Futuremark Releases 3DMark Time Spy DirectX 12 Benchmark

[john_]

benchmarks like time spy are all about high end..

About that. Benchmarks like Time Spy sell cards. The way Time Spy shows that Nvidia cards gain from async, half of what AMD cards gain, is that last little excuse many where waiting before hitting the order button.

 

[EarthDog]

If you would like to see how you compare to the rest of TPU, please follow the instructions in the first post in the link below~!

https://www.techpowerup.com/forums/threads/3dmark-time-spy-post-your-scores.224155/

 

[Caring1]

... benchmarks like time spy are all about high end.. people that download and run them dont do it to to prove how crap their PC is.. which is why its on my gaming desktop and not on my atom powered windows 10 tablet.. he he

trog

Not surprisingly people run benchmarks on all sorts of devices, because they want to know where their system fits in and how it performs in comparison to others.
Have a look at the Futuremark benchmark ticker, you might be surprised, it's not just high end.

 

[efikkan]

Basically the "A-sync" on Timespy is not the real A-sync integrated in Dx12 and Vulkan. It's a code path that can offer the similar effects IN THIS BENCH, and it "happens" to work well on nVidia's hardware. Even Maxwell can have a good time with this "A-sync"

So when the facts doesn't fit your narrative you just tweak them until they do?
The lie about Nvidia lacking the support for Async shaders has gone on for too long, and it's only kept alive by people who know nothing about how GPU rendering and archtectures actually work.

You simply can't claim that Nvidia is "faking it" and is able to achieve improvements like "if they had async shaders in hardware". If that even were remotely true, why don't they "fake it" in all games receive even more performance in all games?
I've not yet seen a proper benchmark which cheats to make something less significant to help a vendor. It's usually the complete opposite; the benchmarks commonly weights some new feature far more than any game. If your claim about Futuremark implementing a separate pipeline to help Nvidia were true, they must be the most stupid developers ever. It would defeat the sole purpose of the benchmark, making no one take Futuremark seriously ever again.

Indeed, it's about DirectX12 right now. I just can understand that everyone is ticked off that FutureMark is selling a "DirectX 12" benchmark, which actually doesn't do anything DirectX 12 related and just says "well, if we throw this work-load at it, we'll let the scheduler decide, which would be kinda like DX12" (talking about FM's response on Steam.)

How is this not a "DirectX 12 benchmark"?
You fanboys have been making up the rumor that AMD is so much better on everything Direct3D 12(based on AMD sponsored games BTW), and suddenly when unbiased benchmarks show up they suddenly don't qualify? You should be ashamed...

 

[sweet]

So when the facts doesn't fit your narrative you just tweak them until they do?
The lie about Nvidia lacking the support for Async shaders has gone on for too long, and it's only kept alive by people who know nothing about how GPU rendering and archtectures actually work.

You simply can't claim that Nvidia is "faking it" and is able to achieve improvements like "if they had async shaders in hardware". If that even were remotely true, why don't they "fake it" in all games receive even more performance in all games?



How is this not a "DirectX 12 benchmark"?

There is enough evidence to point out that Timespy is not a DirectX 12 benchmark.

"Use Dx11 if you can't do this", oh well

Up until Timespy, there has been no single game that can boost nVidia's fps when Async is on. There is even regression on non-Pascal cards. And suddenly Timespy comes out and we see the boost? Oh, nothing shady here.

If you are so confident with your knowledge, take a good look at Timespy documentation. The "A-sync" used here is not the real A-sync integrated in Dx12/Vulkan.

 

[trog100]

Not surprisingly people run benchmarks on all sorts of devices, because they want to know where their system fits in and how it performs in comparison to others.
Have a look at the Futuremark benchmark ticker, you might be surprised, it's not just high end.

ask the average PC user what their 3Dmark score is and all you will get is a blank look.. most people just use their computers not many care where they fit in the grand scheme of things..

trog

 

[ViperXTR]

More stuff from overclock forums desputing the async usage in time spy

http://www.overclock.net/t/1606224/...-less-compute-parallelism-than-doom-aots-also

 

[Prima.Vera]

More stuff from overclock forums desputing the async usage in time spy

http://www.overclock.net/t/1606224/...-less-compute-parallelism-than-doom-aots-also

This is good post. Proves that Futuremark's engine is not using at full potential the AMD cards. Interesting.

 

[john_]

I guess this is already posted somewhere
[computerbase.de] DOOM + Vulkan Benchmarked. - Page 23

an interesting part of the post in the above link is this

I am not disputing that Pascal can get a boost under light loads... never have. As for 3D Mark Time Fly... See concurrent vs parallel execution. All of the current games supporting Asynchronous Compute make use of parallel execution of compute and graphics tasks. 3D Mark Time Fly support concurrent. It is not the same Asynchronous Compute.

Concurrency fills in gaps which are in the execution pipeline. Parallelism executes two tasks at the same time.

Notice the context switch involved?

Parallelism reminds of me a 2 core cpu
Concurrency without parallelism reminds me of a 1 core cpu with Hyperthreading

They can't be scoring the same. If they are, then the benchmark is not reliable.


And there is one more question. Does Pascal really have extra hardware in there, compared to Maxwell? Or, what we see with 3DMark and Pascal cards is what Nvidia was promising in drivers to Maxwell owners, but never delivered?

 

[Exceededgoku]

Also to add flames to the fire, I noticed that there was some stuttering on dual R9 295X2.

Obviously I expect AMD to release a driver update to resolve that, but we were told that stuttering would be resolved by DX12 - although the lack of async is probably at fault here in that regard...

 

[PerfectWave]

All NVIDIA gpu should be labelled DX11 only
yeah Timespy aint the right benchmark for looking at DX12 performance sadly

 

[Prima.Vera]

Doom with Vulkan and AOTS are better benchmarks for D3D 12 imo.

 

[ShurikN]

It's a sad thing for Futuremark when a single Vulkan game is more representative of true DX12 capabilities than their "DX12" benchmark software.
Ironic isn't it...

 

[ViperXTR]

Futuremark responded and released a more in depth detailed information regarding Async Compute in TimeSpy and even included the GPU View shown in Overclockers forums and explains the sequence for each architecture (GCN, Maxwell and Kepler).
http://www.futuremark.com/pressreleases/a-closer-look-at-asynchronous-compute-in-3dmark-time-spy

 

[Deleted member 50521]

Futuremark responded and released a more in depth detailed information regarding Async Compute in TimeSpy and even included the GPU View shown in Overclockers forums and explains the sequence for each architecture (GCN, Maxwell and Kepler).
http://www.futuremark.com/pressreleases/a-closer-look-at-asynchronous-compute-in-3dmark-time-spy

PR damage control seems to kick in full gear now at Futuremark.

 

[Prima.Vera]

Futuremark responded and released a more in depth detailed information regarding Async Compute in TimeSpy and even included the GPU View shown in Overclockers forums and explains the sequence for each architecture (GCN, Maxwell and Kepler).
http://www.futuremark.com/pressreleases/a-closer-look-at-asynchronous-compute-in-3dmark-time-spy

There is a lot of bla-bla-bla written there, but as far as I understood, the benchmark is not using any vendor dedicated optimization for the test, which in my opinion makes the benchmark useless for DX12, where it's all about optimisation. For normal tests you have the rest of DX11/10/9 from the Futuremark.

 

[Fluffmeister]

Sorry @FM_Jarnis , but the AMD fanboys are still upset, could you make it more AMD focused next time?