this post was submitted on 22 Dec 2023
70 points (85.0% liked)
Technology
60071 readers
3589 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
A new start in dGPU is no easy task, but I honestly thought Arc's relative RTRT and compute perf were quite good?
My main complaint would be their Linux support situation for Arc. I'm hoping it will improve over time.
I do not count Arc as decidedly a failure (yet), allegedly the cards are pretty good with the new drivers. It was previous attempts of making cards that supposedly could run pure raytracing, that were supposed to compete with Nvidia in datacenters. But to be fair, 5 years of fumbling is some years ago, I was talking more 5 to 10 years back, where it appeared Intel failed with just about everything.
But Itanium their new server 64 bit CPU is way longer ago, and so is the GPU I think it was Knights Ferry, complete failure with twice the energy consumption and half the performance of Nvidia. Only later production began to fail too, and Intel Core2 was shortly beat by Ryzen on all parameters, and of course Optane failed too.
Now there’s a name I haven’t heard in a long time.
And the Xeon Phi (Knight’s Ferry/Landing) was in the GPU space, but only in GPGPU. The idea was that the Xeon Phi, with an x86-compatible core, could, with less modification, run software that was originally targeted to a standard x86 CPU. Something like 68-70 x86-64 cores.
I had a couple of them when I was taking parallel programming back in the day. Nifty little devices, but largely outshined by distributed multiprocessing for x86-64 and paled in comparison to the power of CUDA. That might be my own bias talking though.
I don't have personal experience, but AFAIK that's what everybody says. They were marketed as compute units, but their compute performance was very poor compared to the competition.
Oh right, my mistake. I do vaguely recall their prior endeavour