To compete with AMD Intel has had to increase the cores. The increased surface area of each chip means fewer are produced per wafer compounded by increased faults as die size increases
There's still no light at the end of the tunnel for PC makers as Intel's CPU constraint – a problem that showed up in anger 13 months ago – is on course to continue for another couple of quarters. At the start of the year, Intel claimed its chip supply drought would end by the summer but it continues to run and run as …
And on top of this, the failure to have any meaningful production on 10nm (ie limited mobile processors, and ASICs) means there is already close to 100% utilisation of 14nm fabs.
Increasing the size of 14nm parts just reduces the number of units Intel produce and there are still questions over competitiveness once AMD release Threadripper.
And Intel's 7nm process is expected in 2022, maybe late 2021 if the process gods smile on them...
Sure and in 2018 Intel generated generated a record $29.4 billion cash from operations, generated $14.3 billion of free cash flow and returned nearly $16.3 billion to shareholders. Expecting record 2019 revenue of approximately $71.5 billion and first-quarter revenue of approximately $16 billion.
All that talk of shortages sure does seem to be helping revenue.
I don’t think Intel are making up their capacity shortage. Intel have always depended on n/n+1 processes for their CPU business with CPUs on n and high value chipsets etc on n+1 with fabs generally being retooled at the end of their n+1 life. With 10nm being almost useless, Intel have ended up with 22nm and 10nm fabs not producing significant product and 14nm at close to capacity. The shortages were the result of Intel having contracted manufacturing, needing multiple CPU lines on 14nm and chipsets moving from 22nm to 14nm. The alternative was Intel dropping new products.
As far as the “it will be all over soon...” it’s wild optimism that they will solve their 10nm process issues and actually be able to produce products the markets want at any volume. The alternative is admitting to the market that the have to write off a $20bn investment in 10nm and have a significant product gap until 2022.
The problem showed up when they started making 14nm modems in volume for Apple last summer. When they signed the deal for that, likely a couple years prior, they believed they would be making a lot of their CPUs in 10nm by then. When they couldn't make 10nm work, they had a lot more 14nm wafers being produced than they planned, and had a contractual commitment for a number of them to go to Apple, meaning a reduced supply for their own CPUs.
Not that they minded, a constrained supply reduced the pressure for them to lower prices in the face of AMD's competition. This way they were able to keep high prices, have an excuse for high prices, and keep their fabs at 100% utilization. Since Apple's demand is somewhat seasonal, they will be selling fewer iPhones next spring/summer than they are currently, meaning less demand for modems from Apple. Plus Apple will be going to Qualcomm for iPhone 12 modems next year.
Between that, and Intel finally squeezing out some 10nm CPUs, there's light at the end of the tunnel. Unfortunately that light is the approaching train of AMD's next gen CPUs made in 7nm coming out next year that are going to crush Intel's pricing power.
Ryzen 2+ (or Ryzen 3) may crush Intel's pricing power; it won't crush Intel's marketing power. Don't forget, at one time, Intel was keeping Dell afloat if they agreed to not sell AMD. Isn't it strange that I cannot buy a Ryzen Optiplex from Dell?
Oh it isn't curtains for Intel by any means, but it will force them to actually compete which they haven't done since the Opteron days. Only this time they may not be able to push aside that competition quite so easily. Intel no longer has a process lead, and will in fact be behind at this time next year as TSMC will be in mass production with their 5nm process (to be first used for Apple's A14, no doubt) and given the problems Intel has had rolling out their 10nm process I don't think anyone believes their claims of being in mass production at 7nm in 2021.
When Dell starts losing sales to competitors selling servers that cost less and perform better using AMD CPUs, they'll be forced to capitulate and that's when Intel will be forced into a price war that will decimate their margins and profits. AMD is used to losing money, they are much better equipped to fight that battle than Intel.
Last time I looked at bulk buying from Dell they seemed to be aware when you were experimenting with others products and would reduce price to keep you finding out if the other products were OK.
I wasn't the buyer but I could look at pricing for a few hundred machines and a couple of weeks later our IT manager would get a much improved offer!
Just a note that Dell *do* sell AMD-based servers - they even have 2nd gen Epyc servers available. The price/performance and performance/watt figures for 2nd gen Epyc are so much better than Intel Xeon equivalents that it's almost a no-brainer now to go Epyc. I can see Intel losing significant server market share in the next few years unless they have some major price cuts.
You’re right - Dell do offer both manufacturers products.
Try getting a quote for a substantial order and see which is cheapest, at least as an existing customer, the Intel offering is likely to undercut the AMD equivalent on either laptops or servers unless you need the high core counts. Intel are offering sizeable discounts over RRP on the 10+ servers or 100+ laptops quotes I’ve seen - I can only assume desktops are the same as I haven’t seen quotes.
Once upon a time, a decade or so ago, Dell did have a few AMD models. I know because I purchased three C521 mini-tower models with dual-core AMD processors. They ran very nicely for a number of years, finally dying off one-by-one as their motherboard capacitors failed. They were good performance at a good price at the time and yes, Dell could do better for their customers if they still offered some AMD models in their lineup.
Truth is, next time I need to replace a desktop unit (currently Dell XPS 8900 i5 Skylake), I'lll be looking for an AMD-powered system first, not a Dell system first.
There are two possibilities, according to Lanci: production issues, though he said Intel should have rectified this by now, or a "problem with the architecture of the CPU. If this is the problem, it is unpredictable."
Speculation* on my part: maybe they're trying to fix all the Meltdown/Spectre/side channel issues in hardware -->
*Pun not intended
10nm fabs have already been built and have failed that is not news. Yields are in the crapper while at the same time Intel is lowering prices, margins are going to get creamed.
7nm Fabs have just broken ground a few months back and might be ready for validation the end of 2020. By the time Intel tapes out and validates any new 7nm microarchitecture AMD will be producing 4th Generation Ryzen on 5nm. Foiled Again.
OEM's are whining about how Intel is essentially screwing them over. Do something about it and stop your whining and complaining.
The funny thing is, AMD wants the server business with high margin EPYC. Desktops and laptops are necessary for consumer awareness but it's news like Frontier and Archer2 that is going to make real money for AMD. In fact given ARCHER 2's expected installation and deployment date building out those EPYC cores for Archer 2 is probably what pushed Threadripper back.
Intel has for years been gouging the consumer by selling expensive and poor performing junk loaded with security vulnerabilities, and OEM's want more?
It takes 4-5 years to create a clean slate micro-architecture. That is what AMD did when they innovated with Ryzen high yield chiplets on an interposer. Intel is still married to low yield massive slabs of silicon. But the real issue is AMD has beacoup patents on Multi-Chip Modules, memory controllers on Stacked Memory and interposers. INtel has to innovate around that, oops that's not innovation but stealing IP.
Biting the hand that feeds IT © 1998–2019