DX12 on fermi

People could turn of effects and set lower resolution. At least, players could enjoy the story of Quantum Break. But no, nVIDIA want people to buy tenth series of their cards instead! I am going to boycott nVIDIA.

 

You could replace it with a 6970 (faster and newer than the 570) from AMD if you want to play DX12 games.....

 

If that's supposed to show how "bad" AMD is, at least AMD's statement from the get-go was that only GCN would get DX12.
And back then, Nvidia got good PR when sites reported it would support DX12 all the way back to its first DX11 GPUs whereas AMD and Intel wouldn't.

Nvidia's website is even still stating that Fermi will get DX12

One is free to claim that Fermi would be useless for DirectX 12, but Nvidia made a statement about Fermi supporting DX12 and absolutely no one that isn't involved with Nvidia should be defending Nvidia for going back on it

 

True, but has there been any official confirmation that they have went back on it yet.

It's a poor show that they haven't addressed the issue either way, but it just seems to be about the principle of instead an actual problem.

 

People could but it wouldn't help them in case of QB in the slightest.

 

On dolphin build 5.0-424 standing in fixed location in metroid prime with the framerate uncapped to show cpu bottleneck more clearly

note:dolphin only uses 2 cores. cpu usage hovered around 25% on a mobile i7 sandy@2.6ghz
gf116 with 192bit bus
358.70
dx11 avg 106.4fps min 101.2
dx 12 avg 109.4 min 107
About 2.8% faster

I think that there is any speed up is interesting given how very immature the dx12 driver for fermi and the dolphin dx 12 renderer are
I imagine the gains will be greater on more capable hardware/better multithreaded applications

will update later with newer driver fps
update:
372.54
dx11 107.9 minimum 104.7
dx12 n/a
appears nvidia continues to reduce overhead even on older gpus
though it is still slower than the halfbaked dx 12 driver

 

What about the OpenGL renderer? For both pcsx2 and Dolphin the OpenGL renderer is regarded as the most accurate and sometimes also performs the best so while the choice of DX12 in Dolphin is nice it's not always usable and sometimes performs worse than OpenGL.

This may very well be true but often you're still better off using the OpenGL renderer for performance and accuracy.

 

the opengl renderer is superior on dolphin it has the most work done on it these days due to its crossplatform appeal.

mostly just used dolphin as a test bed since its pretty much the only known working real world dx 12 application on the dx 12 fermi driver.

372.54
OGL avg 109.4 min 105.6

As you can see it is quite good
interesting note is that cpu usage was higher on ogl sitting around 28%
points to better utlization rather than better overhead, which makes sense since dx11 is more single thread dependant unless you implement command lists which few applications do.

 

Just nonsense what you saying.

 

Just gonna say multi quote does exist,

dolphin isn't particularly demanding on the gpu, and as such even my very slow fermi gpu is cpu bottlenecked at 1080p with ansio and other enhancements on.

Fermi is still supported because they sold it for so long, many products from mainly the 600(m)/700(m)/800m series are fermi based

The average consumer will not know that a 670m is actually a gf114 part not a kepler part ect.

It should also be noted that kepler and fermi are very similar, I think it would be fair to describe kepler as a reconfigured fermi core, It shouldn't take much support fermi on vulkan or dx 12 since the feature set is the same.

http://www.anandtech.com/show/5699/nvidia-geforce-gtx-680-review/2

At this point i think of only 5 reasons as to why nvidia hasn't released a working vulkan/dx 12 driver is:

1, some Hardware erratum that was unforseen
2, lack of interest by developers to support a fermi optimization path.
3, planned obsolescence
4, Diverted resources away from fermi related development to cut costs.
5, Signifigant boost in maximum throughput(unlikely but possible given fermi's relatively high compute performance.)

The whole buisness of nvidia not releasing a public statement on the matter is very poor PR on their part.

You say people don't care but obviously this thread would not have 25,000 views if that was the case.

 

"People need to get used to it or get out at this point." If everyone was like you a gtx 1080 would be 1500$ and get driver support for a year.


"Me, of course I don't care anymore. 5 years is a long time in the tech world. I stop caring about stuff that is five years old."

Good for you , why are you here.

This afterall is the dx12 on fermi thread not the "I don't like what people are saying about nvidia " thread eeright:.

 

Yup, you are completely right. Many users here are kinda living in the dream where we all the new hardware's and company's don't have to deliver any promises what have they made before.

One of the worst thing is, we don't made that promise or promises . Doesn't matter how old is the product after all . Some people just can't really see the difference at all and understand it.
Specially with the claim , the Fermi architecture won't benefit anything from DX 12. Like we have any evidence to prove that theory

 

Fermi is so old you're unlikely to hit CPU bottlenecks, so the main advantage of DX12 just went flying out the window.

I also believe there's actual hardware level incompatibility with WDDM 2.0 which is the real issue

 

WDDM 2.0 is not an issue at all for Fermi architecture. It is simply deprecated.

http://docs.nvidia.com/nsight-visual-studio-edition/5.2/Content/Release_Notes_Main.htm

"Fermi GPUs are no longer supported." on both CUDA e and Graphics tools.

 

Since the Anniversary update, Fermi supports WDDM 2.1


I just wonder, was there really any point in bringing WDDM 2.0 and now even WDDM 2.1 to Fermi when it's not going to get DX12 support?
I always thought creating WDDM 2.0 drivers were the big undertaking and DX12 just the last piece of that puzzle

 

WDDM 2.x is not only about Direct3D 12, it is also related to the desktop compositor.. Fermi is dead, end of story. Moreover, never trust dxdiag.

 

I have learned not to trust anything that comes out of Manuel Guzman's mouth.

 

The point is to allow Fermi cards to work with non-Fermi WDDM 2.x cards in one system. 2.1 is a minor update of 2.0 which is probably the same for all h/w in the driver.

 

Yup, biggest liar ever.

 

What can i say? He's right


Where are you now nvidia?

https://forums.geforce.com/default/...hread-released-8-30-16-/post/4964858/#4964858