The nvidia mesa posts are entertaining. Bet he saw his shadow, and decided to hide for another 6 weeks.
Announcement
Collapse
No announcement yet.
Another NVIDIA Engineer Just Made His First Contribution To Mesa
Collapse
X
-
Originally posted by JPFSanders View PostI appreciate any open contribution they make though.
And in any case, this is a 2 line fix that, based on the reviewer, likely was a customer of RH opening a case which got bounced over to nvidia who understood the internal opcodes that needed to be changed.
- Likes 5
Comment
-
Originally posted by Cotyso View PostWhile I hope that nvidia will open-source their driver, I don't believe they'll start to do it without an announcement. Especially given that they announced xwayland support on the proprietary driver yesterday.(via Eglstreams)
Luckily, we will soon likely be at a point where PCs get squashed into a single chip, putting-add on cards at a massive price disadvantage.
Comment
-
Actually, I'm serious with my assumption about nVidia following AMD's route in open-sourcing their kernel driver part & trying to mainline it upstream (until then shipping it in DKMS-form, like they do now), mainly for the following reason:
NVIDIA's Graphics Driver Will Run Into Problems With Linux 5.3 On IBM POWER
Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite
So, this is probably going to be their big open-source anouncement after all...
Again, the main driver forcing nVidia to take this route is not because of our miserable desktop marketshare, but the big money of enterprise & AMD's recent supercomputer wins with their inferior but open-source compute stack.
Also, this would solve their GPL woes once and for all!
And the (limited) MESA support?
Well, Linux & Linus would simply reject the kernel driver part without an (at least somewhat) usable open-source counterpart...
Anyone got any better theory?
- Likes 1
Comment
-
Michael
It hasn't been clear why Red Hat has been investing so much on getting this open-source NVIDIA compute support improved, especially as to date all GPUs since the GTX 900 Maxwell series suffer from incredibly poor performance stemming from the re-clocking situation around signed firmware images and the Nouveau driver not having the ability to provide proper PMU handling.
AMD just removed GCN2 and GCN3 support in ROCm.- proof GCN2 support is dropped from ROCm: https://github.com/RadeonOpenCompute...ment-755272927 (in fact it was broken since years)
- proof GCN3 support is dropped from ROCm: https://github.com/RadeonOpenCompute...ment-753772733
OpenCL is the only reason to buy an Nvidia (with a proprietary driver) today.
I can see why Nvidia can benefit from having Mesa supporting OpenCL on Nvidia hardware before AMD hardware.
- Likes 2
Comment
-
Originally posted by Linuxxx View PostActually, I'm serious with my assumption about nVidia following AMD's route in open-sourcing their kernel driver part & trying to mainline it upstream (until then shipping it in DKMS-form, like they do now), mainly for the following reason:
NVIDIA's Graphics Driver Will Run Into Problems With Linux 5.3 On IBM POWER
Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite
So, this is probably going to be their big open-source anouncement after all...
Again, the main driver forcing nVidia to take this route is not because of our miserable desktop marketshare, but the big money of enterprise & AMD's recent supercomputer wins with their inferior but open-source compute stack.
Also, this would solve their GPL woes once and for all!
And the (limited) MESA support?
Well, Linux & Linus would simply reject the kernel driver part without an (at least somewhat) usable open-source counterpart...
Anyone got any better theory?
also in compute soon we will have Vulkan compute everywhere and OpenCL and CUDA will be death.
some people even work on many OpenCL kernels to make them run with vulkan instead of openCL.
Because of this apple already mark the OpenCL support as legacy means they drop it.
now nvidia has no other choice if they do not want to lose customers to AMD to improve wayland support and also improve upstream kernel support. and the only way to perform this is what you said: "nVidia following AMD's route in open-sourcing their kernel driver part"
i wrote about nvidas opensource plans for many years but in the past the had no pressure to do so because they had the faster hardware and their driver worked. but this changed now.
as you correctly said AMD won some big customers by this and nvidia lost big: "Again, the main driver forcing nVidia to take this route is not because of our miserable desktop marketshare, but the big money of enterprise & AMD's recent supercomputer wins with their inferior but open-source compute stack."
intel has the same problem in the past intel had all the big data center /super computer sales and now everyone buy amd because a 5950X is much better than a intel 11900K ...
right now it is really the best to buy amd CPU and AMD GPU this way the closed source and walled garden competition lose big.Phantom circuit Sequence Reducer Dyslexia
- Likes 1
Comment
-
Originally posted by illwieckz View PostMichael
…especially since AMD support is very good on anything but OpenCL but people are waiting for OpenCL in Mesa, not waiting for OpenCL in ROCm.
AMD just removed GCN2 and GCN3 support in ROCm.- proof GCN2 support is dropped from ROCm: https://github.com/RadeonOpenCompute...ment-755272927 (in fact it was broken since years)
- proof GCN3 support is dropped from ROCm: https://github.com/RadeonOpenCompute...ment-753772733
OpenCL is the only reason to buy an Nvidia (with a proprietary driver) today.
I can see why Nvidia can benefit from having Mesa supporting OpenCL on Nvidia hardware before AMD hardware.Phantom circuit Sequence Reducer Dyslexia
Comment
-
Originally posted by Qaridarium View Post
nothing of this OpenCL bullshit will happen... because soon we will have Vulkan Compute everywhere.
But you need first to make all libraries/tools/support for Vulkan. That is biggest reason why CUDA also wins vs OpenCL, but Vulkan has almost nothing there.
Vulkan is actually harder mostly.
Adaptation of industry is slow (people will abuse whatever is existing to speed up development process).
- Likes 3
Comment
Comment