Windows 10 build 10586 with WDDM 2.0 support for 400/500-series?

WDDM 2.0, DirectX 12 and Vulkan support for GTX 400/500-series on Windows 10?

Do you think Nvidia have planned on adding WDDM 2.0 (DirectX 12) in time for Microsoft's "Threshold 2" fall update build 10586 on nov 10th for their Fermi architecture or have they intended to remain silent until Q1 2016?

 

What's TH2 has to do with Fermi's WDDM 2.0 drivers? The year hasn't ended yet, we still have two months left.

 

The promise was WDDM 2.0 and DX12 support would arrive "later", the most specific statement was it'll be there in time for the first DX12 games


But who knows, Nvidia is supporting three different architectures, so perhaps Fermi was put on hold due to changes in upcoming builds, which require changing some code.

 

Someone on this forum said that he ran dx12 overhead test on Fermi on one of the latest win10 insider builds with driver probably 358.70 (I don't remember exactly) and in dxdiag he had DirectX 12 and with the same driver on current win 10 build he had dx11.2.

But nobody confirmed that.

 

I got the latest Win 10 build 10586 and latest driver, 358.87. Just tried to run 3dmark overhead api test. Only dx11 runs.

 

nope, no changes to driver model in TH2.

 

Yeah, still WDDM 1.3, DX 12 shows up as it did as before.


Anyway, Nvidia way too silent about the DX 12 support.

 

You expect them to issue a statement on this each day or something?

 

depends nvidia create driver to support wddm 2.0

new gpu-z indicate dx12 support (but nvidia ta with laziness and creates no driver to support dx12 to its former GPUs 400/500).
my sugestion is nvidia open source code driver, to comunity create a driver for
fermi support WDDM 2.0

 

From 358.78 driver thread

 

I am sure the guy was simply bull****ing.

 

Why, whats the point or use of it at this point in time? Not single dx12 game is out of alpha/beta and are nowhere close to being finished this year, like what is it that all of you fermi owners are missing out on, dx 12 api test, infiltrator tech demo with dx12 path enabled that's about it, the first dx 12 games are gonna launch next year q1/2,

 

An even better question would be if any upcoming DX12 games will even target FL11_0 hardware at all considering that the basis of DX12 for the most developers is Xbox One which has FL12_0.

It may well turn out that most D3D12 renderers won't even run on FL11_0/11_1 hardware and thus having Fermi (and Kepler btw) drivers for DX12 will be pointless.

 

It will run on Fermi and Kepler as well. This is how DX 12 works. Anyway, game devs are focusing on FL11 at the beginning.

 

No, it's not how DX12 works. A game requesting a feature which is missing in h/w won't run on this h/w. This is how DX12 works - and it's the same as all DXes always worked. Chances are that most DX12 games will target FL12_0 since this is XBO's feature level and the level which actually makes sense to write a new renderer for. Devs will need to spend additional time to make these games run on FL11_x and it is very possible that some devs will decide that supporting a 4-6 years old architectures does not worth their time.

I'm also wondering a bit what kind of fps do people with Fermi hope to get in DX12 titles considering that even Kepler is struggling to keep at 30 in the latest wave of DX11 games.

 

It is true but DX 12 still runs on these hw's. So, there is a great point to make DX 12 work on Fermi as well. Can use all the advantages what new API can provide, even some of the features are missing. Infact in some point Fermi more "superior" compare to Kepler or Maxwell. In some other points of course it is outdated.

 

All Nvidia cards supporting DX12, except from Maxwell 2, are only feature level 11_0. No studio will be stupid enough to exclude those cards. Whether Fermi will have the required horsepower for the upcoming DX12 games remains to be seen, but they aren't formally excluded.

The Xbox One might be feature level 12_0 but no studio will be stupid enough to require that on PC until many years from now.
AMD's GCN 1.0 and Intel's Haswell and Broadwell are feature level 11_1.

 

Actually fermi has hardware scheduler and this will help better direct x12.

 

Even more, it is Saturday/Sunday(now). I am having drink. It is the last thing i want to go into details.

 

DX12 - the API - can run on these cards as it supports their feature levels with FL11_0. Note that DX12 theoretically can run on Radeons 5000 and 6000 as well but AMD does not intend to provide DX12 drivers for these GPUs. There's a world of difference between and API running on some h/w and applications using this API running on this h/w. The only reason I see to even give Fermi DX12 drivers is that Kepler has the same feature level and because of this there is at least some pressure to provide FL11_0 path in a DX12 renderer which will work on Fermi as well. If Kepler was FL12_0 - there would be no point in supporting DX12 on Fermi basically.

What are we talking about here? Most studios don't support DX12 at all right now. When they'll start - they'll start from XBO's DX12 which is FL12_0. And this is likely to happen next year with PC DX12 versions appearing closer to year's end as they'll require additional effort on part of these studios. We're likely to have a couple of Pascal chips launched at this point. How much Kepler cards will be in active gaming duty by then? What performance will they provide? And who will test their games on Fermi chips which will be over 6 years old by that time?

In all honesty I think that NV may have done a mistake with Fermi's DX12 driver as this is an ancient h/w by now and supporting it and making sure that there are no bugs / issues will be like supporting Windows XP with drivers in 2015

No studio will require DX12 FL12_0 only on PC in 2016. But it is very possible that studios will require FL12_0 support for DX12 renderers and will provide a legacy / fallback DX11 renderer for those without FL12_0+ h/w and/or running on Win7/8.

No it won't. I don't know where this idea have originated from but Fermi's h/w scheduler has nothing to do with how DX12 feeds the GPU, it's the scheduler / compiler rearranging the instruction stream inside the shaders submitted to GPU, not the front end scheduler of job queues (aka "engines"). Fermi doesn't have anything which will make it a better fit for DX12 than Kepler or Maxwell.