Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 


Forgot your password?
Close
typodupeerror
×
Intel Hardware Technology

Intel Reorganizes In Wake of 7nm Woes; Chief Engineering Officer To Depart (anandtech.com) 119

Posted by BeauHD from the cause-and-effect dept.
FallOutBoyTonto shares a report from AnandTech: Coming in the wake of last week's disclosure that their 7nm yields are roughly a full year behind schedule, Intel this afternoon has announced that they are reorganizing the technology side of the company. Key to this change is that Intel is breaking up its monolithic Technology, Systems Architecture and Client Group (TSCG) into several smaller groups, all of which will report directly to CEO Bob Swan. Meanwhile Intel's chief engineering officer, Dr. Murthy Renduchintala, who had been leading the TSCG, will be departing the company at the end of next week. The reorganization is effective immediately.

As a result of this reorganization, TSCG is being broken up into five groups focusing on manufacturing and architecture. These are:

- Technology Development: Focused on developing next-generation process nodes. Led by Dr. Ann Kelleher.
- Manufacturing and Operations: Focused on ramping current process nodes and building out new fab capacity. Led by Keyvan Esfarjani.
- Design Engineering: A recently-created group responsible for Intel's technology manufacturing and platform engineering. Led on an interim basis by Josh Walden while Intel searches for a permanent leader.
- Architecture, Software and Graphics: Developing Intel's architectures and associated software stacks. Led by Raja Koduri (continuing).
- Supply Chain: Handling Intel's supply chain and relationships with important suppliers. Led by Dr. Randhir Thakur (continuing).
This discussion has been archived. No new comments can be posted.

Intel Reorganizes In Wake of 7nm Woes; Chief Engineering Officer To Depart More

Intel Reorganizes In Wake of 7nm Woes; Chief Engineering Officer To Depart

Comments Filter:
  • As their current tech is doomed

    • Re:Also need a manager for ARM licensing (Score:5, Funny)

      by K. S. Kyosuke ( 729550 ) on Monday July 27, 2020 @08:58PM (#60337619)
      Don't these Intel people have some kind of oath? Like, "first, do no ARM"?

      • Re:Also need a manager for ARM licensing (Score:5, Informative)

        by Guspaz ( 556486 ) on Monday July 27, 2020 @09:32PM (#60337693)

        Intel was at one point a major player in selling ARM CPUs with their StrongARM and XScale products. They were found in many PDAs at the time. This lasted from 1997 (when Intel bought the IP from DEC) until 2006 (when Intel sold the IP to Marvell). Intel held on to their ARM license (and still has it to this day) and continued to make some ARM chips for embedded purposes after that (they released one in 2007), but it seems that they stopped shortly after that.

        • Intel was at one point a major player in selling ARM CPUs with their StrongARM and XScale products.

          Then the other ARM implementations got to be an order of magnitude more power-efficient than XScale, and Intel couldn't figure out how to do that, so they sold it. It was the fastest ARM for a moment, but then ARM went in another direction, one in which Intel couldn't follow.

          /former owner iPaq H2215 (PXA255)

        • I'm still looking for a DEC DNARD https://www.vaxbarn.com/index.... [vaxbarn.com]

        • Re: (Score:3)

          by klashn ( 1323433 )
          I did validation and testing on Xscale and the "Programmable State Machine 2" and ARM Coprocessor bus. It was a great entry as my first job out of college. I have a router in which the IXP 425 was included. (codename Westport). Fun times.

      • Re: (Score:1)

        by Anonymous Coward
        There was once a time when everyone slobbered at the notion that Intel might use their industry-leading lithography process to churn our ARM chips. They would have been the smallest, fastest and most energy efficient ARM chips of 2015. If Intel tried it now, everyone would just shrug and buy TSMC's better version.

        • Re: (Score:3, Insightful)

          by 0111 1110 ( 518466 )

          Is there any reason why they would want to do that? I mean aside from just dumping the whole desktop market and going pure mobile which I would hate them for. I get that Slashdot has always been pro AMD but believe me you do not want AMD to have no competition.

          What Apple just did was very stupid and it will bite them in the ass eventually. The last high performance macbooks are going to be worth a lot of money and no one is going to want to buy the ARM ones except Apple fanboys who have never owned anything

          • Comment removed based on user account deletion

            • I don't think there's any question that they can execute a transition -- of course they can. (Arguably they've made this transition already, since iOS/iPadOS are essentially just pared-down versions of MacOS with essentially the same kernel, so they'll just need to write or port a few drivers and libraries.)

              But the question to me is whether this is ultimately helpful or harmful for growth. It's probably a safe bet that Apple silicon will increase their margin, but I'm skeptical that it will drive market s

          • Re: (Score:3)

            by sjames ( 1099 )

            You may have missed a memo or two. ARM based machines are getting bigger and faster every day. ARM is very much current tech.

            • Re: (Score:2)

              by cb88 ( 1410145 )
              They also draw just as much power as thier x86 counterparts... because the ISA decoder probably makes up 1% of the power budget.

              • Re: (Score:2)

                by sjames ( 1099 )

                Actually, they don't draw as much power.

                • Re: (Score:1)

                  by cb88 ( 1410145 )
                  Check any performant implementation of arm, compare it against a similarly performant implementation on x86 on the same node... guess what ARM does not win.

                  Most the the ARM servers out there are many thread machines that are actually competing with Sparc not x86....

            • ARM architecture provides no benefits to the processor. ARM is not more efficient or better in any regard than x86. Hell x86 processors haven't used x86 instructions internally since the original pentium.

              • Re: (Score:2)

                by sjames ( 1099 )

                So how do you explain the steady improvements in ARM (to the point that they're encroaching on the desktop market)?

                • Do you have any evidence that any ARM implementation can compete with the fastest desktop CPUs in terms of performance? The problem is it is a tough thing to benchmark with any real world applications. Has anyone tried?

                  • Re: (Score:2)

                    by sjames ( 1099 )

                    By the same token do you have any evidence it can't? I never made the claim you want proof of, I simply claimed that it is very much current tech and it continues to improve, and is encroaching on the desktop market. It's easy to forget in a world of rabid gamer machines and geeks seeing how fast their workstation can compile the Linux kernel that there's a largish market segment out there who mostly browse the web, email, and do a bit of word processing. When was the last time Intel could manage that with

                    • I agree it is current tech. Of course it is. But it was designed for low power mobile devices. Whether or not it could scale to be competitive with desktop CPUs that were designed for performance alone is entirely speculative.

                    • Re: (Score:2)

                      by sjames ( 1099 )

                      It has already scaled enough to at least compete for the low end desktop. Agreed that beyond that is speculative. Apple may help answer some questions there soon-ish.

    • Re:Also need a manager for ARM licensing (Score:5, Interesting)

      by DrMrLordX ( 559371 ) on Tuesday July 28, 2020 @04:37AM (#60338183)

      That's actually backwards. Intel has designs in-house that could make them competitive today, if they had a process on which to fab them. Golden Cove, which should be out right now, would probably be faster per-clock/per-core than any reference ARM design (A77 or A78) and any AMD design (Zen3 anyway).

      So where is the Golden Cove?

      Intel says they can have it ready as Alder Lake-S by Q4 2021 on their 10nm process. With the current shake-up, it's not clear how that will unfold in the future. It will take a long time to port those designs over to TSMC nodes, and unless they bribe TSMC with foundry capacity, they'll be playing also-ran to TSMC's existing customers - including AMD.

    • None of their yield issues have anything to do with the ISA. Changing from x86 to ARM and doing nothing else would not fix their problems. I always find it amusing how the lowest information types are most likely to be the fanboys.

  • Sure fire the head of tech when their main problem is sales and marketing constantly disabling features for lower grade cpus just to force people into i7s and i9s they don't really need for gaming.

    • Nope. Their main problem is that they no longer have superior process technology, and that's been their only performance advantage over AMD besides cheating on speculative execution in ways that create security defects.

      • Bingo!

      • Even their speculative execution shortcuts weren't good enough (see: 10900k).

      • It was once said Intel only got into the memory game because the other players were slow. This was in Pentium Cache days with cpu risers. Intel should have never lost 'best' status, and how TSMC got there is a mystery. A bit like 4G perhaps - eye off the ball in order to pump quarterly profits. Then you see Intel NOT repairing speculative execution errors in a reasonable timeframe. They lost their bet - but like Hertz and Boeing (and Kodak) values seem immune to rational perceptions. And it looks like Apple

    • Re: (Score:1)

      by Anonymous Coward

      Sure fire the head of tech ....

      Intel is following the same playbook as many organizations, you fire the "manager" since you can't fire all of the people that are actually responsible for failing to deliver. It is true that the manager is ultimately responsible ("the buck stops here"), but if the manager kept the CEO and board aware that the team was not delivering than they are simply the scapegoat to show the market that they are doing something (anything).

  • Soo things will be even later? (Score:3)

    by gweihir ( 88907 ) on Monday July 27, 2020 @09:02PM (#60337631)

    Luckily, I stopped buying their overpriced insecure crap a long time ago.

    • Does anyone still frequently upgrade their CPUs? Seems barely worth doing these days. Maybe some rich techies just because they can. When I am finally ready to upgrade again I will decide between AMD and Intel based on technical merits as always. Hopefully AMD will catch up with Intel soon so that we can have a level playing field and real competition. An Intel dominated market is bad for customers, but so is an AMD dominated one.

      • Catch up with Intel soon? Yeah, I can tell you haven't been buying chips lately.

        They surpassed Intel in most performance metrics last year with Zen2.

        • Re: (Score:2)

          by gweihir ( 88907 )

          And, to my surprise, for some applications, they were superior all along. Try playing some games at 15FPS on AMD (works well) vs. Intel (choppy and unpleasant) or try streaming from the same PC.

        • No they haven't. They have surpassed in performance per dollar terms, but they haven't actually topped them in raw performance. I have high hopes for Zen3 to close this gap. But right now if raw single threaded brute force is important to you, Intel is the only game in town. If your problem is embarrassingly parallel then AMD is more than happy to help.

          • Oh really? Tried beating a 3990WX with a 10980XE lately? Give me a break.

            • Yeah, pretty much every Core i7 beats it on single thread performance. Or did you get so fanboi crazed that you didn't even make it to the 3rd sentence in my 4 sentence post?

              Give me a break.

              If you are able to make sense, follow a conversation, and realise that AMD still has an IPC gap to close you wouldn't need a break.

        • Re: (Score:2)

          by AmiMoJo ( 196126 )

          And at a much lower price (both CPU and motherboard) and with extra features like ECC support for free.

        • For servers and workstations, you'd be correct. For gamers and most productivity tasks, then no. Games and productivity tasks (and many others), single threaded performance is everything, and while AMD has made large gains, they still lag here. They are still behind by about 15% (on average, some tests are closer to 34%) in IPC (Instructions per Clock), and they can't seem to get their clock frequencies up (4.7GHz vs 5.2GHz) which hits them for another 11% disadvantage.

          Considering how far Zen2 came, most

      • Seems barely worth doing these days.

        For posting on Slashdot, no it's pointless. On the other hand upgrading for better framerate in games, better processing of images or videos, or faster compile times absolutely. For some people get paid by work output not by the hour, for those people a new CPU is also a tax deduction. For everyone else, there's enthusiasts.

      • Upgrading [i]because[/i] of your CPU isn't what it used to be; but it's still somewhere between "ubiquitious" and "really common" for companies and similar who operate substantial numbers of computers to refresh more or less on a schedule. Especially with laptops, that get beaten around a bit, fleet-level reliability tends to start degrading eventually, as do batteries, the chassis components of systems used by the most abusive and/or neglectful users, etc.

        What's more of a concern for Intel isn't so much

        • Do you think these serial upgraders will continue to upgrade when there is no measurable performance increase at all in any benchmark whatsover? Because we are almost there. We are at the need some fundamental scientific/engineering breathroughs point now We've hit the wall..

          Intel is definitely behind right now on the PCIe4 issue no question. I am waiting to buy my first SSD until the Samsung 980 Pro is released and obviously I am going to want to get PCIe4 for that thing eventually but I looked at some ben

  • Intel had beaten AMD and the other chip makers and the struggle was over ;)
  • With the combo of the cloud market commoditizing x86 (who cares if it's Intel inside), flat / down pc sales, and having long ago ceded the low end to ARM, they're dead. Just like AT&T, sun and IBM before them, inertia will carry them a long time, but they're dead nonetheless.
    • Comment removed based on user account deletion

      • Re: (Score:2)

        by ELCouz ( 1338259 )

        .. last review I saw of the new 10900k had it running like a space heater for less performance so they really need a miracle to compete.

        Sounds like Prescott 2.0 to me.

      • This is one time their market rigging and crippler compiler isn't gonna save their butts, last review I saw of the new 10900k had it running like a space heater for less performance so they really need a miracle to compete.

        Space heater- yes. But then again- 5.3Ghz. That ain't easy, particularly at their larger process size.
        As for performance? Tsk, tsk.
        Not this stupid fucking argument again.
        Approximately 0% of the fucking market gives a fuck what the aggregate performance of all 875 cores of your AMD part do, because approximately 0% of workloads on all PCs in existence utilize that.
        End of the day, the 10900k outperforms every single Ryzen 9 in tasks that utilize up to 10 cores.
        I know you guys hate to hear it, but it's a

        • By that idiot logic, the 10900k is worse than the 10700k or 9900k. Also "tasks that utilize up to 10 cores" hahah, how many of those do you imagine there are?

          • I think you missed the point.

            For tasks that can use more than 10 cores (how many of these do you ever do?), then the higher core count AMD parts will perform better than the Intel counterparts. However, for all the other tasks, which includes 99.9% of what people use desktops for, including gaming, then the Intel parts will perform better.

            And yes, the 10900k will perform better than the 10700k and the 9900k.

          • Idiot logic indeed.
            How many tasks do I imagine utilize up to 10 cores? Well, all of them of course.
            The better question is, what tasks do you imaging use more, since that's the only place AMD shines.
        • Comment removed based on user account deletion
          • Being Intel parts are superior per-core at 12nm, I too hope for them getting out of the process and utilizing TSMC fabs.
            Imagine how non-competitive AMD will be once they're not 4 process steps ahead of Intel ;)
            • Comment removed based on user account deletion

              • Uhhh...yeah about that just FYI nearly all sites use a clean windows install to do their benches which means you do NOT see the performance impact of spectre/meltdown

                First off, that's flat-out false.
                Nearly all benchmarks have relevant mitigations installed, and they clearly state as such. Do you think they're batting for team blue or something?
                Current and previous Intels are not susceptible to meltdown. (9th and 10th generation Core architecture).
                And finally, Spectre patches apply to AMD as well.
                Getting that much wrong in the first sentence means you need to turn in your shill card, because you fucking suck at it.

                At the end of the day the Core arch has its roots in the Pentium Pro from 1995 [wikipedia.org] and there is just no way to fix all the speculation bugs in that design without starting over, there are just too many issues baked too deep.

                This is almost too stupid to address.
                Your car has it

        • Plus Intel has its Meltdown technology which AMD doesn't. That helps it achive its better single thread performance. :-)

          • Model name: Intel(R) Core(TM) i9-9980HK CPU @ 2.40GHz
            Vulnerability Meltdown: Not affected
            Old news. Its better single thread performance comes from a higher IPC, which is a result of a simply superior overall microarchitecture design (in terms of performance at least, since obviously they've been bitten in the ass in terms of sideband vulnerabilities)
            Should Intel ever get its ass in gear process wise, AMD is so toast it isn't even funny. They have to be 4 steps ahead in proce

    • Re: (Score:2)

      by slazzy ( 864185 )
      Intel could survive. If they put all their engineering into building something truly fast, like 100x better than current tech. Otherwise we'll probably see every laptop and desktop/servers with ARM based chips (and whatever Apple will call their version)

      • No they can't. They can't fab anything beyond 14nm reliably. They poured billions into 10nm and 7nm and failed.

  • Manufacturing and yield have been a problem at Intel since Skylake. Finally, someone is held accountable.

  • thank god Murthy's out (Score:1)

    by Anonymous Coward
    He's been the rot at the center of all this. He pushed out Raja and Jim Keller to eliminate his rivals for future promotions. He had gotten promoted well ahead of his talents. Now the trash has been taken out.

    • Raja is still there. He survived. Keller left of his own accord. He could probably still be there if he really wanted it. He didn't.

  • Re-post from last year: [slashdot.org]

    Intel is insufficiently managed, it seems to me. Three of the MANY examples:

    Intel CEO apparently has no technical knowledge [slashdot.org] (Aug. 21, 2019)

    Intel stuck with $1.45 billion fine in Europe for unfair and damaging practices against AMD. [extremetech.com] (June 13, 2014)

    A Slashdot comment of mine from 14 years ago: More Intel employees should say in public what they have told me in private: Intel CEO Paul Otellini is not a competent leader. He lacks social ability. [slashdot.org] (June 9, 2006)

    • Re:Intel: AMAZINGLY insufficient management (Score:5, Insightful)

      by Luckyo ( 1726890 ) on Tuesday July 28, 2020 @02:50AM (#60338061)

      That 1,45 billion EU fine was a great deal for intel. I recall reading that gains due to those actions were in excess of 6 billion just on sales, and that's before the fact that it carried the company over the moment when they had absolutely nothing that was even remotely competitive with AMD.

      Which drove AMD into problems with them having assumed that they'd get to actually sell what they made. And since they couldn't due to intel forbidding OEMs from doing so to retain meaningful access to intel hardware, AMD had to spin off their fab capacity just to survive the crippling debt they had to take to get it. That punishment was not even a proper slap on the wrist. Was was utterly heinous from consumer standpoint, but if I was an intel stock owner at the time, I'd be voting for massive bonuses for people who managed that event and got intel off that easily. The ridiculous amount of money they managed to make for the company when it had pretty much nothing meaningful to sell was astounding.

      • Quote about Intel from the end of the parent comment: "The ridiculous amount of money they managed to make for the company when it had pretty much nothing meaningful to sell was astounding."

        However: Would you want to work for a company like that? The extremely widespread news of that international incident convinced technically-capable people to find work elsewhere.

        It was as though Intel had bought billions of dollars of advertising to convince people: "Don't work for Intel." That's my interpretation

        • Re: (Score:2)

          by Luckyo ( 1726890 )

          "And then intel spent several years reworking their older CPU tech into CPU tech that would go to dominate the market for almost a decade".

          Your hypothesis doesn't really correspond to what happened after the relevant events.

          • "And then Intel spent several years reworking their older CPU tech into CPU tech that would go to dominate the market for almost a decade."

            Parent comment: "Your hypothesis doesn't really correspond to what happened after the relevant events."

            From talking with Intel employees, it does correspond. The internal lack of healthy organization didn't interfere with that initial path to development. Also, AMD wasn't as healthy back then.

            Intel's ability to "dominate the market" was partly by being self-dest

            • Re: (Score:2)

              by Luckyo ( 1726890 )

              I'm guessing you're a bot, as you appear to be utterly unable to comprehend anything I've said so far. To the point where you're basically providing citations of points I made as if they're diverging from my point, while going for a "he caught a cold, and a decade later got hit by a bus, so he clearly died of cold" argument wit the hyper threading bug in 2017 as if its relevant to anticompetitive practices a decade earlier.

    • "Insufficiently managed" is so vague that it's meaningless.

      None of your links have anything to do with the actual problem, which is that Intel's process technology, which was once well ahead of its competitors, is now equal or worse to them.

      • The "actual problem" is that Intel management has not been able to support the development of Intel process technology.

        If top management doesn't understand the issues, it can be very difficult to accomplish technical advancement.

        Actual problem: Intel top management must hire and support people who are extremely technically knowledgeable. People who have that knowledge have chosen to work for AMD. It has been obvious, for years, that Intel would not be an enjoyable place to work. My opinion.

    • A CEO having no technical knowledge isn't a problem. The role of the CEO is to run the business, the technical knowledge should be closer to manufacture unless you risk micromanagement from hampering progress.

      On the other hand being a crap CEO which he is, definitely IS a problem.

      • Quote from the parent comment: "A CEO having no technical knowledge isn't a problem."

        CEOs influence EVERYTHING in a company. One of many examples: Technically capable people likely won't take a position with a company if they don't feel comfortable with the CEO.

        • What has feeling comfortable got to do with technical knowledge. Not sure about you but I feel far more comfortable working for someone who defers to my expertise and is a good people manager than someone who micromanages, second guesses, and overrules because they too have a technical knowledge (and power).

          The two skill-sets have nothing to do with each other, but one of them is absolutely essential to being the head of a company, and that sure as hell isn't technical knowledge.

  • Note that this is a TSMC 5nm competitor (Score:5, Informative)

    by Solandri ( 704621 ) on Tuesday July 28, 2020 @04:07AM (#60338141)
    Unfortunately there's no standard definition for what part of the transistor "x nm" refers to. Intel, TSMC, and Samsung all use the width of different parts. So Intel's 10 nm process is actually denser than TSMC's 7nm [wikipedia.org].
    • Intel 10 nm = 100.76 million transistors/mm^2
    • TSMC 7 nm = 96.5 million transistors/mm^2
    • Samsung 7LPE = 95.3 million transistors/mm^2
    • TSMC 5 nm = 173 million transistors/mm^2 (which implies it's actually 5.2 nm)
    • Intel 7nm = estimated about 200 million transistors/mm^2

    • It would be a competitor to TSMC's 5nm if it worked.

    • Re: (Score:3)

      by AmiMoJo ( 196126 )

      It looks okay on paper but Intel hasn't been producing in volume for even a year yet, where as TSMC has been doing so for nearly 3 years.

      We also have zero info about the yields but you can bet that TSMC is way above Intel at this stage.

  • Intel not the same since Andy Grove left in 2004 (Score:5, Informative)

    by NotTheSame ( 6161704 ) on Tuesday July 28, 2020 @05:49AM (#60338263)

    You could argue that they have been coasting along ever since then. Their market dominance (in 2004) and their ownership of fabs kept them going for a long time, but they have failed to keep pace with their competition.

    TSMC are worth almost double Intel's market capitalization now ($383B vs $209B). Even Nvidia, who farm out their manufacturing to TSMC, are worth more than Intel now ($256B).

    • Re: (Score:3, Informative)

      by etudiant ( 45264 )

      Sadly true.
      What is astonishing is that the key element of Grove's style was paranoia, the fear that others get there first. He was ruthless in his pursuit of leadership.
      Yet Intel in the past decade has accumulated a long list of failures, expensive acquisitions that went bust as well as technology efforts that went nowhere, without any apparent management repercussions or corrective actions.
      The rot this indicated gradually spread to the core manufacturing segment, where it has been evident for several years

      • Grove was there from the beginning. He was an engineer who understood the technology intimately. He had everyone's respect, and used it to get the most out of people. A lot of people make comparisons between Grove and Steve Jobs, with both their management styles and their willingness to take calculated risks with new products.

        When you take somebody like Grove (and Jobs) out of a company, they can't be properly replaced. With no real leadership, in the long-term, things are likely to go bad.

        • Yup, he was the last of the true engineers as CEO. Next was Barrett, and he came from the manufacturing side. Then Otellini come in and the rest is history. Intel stopped focusing on engineering and started focusing on business. Engineering made them great, and Business will slowly bleed them dry. They need a leader who understand the work. Oh well.

  • Key to this change is that Intel is breaking up its monolithic Technology, Systems Architecture and Client Group (TSCG) into several smaller groups, all of which will report directly to CEO Bob Swan

    An old Dilbert. When centralized, distribute. When distributed, centralize.

    That'll fix it!

  • Intel hasn't hired enough diverse candidates. This is likely how AMD came out on top, look at their CEO. We all know diversity is strength after all. Not just engineers, Intel needs more non-engineers, like ethnographers. Even better if they are of diverse origin. If Intel focuses on diversity statistics, everything else will just work out. The year end focal tool should make it impossible to give any diverse employee less than a successful rating. We need to empower these employees. We can't trust
  • If Intel is serious about competing it should get rid of the female head of Technology Development, Dr. Ann Kelleher and find an aggressively creative male tech leader to head the division. Otherwise Intel will find it has merely very competent management of its continuing decline in technology as women are great managers, but lack creative aggression. Name one major technology today that was invented and pushed forward by a female.

Slashdot Top Deals

If all else fails, lower your standards.

Close