Ashes of the Singularity Coming, with DX12 Benchmark in thread.

Gregster · 25 Feb 2016 at 16:52

Dygaza said:
I have the game Gregster, as I already replied to your last screens with my own. I think you missed them?

Yer, I know bud and even linked it on my site when discussing AoTS and DX12 and a win for AMD :cool:

I got a reply from Stardock confirming that it will go live at 6pm UK time, so I have recorded the current Beta 1 footage running extreme and crazy and will do the same on Beta 2 and then marry them up side by side and see what has changed and look at framerates. I expect come this Beta 2 to get a pasting but all cool

Mauller · 25 Feb 2016 at 16:56

Gregster said:
Yer, I know bud and even linked it on my site when discussing AoTS and DX12 and a win for AMD

I got a reply from Stardock confirming that it will go live at 6pm UK time, so I have recorded the current Beta 1 footage running extreme and crazy and will do the same on Beta 2 and then marry them up side by side and see what has changed and look at framerates. I expect come this Beta 2 to get a pasting but all cool

Does your card appear to have any of the lighting effects missing on the current version?

Gregster · 25 Feb 2016 at 16:58

Mauller said:
Does your card appear to have any of the lighting effects missing on the current version?

I watched that other video and mine looks like there is more lighting going on but it might just be the way he recorded it.

Harlequin · 25 Feb 2016 at 17:10

high setting @ 4k

CAT-THE-FIFTH · 25 Feb 2016 at 17:18

Mauller said:
This is a good point, the Kepler cards which have hardware scheduling see performance improvements in DX12, was shown on anandtech. The main problem is that Maxwell was cut down to bare bones for graphics processing which is why it is such a good over clocker and the main reason for its power savings.

Kepler uses software scheduling.

D.P. · 25 Feb 2016 at 18:20

Mauller said:
Shader code can work well on one vendors hardware but not another's. This was no different in DX11. The difference now is that AMD and Nvidia cannot substitute this code with better code in the driver. NVIDIA was more likely to do this while AMD would work with the dev and modify the code in source.

With DX12 the driver is a thin abstraction layer. It should do nothing but convert api call to hardware calls and back. In other words, the shader code can still be bad but only the AMD source change method will work with DX12.

And code can still be made to work well on both, even if it does not bring out best on any vendors hardware.

Not really true. NVIDIA were actually much more likely to.work with the developers to change the source code. Nvidia have a much more proactive developer relation program. Of course nvidia were also more likely than AMD to work at driver level changes. The difference is nvidia accepted developers made bad code, while AMD would stand by their driver compliance and request developers fix bad code.

And in DX12 there is still the possibility of changing shader code.

Gregster · 25 Feb 2016 at 18:24

No point comparing the old to the new lol. Different maps

D.P. · 25 Feb 2016 at 18:28

Harlequin said:
what like getting DX12 drivers for Fermi as they promised 4 months ago..... oh wait....

They didn't promise 4 months ago, the said when DX12 games were shipping, which hasn't happened yet.They indicated end of 2015 but since DX12 games didn't appear there was no need.
I'm sure the driver team is concentrating on getting Pascal drivers sorted.

LtMatt · 25 Feb 2016 at 18:28

First run
1080P Crazy
0.90 Build
16.2
1x Fury X Stock
5960X @4Ghz

The benchmark is now different, so old results are meaningless.

Vince · 25 Feb 2016 at 18:30

So is the new version released? I can't seem to work out how they roll the versions of this out as the one on my desktop is a different version to the one on my laptop, yet I don't have any updates queued... Random.

Mauller · 25 Feb 2016 at 18:32

D.P. said:
Not really true. NVIDIA were actually much more likely to.work with the developers to change the source code. Nvidia have a much more proactive developer relation program. Of course nvidia were also more likely than AMD to work at driver level changes. The difference is nvidia accepted developers made bad code, while AMD would stand by their driver compliance and request developers fix bad code.

And in DX12 there is still the possibility of changing shader code.

It will be very unlikely as the application directly handles the shader code and submission. The only reason DX11 drivers could substitute shader code is because the driver handled gpu submission and draw call packaging.

If it was the case then nvidia would have just tried altering the shader code for AOTS at the driver level and supplanting it with their code changes that benefit them the most.

Mauller · 25 Feb 2016 at 18:35

AMDMatt said:
First run
1080P Crazy
0.90 Build
16.2
1x Fury X Stock
5960X @4Ghz

The benchmark is now different, so old results are meaningless.

Try with multiple fury matt, mgpu should work now.

LtMatt · 25 Feb 2016 at 18:43

AMDMatt said:
First run
1080P Crazy
0.90 Build
16.2
1x Fury X Stock
5960X @4Ghz

The benchmark is now different, so old results are meaningless.

Mauller said:
Try with multiple fury matt, mgpu should work now.

1080P Crazy
0.90 Build
16.2
2x Fury X Stock
5960X @4Ghz

queamin · 25 Feb 2016 at 18:48

Good scaling there

Dygaza · 25 Feb 2016 at 18:54

So here are my new results:

I7-3770k @ 4.2GHz, Fury X @ 1120/545 (-12 mV voltage)
1080p extreme:

1080o crazy:

I think I need new cpu

Matt!!! Zen!!! Give!!!

Gotta test absolute max cpu scaling aswell. Edit - nevermind there ain't cpu bench anymore. Well quess that saves us from reading some reviews wrongly

LtMatt · 25 Feb 2016 at 18:58

Dygaza said:
So here are my new results:

1080p extreme:

1080o crazy:

I think I need new cpu Matt!!! Zen!!! Give!!!

Gotta test absolute max cpu scaling aswell. Edit - nevermind there ain't cpu bench anymore. Well quess that saves us from reading some reviews wrongly

Nice results, you beat me there.

What speed was your CPU/GPU running at?

Gregster · 25 Feb 2016 at 19:01

Now uploading my vid that comapres Extreme against Crazy and looking at Matt and Dygaza's results, I think I need to do some overclocking

AMD giving me a good kicking in this

Harlequin · 25 Feb 2016 at 19:05

anyone got an FX cpu to test this with?

Dygaza · 25 Feb 2016 at 19:09

AMDMatt said:
Nice results, you beat me there.

What speed was your CPU/GPU running at?

I7-3770k @4.2GHz,
Fury X @ 1120/545MHz (-12 mV voltage).

Added them to my result post.

For some reason I get better performance when I undervolt it , (1120 / -12 mV is faster than 1150 / +60mV)

LtMatt · 25 Feb 2016 at 19:10

Dygaza said:
I7-3770k @4.2GHz,
Fury X @ 1120/545MHz (-12 mV voltage).

Added them to my result post.

For some reason I get better performance when I undervolt it , (1120 / -12 mV is faster than 1150 / +60mV)

I should try with HT on and off and see what differences it makes. I'll give it a go.