cross-posted from: https://lemmy.world/post/13805928
It’s a long vid. I suggest prepping your fav drink before viewing.
It’s re Nvidia’s new gpu architecture for ai, NVlink switch, RAS diagnostics and other Nvidia announcements.
Nvidia knows it’s the star of the backbone of the current ai boom and seems to be going full steam. I’m hoping for more innovations on tools for ai and gaming in the future.
Nvidia has the short term covered, but I’m skeptical they are going to end up leading the AI chip market in about 5 years or so without acquisitions.
Recent research has shown not only efficiency gains but also actual performance gains with binary or ternary weights instead of floats.
This means you don’t need FP calculations or matrix multiplication.
It requires being trained from scratch with that architecture in mind, so it will probably be 12 to 18 months before we see leading models with light weights, but once we do the market may go more towards faster and more energy efficient options that don’t need to rely on Nvidia’s legacy of IP for FP ops.
So while an unmatched king in how things are currently done, the magic phrase that brings any monarch to tears is “this too shall pass.”
God, I wish AMD and Intel could get their shit together.
Happily playing modern games and developing shaders on my AMD GPU. 5120x1440 120 Hz issue free
I wish people would get their shit together and realize they’ve fallen victim to marketing
It’s not marketing, AMD sucks for ML stuff. I don’t just play games. Everything is harder, with fewer features and less performant with AMD.
The situation is mostly reversed on Linux. Nvidia has fewer features, more bugs and stuff that plain won’t work at all. Even onboard intel graphics is going to be less buggy than a pretty expensive Nvidia card.
I mention that because language model work is pretty niche and so is Linux ( maybe similar sized niches? ).
Yeah. Linux boys on AMD are eatin’ good.
Really? I’ve only dabbled with locally run AI for a bit, but performance in something like ollama or stable diffusion has been really great on my 6900xt.
AMD successful at the mid tier?! I’m shocked!
NVIDIA prints money in the enterprise where business will literally lose money over the extra compute, and lesser so in high end gaming with details turned up. AMD simply can’t complete, it’s not marketing, it’s a better product.
AMD has never gotten more than 50% of the market, even in the years where their entire product lineup offered better performance/features for less money. I’m talking about the “good old days” here, where software features weren’t a big factor for consumers and ML was nonexistent. You have to be delusional to think that Nvidia doesn’t hold a very clear mindshare and marketing advantage.
Oh you mean because of the shit buggy drivers they had? NVIDIA has nearly always had the most compelling product at most upper price points. When ATIs product line was briefly straight up faster for a brief period (years lololol👌) and folks choose Nvidia because ATI drivers were a god awful buggy mess.
But yeah everyone is just a little brainwashed leming 🙄.
They still have, I replaced my 3070 with a 7900 xtx and the 7900 is constantly freezing with ring GPU errors and drivers completely effing up the system. I have already replaced it twice, and I am using workarounds to not hit bugs, but they happen every few days…
Which brand gpu?
Powercolor, red devil.
Under 6.7 I was able to find some a combination that was usable for a few days.
With 6.8, timeouts would happen within 30 minutes.
I fiddles with sched_job module option and the system seems stable now.
Seriously. AI aside, if you’re doing anything 3D-related you’re basically shooting yourself in the foot by not going Team Green. The difference in render time/quality is exponentially better. I’d kill to see AMD or Intel pull a Ryzen in the GPU market.
