Guild Wars 2 - Retail Release - CPU and GPU Performance Testing - GameGPU.ru

[AtenRa]

1024x768

Core i7 3930K Minimum = 59fps Average = 85fp

1920x1080

GTX690 Minimum = 63fps Average 87fps

I thought we couldn't have more minimum fps than the highest CPU at 1024x768 because that was the highest your CPU will give you. :whiste:

[hint] In a GPU limited game, a GPU could be faster at 1080p than a CPU at 1024x768. [/hint]

Useless CPU 1024x768 benchmarks. This game is GPU limited even at 1680x1050. If you want to see how the CPU will effect the game at 1920x1080 you really have to bench at that resolution.

 

[boxleitnerb]

1024x768

Core i7 3930K Minimum = 59fps Average = 85fp

1920x1080

GTX690 Minimum = 63fps Average 87fps

I thought we couldn't have more minimum fps than the highest CPU at 1024x768 because that was the highest your CPU will give you. :whiste:

[hint] In a GPU limited game, a GPU could be faster at 1080p than a CPU at 1024x768. [/hint]

Useless CPU 1024x768 benchmarks. This game is GPU limited even at 1680x1050. If you want to see how the CPU will effect the game at 1920x1080 you really have to bench at that resolution.

You overlooked that the i7 for the GPU benchmarks is massively overclocked to 4.8 GHz. Aside from that: Minimums are very error prone as they are a single value. Looking at the average, we see that those values are basically the same. Benchmarks are rarely accurate to the exact fps. A few percent variation is often normal.

While the game indeed is GPU limited, the CPU benchmarks at 1024x768 (better 1280x720) are still important. They show for example that an FX8150 would begin to bottleneck anything faster than a 670. They also show how much CPU power you need for a given level of fps. If you need a constant 60fps, you will have to get something in the range of a 2600K and SLI/CF.
And again: If you need more fps, it is no crime to reduce details. Don't be so arrogant and assume that everyone will be playing the way you are, not touching any options and being content with the fps that are provided out of the box.

Finally, if you benchmark a CPU in games, you usually test several games. Some will be more GPU limited, some more CPU limited. You have to have the same testing methodology for all of them for consistency reasons. And if you don't know the benchmark sequence by heart and can truly be sure if it indeed is completely GPU, you test at low res to be sure. If you drop the lower res CPU benchmarks just because you have some GPU limited games in your lineup, you completely neglect the other games, making benchmarking them basically useless.

As for your hint: No!

If the CPU does 50fps at 1024x768 and 46fps at 1920x1080, that is all your GPU will ever be able to push unless you lower details that reduce CPU load.

I have to ask again:
Is it so difficult to check both, CPU and GPU benchmarks in their respective optimal settings and combine the results? I think most people are intelligent enough to do that. Lazy maybe, yes. But the basic principle here is simple and easy enough to understand.

AtenRa, for your own good, please stop posting

P.S.
How are the Shogun 2 and Civ 5 benchmarks coming? And don't tell me you didn't have time - you had the time to post this nonsense and other stuff since then.

 

[AtenRa]

You overlooked that the i7 for the GPU benchmarks is massively overclocked to 4.8 GHz. Aside from that: Minimums are very error prone as they are a single value. Looking at the average, we see that those values are basically the same. The 690 can do 92fps at 1680x1050, the 3930 can push 89fps - close enough. Benchmarks are rarely accurate to the exact fps. A few percent variation is often normal.

3930K output 85fps vs 92fps for the GTX690 at 1680x1050. That is not close enough


When you where talking about the minimum fps before the wasn't error prone ?? If the game is GPU limited, the minimums you will get could be set by the GPU and not your CPU at 1080p.

While the game indeed is GPU limited, the CPU benchmarks at 1024x768 (better 1280x720) are still important. They show for example that an FX8150 would begin to bottleneck anything faster than a 670.

Can you backup this claim ??? because you have never provided any proof.

And again: If you need more fps, it is no crime to reduce details. Don't be so arrogant and assume that everyone will be playing the way you are, not touching any options and being content with the fps that are provided out of the box.

You will still be GPU limited but it seams you dont understand that.

Finally, if you benchmark a CPU in games, you usually test several games. Some will be more GPU limited, some more CPU limited. You have to have the same testing methodology for all of them for consistency reasons. If you drop the lower res CPU benchmarks just because you have some GPU limited games in your lineup, you completely neglect the other games, making benchmarking them basically useless.

NOBODY plays at 1024x768 anymore, it is useless to bench in that resolution. Bench at 1080p and see how the CPU effects the game. It is simple as that.

As for your hint: No!

If the CPU does 50fps at 1024x768 and 46fps at 1920x1080, that is all your GPU will ever be able to push unless you lower details that reduce CPU load.

The graphs shows otherwise, GTX690 is faster than 3930K at 1680x1050 and at 1080p. It seams you dont understand that you are GPU limited, that means that a faster GPU will produce more fps.

Your CPU may provide only 50fps at 1024x768 but it may be enough to feed a faster GPU that will render more than 50fps at 1080p.

But unless you bench at 1080p you will never know

AtenRa, for your own good, please stop posting

I will be the judge of that, you keep watching your own doing.

P.S.
How are the Shogun 2 and Civ 5 benchmarks coming?

I dont have time for that right now, i prefer to spend my free time with my baby daughter.

 

[boxleitnerb]

3930K output 85fps vs 92fps for the GTX690 at 1680x1050. That is not close enough

Overclocked i7@4.8 GHz...cough.

When you where talking about the minimum fps before the wasn't error prone ?? If the game is GPU limited, the minimums you will get could be set by the GPU and not your CPU at 1080p.

They could...but they also could not. Depends on the CPU/GPU combination. Don't forget that in a test only one such combination is analyzed. The average reader may not have a 3930K@4.8 GHz...
Generally I put little faith in minimums. They might be roughly plausible, but I would not trust them to be perfectly accurate.

Can you backup this claim ??? because you have never provided any proof.

The benchmarks back that up pretty well. The FX can do 49fps avg at 1024x768 and would do about 45fps at 16:9 resolutions. There are clearly cards that can push more than 49fps avg. Look at the benchmarks in the first post.

You will still be GPU limited but it seams you dont understand that.

Example:
Someone has an 2500K and a 7870 GE and plays at 1080p. He gets around 40fps as shown in the GPU benchmark. However, for him that may be not enough, he may want 50 or 60fps for enjoyable gameplay. So he lowers details/resolution a bit and reaches 60fps. His 2500K can do that on average. If this person had an FX8150, he would not be able to do that. He could lower the details all he wanted, he would not be able to go above 45fps in that particular scene that was used for the CPU benchmark. He would be CPU limited.

NOBODY plays at 1024x768 anymore, it is useless to bench in that resolution. Bench at 1080p and see how the CPU effects the game. It is simple as that.

No, it is definitely not as simple as that.
It isn't about playing at that resolution, it is about showing what a CPU can ultimately do when you let it (by using a faster card, SLI/CF and/or lowering GPU load). See the example that I provided above.

When benchmarking at 1080p, you destroy valuable information and you assume that any person is happy with the provided fps. They may not be. And you also forget that not all games are GPU limited at 1080p.

The graphs shows otherwise, GTX690 is faster than 3930K at 1680x1050 and at 1080p. It seams you dont understand that you are GPU limited, that means that a faster GPU will produce more fps.

Your CPU may provide only 50fps at 1024x768 but it may be enough to feed a faster GPU that will render more than 50fps at 1080p.

But unless you bench at 1080p you will never know

Again you didn't notice that the 3930K was overclocked to 4.8 GHz.

As for your bold-font statement:
That is pure nonsense. A GPU can only render as many frames as the CPU can prepare for it at any given time. You can easily test that yourself when overclocking your GPU or going SLI/CF. You will never ever go above what the CPU can do.

I will be the judge of that, you keep watching your own doing.

I just don't want you do damage your reputation any further

I dont have time for that right now, i prefer to spend my free time with my baby daughter.

While I congratulate you on your family life...that is BS and we both know it. You prefer to spend your free time posting in the forums. Should I look up how many posts you did since then? You don't want to benchmark because it would prove you wrong. I have no doubts about that.

Edit: 100 posts since mine on 8-27-2012. Not including time to read on the forums to write these posts in answer to others. Please have at least the courtesy and not lie to me.

 

[balane]

Good sized new game build was downloaded this evening. Maybe it's in my head but the areas with frame slowdown (Supersampling.) don't seem nearly as bad. I've only tested the grassy PVE areas though.

 

[PrincessFrosty]

When you where talking about the minimum fps before the wasn't error prone ?? If the game is GPU limited, the minimums you will get could be set by the GPU and not your CPU at 1080p.

I think his point is that the minimum frame rate is only a single value in a very large set of values, and that it's possible for a very low minimum to occur that lays far outside the average frame rate, it can be a very misleading metric.

NOBODY plays at 1024x768 anymore, it is useless to bench in that resolution. Bench at 1080p and see how the CPU effects the game. It is simple as that.

Ugh. CPU Benchmarks are run at 1024x768 with low game settings for a very good reason, by lowering the resolution and by lowering visual settings you reduce workload on the GPU which shifts the bottleneck from the GPU to the CPU.

If you benchmark a bunch of different CPUs using high screen resolution and video settings, you force the system to be GPU limited then CPU performance is going to incorrectly appear uniform across all CPUs, which is misleading.

The point of benchmarking like this is to compare the relative speeds of the CPUs to show what sort benefit one CPU has over another in this particular game.

Your CPU may provide only 50fps at 1024x768 but it may be enough to feed a faster GPU that will render more than 50fps at 1080p.

No, if you've taken the GPU out of the equation by setting your graphics very low, and you've found the maximum frame rate that CPU can achieve then it won't get a higher frame rate with a better GPU.

Remember that your frame rate depends on both the CPU and GPU, the CPU to calculate the game state for every frame and the GPU to render the image, your actual frame rate at any point in time cannot exceed the slowest of these 2 components. If your CPU can calculate 50 new game states every second and the GPU can render 100fps you'll get 50fps output.

By benchmarking at a higher resolution you stress the GPU more, for example say you get only 30fps because of a slow GPU, and you test 3 CPUs that are capable of 50fps, 60fps and 70fps, in all these circumstances the final frame rate will be 30fps which makes it appear as if there's no difference between these CPUs, which is misleading.

This has been common best practice for benchmarking CPUs and GPUs in games for as long as I can remember:

-To test CPUs you remove the GPU bottleneck with artifically low graphics settings
-To test GPUs you remove the CPU bottleneck by using an insanely fast CPU often OC'd

 

[Anubis]

not sure what is blocking it but i cant resolve GameGPU.ru