Ashes of the Singularity, Rise of the Tomb Raider & Hitman CPU Tests – DX11 vs DX12

[Sweepr]

Low-end Skylake Celeron G3920 is where the biggest gains were seen, ~42% boost (Fury X).
Curiously DX12 also improves Geforce GTX980 Ti's results here.

Overclocked Core i7-6700K - ~25% higher framerate (Fury X).

http://pclab.pl/art67995.html


Nice to see DX12 improving the performance of Intel dual-cores as well as more expensive chips with higher core/thread count. Quantum Break results should be interesting.

 

[monstercameron]

Interesting lack of commentary about the amd platform.

 

[Sweepr]

Interesting lack of commentary about the amd platform.

Plenty of websites already tested AMD hardware and Core i5/i7 @ AotS, but this was the first time I saw Celeron results.

 

[dark zero]

All the chips, even the FX got benefits on this game.. Except GTX960 who all the Intel chips got nerfed.

Even the Celeron looks decent now.

Where are the people who said that DX12 is not the future?
This is the proff that DX12 indeed is the future along with Vulkan.

 

[monstercameron]

Plenty of websites already tested AMD hardware and Core i5/i7 @ AotS, but this was the first time I saw Celeron results.

Fair enough.
Darn, I wanted to start trouble, y u no take bait.

Ot, I kinda want to see athlon 5350 and the athlon 860/845 results. I've only seen am3+ results.

 

[dark zero]

Fair enough.
Darn, I wanted to start trouble, y u no take bait.

Ot, I kinda want to see athlon 5350 and the athlon 860/845 results. I've only seen am3+ results.

Not only Athlon 5350, but also Bay Trail and Phenom II.

 

[inf64]

Overclocked Core i7-6700K - ~25% higher framerate (Fury X).

Hmm looking at last 2 charts I see basically the same fps on OC and non-OC 6700K + FuryX (both DX11 and 12).

 

[bystander36]

All the chips, even the FX got benefits on this game.. Except GTX960 who all the Intel chips got nerfed.

Even the Celeron looks decent now.

Where are the people who said that DX12 is not the future?
This is the proff that DX12 indeed is the future along with Vulkan.

Are you sure anyone ever said DX12 is not the future? At most, people might say it will take a few years before it is mainstream. We've all known it was the future.

 

[Sweepr]

Hmm looking at last 2 charts I see basically the same fps on OC and non-OC 6700K + FuryX (both DX11 and 12).

The comparison there was DX11 vs DX12 @ OCed Core i7-6700K.

 

[Shivansps]

The Fury X performs the same on every CPU wtf.


Scratch that, every card performs the same.

 

[inf64]

The comparison there was DX11 vs DX12 @ OCed Core i7-6700K.

On a stock 6700K the difference between DX11 and DX 12 is 26%, OCing 6700K makes no difference. Heck, even 6600K @ stock score almost the same, ~ minus a frame or two.

 

[superstition]

The last Ashes results I saw were pretty lame-looking in terms of multithreading, with a 3.5 GHz Haswell i3 soundly beating an FX 8370 in DX 12. The i3 got a huge boost over DX 11 but the AMD didn't get so much.

 

[Abwx]

The last Ashes results I saw were pretty lame-looking in terms of multithreading, with a 3.5 GHz Haswell i3 soundly beating an FX 8370 in DX 12. The i3 got a huge boost over DX 11 but the AMD didn't get so much.

http://www.computerbase.de/2016-02/ashes-of-the-singularity-directx-12-amd-nvidia/5/

 

[superstition]

http://www.computerbase.de/2016-02/ashes-of-the-singularity-directx-12-amd-nvidia/5/

Interesting. Same results as before (3.5 GHz i3 beating the 8370 with a 980) but when moving to Fiji they end up tying.

Still pretty lame that a 4.0/4.3 4 module part is trailing a dual core running at 3.5.

 

[Abwx]

Interesting. Same results as before (3.5 GHz i3 beating the 8370 with a 980) but when moving to Fiji they end up tying.

Still pretty lame that a 4.0/4.3 4 module part is trailing a dual core running at 3.5.

Difference is negligible, and the game is thread count limited anyway, apparently it doesnt use more than 4 cores, the FX6300 does better than Kaveri surely due to its L3 cache.

 

[deasd]

Interesting. Same results as before (3.5 GHz i3 beating the 8370 with a 980) but when moving to Fiji they end up tying.

Still pretty lame that a 4.0/4.3 4 module part is trailing a dual core running at 3.5.

This game is using 4 threads if I figure it right. Not very surprised that Piledriver on par with Haswell i3.

 

[dualsmp]

Nvidia doesn't see any improvement until it gets to the Celeron processor. I'm guessing since Nvidia drivers take advantage of multiprocessing better than AMD there is little improvement on most processors when moving from DX11--> DX12.

 

[deasd]

Nvidia doesn't see any improvement until it gets to the Celeron processor. I'm guessing since Nvidia drivers take advantage of multiprocessing better than AMD there is little improvement on most processors when moving from DX11--> DX12.

Most acceptable reason is NVIDIA doesn't have effective shader async compute on DX12 at current SKUs.
​

https://www.reddit.com/r/nvidia/comments/3j5e9b/analysis_async_compute_is_it_true_nvidia_cant_do/

Since this game seems favor Intel + NVIDIA from most of other review, the case above is good to explain why DX12 eliminate the gap between AMD/NVIDIA dramatically in this game.

 

[Denithor]

980 Ti is painfully gimped on 2C/2T Celeron in DX11, almost fully unleashed by DX12. Once you get to a 4T chip the CPU fails to have ANY further impact whatsoever (<1 FPS difference between i3 and OC i7). This looks like classic GPU-bound behavior, throwing more and more CPU power behind it doesn't increase throughput in the least.

FuryX on the other hand, different story. Huge improvements going from DX11 to DX12 in every case, biggest gains down in Celeron territory where it was obviously the most CPU-bound. Going up in CPU power has a minor impact on DX12 performance, 3% gain from Celeron to i3, basically no change from there until the jump to 4C/8T (5% increase). This says the FuryX is still slightly CPU-bound even at the top end.

Now, looking at CPU impact on FuryX in DX11, it's an even more CPU-bound situation. There's a 20% improvement going from Celeron to i3 and then 3-5% increases with each additional step up (in either speed or core/thread count).

So final conclusion, I'd say FuryX under DX12 is far less CPU-bound than it is under DX11.

 

[Sweepr]

As an example to illustrate the point, below is a screenshot of a scene in the game running on an Intel i7-2600 processor with 1333Mhz memory, paired with a GTX 970. Using DirectX 11 at High Settings we would only get 46 fps.

DX11

DX12

http://tombraider.tumblr.com/post/140859222830/dev-blog-bringing-directx-12-to-rise-of-the-tomb

+30% better performance, tested with Sandy Bridge Core i7-2600K from 2011.
Looks like DX12 should benefit Intel's dual-cores and older quad-cores.

 

[moonbogg]

Hmm, maybe I can keep this 3930k for 5 more years.

 

[el etro]

It far excced my expectations. DX12 is a beast.


CPU dependancy is over. Save half of your CPU budget to buy a better GPU.


Like trading a $340 i7 for a $200 i5 and then pick a high tier GPU.



PS: by the way, i will give dolphin API results on RE4 mercenaries, i5 2310 plus HD7950 Boost:

 

[moonbogg]

It far excced my expectations. DX12 is a beast.


CPU dependancy is over. Save half of your CPU budget to buy a better GPU.


Like trading a $340 i7 for a $200 i5 and then pick a high tier GPU.



PS: by the way, i will give dolphin API results on RE4 mercenaries, i5 2310 plus HD7950 Boost:

Even if CPUs get effectively much faster, developers will get equally more lazy and terrible at their jobs, or they will do a good job and use all that CPU power, and the end result will be CPU dependency all over again.
We are getting closer to the launch of broadwell-E and I can't wait so that those chips can die and be forgotten and then we can get to Skylake-E and have something worth buying. 8 core skylake overclocked + DX12 = beast with bloody teeth!

 

[Blue_Max]

Looking at the nVidia stuff, I'd say this doesn't benchmark the CPU's at all - the game setting is so high, it's pure video card stats. Even the GTX 980TI barely moved at all from top CPU to the lowly dual-core.

But it's nice to see AMD getting some big performance gains. Interesting how their gear seems to work so much better under it!

 

[Sweepr]

17% better performance with Haswell Core i3 (DX11 vs DX12) @ Soviet Base in Rise of the Tomb Raider. Probably one of the most CPU intensive portions of the game.