DX12 ... why you want it for P3D

[Guest]

http://wccftech.com/dx12-revealed-compared-dx11/

 

As you can see by these charts D3D12 should lead to a stutter free experience (especially for those of you working the P3D platform and sticking with it).  Note:

 

 

Performance in CPU limited scenarios will definitely see a great improvement.

 

Now lets see if my "huge" influence (I'm joking here of course) with LM can get them to adopt DX12 as it seems like a perfect fit for the P3D platform.

 

This is probably the most "exciting" news I've seen of recent given I was pretty disappointed with Intel's Broadwell-E and have pushed off a hardware upgrade as I wait for Skylake in late Q3 2015 before making the hardware upgrade.

 

However, I think the DX12 and Skylake combo will put an end to performance bottlenecks ... maybe even make the need for SLi moot.

 

Cheers, Rob.

 

 

 

[ChaoticBeauty 6,692]

DirectX 12 would bring huge gains to CPU-limited scenarios in Prepar3D, considering DirectX 11 contains a 20% CPU overhead.

 

Hopefully they'll start experimenting with it in 2.x as well instead of waiting for the next major version.

[Guest]

 

 

Hopefully they'll start experimenting with it in 2.x as well instead of waiting for the next major version.

 

No idea what LM will do with DX12 ... of course for full DX12 hardware support we need nVidia Maxwell based GPUs (i.e. 980GTX), however existing DX11 cards will be able to run DX12 API but might be missing a few hardware accelerated features. 

 

It's a perfect match for P3D v2.x -- but future is looking good.

 

Cheers, Rob.

[ChaoticBeauty 6,692]

No idea what LM will do with DX12 ... of course for full DX12 hardware support we need nVidia Maxwell based GPUs (i.e. 980GTX), however existing DX11 cards will be able to run DX12 API but might be missing a few hardware accelerated features. 

 

It's a perfect match for P3D v2.x -- but future is looking good.

 

The features will be missing in GPUs before Maxwell 2.0, but all GPUs down to Fermi will benefit from the massively decreased overhead, and just that is enough to make Prepar3D shine.

[CrownCityMisfits 58]

No idea what LM will do with DX12 ... of course for full DX12 hardware support we need nVidia Maxwell based GPUs (i.e. 980GTX), however existing DX11 cards will be able to run DX12 API but might be missing a few hardware accelerated features. 

So my GTX680 would still benefit from DX12? What kind of accelerated features would I miss?

 

I built my first PC from scratch about 18 months ago, and I am still trying to learn how everything works under the hood. I really appreciate the info, even if obvious to most.

 

Aaron

[Guest]

So my GTX680 would still benefit from DX12?

 

900 series is Maxwell, 700/600 series is Kepler, 500 series Fermi ... Maxwell will obviously get the most benefit, but our 700/600 series cards should see "some" benefit, how much is yet to be determined.

 

I'll probably dump my Titans when nVidia Volta cards are released ... but that's 2016 ... but timing might be right given Skylake CPUs in Q3 2015.

 

So by early 2016 we should have some pretty impressive CPU/GPU improvements (as in actually 2X performance over current hardware TOL hardware, not just 10-20%).  I think the benefit here is that DX12 will actually make a pretty significant difference for those platforms that will support it and API transition from DX11 to DX12 doesn't appear to be as difficult as I thought it might be (from the few devs I've talked - no not LM).

 

Of course it looks like Microsoft will not offer DX12 API for Win7 so we'll most likely have to OS upgrade to Win10.

 

Anyway, it's a future that actually looks poised to provide some impressive benefits for P3D performance.

 

Cheers, Rob.

[ChaoticBeauty 6,692]

900 series is Maxwell, 700/600 series is Kepler, 500 series Fermi ... Maxwell will obviously get the most benefit, but our 700/600 series cards should see "some" benefit, how much is yet to be determined.

 

I'll probably dump my Titans when nVidia Volta cards are released ... but that's 2016 ... but timing might be right given Skylake CPUs in Q3 2015.

 

Two small corrections, 400 series is Fermi as well (500 series is the Fermi refresh), and Volta is coming out in 2018. 2016 is the year for Pascal.

[Guest]

Stand corrected, thanks.  Pascal will be 2016 but also 1TB/s bandwidth (same as Volta) which is still over 3X more than my now archaic Titans (probably be lucky to get $100 each for them).  PCIe 4.0 also which is still too slow for a Pascal or Volta ... interesting to see how NVLink works.

 

I'm gonna guess some initial growing pains so might be prudent to wait until mid 2016 for revisions and drivers sorted out.

 

Cheers, Rob.

[ChaoticBeauty 6,692]

Stand corrected, thanks.  Pascal will be 2016 but also 1TB/s bandwidth (same as Volta) which is still over 3X more than my now archaic Titans (probably be lucky to get $100 each for them).  PCIe 4.0 also which is still too slow for a Pascal or Volta ... interesting to see how NVLink works.

 

I'm gonna guess some initial growing pains so might be prudent to wait until mid 2016 for revisions and drivers sorted out.

 

You are right, Pascal is bascially a rebranded Volta, which had its name moved to the 2018 architecture. One interesting thing to note is that Pascal will also introduce the features that were cut from Maxwell like the shared memory.

 

As for NVLink, let's see how hesistant Intel and AMD will be to add it in their motherboards.

[Guest]

It will be interesting ... but good times ahead ... and I'm sure we'll all encountered some turbulence along the way.

 

Cheers, Rob.