The Intel Execution Thread [2020]

Status
Not open for further replies.

BRiT

(>• •)>⌐■-■ (⌐■-■)
Moderator
Legend
Supporter
Been hearing some rumblings that something big is going to happen an Intel, but no details and just vague hints. Did anyone see or hear anything likewise?
 
It being Good or Bad depends if you're located in a camp and then which camp.


Exclusive: Intel’s Ari Rauch Asked ‘To Look For Other Opportunities’ [Updated]

So something pretty big just happened. I have just learned from multiple sources that Ari Rauch, VP & GM of Intel Visual and Parallel Compute Group has been asked to 'look for other opportunities'. Intel's graphics group is at the moment reporting to Raja Koduri. Ari Rauch has been the Vice President and General Manager of the Intel Visual and Parallel Compute Group since 2015 and was also responsible for the Intel Xe efforts (although this development is unlikely to impact that).
 
Well, time to spin off that manufacuring arm already.
Sadly don't recall who initiated that rumor (other than it was by some1 I've heard of before) and it was 2018 but I'm still expecting it to happen. Still makes a lot of sense to me that they'd do that
 
New Well, time to spin off that manufacuring arm already.
That news seems to be about Raja taking over all the GPU business from Intel. I don't see anything about Intel spinning off their manufacturing.
 
Well, time to spin off that manufacuring arm already.
Sadly don't recall who initiated that rumor (other than it was by some1 I've heard of before) and it was 2018 but I'm still expecting it to happen. Still makes a lot of sense to me that they'd do that
Why? They don't exactly have any unused manufacturing capacity right now.
 
Why? They don't exactly have any unused manufacturing capacity right now.
There are many possible reasons. Intuitively, Intel business model worked flawlessly (and devastatingly for the competition) when they could design cpus & deploy processes in lockstep. This is no longer the case.
Further, the economics of a bleeding edge foundry may be yet set to explode again. Or it may be particullarly for Intel's manufacturing, their very poor execution in the last 3 years may have already sucked a lot of cash (while delaying some products - but that only impacted stock price at most, fortunatelly)

The rumors of Intel spin off originated in 2018 and while I didn't save the source as I said, it was around the times when some of the fiasco of their 10 nm node troubles got leaked. Notably the fact that Intel management ignored the semi tool vendors that said how their competitors are well ahead and had already solved most of the issues Intel was facing at the time
 
There are many possible reasons. Intuitively, Intel business model worked flawlessly (and devastatingly for the competition) when they could design cpus & deploy processes in lockstep. This is no longer the case.
This hasn't been the case due to issues they've ran into with 10nm process specifically. We have no reasons to just assume that it will happen again in the future.

Further, the economics of a bleeding edge foundry may be yet set to explode again.
Sure but how would spinning off a production capacity which is 100% filled (and there's a shortage) with your own orders help with economy?
It's not like there's a huge choice of options below 14nm right now and even by going fabless Intel won't have many places to go for production - and this will definitely stay this way for the foreseeable future.
Why not just add 3rd party fabs into your production mix in addition to your own factories?

Spinning off of fabs for Intel would make sense in two cases:
A. They aren't fully loaded with your own orders and you need 3rd party orders to keep them running. This isn't the case obviously.
B. They are significantly technologically behind all of competition on the market. Thus far they aren't really behind even TSMC and it's unlikely that they will stay behind for more than a couple of years.
So why spin them off right now?
 
This hasn't been the case due to issues they've ran into with 10nm process specifically. We have no reasons to just assume that it will happen again in the future.
Disagree. The status quo is now having multiple nodes and cpus all branded under the same series 9xxx etc umbrella
Returning to tick tock would be the change

Sure but how would spinning off a production capacity which is 100% filled (and there's a shortage) with your own orders help with economy?
It's not like there's a huge choice of options below 14nm right now and even by going fabless Intel won't have many places to go for production - and this will definitely stay this way for the foreseeable future.
Why not just add 3rd party fabs into your production mix in addition to your own factories?

Spinning off of fabs for Intel would make sense in two cases:
A. They aren't fully loaded with your own orders and you need 3rd party orders to keep them running. This isn't the case obviously.
B. They are significantly technologically behind all of competition on the market. Thus far they aren't really behind even TSMC and it's unlikely that they will stay behind for more than a couple of years.
So why spin them off right now?

None of the reasons you invoke.
I don't agree that Intel won't be able to fab their chips elsewhere, so there's that
I did mentioned already bits about foundry economics. It's just that foundries move at different paces now (slower). The investments are larger, recovering of investments takes longer. So it's increasingly likely that a heavy investment year of the foundry would drown all the financial performance of the entire company.
Also, 'sell' high and don't do an AMD. If there's a risk of the foundry business lagging behind others (which is not currently the case), perhaps it's better to handle that now rather than when the company will be in real trouble
 
Disagree. The status quo is now having multiple nodes and cpus all branded under the same series 9xxx etc umbrella
Returning to tick tock would be the change
Don't really see what "tick tock" has to do with branding but anyway - this is a result of just one thing - issues with 10nm process. You think that this will still be the case with 10nm vs 7nm and so on?

I don't agree that Intel won't be able to fab their chips elsewhere, so there's that
Never said anything like this. They can and they do actually, right now.

I did mentioned already bits about foundry economics. It's just that foundries move at different paces now (slower). The investments are larger, recovering of investments takes longer. So it's increasingly likely that a heavy investment year of the foundry would drown all the financial performance of the entire company.
And the foundry isn't a part of the company? It's an interesting way of looking at a part of the business which was one of two (sometimes just one even) drivers behind the success of the company.

Also, 'sell' high and don't do an AMD. If there's a risk of the foundry business lagging behind others (which is not currently the case), perhaps it's better to handle that now rather than when the company will be in real trouble
Is there such a risk at Intel's foundries right now? Also how is it "high" at the moment with all the issues with 10nm on fixing which Intel spent billions already?

I'd argue that it's the wrong time to sell off the foundries as a) they've made a huge investment into their 10nm and 7nm processes already and b) it looks increasingly likely that process advantages will soon be less about gate sizes and more about 3D integration in which case having your own foundries may actually be a huge benefit for your products.
 
Don't really see what "tick tock" has to do with branding but anyway - this is a result of just one thing - issues with 10nm process. You think that this will still be the case with 10nm vs 7nm and so on?

Yes. Maybe not the issues, but the lack of tick tock-like cadence


Is there such a risk at Intel's foundries right now? Also how is it "high" at the moment with all the issues with 10nm on fixing which Intel spent billions already?

The risk is non zero, for sure. Obviously I cant quantify it in any meaningful way.
Would be 'higher' than if the risk materialises (which I did suggest by bringing up AMD's situation which was far worse than Intel's now)

_______

Do note however that I posted this "Intel will spin off its fabs thinggie" as a reply to the OP asking what major thing could happen at Intel. I thought : well, there's always this
 
Yes. Maybe not the issues, but the lack of tick tock-like cadence

I don't think so.

Intel pushed things too far on their 10nm process. The desire to have pitches smaller than 40nm on the bottom metal layers meant they had to go with quad patterning, they also had to go with all-cobalt wires, because copper wires don't scale down. Contact on gate was also introduced.

In the planning phase, these changes would have had a projected negative impact on yield, but once they started implementing these impacts turned out to be much worse.

The low yield means only smallish dies are feasible, that's why there are no 6-core 10nm SKUs. The cobalt interconnect means limited effective drive current (cobalt has four times the resistivity of copper) which means lower f_max, which means they can't compete with 14nm desktop SKUs.

To counter the lower f_max, Intel is going wider; More execution units, bigger schedulers, much bigger ROB.

Intel will bounce back once they transition to 7nm (Samsung/TSMC 7nm) with EUV scanners for the FEOL and the bottom metal layers; single patterning instead of quad patterning, manhattan routing will be back in vogue at Intel, etc.

EUV is the great equalizer, it will solve all the current problems at Intel, but of course come with its own set of problems.

With a very wide microarchitecture, Intel is set up for a big comeback once they get the production facilities converted to EUV, IMO.

Cheers
 
one of the things killer was famous for was having a network chip that could run linux maybe intel want that tech ???
 
Hyper-scale Infrastructure Services Accelerate
May 21, 2020

In the top four clouds, all accelerator chips (regardless of type or manufacturer) have been attached exclusively to Intel Xeon processors for at least the past year, with the very recent exception of AMD’s EPYC on Microsoft Azure. Azure was the first to break ranks in February with its first production deployments of AMD Radeon Instinct MI25 GPUs with AMD EPYC v2 Rome processors.

Processor share of accelerated instance types within each cloud in March 2020 (Source: Liftr Insights)
Teich_1.png

https://www.eetimes.com/hyper-scale-infrastructure-services-accelerate/
 
Last edited:
Wow indeed. Speculation is going to run rampant. That's a huge loss for Intel.
 
Status
Not open for further replies.
Back
Top