Jim Keller leaves AMD - AMD claims Zen "on track"

[PlayStation 2]

Well the A4 was a single core Cortex A8. and The A5 was a great upgrade, it brought a dual core Cortex A9 with out of order execution. The A5 is the reason multi core SOCs exist.

 

ok i was a bit off qq

[Dimwitted]

Intel doesn't really spend much more than AMD on producing a new CPU architecture. Most of Intel's R&D goes into its foundry construction, equipment purchases, and related expenses. After that you have Infiniband/Omnipath, USB, Thunderbolt, PCIe, Ethernet, and a plethora of standards Intel is basically the main innovator for, and then about 1.5 billion out of the whole 100 billion USD R&D budget i spent on new CPU/GPU architectures each year.

And what has this whole thread been about? Foundries' fuckups that cause people like Keller to bail before product can hit the floor. It's not pretty but Intel seems to have the *least* amount of problems. One reason is that they are very far ahead of anyone else, allowing them the luxury of working out the issues for several years before going "live". The fact that AMD isn't pushing them has helped immensely as well. It's why they've already broken ground on a 10nm fab to be ready for service in a couple of years. Absolutely no other commercial fab is even going there yet. Only IBM is doing R&D with smaller processes but IBM has no intention of commercially building any chips.

[MageTank]

And what has this whole thread been about? Foundries' fuckups that cause people like Keller to bail before product can hit the floor. It's not pretty but Intel seems to have the *least* amount of problems. One reason is that they are very far ahead of anyone else, allowing them the luxury of working out the issues for several years before going "live". The fact that AMD isn't pushing them has helped immensely as well. It's why they've already broken ground on a 10nm fab to be ready for service in a couple of years. Absolutely no other commercial fab is even going there yet. Only IBM is doing R&D with smaller processes but IBM has no intention of commercially building any chips.

Umm, Keller finished his job. He left back in the K8 days before it launched on the market too. K8 was very successful. His job was specifically listed as redesigning the core, and tasking a team to see it through. He did both and moved on. The foundries had nothing to do with it.

[looncraz]

I think you're right. I think I'm conflating the GPU issues that TSMC is having @16nm with GloFo problems @14nm. Either way, neither is coming up roses. And I just heard that Glofo is now saying 1st qtr 2017. That's waaaay too long. I don't know if AMD can hang on that much. It's too bad that they don't have the resources to whip out a FM4 chipset and refresh their line with new m/bds. See if there was something that could supercharge the current lineup. Of course, if it was that easy, they would have done it by now.

 

It just goes to show just what the powerhouse Intel is. They make it look easy.

 

GloFo has planned to ramp 14nm LPP in 2016 for as long as they've licensed the tech from Samsung.  Samsung just barely is beginning to ramp the process.

 

Intel is definitely a powerhouse, but that's because they have two teams working on two options all the time.  If a node fails to be available on time, they can just produce the non-shrunken design on the existing node.  That is the advantage of their tick-tock strategy.