Friday, September 9th 2022
Intel Expects to Lose More Market Share, to Reconsider Exiting Other Businesses
During Evercore ISI TMT conference, Intel announced that the company would continue to lose market share, with a possible bounce back in the coming years. According to the latest report, Intel's CEO Pat Gelsinger announced that he expects the company to continue to lose its market share to AMD as the competition has "too much momentum" going for it. AMD's Ryzen and EPYC processors continue to deliver power and efficiency performance figures, which drives customers towards the company. On the other hand, Intel expects a competing product, especially in the data center business with Sapphire Rapids Xeon processors, set to arrive in 2023. Pat Gelsinger noted, "Competition just has too much momentum, and we haven't executed well enough. So we expect that bottoming. The business will be growing, but we do expect that there continues to be some share losses. We're not keeping up with the overall TAM growth until we get later into '25 and '26 when we start regaining share, material share gains."
The only down years that are supposed to show a toll of solid competition are 2022 and 2023. As far as creating a bounceback, Intel targets 2025 and 2026. "Now, obviously, in 2024, we think we're competitive. 2025, we think we're back to unquestioned leadership with our transistors and process technology," noted CEO Gelsinger. Additionally, he had a say about the emerging Arm CPUs competing for the same server market share as Intel and AMD do so, stating that "Well, when we deliver the Forest product line, we deliver power performance leadership versus all Arm alternatives, as well. So now you go to a cloud service provider, and you say, 'Well, why would I go through that butt ugly, heavy software lift to an ARM architecture versus continuing on the x86 family?"
Finally, Pat Gelsinger has emphasized that the company will continue to exit more businesses where it doesn't thrive. Just like it did with Optane memory, we could see Intel pulling out of other markets that don't necessarily align with the leadership's vision. Given that the CEO appointed a new leadership group and performed major company structure reforms, it is interesting to see what comes out of this. Just a few days ago, we saw the appointment of Shlomit Weiss as senior vice president and Co-GM of the Design Engineering Group.
Source:
via Tom's Hardware
The only down years that are supposed to show a toll of solid competition are 2022 and 2023. As far as creating a bounceback, Intel targets 2025 and 2026. "Now, obviously, in 2024, we think we're competitive. 2025, we think we're back to unquestioned leadership with our transistors and process technology," noted CEO Gelsinger. Additionally, he had a say about the emerging Arm CPUs competing for the same server market share as Intel and AMD do so, stating that "Well, when we deliver the Forest product line, we deliver power performance leadership versus all Arm alternatives, as well. So now you go to a cloud service provider, and you say, 'Well, why would I go through that butt ugly, heavy software lift to an ARM architecture versus continuing on the x86 family?"
68 Comments on Intel Expects to Lose More Market Share, to Reconsider Exiting Other Businesses
There's an inconvenient fact about photon(ic)s mentioned here, I never thought of that before but now it seems obvious:
semiengineering.com/cognifiber-photonic-computing/Then there's manufacturing. No matter what great new materials and structures scientists develop, be it in electronics or photonics or quantastics, someone must be able to pack billions of components on a chip, then mass produce that, achieve good yields and uniform characteristics, achieve an acceptable price/performance, guarantee long term stability of the product (so no OLED-based photonics, I guess), be able to simulate and optimise a chip design, and so on.
Fabs as we know them can only do so much. And a bit more in the foreseeable future with high-NA EUV and higher power light sources. But what manufacturing breakthrough is expected after that?If AMD decides to go E-core as well, they'll do it in their own way, radically different from Intel's way. One leak(?) stated that cores may come in 1P+1E pairs - maybe Bulldozer cores done right? If that materialises then it surely will behave much differently than the Lakes, with serious advantages as well as disadvantages.
A quick history overview: Pentium 3 (3 pipelines, normal clocks) vs. Pentium 4 (2 pipelines, insane clocks).
Since intel could not wind up the Pentium 4 clocks further (material restrictions), the project failed, so they just took Pentium 3 design added a very good branch prediction unit, developed at their intel Izrael branch and there it was the "Core" line (plus some improvements here and there).
They can't just fall back on any good designs these days because they do not have any, all of the good from Pentium 3 and 4 are mixed in these newer CPUs with improvements and added security holes (you know, for security).
The only thing holding them back is their production process lack of progress, first being stuck for ages at 14nm, now they have somehow patched together that 10nm process that they lie is "7" which it is not and are kind of still stuck at it as with 14nm, so nothing strange they crank out 200W+ wonders and lie and call them 65W And 125W all the while AMD uses TSMC's true 7 and lower nm processes (automatically cooler chips because of smaller transistor size).
Intel needs smaller production process, oddly enough they seem incapable of deploy it in so many facilities that they have. They even produce some of their chips at TSMC, quite a shameful practice for a company that used to lead in process advances. Quite an ironic punishment for their past practices...
The advantage of P-cores is much better performance when running one thread per core. That's great but why would you need more than so many of them?
The advantage of E-cores is better performance per mm2 of silicon, which is reflected in better performance of another Intel product, known as INTC.
I assume, based on available data for desktop ADL (which is scarce), that two threads on
an E-coretwo E-cores execute as fast or a little faster (on average) than two threads on one P-core.Issues with P+E scheduling remain but it's a matter of software, which can be improved and tuned over time. Hopefully there will be a couple tuning knobs left to the end user to turn, too.
Heat density be damned, we'll sell you a larger cooler.
Edit: because it was unclear before
SMT does only usually amount to +30% the performance of one core.
Here's what I mean, in terms of relative performance.
1 thread on 1 P-core = 1 (low-threaded CPU load, so SMT is not necessary)
1 thread on 1 E-core = 0.65
2 threads on 1 P-core = 1.3 (SMT used because high-threaded CPU load) (probably a bit less on average)
2 threads on 2 E-cores = 2 x 0.65 = 1.3 (probably a bit less because of shared L2 cache)
On a CPU with 8P+8E cores, the scheduler should put 8 most demanding threads on the P-cores, one on each; next 8 threads on E-cores; and next 8 on P-cores using SMT. (I do realise this is an oversimplification, threads are dependent on other threads, power consumption matters too, etc.)These things are not mutually exclusive. Is it good Intel giving CPU buyers more performance per mm2 of silicon, or bad intel giving INTC buyers more performance because of less silicon used? Or both?