Jim Keller joins Intel

Part of me hopes that Raja and Keller are trojan horses, sent to destroy intel from inside.

(intel’s shady business practices since the 80s warrant it)

:smiley:

4 Likes

I guess I wasn’t far off.

I am pretty sure he when from AMD Zen to Tesla working on AI chips that AMD where to build for Tesla.

Tesla dropped Nvidia after they hired Keller. Keller was to design the new AI driving chips and AMD was to build them for Tesla.
I guess Nvidia pissed off another company again. LOL

Keller is gonna work on Car Lake.

2 Likes

SIGH

cough

1 Like

You need water or something? This coughing is contagious

Maybe, but I think we both need antidepressants:

1 Like

Why? What’s wrong with you?
I may need some for completely different reasons though. But when it comes to hardware I am only sad, that Radeon doesn’t seems to go anywhere… And more and more people are leaving.
As for Intel - they are getting quite the team… I should start chearing for them…

Well I was being a bit facetious.

Work is driving me to depression though… but that’s a different story.

If we are going into this whole depression stuff we may need an entirely new thread…
Do you think Raja and Chris Hook will do better job in Intel because of the better financial situation? I mean RND budget and everything?

Probably. The thing is RTG could have developed a better purpose specific device however they were tasked with creating a one for all use cases product.

They might aswell create a newer architecture that can hardware-emulate x86 for all purposes (the budget is without a doubt there. Even though i would rather have amd do something like that)

Thats highlly likelly yes.
Higher budget means more room for development en testing.

And its not that strange that Intel hires those guys really.
Because those people work in the field for a pretty long time allready.
So they can use their experience, which intel doesnt really have much to invest in.
If you hire totally new people, they also have to invest in educating.
So its just a pretty smart move from intel, hiring some of the most experienced people in the field.
But the AMD diehard fans are probablly going to seek something bad behind this.
But in reallity this is just a normal day in buisness life really, people come people go.

Given AMD are in 2 (or is it 3?) of the big consoles, and console gaming dwarfs PC gaming, i suspect AMD are betting on their hardware features becoming used eventually and getting the performance back that way. I have no doubt that AMD will be in the PS5 and next XBOX as well. Plus they’re in Apple’s desktops for the foreseeable future (until cuda on apple dies at the very earliest, and also until apple forgive NV for fucking them over on RMAs).

Vega for example has some interesting hardware features that aren’t currently used much. When they are (e.g., rapid packed math in Wolfenstein 2) the performance is great. But the code needs to be written for it.

The way that will occur is via widespread adoption of the Vega (and AMD in general) GPU platform - in their APUs, in the major consoles, in intel’s G series products as well as on PC discrete GPUs. Even if discrete Vega GPUs don’t sell that well, the architecture is going to be in a very large number of devices.

Give a console developer a feature like rapid packed math that is close to “free performance” if they do a little optimization of their data and they’ll be on board. The console developers are used to jumping through far bigger hoops to extract more from the fixed hardware spec on offer.

And that code will end up being ported to PC.

edit:
AMD’s hardware design is solid. As above though it is not optimized for current software, the drivers aren’t great on release and Vega in particular got kneecapped by HBM2 pricing and availability.

HBM2 was projected cost at half of what it ended up costing when vega was designed. Which pushed the price of the card(s) up significantly.

I’ll be very interested to see what a 7nm Vega refresh with 2x speed HBM2 on it (which is what it is rumored to have - even faster HBM than that exists already) can do.

TLDR:
Vega is fine for the long-game AMD is playing. Software will catch up and architecure tweaks will improve it.

However, If you’re a gamer who wants the highest FPS possible for the lowest price on today’s existing software, it’s not the best buy. But that’s not what AMD built it for.