Fermi support?

Earlier marketing materials stated that Vulkan would be supported on Fermi GPUs. The beta driver page says that Kepler or Maxwell is required. Has Fermi Vulkan support been officially dropped or will it be coming later?

I am curious about that, too. Last time that Fermi support was promised was at SIGGRAPH last August so not that long ago afaik and the API was in a sort of soft freeze already at that point if I remember correctly.

Unfortuantely I could not test if there maybe is experimental Fermi support in the current driver, because my Linux distro is too bleeding edge. Downgrading the kernel to the latest supported in the 355 branch would be possible, but I’d also have to downgrade to xorg-server 1.17.

I’m just guessing, but Fermi drivers also still lack D3D12 support and therefore they want to kill 2 birds with one stone with a driver which supports both APIs because of their similarities. But as I said that’s just an assumption.

With DirectX12 it seems to be more complicated anyway. There are so many FeatureLevels and only Maxwell 2 seems to have full support. With Vulkan 1.0 the base requirement in terms of hardware features was OpenGL ES 3.1 which is a subset of OpenGL 4.3, which is supported on all Fermi and newer cards.

http://www.geforce.com/hardware/notebook-gpus/geforce-gt-620m has Dx12 support but no Vulkan support.

That would be really great.
I have an old 560 Ti and still waiting for the new Pascals.
I could continue building my own graphics engine which is somewhat limited with OpenGL yet.

Not too long ago DirectX12 on Fermi was obviously still planned for the end of 2015.

So maybe there is still hope. Now would be a stupid time to be in the market for a new GPU. The only thing that bugs me about my 580 is that 1.5 GB VRAM is getting pretty close to full with my 1440p monitor that I purchased in the last 5 years. Still has enough horse power for anything I play on Linux. Worst case NVIDIA rolls out the lower end Pascal cards first and we have to wait months and month to replace our Fermi cards with a 1080 Ti.

Just not really an option for me. Nvidia stated Dx 12 and Vulkan support. We need to have it, especially blakout24 said, it is kinda easy to implement Vulkan support to Fermi. We already had an very alpha DX 12 driver long ago. I managed to ran 3Dmark API overhead test in Dx 12 mode, so just don’t really see the matter behind all of those delay, what we experience over 6 month now, according to original DX 12 support release date.

+1 for this question. Need to know if driver for my GTX 570 will be provided or if I should wait for an open source driver.

No guarantees though that Fermi will have full Vulkan support in final release drivers.

The oldest cards that AMD supports is the HD7000 series from December 2011. My 580 was launched December 2010, if there is no technical reason I would support them just to be better than AMD. ;)
If Vulkan drivers really are so small (PowerVR guys wrote one in a month with 2 guys) and there is no heuristics tuning anymore, because that’s the job of the application the maintenance burden of supporting older cards that have all the needed hardware features should be minimal.

I thought every graphics card that supports OpenGL 4.x will support Vulkan.
I have a NVidia NVS 5200M in my notebook, with GF117 (also Fermi).

According to the developer these entries have been faked. He checked back with NVIDIA and there definitely isn’t Fermi support in the driver. However, he says “Es gibt momentan definitiv (noch) keinen Support von Vulkan auf Fermi.” which means there is no support (yet). This would imply that he knows more than we do and that support will come.


Would be nice if NVIDIA could give use a simple yes or no.


"Had to register to post here, since I was the one that uploaded that entry.

It’s not a compromised database, that entry is uploaded directly - unaltered - from the utility that they had for uploading them.

Notice that my GTX 570 reports not a single image format supported, all the Vulkan applications also fail to initialize the display. Most likely Nvidia didn’t make a whitelist for enabling the Vulkan system.


It’s not fake but it is non functional anyway on Fermi.

Per the Q&A for the Khronos Vulkan webinar ( https://youtu.be/nGkpPp2tGSs?t=2784 ) nVidia is currently not planning on supporting fermi due to an install base issue, not a technical issue.

As a long time nVidia supporter and GTX 580 linux user this greatly saddens me as there is not a new video card coming for me for the next several years.

I don’t think the install base is that small. In the Steam HW Survey cards like the 550 Ti are still near the top of the list. Also other GT 600 cards that are actually Fermi based.


Doesn’t make much sense to support many low spec GT cards that really won’t turn into gaming cards even with Vulkan, while cards that still can get stuff done like a 590, 580, 570 etc aren’t supported.

Also there definitely is preliminary work for the Fermi architecture in the driver. Otherwise I wouldn’t be able to enter the driver with vkGetPhysicalDeviceProperties and return a deviceid, GPU model string and FramebufferSizeLimits etc.

So why not finish what was started?

It sounds like they want to push us into buying new cards. I’m eventually going to get a Pascal card anyway since, like you, I’m hitting on VRAM issues.

My GTX 580 is still faster than popular cards like the 750 Ti or the Tegra Chips.
If the Fermi install base is an issue why is it part of the mainline driver, which coincidentally supports all cards that could support Vulkan, while the 340 legacy branch supports the non-DX11/non-OpenGL 4.3 cards that only get bugfixes. With that logic it shouldn’t get any new features, but it will likely get the refactored driver that supports the KMS API and their “I can’t believe it’s not GBM” EGLStreams thingy. But of course no Vulkan, because … reasons! If my card ever dies from its high overclock I’m going to take a good look at what AMD has to offer. They are landing tons and tons of code in their AMDGPU mainline kernel driver. With every major middleware engine implementing Vulkan performance on Linux shouldn’t be a problem anymore and for legacy OpenGL Games Mesa should be fast enough.

I fixed the SIGGRAPH slides:

In case someone still believes the lame “install base” excuse.

And even Linux is still a tiny market. I’d go as far and say that there are probably more Fermi users on Windows still than the 1% Linux Market Share on Steam.

Well if what is said in that Q&A is to keep, for me that’s must be a bad joke.

I don’t want to be offensive but I buy an used “MSI GeForce GTX 580 Lightning Xtreme Edition 3 GB”, it’s old but a beast, it overpower lots of new GTX series.

I just buy it because the deal was great and mostly because it was promised in SIGGRAPH last August that Fermi was getting Vulkan support otherwise I will never get that card because of the release date.

I understand if they simply ditched Fermi but announce the support and after that get word back saying that the customers that have Fermi now doesn’t count, well, is just an awful attitude.

I know that Nvidia Linux support was always great but now, for me, is just thanks for nothing. In the future, when the time to get a new card came, like it was said before, “I’m going to take a good look at what AMD has to offer”. I never considered leaving Nvidia but my opinion just changed, I’m really, really disappointed.

Yeah as if the Fermi install base suddenly changed since last August. :D
I can understand that companies want to make money and would rather sell new products than to support old ones, but you definitely don’t get people to buy your new products by announcing something, not keeping your word and than back paddling on lame excuses and expecting people to not notice what is blatantly obvious.