I’ve been running PopOS with an RTX 3080 for years now and I’m absolutely happy and zero chance of switching back to this Microsoft trash.
I have a problem with how AMD handles their software called “Rocm”. It’s basically AMD’s version of CUDA and it’s a complete mess. It’s ambiguous which cards are supported and which aren’t. They have gotten better with this problem over the past few years. But for example, their latest private customer graphics card is only supported on Ubuntu. Other products are only supported on other distros. Some cards, who aren’t even listed as supported, are very well supported in all distros. That’s what I mean, it’s a mess. Essentially, the only way to find out is take the bullet and plug it in, see what happens. I mean, Nvidia is a trash company that makes Apple and Microsoft look like saints. But at least, if you buy one of their products, you know it runs CUDA. No support matrix needed. Everything, no matter how old it is, supports CUDA. The company and their graphics cards are still trash though. Unless you buy the most expensive model, of course. Then all of your problems miraculously go away. cries in 10GB VRAM
Where did you read this? As far as I know and has open sourced all their graphics drivers and they’ve all been wrapped into Mesa, which is available on all distros
No, this isn’t “your bad”. It’s AMD’s. If they actually were interested in providing competition, they would have a money printer on their hands. But well, carrying on as usual without any sort of investment or risk of failure is also a modus operandi that many companies follow quite successfully. If Intel doesn’t kick their arse, they won’t lift a finger. Simple efficiency, right?
Obligatory switch to Linux notice. The AMD drivers for linux put the nvidia drivers for windows to shame.
I’ve been running PopOS with an RTX 3080 for years now and I’m absolutely happy and zero chance of switching back to this Microsoft trash.
I have a problem with how AMD handles their software called “Rocm”. It’s basically AMD’s version of CUDA and it’s a complete mess. It’s ambiguous which cards are supported and which aren’t. They have gotten better with this problem over the past few years. But for example, their latest private customer graphics card is only supported on Ubuntu. Other products are only supported on other distros. Some cards, who aren’t even listed as supported, are very well supported in all distros. That’s what I mean, it’s a mess. Essentially, the only way to find out is take the bullet and plug it in, see what happens. I mean, Nvidia is a trash company that makes Apple and Microsoft look like saints. But at least, if you buy one of their products, you know it runs CUDA. No support matrix needed. Everything, no matter how old it is, supports CUDA. The company and their graphics cards are still trash though. Unless you buy the most expensive model, of course. Then all of your problems miraculously go away. cries in 10GB VRAM
Touché. While Mesa is rock solid, using AMD cards for compute is indeed a fool‘s errand.
Where did you read this? As far as I know and has open sourced all their graphics drivers and they’ve all been wrapped into Mesa, which is available on all distros
https://rocm.docs.amd.com/projects/install-on-linux/en/latest/reference/system-requirements.html
Click on the footnote next to the 9070 XT.
Edit: Also, yes, they have published FOSS drivers and integrated them into Mesa, I know that, but rocm is a separate beast entirely than drivers.
Thank you for clarifying, my bad😅
No, this isn’t “your bad”. It’s AMD’s. If they actually were interested in providing competition, they would have a money printer on their hands. But well, carrying on as usual without any sort of investment or risk of failure is also a modus operandi that many companies follow quite successfully. If Intel doesn’t kick their arse, they won’t lift a finger. Simple efficiency, right?