[gamegpu] Dragon Age Inquisition performance

[frozentundra123456]

The whole point in that test is to see how well the CPU does. They test the CPU by reducing all GPU demand, which is done by turning down settings, reduce the resolution and using the fastest GPU they can. In this way, it insures the only thing holding back performance is the CPU, so they can see how well it performs.

It is not a GPU test. Oh, and just so you don't get the wrong idea, resolution changes do not change the CPU load, only the GPU load.

This is the age old dilemma. Obviously, a gpu limited test is not a true test of cpu performance, but artifically lowering the resolution to 720p is also pretty useless, as it is an artificial situation that few if any will ever use.

I think the ideal solution, but perhaps not practical because of time and cost, is to test a cpu with a variety of gpus, and at different resolutions. Game.gpu used to do their cpu tests at a more realistic 1080p, but with a powerful gpu, or even SLI that struck a balance between cpu and gpu usage. To me that was much preferable to lowering the resolution to 720p.

 

[frozentundra123456]

Sure it can. It just shows the game uses all the cores you give it, and hyperthreading.

It is possible for an 8350 to be faster than a 4670k in a highly threaded game. I will not deny that. However, if you look closely at the graphs, it looks like the 4670k result with the R9 290X is anomalous, perhaps just a testing error. I say that because the 4770k on the same graph gets 94 FPS, which seems about right relative to the other cpus in this particular game. But the 4670k gets only 67FPS, meaning the 4770k is 40 percent faster, which is far above the increase expected or ever seen from hyperthreading. And the 4670k is also only 10% faster than the i3, which also does not make sense in a heavily threaded game like this.

 

[toyota]

This is the age old dilemma. Obviously, a gpu limited test is not a true test of cpu performance, but artifically lowering the resolution to 720p is also pretty useless, as it is an artificial situation that few if any will ever use.

I think the ideal solution, but perhaps not practical because of time and cost, is to test a cpu with a variety of gpus, and at different resolutions. Game.gpu used to do their cpu tests at a more realistic 1080p, but with a powerful gpu, or even SLI that struck a balance between cpu and gpu usage. To me that was much preferable to lowering the resolution to 720p.

why people try to complicate such a simple test is beyond me. lowering the res shows exactly what the cpu is capable of if not limited by the gpu. that is the WHOLE POINT of a cpu test.

lets look a simple hypothetical test:

you look at "cpu test" for a game maxed at 1080 with a little AA and see your 680 getting only 36 fps and a 780 ti getting 50 fps which is clearly a gpu limited. what the hell would that accomplish? nothing at all. you have a 2100 i3 and see a real cpu test at 1280x720 showing that cpu can only get 50 fps. well that clearly tells you that you will need a faster cpu if you want 60 fps as lowering the graphically demanding settings or grabbing another card will not give that to you. its just that SIMPLE.

 

[bystander36]

This is the age old dilemma. Obviously, a gpu limited test is not a true test of cpu performance, but artifically lowering the resolution to 720p is also pretty useless, as it is an artificial situation that few if any will ever use.

I think the ideal solution, but perhaps not practical because of time and cost, is to test a cpu with a variety of gpus, and at different resolutions. Game.gpu used to do their cpu tests at a more realistic 1080p, but with a powerful gpu, or even SLI that struck a balance between cpu and gpu usage. To me that was much preferable to lowering the resolution to 720p.

In the vast majority of situations, resolution has 0 effect on CPU performance. 0, zippo, nadda. So yes, if you want to see how CPU's perform with a game, you just reduce the resolution to as low of a point as possible.

It is useful to find out if a game is CPU bound, and for those who want higher FPS. If you want to play at nothing less than 75 FPS, you can look at the benchmark, and see what you'll have to purchase. Getting GPU performance just requires to look at the normal benchmark and maybe reduce settings to get your goal.

This is the point of the test. It is useful, even if you don't care.

 

[alcoholbob]

This game is very GPU intensive Mantle isn't going to do squat unless you play at low resolution with a very weak CPU. Which doesn't apply for most of us here who even wants to run this on high graphics.

Also, my 4K DAI dreams are dashed. Going to need a lot more grunt to handle future games at 4k.

It's max at 4xMSAA on stock cards. Turn off AA and add 20% overclock and a 290x or 980 can easily hit 30+ fps. Hell with vsync on you are stuck with 30fps anyway when you are below 60, so its not too bad.

 

[Carfax83]

Oh, and just so you don't get the wrong idea, resolution changes do not change the CPU load, only the GPU load.

Actually it does change the CPU load. Lowering the resolution increases the burden on the CPU, because now it has to keep up with the GPU which is rendering frames at a faster pace.

Conversely, increasing the resolution lowers the burden on the CPU, as the GPU is taking longer to render each frame so the CPU doesn't have to work as hard.

 

[Enigmoid]

Actually it does change the CPU load. Lowering the resolution increases the burden on the CPU, because now it has to keep up with the GPU which is rendering frames at a faster pace.

Conversely, increasing the resolution lowers the burden on the CPU, as the GPU is taking longer to render each frame so the CPU doesn't have to work as hard.

Well sure but it doesn't affect the maximum number of frames that the CPU is capable of outputting.

 

[toyota]

Actually it does change the CPU load. Lowering the resolution increases the burden on the CPU, because now it has to keep up with the GPU which is rendering frames at a faster pace.

Conversely, increasing the resolution lowers the burden on the CPU, as the GPU is taking longer to render each frame so the CPU doesn't have to work as hard.

nope. I even checked Crysis 3 and cpu usage is basically the same at 1280x720 as it is at 1920x1080.

 

[Carfax83]

Well sure but it doesn't affect the maximum number of frames that the CPU is capable of outputting.

Yes this is true.

nope. I even checked Crysis 3 and cpu usage is basically the same at 1280x720 as it is at 1920x1080.

That can't be right. I've tested this myself personally, not with resolution but with V-sync.

Disabling V-sync causes an increase in CPU activity because the GPUs aren't being limited by the monitor's refresh rate, and can attain higher frame rates.

Frame rate is very dependent on the CPU, because it's the CPU that's sending data to the GPU to render. So a faster rendering GPU requires a fast CPU to make sure that it's not being bottlenecked.

The only way I can see your example happening is if the GPU was dog slow at both 720p and 1080p and the resolution decrease barely affected it's output.

 

[toyota]

Yes this is true.



That can't be right. I've tested this myself personally, not with resolution but with V-sync.

Disabling V-sync causes an increase in CPU activity because the GPUs aren't being limited by the monitor's refresh rate, and can attain higher frame rates.

Frame rate is very dependent on the CPU, because it's the CPU that's sending data to the GPU to render. So a faster rendering GPU requires a fast CPU to make sure that it's not being bottlenecked.

The only way I can see your example happening is if the GPU was dog slow at both 720p and 1080p and the resolution decrease barely affected it's output.

what does disabling vsync have to do with this? how about actually testing resolution and seeing for yourself? I checked it and what he said was true. If it was not then I would say so.

 

[bystander36]

Actually it does change the CPU load. Lowering the resolution increases the burden on the CPU, because now it has to keep up with the GPU which is rendering frames at a faster pace.

Conversely, increasing the resolution lowers the burden on the CPU, as the GPU is taking longer to render each frame so the CPU doesn't have to work as hard.

For any given FPS, the CPU load is the same, no matter the resolution. The maximum frames the CPU can deliver is not effected by the resolution at all. So if you are trying to see how many FPS your CPU can achieve, you simply turn the resolution as low as possible.

 

[Carfax83]

what does disabling vsync have to do with this? how about actually testing resolution and seeing for yourself? I checked it and what he said was true. If it was not then I would say so.

What does V-sync have to do with this? It's the same principle.

V-sync limits the GPU's performance by artificially imposing a frame rate cap. When that cap is removed, the GPU can process frames at a faster rate, but only if the CPU is up to the task. Faster frame rates requires more processing power from both the CPU and the GPU.

That's why it's important to have a balanced system. Pairing a Core 2 Duo with a GTX 980 SLI system would be profoundly stupid..

Similarly, resolution increases the demand on the GPUs' resources, and lowers frame rate output. So when the GPU is rendering at a slower pace, the CPU doesn't have to work as fast to send data, which lowers CPU activity..

Conversely, lowering resolution decreases the demand on the GPU's resources, and increases the demands on the CPU because the GPU is rendering frames at a faster rate.

 

[Carfax83]

For any given FPS, the CPU load is the same, no matter the resolution. The maximum frames the CPU can deliver is not effected by the resolution at all. So if you are trying to see how many FPS your CPU can achieve, you simply turn the resolution as low as possible.

That's different from what you said earlier. You said,"resolution changes do not change the CPU load, only the GPU load."

You didn't say anything about any given FPS or maximum frames. You're right, that the maximum frame rate remains the same regardless of resolution, but the LOAD is variable and is reliant on how fast the GPU can render frames.

Maybe I'm being pedantic, or perhaps you just didn't clarify yourself earlier.. Anyway, I'm done here. Carry on ^_^

 

[toyota]

lol okay keep typing all you want to try and act like its the same thing. during that time you could have actually lowered the res and saw for yourself that cpu usage is the same.

 

[bystander36]

That's different from what you said earlier. You said,"resolution changes do not change the CPU load, only the GPU load."

You didn't say anything about any given FPS or maximum frames. You're right, that the maximum frame rate remains the same regardless of resolution, but the LOAD is variable and is reliant on how fast the GPU can render frames.

Maybe I'm being pedantic, or perhaps you just didn't clarify yourself earlier.. Anyway, I'm done here. Carry on ^_^

I do believe most people, or rather, all but you, understand that the "for any given FPS" was implied.

 

[toyota]

ok we can end this since I am the only one who is actually testing here. cpu usage is the SAME just like bystander36 said.


31% here at 720


imagehosting


31% here at 1080


image hosting 12mb

 

[Carfax83]

What was your frame rate? You must have V-sync enabled, that's the only thing I can think of that would explain why the CPU usage is the same.

Use Fraps so you can see what frame rate you're running at..

 

[toyota]

What was your frame rate? You must have V-sync enabled, that's the only thing I can think of that would explain why the CPU usage is the same.

Use Fraps so you can see what frame rate you're running at..

lol please stop with the excuses. are you incapable of testing this yourself?

 

[Carfax83]

lol please stop with the excuses. are you incapable of testing this yourself?

I've done it myself. How hard is it to download Fraps and have the FPS counter running?

 

[toyota]

I've done it myself. How hard is it to download Fraps and have the FPS counter running?

you are making no sense at all. whether vsync is on are off has nothing to do with the fact that 1080 is still way more graphically demanding than 720 yet cpu usage does not go down.

 

[bystander36]

He's just trying to muddy up the conversation in stating that at a lower resolution, you'd get higher FPS, so the CPU usage goes up. Of course that is assuming that you aren't CPU bound.

The point is, it takes the same CPU usage to produce a frame, no matter the resolution. Yes, I was careful with my words so someone doesn't get confused.

 

[Carfax83]

you are making no sense at all. whether vsync is on are off has nothing to do with the fact that 1080 is still way more graphically demanding than 720 yet cpu usage does not go down.

Dude, just try it. If you had V-sync on at the 720p screenshot, then your FPS would be locked to 60. And depending on what settings you're using, if your 1080p screenshot had V-sync enabled as well, then it would also be locked to 60 FPS, which explains why the frame rate is the same..

Frame rate and CPU activity are linked together from my experience.. So the only way the CPU load remains the same is if the frame rate is also the same, and that's hard to believe unless V-sync was enabled.

 

[Keysplayr]

http://gamegpu.ru/rpg/rollevye/dragon-age-inkvizitsiya-test-gpu.html

Mantle vs DX11

Anybody notice anything wrong or "sketchy" with the first 3 graphs in the OP?

 

[toyota]

Dude, just try it. If you had V-sync on at the 720p screenshot, then your FPS would be locked to 60. And depending on what settings you're using, if your 1080p screenshot had V-sync enabled as well, then it would also be locked to 60 FPS, which explains why the frame rate is the same..

Frame rate and CPU activity are linked together from my experience.. So the only way the CPU load remains the same is if the frame rate is also the same, and that's hard to believe unless V-sync was enabled.

ok you were actually correct. I did not realize I had vsync on nor did I think it mattered. at 720 I was at 83% cpu usage and at 1080 I was at 63%.

1280x720 83% cpu usage


free upload


1920x1080 63% usage


free pic

 

[KaRLiToS]

Basic 101

Of course Vsync will have an impact on every usage of the PC parts. CPU or GPUs.