If your reading is correct then we're talking passed each other because I'm talking market positioning (which I thought was clear from my post but maybe not).
What Adroc is saying is that AMD already ha caught up with Nvidia, and that they do not need to catch them.
From technology point of view - he might be correct.
I mean what even is market positioning in this context?If your reading is correct then we're talking passed each other because I'm talking market positioning (which I thought was clear from my post but maybe not).
That was an adoption of different workloads.
ML is a single one, you run chunggus model on a large matrix math farm.
Throw compute and membw at the problem is the core precept of anything and everything ML.
We've managed to get usable results out of xformers by making bloatier models.Time will tell but I don't expect (nearly) all of the AI spending to be solely on giant farms forever.
I mean what even is market positioning in this context?
Intel still ships more server CPU units, just that none of that matters.
The point of AMD having a very even roadmap is for that to never happen again.Theoretically, if AMD had a major slip up and Intel came with another Conroe type moment, starting from already being the biggest player in the market does matter.
The point of AMD having a very even roadmap is for that to never happen again.
Nvidia?Of course, no one creates road maps to failure but no company executes on their roadmaps to perfection either.
yeah which is why you work to minimize them into irrelevance.Every tech company has had major slip ups and it will happen again.
Nvidia?
yeah which is why you work to minimize them into irrelevance.
And it didn't matter in the end.Nvidia has stumbled before too.
And it didn't matter in the end.
It's all about the roadmap and how well it's paced.
Intel issue is that they stumble constantly.
No, AMD caught up in technology. It's a two way battle now every gen and every launch for who offers the baddest boys, best perf/w and best perf/$. And this is Nvidia's battle to lose since they won't be able to retain their current (that is pre-MI300) monopoly on this market and that will be felt in revenue and margins. The real wildcard in all this is how the AI hype train will continue.So you're saying that no one can design a better AI chip/module/system than NV from now until the end of humanity?
For those using
(I was looking for some Thread Reader tool but it seems they have been musked out.)
Edit: Ian's choice pick:
I know. It's always for those that can't wait.A full transcript should be made available later this evening.
Overpriced stock maybe - people always overstate the growth potential. The projection pretty much shows MI300X isn't expected to be a runaway success.Ehh. I feel like AMD is overhyped in every sphere. Move on.
Ugh, no, GPGPU revenue is literally the only Q1 upside.I ve seen much better projections for the MI300(x), the expected numbers are surprisingly low for some reason, price of such a device is 20 000$, so they dont expect to sell more than 20 000 units in Q1 2024 and given that it cost 0.5x the competition offering this is a very weak outlook.
Ugh, no, GPGPU revenue is literally the only Q1 upside.
Everything else is bad, especially gaming.
No? no.that their offering cost only 0.5x Nvidia
They're already above that number.400M/quarter is negligible quantity
Did you listen to the ER call orone would had expected 1bn/quarter at least for a start.
I ve read somewhere 20 000$ for the MI300, to compare to 40 000$ for a H100, it that s not 0.5x i dont know from where you got your own numbers.No? no.
They're already above that number.
Did you listen to the ER call or
H100 isn't 40k at major customers.compare to 40 000$ for a H100,
Everything else is bad, especially gaming.