Intel Board Member Quit After Differences Over Chipmaker's Revival Plan (reuters.com) 52
An anonymous reader shares a report: The sudden resignation of a high-profile Intel board member came after differences with CEO Pat Gelsinger and other directors over what the director considered the U.S. company's bloated workforce, risk-averse culture and lagging artificial intelligence strategy, according to three sources familiar with the matter. Lip-Bu Tan, a semiconductor industry veteran, had said he was leaving the board because of a personal decision to "reprioritize various commitments" and that he remained "supportive of the company and its important work," in a regulatory filing on Thursday.
The former CEO of chip-software company Cadence Design joined Intel's board two years ago as part of a plan to restore Intel's place as the leading global chipmaker. The board expanded Tan's responsibilities in October 2023, authorizing him to oversee manufacturing operations. Over time, Tan grew frustrated by the company's large workforce, its approach to contract manufacturing and Intel's risk-averse and bureaucratic culture, according to the sources, who were not authorized to speak publicly. The circumstances around Tan's exit have not previously been reported. The departure of the industry veteran, who is well-regarded by investors, over Intel's strategy illustrates the uncertainty of its turnaround efforts. Tan leaves as the company endures one of the bleakest periods in its five-decade history that has left it vulnerable to a potential activist shareholder attack, former executives said. Intel has hired investment bank Morgan Stanley to prepare a defense, according to sources familiar with the matter, confirming an earlier report.
The former CEO of chip-software company Cadence Design joined Intel's board two years ago as part of a plan to restore Intel's place as the leading global chipmaker. The board expanded Tan's responsibilities in October 2023, authorizing him to oversee manufacturing operations. Over time, Tan grew frustrated by the company's large workforce, its approach to contract manufacturing and Intel's risk-averse and bureaucratic culture, according to the sources, who were not authorized to speak publicly. The circumstances around Tan's exit have not previously been reported. The departure of the industry veteran, who is well-regarded by investors, over Intel's strategy illustrates the uncertainty of its turnaround efforts. Tan leaves as the company endures one of the bleakest periods in its five-decade history that has left it vulnerable to a potential activist shareholder attack, former executives said. Intel has hired investment bank Morgan Stanley to prepare a defense, according to sources familiar with the matter, confirming an earlier report.
Re: (Score:3, Interesting)
When the AI bubble bursts, Gelsinger will then look like a genius. Don't get me wrong, AI will continue to improve and grow, but not at the pace of current investments. Longer term (lack of) revenue doesn't lie. The breakthrus would have to be C3PO-like within a decade to justify current AI investments. Not gonna happen.
Tan belongs in a smaller startup.
Re: (Score:2)
Intel as we know it may not last that long.
Re:Good riddance (Score:4, Interesting)
The rent-a-fabs have no trouble finding someone to buy time on their older fabs. What are the rent-a-fabs making on their older fabs? A whole lot of specialized compute chips. Googles first TPU was built on an old 28nm process while desktops were on 14/16nm.
Intel needs to find a way to run their older fabs 24/7 also. They cant get in on the rent-a-fab game but they can at least try Intel-branded solutions that compete with the game the rent-a-fabs are playing.
AVX-512 isnt even on the right playing field as its proven to be too wide to be useful for a lot of stuff (AoS) and too narrow to compete with even low end gpus (SoA)
Intels been fucked ever since they got blind-sided by chiplettes, but their fundamental problem existed even prior to then. A vertically integrated fab company cant compete against the rent-a-fabs long-term. Its over.
Re: (Score:3)
Intel has been so obsessed with process shrinking for so long that they have forgotten a whole world of chips that NEED to be made with larger features in order to have sufficient temperature and voltage tolerance for their intended environment. In a world of ARM based microcontrollers, there's a reason the good old 5V ATMega still sells.
Re: (Score:1)
Maybe they thought the market would keep rewarding the leading process shrinkers, leaving specialized mid-density chips to slower competitors. The market changed, but could change back to reward shrinkers again.
It's kind of like you specialize in X and do X well. But demand starts spiking for Y. However, it's hard to tell if Y will stay high or is a blip. If you change to fit Y, you may miss the swing back to X, making you a come-from-behind Y'er and giving up lots of market share in X.
Re: (Score:2)
Intel used to do X and Y. X offered higher margins so they dropped Y (but not the FABs that did Y). The employees who did Y got moved to X.
It seems X is running out of gas now.
Re: Good riddance (Score:2)
"Intel needs to find a way to run their older fabs 24/7 also. They cant get in on the rent-a-fab game"
Why not? Did we not talk a lot about demand for older processes from automakers and such?
Re: (Score:2)
"Did we not talk a lot about demand for older processes from automakers and such?"
The problem is that to Intel older processes with spare capacity are like 22nm or 14nm, while the automakers are on 90nm or even larger and balking at the NRE to process shrink existing products. (Granted, the quality management requirements are much higher for an automotive product than for general computing so it's greater effort to process shift.)
The automakers got by on buying marginal capacity on lines paid for by process
Re: (Score:2)
Until the billionaire realizes that nobody is buying his product anymore.
Boeing's strategy played well for a time, too.
Is there an actual problem? (Score:5, Interesting)
Re: (Score:3, Interesting)
Re: (Score:2)
And even on the x86 desktop side of things a lot of vendors are looking at AMD chips, particularly with the issues Intel has been having with their 13th and 14th gen chips frying themselves.
Desktop Windows computers are quickly becoming a niche market for hardcore gamers and business users, and even in that market AMD has gotten up to 20% market share.
Intel needs to stop acting as if they're the premium choice in chips and start actually innovating again, and expand into a few other market segments. They'v
Re: Is there an actual problem? (Score:2)
This is it. People are predicting an AI bubble, but as the tech matures, itâ(TM)s getting cheaper to run.
Itâ(TM)s insane amount of money being spent, but at the end we will have a working star-trek computer interface.
GenAI is just basically the new UI.
Re: (Score:2)
Re: (Score:2)
Why not? Intel CPUs still work reasonably well (with a few notable exceptions, of course). They can be recommended. Unlike Intel's stock.
Did I miss the merger news? (Score:3)
Did intel merge with boeing?
Re: Did I miss the merger news? (Score:5, Insightful)
Intel has the opposite problem. Boeing is failing because they divested and outsourced. Intel is failing because they didn't. They want to control their own process technology but they apparently aren't capable of keeping up with TSMC. They should have given up on that long ago and gone to just doing design. Let the specialists handle the process. AMD is currently eating their lunch with that formula. AMD has better single thread, better multi thread, AND lower power consumption. They have conclusively surpassed Intel in end results and Intel's sales are falling.
At this point Intel needs a process miracle. Maybe they will pull one out, but I doubt it.
Re: Did I miss the merger news? (Score:2)
Hasn't AMD been better already in the early naughts (at least) with their 32-bit Athlons, or was there time in between when Intel managed to catch up? I bought my 1st AMD in 02, and didn't really follow after, but was always under that impression. After that I've bought used PC's, and they have been quite random...
Re: Did I miss the merger news? (Score:5, Informative)
Up until the P4 era, everything revolved around clock speed so intel went big on clock speed. You could say that their clock speed was bursting.
AMD decided to go big on execution units. 3 ALU's completely symmetric, an AGU (address generation unit), and an FPU.
The Athlon became King.
And it remained King until Intel scrapped the P4, introducing the first Core chips, based on the older P3 designs, at the same time as they invented a novel transistor arrangement that gave them a pretty big leap in transistor density (they then started the "lying" about process size) over their competitors. They rode that advantage down to "14nm" and then slammed hard into the yields problem made worse by their novel trigates.
AMD solved the yields problem, you could say in a novel way, but actually they just took a page from Intel who used chiplettes themselves to compete with AMD's first dual-cores.
Re: Did I miss the merger news? (Score:2)
Thanks for the info :)
Re: (Score:1)
Not quite accurate. Intel's yield problems for their 10 nm node were due to challenging patterning processes and not committing to EUV (which TSMC did); not specifically by their transistor design (which by that time everyone moved to some kind of finfet architecture). AMD "solved" their yield problem by divesting their manufacturing entirely (Global Foundries) and then moving to TSMC for production.
Re: Did I miss the merger news? (Score:4, Interesting)
For much of the naughts it was a horse race. This week Intel by a nose, next week AMD.
Then Intel made the 1st big fumble with Itanic (Itanium) while AMD did x86_64 (which Intel now licenses). Apparently Intel's best comeback move was trying to sandbag AMD by rigging their compiler to detect non-Intel CPUS and deliberately tank performance.
Lately, for every time AMD stumbles, Intel falls face first in the mud.
The REAL story: Once upon a time, Intel tripped over a success in the making and once they picked themselves up, they made bank and their PR flunkies convinced the world that they meant to do that. The x86 was born (It was supposed to be a channel controller from a way over-complicated CPU that is long gone).
They rode that wave for 20 years virtually unopposed. Then they got real competition. The market allowed them a few stumbles because it still believed that Intel meant to do the x86 and thought they would soon do it again. They never did. Their management doesn't realize that the market is realizing that, so they think they can still get away with the Intel tax.
Re: Did I miss the merger news? (Score:2)
Thanks for you too, sounds familiar :)
Re: (Score:2)
The x86 actually was the logical continuation of Intel's x80 line, which in turn was a refinement of the Datapoint Corp. - designed early microprocessor known as 8008. x86 was since the beginning developed as a CPU line. It might be true that i8086 was not the first x86 CPU Intel worked on, and it might have been born from the debris of another, more ambitious, ultimately failed x86 CPU project, but still, i8086 was created as a CPU, to allow Intel to have something to fight the competition from Motorola, Z
Re: (Score:2)
Yes, the 8086 is a descendant of the 8008 and 8080. It was also originally intended to be the channel processor for the iAPX 432. Then, engineers noticed that things ran a lot faster if the program was compiled for and run on the 8086. Meanwhile, Zilog had long one-upped the 8080 with the Z80.
iAPX 432 flopped, though Intel salvaged some of that effort by removing the heavyweight CISC part and making the rest into the i960. Then Intel gave up the i960 to DEC to get in to the StrongArm, which they abandoned a
Re: (Score:3)
AMD did all right, but was the 'knock off' alternative. This was up until Intel went netburst, and AMD was able to capitalize on much more efficient product. Intel also went Itanium, and AMD went 'amd64', which was *much* more desired in the industry. They also went NUMA with integrated memory controllers.
On the desktop, Intel got more competitve with Penryn, and on the desktop pretty much closed the gap with Nehalem in server space. Then AMD went Bulldozer and Intel enjoyed a long period of being prett
Re: Did I miss the merger news? (Score:5, Informative)
"Let the specialists handle the process". That makes no sense, Intel is one of the three companies left that has leading edge tech (TSMC/Intel/Samsung). Intel IS a specialist in this area. Also Intel design already outsources to TSMC/Samsung; Lunar Lake/Arrow Lake makes lots of use of non-Intel Silicon, and early performance figures look very good (as they are using the TSMC tech like AMD/Apple). With that in mind, Intel is already planning on moving Panther Lake back to Intel process - which should tell you something about what they expect of their future process (otherwise they could just continue stick with TSMC).
Long ago AMD split into AMD-design and Global Foundries. GloFo failed to keep up with the leading process edge, leading to lower competition and choice. It does the world no good if there are fewer and fewer players in the leading edge semiconductor space; competition is good for consumers. TSMC alone cannot fulfil the total demand for leading edge process.
Re: (Score:2)
Intel is one of the three companies left that has leading edge tech
That would be true if Intel had leading edge tech, but they do not. They have repeatedly failed at bringing out new processes and had to abandon them, and they have clearly had a lithography failure causing defects in their recent chips meaning that they are in the middle of failing at a new process again.
Re: (Score:1)
Intel is roughly on par with Samsung and about a half node-node behind TSMC currently. Yes they had lithography challenges on their 10 nm node, but that is old news; they have since moved to utilizing EUV on their current 4/3 nm node and are successfully producing high volume products with it (Meteor Lake, Sierra Forest). If they are in the middle of failing a process, as you claim, then we will learn of that if Intel delays their Arrow Lake i5 sku or Panther Lake, which is the next set of products on their
Re: (Score:2)
Re: (Score:3)
Re: (Score:3)
I suspect that he did see the writing on the wall and got vocal about it and then when nobody agreed with him then it was time to leave.
This is a sign of a company going downhill - the blame game starts and those with some sense bails out.
As of today the market cap of Intel is about 85Bn and AMD is 242Bn. Meanwhile Nvidia is 3170Bn. So if Intel tanks - what will happen? The pucker factor in the computer industry would be immense.
At my workplace there was a decision some months ago to go for AMD servers inst
Re: (Score:2)
nvidia negotiating to get an amd64 ISA license so that they can turn the fabs back on.
nvidia probably isnt looking to buy anyones fabs, let alone intels. TSMC and NV CEOs are literal cousins or whatever.
Re: (Score:3)
No, Intel is jumping on the AI bandwagon, and spending a boat load of money to do so, but is coming up short, late and underwhelming compared to the competition. Example on the client front, when Microsoft declared AI branding and standards for PCs, Qualcomm spent the effort to get there first, AMD is not far behind, and Intel is going to be there eventually. They all spent money and effort to do so, but Intel was last to the party. It might not matter, but to the extent it could have, Intel was last. I
Re: (Score:2)
It is Nvidia's CUDA and high performance tensor cores that are largely driving their success. Intel (AVX), and AMD have hardware that should be able to handle AI workloads, but there is a big gap in software support. Operating systems should provide a hardware abstraction layer so that an API call to the OS handles AI workloads on a variety of hardware.
Intel dropped the ball by not succeeding in driving the software e
Re: (Score:2)
AVX-512 does lack software support, because it is strictly inferior to even using an integrated gpu.
The CPU is behind a set of caches that are optimized for latency.
GPU's have cache hierarchies optimized for bandwidth.
Its not a surmountable difference. A dedicated chip with a dedicated memory controller will always destroy a general purpose cpu in bandwidth, so much so that completely different kinds of memory chips need to be used
Things get better over time of course. At some point a gpu placed
Re: (Score:2)
This is from the land of coulda, woulda, shoulda, but if I were at Intel I would be asking why not and where are we on getting a hardware abstraction layer and api into Linux and WIndows.
The alternative is watching the world move on to Nvidia and Arm. I like competition, but right now, Nvidia has very little of t
Re: (Score:2)
The entire "bridge" that co-processor sockets were on is gone. You've probably heard of "north bridge" and "south bridge." The old north bridge is gone from motherboards. What used to be there was the memory controller and coprocessor. Both of those are now under the lid of the cpu. The "new" north bridge is the faster/wider PCI and the south bridge is the slower/narrower PCI.
Its still evolving. Intel
Re: (Score:2)
EVERYONE except NVidia is coming up short, late and underwhelming compared to the competition, i.e. NVidia.
Re: (Score:2)
Intel has jumped on the Ai bandwagon. It just isn't making them any money yet, and may never.
Gaudi, Loihi, and Ponte Vecchio/DGx have all failed to move hardware and make money, at least up to this point.
They have a *major* problem with the last 2 gens (Score:2)
That's going to mean a few years of either uncompetitive products or barely functional ones or both. That can quickly lead to a death spiral. Especially with Intel facing competition from Arm on Windows and Macs using their own CPUs.
Re: (Score:2)
The situation is that Intel thought that 'cheating' on their designs to achieve better benchmark scores was a smart move. It has led to millions (billions?) of CPUs becoming unreliable and failing which has induced a fear of having to deal with the liability that such shenanigans have made inevitable.
In other words, no matter how you spin it, Intel is going to go through some VERY rough times. They deserve to fail.
Excess Workforce. (Score:1)
I would not buy anything Intel right now (Score:4, Insightful)
I'm sure they'll still be money for stock buy backs and C-level bonuses though.
Re: (Score:3, Funny)
We need Intel to survive (Score:2)
Back a few years, AMD was a mess and Intel had no serious competition in the x86 space. If Intel flounders, AMD will have no serious competition in the x86 space. We need these two to keep each other in check.
Re: (Score:2)
Re: (Score:1)
"arm is winning" is computing's version of "fusion is 10 years away". even superior hardware isn't going to supplant x86. replacing x86 is a software problem. not really a hardware one. and the value of updating the software for arm is not there for end users.