It can, if you’ve designed from the start for it to be scaled.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Incoming Intel CEO Derides Company's Inability to 'Deliver Better Products' Than Apple's M1 Chip
- Thread starter MacRumors
- Start date
- Sort by reaction score
This happens a lot to large companies who primarily service other large companies. While their customers may want to see better progress, they also have a large amount of infrastructure built around present design trajectories. Apple doesn't have to answer to any large OEMs. I guess that doesn't explain away the failure of the Itanium though.
Intel has arguably focused on the use of x86 in data centers in recent years, and most of those machines run some flavor of Linux.
I think the OP’s point was being made in context the smartphone market. Where, Microsoft’s market might not be “nothing”, but, in the words of Douglas Adams, it’s as near to nothing as makes no odds. And, any Windows smartphones you may actually see are merely the products of a deranged imagination.
Yeah, the majority of “work” being done on Apple computers is being done using a laptop form factor. Approximately 80% according to Apple, probably more than that by now (that was a few years ago).
Wait, where is Motorola? They were so good at going slow after they were beating Intel while making the PowerPC 604/604e. People on MacRumors were complaining about the 133 MHz upgrade speeds of the G3 series.They delivered mediocrity for so long and not even on time. Now they are casting aspersions at a company that is eating the lunch of almost every other chip house...
Motorola spun their semiconductor stuff off as Freescale, who in turn were acquired by NXP (who are a spin-off from Philips), and are now mostly an ARM shop.
There is little evidence that they are struggling. There are always rumors, but yields change from week to week and are closely held secrets, so you can't really trust outside analysts. A good indicator are profit margins, but even that is difficult for diversified companies like Intel. In any case Tiger Lake laptops are readily available, including top sellers such as Dell XPS. Intel even announced Tiger Lake-based NUCs (I'll probably get me one of those as my next Linux tinker box).
Hm, they are actually holding their own as it is even with a process and core handicap. Once the 8-core Tiger Lake is available, it will probably beat everything in the ultraportable laptop segment in terms of performance. In terms of power consumption the M1 will reign though for now. The Big/Little design alone gives them a big advantage with normal laptop workloads.
I don’t get it. All of those things STILL require engineering know-how across many disciplines.
It was a taunt, so jibe would be accurate. But, then again, you could claim the world is an oblate spheroid and get 12 pages of comments. There will always be someone to say “no it’s not” even if it’s just for the lulz
There would also be a call from the younger ones to quote sources because they didn't know the history.It was a taunt, so jibe would be accurate. But, then again, you could claim the world is an oblate spheroid and get 12 pages of comments. There will always be someone to say “no it’s not” even if it’s just for the lulz
I don't know my friend. I have a hard time interpreting Intel's behavior during the past 5-6 years. Fact is they stalled at 14nm, and I have a hard time believing they did it because of incompetence. But they did.
Regarding tigerlake, if i am not mistaken this would be at Intel's 10 nm which equates to 100 mtr/mm2. Still a far cry from TSMC's 5 nm which is 170 mtr/mm2.
What happened to Intel''s tik tok? I am aftaid they made it tik tik tik tik.. maybe tok? That's sad not only for them but for all of us.
Intel used to be e ahead of everyone regarding manufacturing process. When you are at your peak with all talent and research on your side, how is it possible to fail, literally throwing such advantage out of the window? Do you see now what I am saying?
Last edited:
A number of things happened. One, Tik Too didn’t work any more; they couldn’t scale down that fast. Two, despite that, they wanted to cling to the idea that a new microarchitecture should happen after a process shrink.
Because the process shrink to 10nm was delayed (and still isn’t finished), their chips were all Skylake derivatives.
They bet big and also had hubris, which from a ca. 2005-2015 perspective makes sense (if you ignore mobile).
Only with the releases or Ice Lake, Tiger Lake, Rocket Lake (a backport of some of Tiger Lake to 14nm) are they discovering from this disaster, which lost them about half a decade.
Hubris. Success hides problems.
You let all the people go you think you don’t need. And then suddenly, you need them back, fast.
idk how long it takes to create a new design or improve on the current. Maybe 4 years is the minimum and is considered fast not slow. Maybe @cmaier can give better insight.
I never heard of the guy before, but someone who calls Apple "lifestyle company" does not sound too "aware" of the situation. Apple creates like 1 of 2 major operating systems on all handheld devices, and the second major OS after Windows in market share. They are also the only computer company that designs their own CPUs.
We will see how good of an engineer he is, hopefully he is not like that one guy who went on stage and said "palm pre/webos eats iphones for breakfast"
He worked for Intel for 30 years before he moved on to other companies. And then Intel brought him back. I bet he's pretty good.
If you boil down the quote... he basically said "we have to be as good as Apple"
I take the "lifestyle" comment to mean "a company who makes all sorts of consumer products"
It sounds like he's frustrated that Apple has become a great chipmaker among all the other things they do... while Intel has been struggling as *solely* a chipmaker.
That's my read on it.
People here are focusing too much on the "lifestyle" comment... but missing his overall message.
Would we be having this discussion if he just came out and said the word "Apple" ?
We know who he was referring to. The Intel employees knew who he was referring to. So imagine if he said this instead:
"We have to deliver better products to the PC ecosystem than Apple. We have to be that good, in the future."
It would still be the same message.
Supposedly, Apple's chip team has a pipeline of roughly three years. I.e., when we saw the A14 in fall, they were already beginning work on the A17.
Everything doesn't scale and and to use the words "obviously" and "trivially" in chip development is to dismiss that.
Even if you design a scalable architecture, you will run into PPA (power, performance, area) issues.
If if you deigned to scale you can run into the recticle limit of the process, you can run into distance limits as you increase size the require higher drive strengths in gates which translates into power consumption that is not linear but exponential. You then run into timing issues and layout problems. If you add pipeline stages, then you have latency issues.
I've been doing processors, DSPs, ASIC and SoC design for 30+ years.
This is what I do everyday.
I've worked in every technology from ECL, CMOS, BiCMOS, etc. in mainframes to consumer products and at every process node from .35um to 7nm.
There is absolutely nothing "obvious" nor "trivial" about scaling a complex chip design.
How long have you been designing chips?
I was specifically responding to a claim about the current yield of their 10nm process.
There is no question that they ran into trouble a few years ago. I don't know the details, but contacts at Intel tell me it's because their initial goals were overly aggressive and they had to make several corrections which delayed their roadmap.
The node numbers are misleading. Intel's 10nm process actually has a higher transistor density than TSMC's 7nm (which is what current AMD CPUs are using). Intel's 7nm process will probably be up to par with TSMC's 5nm.
My interpretation would be "we have to be better than any company that doesn't specialize in semiconductors".
Sounds roughly in the ballpark.
A company is always looking several generations ahead.
If you are using the sort of design methodology that derived from DEC, AMD, and Exponential, which is where many of those folks are from, then about 2.5 years is the expected lead time. The first 6 months are defining things at high level, the last 6 months are meeting electrical and timing constraints, and the middle period is just continuously iterating on the design until it passes logical verification testing and has electrical and timing constraints “close enough” that there is not likely to be a need to modify anything at the microarchitectural level to compensate.
What’s interesting is that if you use “asic” methodology, where you use software to synthesize, automatically place gates, and automatically route all the wires without much intervention, you can shave 9-12 months off that time. But your chip will be 20% slower and burn 20% more power.
Also a good interpretation.
I think Pat Gelsinger was simply trying to rally the troops at Intel. I don't believe he was trying to denigrate Apple.
He knows what Apple has been doing. The engineers at Intel know what Apple has been doing. Hell... everyone in the entire semiconductor industry has been paying attention to what Apple has done over the last decade starting with the A4
I bet these kinds of pep talks are happening at other companies to light a fire under their employees' asses... 🔥
Altra performance is generally a little superior to Graviton 2 (no slight to Graviton 2, since it shipped a year earlier) and you CAN buy Altra.
Also a good interpretation.
I think Pat Gelsinger was simply trying to rally the troops at Intel. I don't believe he was trying to denigrate Apple.
He knows what Apple has been doing. The engineers at Intel know what Apple has been doing. Hell... everyone in the entire semiconductor industry has been paying attention to what Apple has done over the last decade starting with the A4
I bet these kinds of pep talks are happening at other companies to light a fire under their employees' asses... 🔥
Two pep talks I remember vividly from my days at AMD:
PEP TALK 1:
Some very talented folks had been quitting from the Sunnyvale design team. Among other problems was that our compensation was highly dependent on stock price, and we had been given large numbers of options at $42/share. But nearly as soon as the options were granted, the stuck price plummeted, and it was probably around $10/share by the time Sanders decided to give a pep talk. (It was a long time ago, so all numbers may be off).
Paraphrasing, but I‘ll never forget the gist of what he said when he showed up in his custom suit with pinstripes that spelled out his name in tiny letters: “I’m upset, too. My wife is demanding that we refurnish the house, and with the stock in the tank I have to tell her we need to wait.” (Keep in mind that there was no way any of those folks could afford an actual house in Silicon Valley unless provided huge bonuses or getting stock windfalls)
PEP TALK 2:
K8 design work had begun, but seemed to be a bit of a morass, and most of us were still stuck working on spins of K6. Sunnyvale didn’t get along all that great with Austin, at least in some cases, and most of the K6 team came from NexGen and were not AMD natives. People started quitting in droves. Some left the industry, and others just wanted to go work someplace with more exciting work, better pay, etc. Dirk Meyer, when he was in charge of the Austin team and before he became CEO, came up to give us a pep talk. He said “if you don’t like it here, quit.”
Within a couple weeks, the remaining design team was down to just a couple dozen people, and probably only around 18 who had real experience designing chips.
This resulted in a meeting at a french restaurant to see if the rest of us would quit or stay, which turned into the idea that became x86-64, and opteron and athlon 64, designed by a very small team of people who were all-pep-talked-out.