHz)
CPU Cooler: AMD FX WaterCooling
Case: ThermalTake Armor A60
Mobo: ASUS Crosshair V Formula
VGA: MSI HD6950 Twin Frozr III (
RAM: Kingston HyperX 2133MHz 16GB kit (2x 4GB)
Ram was running at 1926MHz 9-11-9-27 (default CPU frequencies at 1333MHz)
HDD : Seagate 1TB 7200rps SATA-3
PSU: ThermalTake Smart 730W 80+
Win 7 64bit SP1
Cat : 12.8


BoarderLands 2 Map tested : Three Horns Divide
FoV = 76

 

[RussianSensation]

(1) Nvidia runs way better in this game (no surprise, as they were very involved in the production of it). This is most evident at 2560x1600.

The reviewer only tested the 925mhz 7970 and failed to include a 15% faster HD7970 GE. Looking at GTX670/7970GE/680, the performance difference at 1600P is very small. All > 60 fps. The performance advantage NV has in this game is basically immaterial for high-end GPUs. Considering you can now get a 1Ghz 7970 for $380 on Newegg and 1.05-1.1ghz 7970s for $450, TechSpot's omission of the HD7970 GE card is a strange one. You can see GTX670 > 7970 in this game but 7970 GE is 15% faster than the 7970 in this game.

Silverforce11 already linked the benchmarks at GameGPU that included the faster HD7970GE. If you mostly play games that run better on GTX600 cards, you get a GTX600 card If you play BF3 and BL2 for 90% of your time, you don't buy an HD7850 over the GTX660. Then again at current prices HD7870 competes with something like a 660 for example. 7850 doesn't really have an NV competitor yet until 650Ti launches next month.

(5) The HD7850, a card I'm seriously considering for my HTPC, is again underperforming the HD5870. I really think AMD has more work to do in getting the most performance out of that card.

At current prices it makes sense to look at GTX660/7870. HD7850 is fast once you overclock it 25-30%. If you are not going to overclock and specifically play games where NV cards have a 10% advantage out of the game such as BL2, of course you don't get a 7850 for such a system . At stock speeds, 7850 is not anything special and never has been. The main reason 7850 was popular was that it overclocked ~ GTX580 speeds for $250 and consumed 1/2 the power doing it, while having 2GB of VRAM.

Overall for 1080P, this game doesn't need anything faster than HD6950/GTX560Ti which has been shown by both TechSpot and GameGPU.

Again, NV cards have a slight advantage but this is not one of those games you go out and drop $300-500 on a brand new generation GPU as it's getting ~60 fps on last generation $250 GPUs from the likes of 6950/GTX560Ti. Plus as you have noted, it looks like overclocking your CPU nets you far more performance increase than even going from a GTX660Ti to a GTX680 in this game. Clearly a largely CPU limited game beyond $300 modern GPU.

An OC'd 5850 would easily beat a stock 7850.

Why would you compared an overclocked 5850 to a stock 7850 though? If you were willing to overclock the 5850, why wouldn't you overclock the 7850?
Also, you can easily find a good HD7870 for $215 now (see coupon on the front page for $170 price range.

I am not sure why you keep noticing only the situations where a stock 7850 loses to the 5870 but ignore the situations where an overclocked 7850 would be much faster than a 5870 (Batman AC, Crysis 2, Civilization V, Anno 2070, Arma II, Dirt Showdown). I think if you mostly play games that run faster on NV cards, you should be looking at GTX660 then or at the very least be looking for a sale on a 7870 if you don't want to overclock the 7850 by 25-30%. Generally speaking HD7850 is not a great upgrade from a 5850 in the first place. I would say HD7950 @ 1100mhz is a good upgrade.

I also think if you waited this long to upgrade your HTPC, might as well wait 3-4 more months and see if you can pick up an HD8850 instead of an HD7870. Either way it's somewhat expected that this game would run slightly faster on NV cards since BL1 performed better on NV cards as well.

 

[Yuriman]

CPU: AMD FX 8150 (overclocked at 4.6GHz)
CPU Cooler: AMD FX WaterCooling
Case: ThermalTake Armor A60
Mobo: ASUS Crosshair V Formula
VGA: MSI HD6950 Twin Frozr III (
RAM: Kingston HyperX 2133MHz 16GB kit (2x 4GB)
Ram was running at 1926MHz 9-11-9-27 (default CPU frequencies at 1333MHz)
HDD : Seagate 1TB 7200rps SATA-3
PSU: ThermalTake Smart 730W 80+
Win 7 64bit SP1
Cat : 12.8


BoarderLands 2 Map tested : Three Horns Divide
FoV = 76

Thanks for the additional data. Looks perfectly playable on your 8150. I wonder if it would still be playable on a 6xxx or 4xxx?

I've not played borderlands 2, what was the player count and were the gameplay patterns similar?

 

[RussianSensation]

We already know games get CPU limited at low resolutions with no AA, and this one is no different.

It is different since even at 2560x1600 this game is still CPU limited. It continues to benefit from a much faster CPU (and from a faster GPU). Meaning, it's both CPU and GPU limited.

In this particular case the game was tested with FXAA. Termie's comment that this game is very much CPU limited is evident from this:

If you swap a GTX680 for an HD7970 GE at 2560x1600, you are looking at 5-6 fps difference. If you overclock a Core i5 2500K to 4.5ghz, you are looking to gain 15-20 fps. That's the definition of a CPU limited game. No one denies that on average for most games, GPU is more important. For this game above a certain level of GPU speed, i.e., GTX670/HD7970, you can still benefit from a faster GPU no doubt, but at the same time the game is also very much CPU limited.

Essentially what you have is a CPU+GPU limited game. However, the main point here is you'll benefit far more from overclocking the i5 / i7 CPU than upgrading from a GTX670/7970 to a GTX680/7970GE for example, even at 2560x1600.

Really this is a textbook definition of a CPU limited game.

Let's not confuse the terms CPU limited and CPU bottlenecked. It doesn't mean the same thing. Battlefield 3 for example is still CPU limited but the GPU is the more limiting component, until you go with 2x GTX670/680 cards at which point something like a stock i5 2500k becomes a huge bottleneck in BF3. At that point, you'd call the game CPU bottlenecked, unless you apply super-sampling and cripple the GPU performance to 20-30 fps.

CPU limited: If you add a faster GPU, performance still improves. If you add a faster CPU, performance also improves even with the same GPU. If the performance continues to improve with a faster CPU (especially by 15-20 fps with a CPU overclock, faster IPC cores), the game is CPU limited. You can have a game that's both CPU and GPU limited - BL2, BF3 multi-player are 2 such games for example.

CPU bottlenecked: If you add a faster GPU, performance barely improves or doesn't improve at all until you overclock the CPU/or swap for a much faster CPU. This occurs when the performance is almost entirely dependent on the CPU speed with a modern high-end GPU. This can be observed in games such as WOW and Starcraft 2 on a FX8150 + GTX680. A core i5 @ 4.5ghz + GTX660Ti would smoke the former system in those 2 games.

Is this game entirely CPU bottlenecked with modern i5/i7 processors? Of course not. Is it CPU limited? Big time since it benefits from a faster CPU even at 2560x1600, unlike Witcher 2 EE, Crysis 1, Metro 2033, which are basically 99% GPU limited games with modern processors. Since the GPU dependence is so severe in the Witcher 2 EE, Crysis 1/Warhead and Metro 2033, for simplicity for often call those types of games GPU bottlenecked.

Generally speaking the less GPU intensive a game is, the more CPU limited it is. BL2 runs at nearly 60 fps on mid-range GTX560Ti/6950 cards at 1920x1200, which is a recipe for a CPU limitation on much faster 7970GE/680 cards with a stock i5/i7 CPU, unless you invoke SSAA.