I think the benchmark has problems

from AshesoftheSingularity Forums

...and I challenge you to prove me wrong. Check out these benchmark results:

These results indicate I am GPU-bound, that is, quote "If this number is near 100%, then a faster CPU would make no difference in the score". The logical inference is that a faster GPU is called for. That's the conclusion Frogboy came to here. Let's explore that. Check out ASADDF's benchmark:

He has the same video card as me, his test has higher graphics quality, his test is not GPU-bound, and he gets 50% higher FPS than me. You might say "Yeah, but he has a better processor and more memory". True, but wouldn't that imply that my CPU and memory are the bottle-necks, not the GPU? You might say "Yeah, but you are comparing apples to oranges, you have different systems running different software, it's not a fair comparison." Fair enough, but just taking raw numbers, it doesn't make sense for my system, with a weaker CPU, to be GPU-bound and his system, with a strong CPU, to not be GPU-bound.

Note: I'm not suggesting that the benchmark itself is at fault. I think either the data gathering, the results, or the interpretation of the results is wrong.

So, what's the word, hummingbird?

33,597 views 10 replies
Reply #1 from AshesoftheSingularity Forums Top

can it be maybe his card is oc ?

Reply #2 from AshesoftheSingularity Forums Top

He did not indicate such in other topics where he discussed his results.

Offworld Trading Company
Reply #3 from Sins Forums Top

As I asked in the other thread, there is a 2GB and a 4GB version of the 770 GTX, do you guys have different ones?

 

Reply #4 from AshesoftheSingularity Forums Top

We both have the 4GB version.

Reply #5 from Stardock Forums Top

Same drivers? And those are two different game versions.

Reply #6 from AshesoftheSingularity Forums Top

One is a cpu test, the other is a full system text as labeled at the top of your screenshots. :)

Reply #7 from AshesoftheSingularity Forums Top

Ok this is really weird

My benchmarks Changed Completely here

This is Full System test                                                                                  CPU Test
Full System TestCPU test

 

I am not sure what changed, the new AOTS build(0.60)? or the New Nvidia Drivers(358.50)?

That's DX11 benchmark
DX11 0.60

So Weird.. no changes whatsoever.

Ohhh well not sure Anymore, i did test it with 8GB and 24GB of Ram and its the same Thing, With 8GB its even faster lol.

Reply #8 from AshesoftheSingularity Forums Top

The CPU test with the latest build is just weird, it shows the result the same way as the Full System Test (instead of a CPU test from the previous versions), except the percentage GPU bound part, and the graph (which is also weird). When I run a cpu test, it seems everything is moving/flying/shooting/etc. much faster than with Full System Test. Also, the benchmark crashed for me, with the latest Catalyst (15.10), but that is probably a driver issue, since it's fine with 15.9.1.

Reply #9 from AshesoftheSingularity Forums Top

Okay what I think I learned from all this is that my 4.5 year old CPU is good enough for this game/benchmark in its current configuration because DX12 uses the CPU more efficiently. As such, my bottleneck is my 2.5 year old video card. If I got one of the high-end cards, Then my CPU would likely be the bottleneck. The reason my DX11 and DX12 numbers are so close, aside from nVidia working on more DX12 driver enhancements, is that one thread is enough to keep my video card pegged with DX11. So with a new card I might see some improvement in the numbers under DX11, but most likely it'd be less than the same configuration with DX12. Sound about right?

 

ASADDF, I appreciate you taking the effort to do the 8GB RAM test for my benefit. Above and beyond, so thanks!

Reply #10 from AshesoftheSingularity Forums Top

Quoting eviator, reply 9

Okay what I think I learned from all this is that my 4.5 year old CPU is good enough for this game/benchmark in its current configuration because DX12 uses the CPU more efficiently. As such, my bottleneck is my 2.5 year old video card. If I got one of the high-end cards, Then my CPU would likely be the bottleneck. The reason my DX11 and DX12 numbers are so close, aside from nVidia working on more DX12 driver enhancements, is that one thread is enough to keep my video card pegged with DX11. So with a new card I might see some improvement in the numbers under DX11, but most likely it'd be less than the same configuration with DX12. Sound about right?

If you have two CPU test results (a dx11 and a dx12) from a previous version of the game, you can compare them. Both have to be from the exact same build. If your cpu is capable for more with dx12, then you should see the benefit of a better gpu. Game saves log files about the benchmarks to Documents/My Games/Ashes of the Singularity/Benchmarks. If you don't have such cpu test results, then you might be able to do the comparison with the current build as well. Run two Full System Tests (dx11 and dx12) with the lowest possible graphics settings, no AA, no temporal AA, no vsync, etc.