Revenue increase ?
How much of the revenue increase is because of higher prices caused by the shortage?
(Shades of the old DRAM price gouging - drop the supply 10% - increase the price 50%)
Intel on Friday published a letter from CFO and interim CEO Robert Swan reassuring customers that the chip biz will be able to make enough processors to satisfy its customers. Currently, Intel chips are in short supply. The company's inability to deliver 10nm processors is well known; it played a role in the resignation of …
Has nothing to do with an "unexpected recovery" in the PC market. It is because they are making all the cellular chips for Apple's latest phones. When they made the agreement to do that, they believed they'd have 10nm working by now so devoting 14nm fab space to those cellular chips wouldn't be a problem.
Cascade Lake Xeon (Purley refresh) is being produced right now - originally, we'd be on 10nm by now. However, since this did not happen Intel is left with a huge mess, they need to produce huge quantities of new Xeon SP for hyperscalers, clouds, etc. and all that on the same 14nm process as everything else. Tough.
Mind you, these Xeons are tough, since they are huge monolithic 28-core monster dies, so few chips per wafer + lots of possibilities for failure (of course, Intel does bin those failed chips as cheaper 18/22/24 cores, but still).
"The surprising return to PC [market] growth has put pressure on our factory network," said Swan. "We’re prioritizing the production of Intel Xeon and Intel Core processors so that collectively we can serve the high-performance segments of the market. That said, supply is undoubtedly tight, particularly at the entry-level of the PC market."
Meltdown, Spectre (all variants), and people are still buying this half-arsed crap in droves?
Bullpats. Artificial scarcity disguised as capacity issues in an attempt to protect the cash cow from a well-deserved rogering by the silicon buying public.
While Intel screwing up would undoubtedly be Very Bad™ for technology, I have to admit a large part of me would like to see them smart a bit as payback for this and the weaselly way they tried to alter the microcode licence earlier this year.
"Artificial scarcity disguised as capacity issues in an attempt to protect the cash cow from a well-deserved rogering by the silicon buying public."
It's not artificial scarcity though - Intel have screwed up and can't deliver more chips due to being a fab down (i.e. 10nm). While it maybe caused by contractual issues (i.e. parts for Apple), they can't just stand up another fab inside of two years and economically it is unlikely to be profitable for Intel.
In terms of it being Intels cash cow, we have to wait and see. If AMD market share increases significantly, Intel may struggle to recover come 7nm. If Azure proves AMD CPU's work well in the cloud (i.e. through the increased cores/memory at the expense of higher power usage), it may benefit Azure and AMD significantly.
If Azure proves AMD CPU's work well in the cloud
The problem historically (I don't know if this is still the case, but I believe it is) is that the Intel Xeon chips gave better performance per-core than the AMD Opterons. Now, that wasn't a big problem on the surface because the AMD chips gave more cores for the same purchase price and power consumption, so they worked out well.
Enter per-core licensing.
For Microsoft running their own DCs, I'm pretty sure they'll cut themselves a good deal on the licensing. But for the rest of us, for Oracle users, for anyone not rolling their own or running a FOSS stack that hurts over time. And that's why Intel still rule the datacentres.
My argument around AMD being a good fit for cloud is purely around efficiency vs the existing options at scale.
As core licencing will tie back to customer usage, it will be unlikely to affect Azure outside operational efficiency vs other platforms.
There's also the potential for Meltdown to have altered the real world balance with current gen AMD chips and if so, the efficiency gap maybe gone...
Yep. I agree with you 100% on both counts. Just pointing out that every time I've bought servers, per-core performance and more recently license costs associated with the extra cores have been a significant factor.
I would absolutely love to see AMD going toe-to-toe with Intel in the server market. Save getting gouged for these Xeons...
And whilst I'm writing my letter to Santa, if they could use the same sockets that would be nice too. Being able to swap Intel to AMD on a whim was great. But it was also twenty years ago.
"Meltdown, Spectre (all variants), and people are still buying this half-arsed crap in droves?"
"Bullpats. Artificial scarcity disguised as capacity issues in an attempt to protect the cash cow from a well-deserved rogering by the silicon buying public."
No. If Intel can't or won't produce enough processors, the OEMs still need to sell *something* so the OEMs will go for the alternatives. The big cash cows are the very expensive Xeons after all, but the supply problems include relatively inexpensive i5 / i7 models. Since AMD finallty has very competitive processors, why would Intel risk losing desktop/laptop market to AMD?
HPE has already put up a notice that some Xeon models have supply problems so HPE is recommending AMD EPYC servers as solution. Hardly something Intel wants to hear, don'tcha think?
I presume they mean Skylake chips. As you may recall, MS blocks later silicon from receiving Windows updates. Of course, that didn't stop me building a Ryzen system last year (with what MS considers "blacklisted" silicon), popping Windows 7 onto it, then using wufuc to get around MS's idiotic block. However, companies generally won't do this, and hence there is still a market for devices with Skylake chips that are officially supported by MS under Windows 7 still.
I've seen businesses not purchase cheaper CPU's because they don't want to be running "unsupported" hardware.
I was going to rubbish it, then I saw vmware bork my vmplayer install with an update which decided that, after many years of running happily, my laptop wasn't good enough, despite adding no detectable beneficial features.
That isn't a business risk I would take, though I did *cough*howmuchiswindowsactuallycostingyou*cough*
"36nm metal pitches with self-aligned quadruple patterning and multiple block layers is probably the problem."
I hadn't, myself, yet identified this as being the particular problem, but I'm sure I would have got there in the end ... Maybe around the time Intel got round to implementing a 3pm plasma pitch with ai-aligned nontuple tessalating and dancing block layers system.
The company's inability to deliver 10nm processors is well known;
The company's inability to deliver 10nm processors that are not subject to all sorts of vunerabilities and with any recognisable performance improvement over what was release 2 or even 3 years ago is well known;
There fixed it for you Intel.
Why the shortage and price hike? How to meet 2018 revenue outlook?
Higher core count, (lower die count)*(lower yield), (lower)**2 product volume and higher cost!
Revenue = (Volume)*(Price), V decreased, so P increased to meet the full year revenue outlook.
So the shortage is due to Intel's technical incompetency of not having 10nm to offset the core count increase, and the price hike is for meeting the 2018 revenue!
Biting the hand that feeds IT © 1998–2020