Well, if this is the case, lets hope amdgpu has become more finalized by the time polaris comes out. If buying nvidia no longer gives me better support, then I might as well look back at AMD. I’m not spending up to $2000 on a new video card if it’s life is < 5 years.
What happened to your promised support for DirectX 12 and Vulkan on Fermi (FL11_0) 400/500-series GPUs?
You can’t just make stuff up like that and then leave us in the cold.
I thought you were the good green martians…
Get back to work!
The last 355/356 series of Vulkan drivers is their response
This is the changelog
I’m reading between lines. It says: Fuck you stupid customer.
“It’s not an engineering issue, it’s an install base issue…”
Source: Khronos Group Vulkan Webinar at 46m25s
This has to be the lamest excuse not to include Vulkan support for Fermi (FL11_0) 400/500-series GPUs if it’s indeed supported by hardware.
I most certainty won’t upgrade anytime soon (just to show you how much we care for backward compatibility…)
and/or not allow for this to become the new standard industry businesses practice.
I’m not buying a new video card for just a driver, too bad NVIDIA, let’s hope you change your mind later and add the support or let’s wait for someone to make a patch for the drivers :P
Definitely won’t be buying nVidia next time due to their tactics. Plus, AMD cards are better suited to Vulkan/DX12 it appears. Just look at the initial DX12 benchmarks (Doom etc…)
Cool, now I have a reason to get me an AMD APU. Thanks Nvidia!
Really unbelievable, they used the word Fermi in the changelog but just to say something who let us down. Thanks
To me seems that they’re like: “Come on team, Fermi graphics cards still listed in the driver, hurry up remove that, let’s not giving them any hope”.
Long time ago I always put this company in a pedestal, I had many cards like FX5200, FX5700, FX7300, 8600GTS, 8800GTX, etc etc etc and now an GTX580 Extreme 3Gb. I never had an AMD GPU, well seems that next time is that time.
You lost a customer, but I know that I don’t matter because I belong to a “small base”…
HD4000(DX10), only 4 years of driver support.
HD5000/6000(DX11), only 6 years of driver support.
Nvidia Tesla(DX10 legacy GPUs), 10 years of driver support from 2006/2007, only ending in April 1 2016.
If you think you can get better driver support from AMD, a company well known to ditch old hardware support quickly & open source driver doesn’t even currently support older non Tonga/Fiji hardware, feel free to sell off your Nvidia cards now.
Actually AMD does support Sea Island GPUs with AMDGPU there is a Kconfig switch for that. Having an open source mainlined kernel module is also much better for long term support. For NVIDIA there isn’t even really an option to have a community made Vulkan driver for nouveau, because reclocking on NVC0 (Fermi) doesn’t work and a driver for a card that always runs at 50 Mhz is pretty pointless. Also makes it much easier to support modern display server/protocols like Wayland. NVIDIA announced some plans in 2014 but nothing ever came out of it.
NVIDIA is also getting slapped in DX12, that doesn’t look too good for Vulkan:
I will keep my 580 until I hit any walls probably 2 years. When I was young I had the time, but not the money to keep buying new GPUs. Now I have the money to blow on high-end GPUs, but not the time so I just keep using my Fermi until it makes no sense anymore.
Really sad to see that NVIDIA is too scared to properly communicate with their customers and rather brings out a new driver changelogs and hope that Fermi users get the hint.
This is why I paid a price premium for Nvidia cards up to this point. Because of the track record of support. But the past does not change what is happening currently. I understood that when I bought my Nvidia cards, I was paying more for a card with les sperformance, but support is much more important. In the end, it is always a gamble to learn from. I also dislike the mentality that if they can have reduced support, so can we. Both companies right now as of this writing only support vulkan back to release dates of early 2012. This is where Amd went with GCN architecture and Nvidia went with keplar.
Keep in mind, Nvidia is much more successful at this point and originally announced it would support the 400/500 series. People are brining up numbers of 10% user base still use fermi.
Now to talk about the amd side. There is no secret that amd isn’t fairing the best right now, but they have been silly with the drivers up to amdgpu. Up to now they have been programming windows, linux, linux opensource, etc… all separate drivers with a mixmash. Nvidia had done the drivers a better way upto this point. But amdgpu is supposed to fix this. (Yes I am primarily talking about linux at this point, but thats where I do my developing). If amdgpu can do what its goal is, then I see long time support being corrected. People are using it right now, and it will have time to mature by polaris’s release date.
But as of right now, Vulkan support on both companies goes back to the same date. And depending on amdgpu’s outcome, it can make support for future technologies a possibility.
Right now, I am looking at buying a used 970 (can’t justify a new one for 6 months use) until the new video cards come out. And yes its an nvidia because my linux distribution isn’t the nicest to get along with the aging amd drivers (which will change with amdgpu). But, when I weigh in the options, Nvidia will no longer be on the pedestal I put it on in the past for support. I will also take into account the amdgpu initiative and see how that has progressed.
I know very well the duration of the support from Nvidia and from AMD. I’m a Linux user since the time that was almost impossible to put an ATI card working in Linux.
The question in this topic never was that!
The question is that is the second time that Nvidia played us. I made a deal for a Nvidia card because they said, or better, they made a fuss about how awesome they are and even Fermi got they’re love for Vulkan support. Same for DX12 and that is even in the product sheet, in the Nvidia website. Man in that time I simply think to my self: “that company is really awesome”.
Gess what, the numbers got a giant change (it’s the onlyyyy explanation) and no more love for us.
Come on you don’t do that kind of stuff you don’t treat millions of costumers as a “small base”.
If in the future I going to buy an AMD is not because they conquered my support but rather because I trust Nvidia word, I spend money and they simply drop me like a peace of bad meat. So maybe I am going to try AMD and they’re bad drivers, they gave life to Vulkan so now I think that they deserve a shot.
I would expect an explanation why NVidia announced Vulkan for Fermi, and now dropped it. The userbase shouldn’t have changed that much since Aug 2015. Seems to me like a short term decision from the management.
I would too but I had to sit because I think that the waiting is gona be huge.
Meanwhile look more fire:
Meanwhile at NVIDIA HQ:
“If we wait long enough Fermi users will disappear on their own.”
This guy hit the nail on the head:
What people are forgetting in that forum, is that vulkan is combining the api’s opengl and opengl es into one lower level api. This means, program once and it can work under different operating systems and architects all together(of course there will probably be a couple gotchas, but still no where near reprogramming the whole 3d section). This also means, the possibility exists to run as many fermi and non-fermi cards you can fit onto a motherboard and have them work together.
While I understand fermi is coming to a close in performance with the future hardware release on triple A titles, a lot of games aren’t triple A. And you can’t tell me fermi can’t hold it’s own against a mobile device. With vulkan’s release, mobile games will be easier to program to also run on the pc.
I’m looking at the lack of vulkan support on fermi effectively killing it as it won’t even be able to run the most basic of programs utilizing it. And the icing is, it is a meeting room decision for more profit by forcing consumers to purchase new hardware that is killing it; even when their old hardware is fully capable of handling it.
Here is 1 scenario of many I would like to see fermi’s still being used. Lets say I make a racing game and I have a 970 and a 560 ti. I could have the 970 do the normal rendering as expected, but I could leave the mirrors up to the 560 ti. The 560 ti could run lower resolution on the mirrors, lower quality as objects are in the distant, even a lower update rate and that would still be awesome. The newer hardware doesn’t trump the older hardware enough to throw it away in a time an api is released where they can work together.
A scenario 2 of many. You could go stupid crazy with a hud running on a 560 ti overlayed onto the 970 rendering the game.
These suggestions are offloading on a basic level to help demonstrate the capabilities, vulkan will be able to go to a level much lower then that.
There is no support for Multi GPU as in GPU0 and GPU1 share resources and work together in Vulkan 1.0.
I was under the impression with the lower level access, I could have the secondary card render it’s image and then have it transferred to the first gpu and have it loaded as a texture at the very least. In this mode, it wouldn’t be as fast, but a frame or two behind in secondary objects like car side mirrors would be fine.
But I was also thinking I could go lower level then that. I wouldn’t want to have gpu1 regularly using gpu0 ram anyways do to the latency and relative slow speed of the pci-e bus (especially since it would be expected some cards to be running in 4x mode).
Alas, I don’t have vulkan support to try my theories out.
Oh Vulkan, Vulkan, wherefore art thou vulkan support.
Well , so far so no respond for the question. To be honest , I didn’t expect much from NVidia representative. Looks like developers are left with the dark with Fermi cards.
I don’t understand why Nvidia reps are “operating” if they can’t give simple answer for this question or questions.