• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

DirectX 12Futuremark 3DMark Time Spy Benchmarks Thread

C

csbin

Senior member
http://www.anandtech.com/show/10486/futuremark-releases-3dmark-time-spy-directx12-benchmark



TimeSpyStats_575px.png


Explicit1_575px.PNG



82847.png
 
Last edited:
3449 overall, 3259 graphics on stock 6700K/GTX 970.

4013 overall, 3868 graphics on stock 6700K/GTX 970 @ 1500/7800.
 
Last edited:
I wonder if my system will be any good with this benchmark.

Guess I'l have to try it out when I get home.
 
This may end up not being very representative of actual game performance. At which point it would be a pointless benchmark.
 
Wound up with 5675 GTX 1070 Amp with default clocks and settings.

4790k @4.4GHz

http://www.3dmark.com/spy/8372

Guess I can try a little OC and see what happens.

On another note why don't they at least make visually pleasing benchmark apps? Seems like the overall scores just don't match the visual fidelity as one calls it.

Seems like this might be good for dialing in OC settings at least.

Left core at stock speed. Upped power limit slider to max. Upped memory clocks to 9026.

Score increased to 5869

http://www.3dmark.com/spy/9958

Going to see how high memory will be stable before I mess with the core.

Set my 4790k back to my 4.6GHz overclock. Forgot I dialed it back down during the DPC latency scare testing.

Upped the memory to 9072
Score upped to 5992

http://www.3dmark.com/spy/10795
 
Last edited:
crisium said:
480 only having 4 ACE's hurting it? Doesn't see the Fiji gains.
Click to expand...

Forgot where I read it, but someone said 480 has 8 ACEs, the 2 HWS have 2-ACEs each.

Wish I could remember where I read it, so I could link you a more technical explanation haha.
 
Bacon1 said:
Isn't that still 4?

2 HW * 2 ACE each = 4.
Click to expand...

P4.png


4 ACE + 2 HWS (functionally 2 ACE in one HWS , sort of that is).


But the strange thing is, if i look up the fury article on anandtech, it has 8 ACE.
http://www.anandtech.com/show/9390/the-amd-radeon-r9-fury-x-review/4
But this article is from July 2015 and the extremetech article is from Februari 2016. I guess AMD did not release all information when Anandtech wrote the article. Or i am overlooking something. :hmm: .. Or we see here microcode updates in effect ? The ACE from GCN gpu's can be updated with microcode. At least, that is what i have been reading about.


FijiBlockDiagram.png


And if i look up the article on extremetech, it has Fury has 4 ACE and 2 HWS.

http://www.extremetech.com/gaming/2...ashes-of-the-singularity-directx-12-benchmark

It is mentioned here (assuming the information is correct) :

A GPU that supports asynchronous compute can use multiple command queues and execute these queues simultaneously, rather than switching between graphics and compute workloads. AMD supports this functionality via its Asynchronous Compute Engines (ACE) and HWS blocks on Fiji.

Asynchronous computing is, in a very real sense, GCN’s secret weapon. While every GCN-class GPU since the original HD 7970 can use it, AMD quadrupled the number of ACEs per GPU when it built Hawaii, then modified the design again with Fiji. Where the R9 290 and 290X use eight ACEs, Fiji has four ACEs and two HWS units. Each HWS can perform the work of two ACEs and they appear to be capable of additional (but as-yet unknown) work as well.
Click to expand...

Fiji :

AMD-HWS.jpg


Polaris :

P4.png



EDIT:
Forgot to mention that the RX 480 does really well while comparing with the GTX 1070, the gpu the RX 480 was not meant to compete with.
No need to cry, the RX 480 is doing really well.

But why no tests at 1080p , the target resolution for the RX 480. 😡
 
Last edited:
cn66uiD.jpg


Using daily OC settings. I could push it a hair further and hit 10k, but screw it. Not bad I guess.

Overall: 9929
Graphics: 11275
CPU: 5923
 
Last edited:
That first slide is nuts.

TimeSpyStats_575px.png


This is basically throwing more at the scene with minimal visual quality improvement. -_-

I hope games are not going to be that wasteful.
 
Sweepr said:
So now we know what to expect from neutral titles. 🙂
Click to expand...

Doom is a neutral title btw, no GameWorks or GE. 🙂 So is Quantum Break.

It is actually interesting to note, Ashes & Warhammer, despite being AMD sponsored, runs quite well on NV GPUs. The gap is nowhere near as close as in those two neutral games above.

I wonder if there's any CPU utilization results reported by Time Spy.

If NV has finally "enabled Async" in their drivers for Pascal and not Maxwell (like they promised), could well be a CPU off-loading going on.

Fundamentally, DX12 does not specify how those separate queues (Graphics, Compute, Copy) get processed and one of the original theory when the Maxwell fake Async scandal erupted is CPU off-loading is possible for some tasks.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top