In my social bubble people actually do care.
Sounds like Required Reading at school. Some students actually read it. The other students just asked the former ones questions about it.
For the better, only for the better of course.
What AMD needs is to remind investors at every opportunity that there are other potential players than Nvidia. Somehow, Consumer Technology Association gave them another such opportunity at CES 2026. — Edit 2, to tie that back into Zen 6 speculation: In other words, my guess is that what the press release says is pretty much exactly what kind of information will be given by way of this keynote, hence, nothing substantial about Zen 6 (the architecture and the generation of products).
Not everyone's English comprehension is great especially those using it as a second language. Plus, an LLM can summarize the stuff in a neat way and print it in the form of a table, for example. Gotta use whatever advantage you have these days. It's a cruel, unsympathetic and cutthroat world out there.
Probably both.
This is a take that focuses on the immediate term and ignores the very real long term degradation of one's own skills by offloading these tasks to an AI.
You're also offloading stuff to a very unreliable system.
There are also ways to spread a Mixture-of-Experts model among GPU and CPU, where the GPU+VRAM only contains and processes the central logic and then the CPU+RAM handle the experts.
I suppose it depends on your philosophy and work ethic. Some of us are lazy and aren't going to be changing our ways so AI for us is more like, oh nice, more laziness!
Maybe you should use it and help improve it for everyone else rather than impede progress by telling others to not even experience it? It's not like if it says "chickens have green blood", people will start believing that. If someone is that clueless, they are going to be doing and believing stupid stuff without AI anyway.
That's the fascinating thing. New techniques are being discovered all the time. Wish I could get a peek into the future and see how much of an alien race we become
Or at least until someone makes a post related to Zen6 in this thread.
Blame adroc for that. He could tell one of his many friends to create an account here and leave a clue every other day to keep the relevant discussion alive
Will Zen 6 SMT be even more effective?
No.
They already do.
No idea.
Yes.
idk.
moar.
idk.
Well the IHS is the same thiccass slab as anything else AM5, so...
pipe down.
Discrete GEMM blobs are on their way out, courtesy of our Lord and Savior gfx13.
Well for gfx13 parts yes.
on DT? maybe.
If you mean Zen5, then Software optimization guide and C&C disagree with this statement. Unless for some reason, 4 inst/C in ST mode is achieved using both decoders and throughput limitation comes from elsewhere.
So does he already have a sample?
You mean they should slow down some of them? I mean AMD to match Intel?
Turns out it's not true.
yea. I hate icaches.
No, but you'll get to enjoy his ramblings.So does he already have a sample?
He means 2clk FADD regression.
Do you know if there is a write-up somewhere. I thought the dual fetch on ICache was to help with that, curious if I read it wrong or they hit a funny problem somewhere.
Looking forward to.
I guess min 2clk latency for otherwise 1 clk instruction from the "scheduler full problem". 2CLK FADD always would be a nice boost, now it's 2CLK only on happy path, 3CLK otherwise but this is not affected by the "scheduler full problem".
Facebook
Twitter