Intel to Outsource 14nm Chip Production to TSMC Due to Tight Supply

[LePhuronn]

Epyc uptake in data center is still sluggish: "Mercury estimates Epyc revenue was $57.66m in the second 2018 quarter vs $36m in the prior quarter."
For reference Intel's data center revenue in Q2 2018 (of which a large part is their Xeon CPUs) is $5.5 billion, growing 6% QoQ or 27% YoY. Don't expect any miracles yet.

My maths isn't the greatest, but isn't that a 60% increase in EPYC revenue for Q2? If we see Q3 revenues reach about $90m then you can argue 60% growth QoQ, that's 10 times what Intel are seeing. I'd say that was miraculous already.

Now yes, we are comparing AMD's revenue millions to Intel's revenue billions, but AMD literally started from nothing (0.3% server share in 2017 or something?). And for the likes of Dell, HP, Microsoft and Cisco to take the plunge on EPYC 1, I don't see it being a one-off purchase.

 

[AndreiD]

Of course that's good growth for AMD given their starting point, but it's still peanuts compared to Intel's revenue and absolute growth in $.
It's going to be interesting to see if Epyc's growth is sustained during the next quarters, then we'll know for sure if AMD has a winner on their hands.

 

[LePhuronn]

but it's still peanuts compared to Intel's revenue and absolute growth in $.

Although it occurs to me that number could be skewed by the fact the Intel stuff probably costs 10 times what AMD does

Stupid example: AMD revenue was $10,000, Intel revenue was $100,000, but they both sold 10 CPUs. Absolute figures can be misleading...

 

[CAT-THE-FIFTH]

Wow:

https://twitter.com/CDemerjian/status/1040586902540378113?s=19

 

[Distracted]

Wow:

https://twitter.com/CDemerjian/status/1040586902540378113?s=19

No source. Speculation?

Sure it might end up being true but i dont think the tweet has anything to back up the claim right now does it?

 

[CAT-THE-FIFTH]

No source. Speculation?

Sure it might end up being true but i dont think the tweet has anything to back up the claim right now does it?

He has been correct on the 10NM issues for quite a while,but still I get what you are saying. I suppose if what is happening is true Intel will probably release some bumpf to indicate this at some point.

AMD has a golden opportunity here TBH,although I just hope TSMC 7NM doesn't end up having its own set of issues too,even if it is only volume. Its my major concern with AMD,as with GF throwing in the towel,TSMC is going to be oversubscribed for 7NM I suspect!

 

[AndreiD]

He's thrown a lot of rumours on 10nm, some correct, some not, but Intel relaxing their 10nm has been rumoured for a while now.
Even if they relax it to what Intel would call 12nm, it would still be similar density to TSMC/Samsung 7nm and most likely higher performance. Going to be interesting to see if we actually get to see "relaxed 10nm" non-mobile consumer products in 2019.

 

[CAT-THE-FIFTH]

He has an update on the previous article:

https://www.semiaccurate.com/2018/09/17/more-on-intels-10nm-process-problems/

SemiAccurate has a little more on Intel’s 10nm woes, this one is actually goodish news for once. It involves a technical point that we told you about earlier and how it is used.

In our earlier exclusive outing of Intel’s 10nm problems we laid out 4-5, depending on how you count, issues with the process. Since then we have gotten new information on one of them, more specifically how it is used and where. The up side to this information is that the new downgraded ’10nm’ process from Intel will not take as big a hit from the removal of this tech as SemiAccurate said earlier, but it will still take a hit.

 

[Rroff]

He's thrown a lot of rumours on 10nm, some correct, some not, but Intel relaxing their 10nm has been rumoured for a while now.
Even if they relax it to what Intel would call 12nm, it would still be similar density to TSMC/Samsung 7nm and most likely higher performance. Going to be interesting to see if we actually get to see "relaxed 10nm" non-mobile consumer products in 2019.

Intel's 10nm is like 2nd generation 7nm (or intended to be) in some respects but not others - a "relaxed" version wouldn't be overall higher performance than TSMC first generation 7nm though it might perform a bit better for certain areas relevant to CPUs. Which is where it seems they've tripped up as those areas don't seem to be accomplishable without EUV and EUV at Intel is likely atleast another 2 years away. (For reasons way beyond my level of knowledge to explain it is unlikely they are going to go for some 12nm like downgraded version though).

If Intel do a "10nm" that isn't 10nm downgrade though I'd have thought they'd more likely give their 14nm another spin - despite being the "plus plus" variant there are still optimisations they've skipped - I assume due to the assumption it wouldn't be worth it when they were moving on to their 10nm - there is still around 15-16% performance that can be wrung out of it. (Kind of like TSMC's 12FF to their 16FF).

 

[CAT-THE-FIFTH]

Take with a ton of salt:

https://mobile.twitter.com/witeken/status/1042043796660387840

2020 for desktop 10nm parts??

 

[NinjaCool]

Intel to Outsource Entry-level CPU and Chipset Fabrication to TSMC
https://www.digitimes.com/news/a20181030PD205.html