GeForce + Radeon Previewing DirectX 12 Multi-Adapter with Ashes of the Singularity

[D.P.]

Valid point, but quite outdated review. Afterall we're talking about DX12 which gives devs the control , what to split, and when. Before it was done by driver.

Also keep in mind that you don't need to split everything equally (you can't). So you can't get as good scaling as you get with AFR. But you do get the smoothness, and benefit of adding cards that ain't equally as fast.

It's not easy to do, but deffinately worth the effort.

DX12 doesn't really change anything. Devs have more control but they can't resolve the fundamental issue that you don't know how long a portion of the view port will take to render until it has been rendered and you measure it.

For example in a Doom 3 corridor shooter type games you have a horizontal split, both ceiling and floors are equally complex so the split it in the middle. A big monster crawls over the ceiling. In the basic form of SFR the driver would detect that the upper half of the screen is taking longer to render to will start sliding the split point up, but it has no idea how far will create equal load so can only shift a little. Next frame it is still taking longer so it shifts again, again, and again catching up but not without signficiant lag in the load equalizing. A split second later the monster jumps to the floor and now the work load is massivley screwed again and the lower GPU is being over loaded.


If the Dev is in control then they can know that there is a complex monster with loads of fancy fragment shader on the ceiling to render so the split needs to be higher up. The problem is again the developer wont really know where that split should be because they can only really guess how much longer the monster will take to render vs the floor, and the timing difference will highly depend on the GPU as well and a load of parameters of the shader such as lighting conditions. Its a complex non-linear optimization problem the devs simply wont resolve. You can make the scenario arbitrarily complex, e.g. a different monster with different shaders, different tessellation, different lighting etc. now crawls along the floor and something else crawls up the left wall.

You basically continuously drop back to single GPU performance, but with increased lag and jitteryness.


the 3DFX scanline mode worked better because there was an even load distribution but that only worked for the basic texturing GPUs were capable of then, and all Geometry was done on the CPU (3DFX's downfall and Nvidia's killer technology). With the modern rendering pipeline that gets very inefficient.

AFR is much simpler to get right and the only way to consistently get over 80% scaling and a smoother gaming experience.

 

[Kaapstad]

The 5960X isn't a great performer in Ashes, Skylake beats it:

Thankyou for making one of my points for me.

8/16 cores/threads v 4/8 cores/threads in DX12 should be an absolute non contest and just goes to highlight my point that AOTS game/bench is seriously flawed.

And what you are forgeting from the graph you posted is the 5960X is a pretty reasonable overclocker despite having 8/16 cores/threads. We don't all run our Haswell-E CPUs @stock like the people who produced that graph were.

 

[chroniclard]

AOTS bench 2 is coming soon.

 

[Dave2150]

Errm, DX11 is very poorly threaded and at best the 6700k has 20% higher clock speeds(3.5 vs 4.2), in reality from a very quick look at reviews I would expect the 5960x to not get as high out of it's turbo and the gap between stock/turbo is also much greater on the 5960x, even greater when you consider the clockspeeds 3Ghz to 3.5Ghz is just over 16%, 4 to 4.2Ghz is 5%.

So with a 20% minimum but maybe closer to 30% clock speed difference in DX11 the 6700K is all of 15% faster than the 5960x, under DX12 it's about 4%. With far great overclocks available(due to lower base clock) which is the 'faster' chip. If both clocked to 4.5Ghz, the 5960x would be spanking the 6700k in DX12 with ease.

6700k does 4.7 on air... Dont think any 5960x is doing that. On water the 6700k can do 5.1Ghz.

 

[Kaapstad]

6700k does 4.7 on air... Dont think any 5960x is doing that. On water the 6700k can do 5.1Ghz.

Why don't you talk to AMDMatt, his 5960X is quite a bit faster than nearly all 6700k CPUs.

What's more he uses an AIO cooler rather than a custom loop.

 

[Dygaza]

Remember Skylake is faster in clock to clock aswell, it's newer arch with some nice improvements. And we are only talking about few fps here, in older version of benchmark. Remember there has been roughly 50% cpu performance uplift in game since that test came out.

 

[Kaapstad]

Remember Skylake is faster in clock to clock aswell, it's newer arch with some nice improvements. And we are only talking about few fps here, in older version of benchmark. Remember there has been roughly 50% cpu performance uplift in game since that test came out.

There is not that much difference in per core performance between Haswell-E and Skylake but the 8/16 cores/threads of a 5960X will totally thrash the 4/8 of a 6700k in things like DX12.

Sadly AOTS is not a very good of how DX12 should perform.

 

[Andygully]

Can see nvidia locking this down. But what interests me with this is when you decide to upgrade say from a 970 to a 980ti you could drop the 970 to the second slot and have the extra performance from it.

 

[r7slayer]

6700k does 4.7 on air... Dont think any 5960x is doing that. On water the 6700k can do 5.1Ghz.

Really so your saying on water most can do 5.1? Because i'm not seeing this. This seems like that same claim that the 4790k would do 5Ghz but you only saw a select few reaching these frequencies. So that's more pot luck and you have to be lucky to get those high freq. Silicon lottery basically so saying a 6700k can do 5.1 like that is insinuating its easily achievable when i bet its not.

4.5 is probably more achievable on a 6700k but still not guaranteed to hit this freq.

 

[Gerard]