Deus EX: Mankind Dividied system specs revealed

[Carfax83]

DX12 performance with previous patch v1.10:

Now with the current V1.11 patch:

While the in game benchmark is crap and not really indicative of actual in game performance (which is much higher), it's still a good reference for performance monitoring from patches and driver updates. As you can see, this latest patch increased performance substantially

 

[AtenRa]

DX12 performance with previous patch v1.10:



Now with the current V1.11 patch:



While the in game benchmark is crap and not really indicative of actual in game performance (which is much higher), it's still a good reference for performance monitoring from patches and driver updates. As you can see, this latest patch increased performance substantially

Nice thanks,

Can you do some more CPU testing ?? test DX-11 vs DX-12 on the latest patch with 2C 4T, 4C 4T and 6C 6T

 

[2is]

On my system (specs in sig) with the latest patch and latest drivers, I drop 2 fps across the board (min/max/avg) when enabling DX12. Looks like DX11 is still the way to go for Maxwell cards.

 

[Carfax83]

Nice thanks,

Can you do some more CPU testing ?? test DX-11 vs DX-12 on the latest patch with 2C 4T, 4C 4T and 6C 6T

Man, that's too much benchmarking!

 

[Carfax83]

On my system (specs in sig) with the latest patch and latest drivers, I drop 2 fps across the board (min/max/avg) when enabling DX12. Looks like DX11 is still the way to go for Maxwell cards.

The performance between DX11 and DX12 is still very close, with DX12 now being slightly faster than DX11 for Pascal. The big advantage though with DX12, is that the hardware resources are used much more effectively, which translates to more consistent and smoother frametimes and less hitches and stutters.

This should be noticeable on your rig. Don't just look at framerates when determining performance, as that's just one aspect of overall performance. Frametime consistency is just as, if not more important.

 

[2is]

The performance between DX11 and DX12 is still very close, with DX12 now being slightly faster than DX11 for Pascal. The big advantage though with DX12, is that the hardware resources are used much more effectively, which translates to more consistent and smoother frametimes and less hitches and stutters.

This should be noticeable on your rig. Don't just look at framerates when determining performance, as that's just one aspect of overall performance. Frametime consistency is just as, if not more important.

Played about an hour on both DX11/DX12. DX11 is still a better (smoother) experience on my machine.

 

[AtenRa]

Man, that's too much benchmarking!

ehehe ok, could you do DX-11 vs DX-12 with 2C 4T ?? but please disable CPU cores through BIOS to simulate Core i3.

 

[Carfax83]

Played about an hour on both DX11/DX12. DX11 is still a better (smoother) experience on my machine.

Weird.. I'm getting a 5 or 6 FPS increase with DX12 in Prague now compared to DX11, and DX12 is much smoother.. I wonder if this has to do with asynchronous compute? Asynchronous compute can be enabled or disabled in the registry. I might mess with that setting later and see if it affects anything..

In Gears of War 4, AC is good for a 4 to 5 FPS increase, plus I've noticed that the framerate is more stable.

 

[Carfax83]

ehehe ok, could you do DX-11 vs DX-12 with 2C 4T ?? but please disable CPU cores through BIOS to simulate Core i3.

Do you mean with the in game benchmark? If so, prepare for disappointment, but I can't say I'm surprised given that the in game benchmark is completely GPU bound and not indicative of actual game performance. I disabled every core except 0 and 1 in the UEFI like you asked, though I should note that it's not an accurate simulation of the Core i3, as these cores have access to 15MB of L3 cache, so that will definitely increase performance over a real Core i3.

That said, I also ran around Prague a bit for the 2C/4T test, so I will provide my subjective opinions which should be much more useful. With 2C/4T, the framerate was all over the place, and was definitely unstable. On average I would say it was about 10 FPS less, but the worst thing by far was the framerate drops. In some parts of Prague, the framerate would drop by 20 FPS or more suddenly, and then go back up which I believe is due to the slower asset streaming. This occurred in both DX12 and DX11, but more so the latter. As I've said before, DX11 also suffers from way more stuttering and significantly higher frame lag compared to DX12, and this was amplified even more with just 2 cores enabled.

DX11 2C/4T:

DX12 2C/4T:

 

[AtenRa]

Do you mean with the in game benchmark? If so, prepare for disappointment, but I can't say I'm surprised given that the in game benchmark is completely GPU bound and not indicative of actual game performance. I disabled every core except 0 and 1 in the UEFI like you asked, though I should note that it's not an accurate simulation of the Core i3, as these cores have access to 15MB of L3 cache, so that will definitely increase performance over a real Core i3.

That said, I also ran around Prague a bit for the 2C/4T test, so I will provide my subjective opinions which should be much more useful. With 2C/4T, the framerate was all over the place, and was definitely unstable. On average I would say it was about 10 FPS less, but the worst thing by far was the framerate drops. In some parts of Prague, the framerate would drop by 20 FPS or more suddenly, and then go back up which I believe is due to the slower asset streaming. This occurred in both DX12 and DX11, but more so the latter. As I've said before, DX11 also suffers from way more stuttering and significantly higher frame lag compared to DX12, and this was amplified even more with just 2 cores enabled.

DX11 2C/4T:



DX12 2C/4T:

Excellent, thanks a lot

Yes i was more interested on the actual Game Play rather than the build in benchmark, thank you for providing both.
Benchmark numbers dont always tell the entire story, no matter the fps you are getting with a dual core even if it has HT like the Core i3, its not what you want for the new 2016 and onward games. And yes im aware this is not a real Core i3 because of the larger L3, but it will do ok for simulating a Dual Core + HT.

I have seen this happening so many times in the past but people still only looking at the benchmark numbers in reviews and believe that Dual Cores are fine because they are producing 50-60fps or more.
What is even more sad is that reviews (even from highly respectable sites) dont include frametime analysis. It is highly misleading not to inform the readers about high frametimes that results in to stuttering, every CPU/GPU review should include them.

 

[2is]

Weird.. I'm getting a 5 or 6 FPS increase with DX12 in Prague now compared to DX11, and DX12 is much smoother.. I wonder if this has to do with asynchronous compute? Asynchronous compute can be enabled or disabled in the registry. I might mess with that setting later and see if it affects anything..

In Gears of War 4, AC is good for a 4 to 5 FPS increase, plus I've noticed that the framerate is more stable.

I think it has to do with Pascal vs Maxwell

 

[Carfax83]

I have seen this happening so many times in the past but people still only looking at the benchmark numbers in reviews and believe that Dual Cores are fine because they are producing 50-60fps or more.

Since the advent of the PS4 and Xbox One, game developers had to increase the parallelism of their game engines to fully exploit the CPUs in those consoles. As a result, CPUs with less than four cores/threads can really get piled on because task based parallelism will run several different tasks on the same thread. So the less threads that are available, means that the CPU can easily be overwhelmed.. That's probably what was causing the massive framerate drops I was experiencing..

A true dual core would likely be totally unplayable with Deus Ex MD..

What is even more sad is that reviews (even from highly respectable sites) dont include frametime analysis. It is highly misleading not to inform the readers about high frametimes that results in to stuttering, every CPU/GPU review should include them.

Probably because of the amount of work involved with frametime analysis. Benchmarking is already boring and time consuming, and now with frametime analysis you have to do even more work

 

[Carfax83]

I think it has to do with Pascal vs Maxwell

Yeah I think you're right. I turned off asynchronous compute for Deus Ex MD and the performance loss was negligible. I think asynchronous compute is only used for PureHair simulation in Deus Ex MD PC, and that's it. In games like Doom and Gears 4, it has a much bigger impact because more compute is being overlapped with the graphics..

Would be interesting to see how much AMD gains from asynchronous compute in Deus Ex MD. Any AMD users want to test it out? Open regedit, go to HKEY_CURRENT_USER, SOFTWARE, Eidos Montreal, Deus Ex MD, Graphics...