back to article AMD chief architect Jim Keller quits chipmaker – again

AMD's top microprocessor architect Jim Keller has walked from the chip vendor, just months after the company unveiled a shift in strategy to reinvest in high-end FX CPUs. Keller quit on Friday, having returned to AMD in 2012 after most recently working at Apple. It's unclear why the industry veteran has walked away from the …

  1. ecofeco Silver badge


    When you make the kind of money he makes, you really don't need a full time for years and years.

    1. Anonymous Coward
      Anonymous Coward

      Re: Mystery?

      But that was true before he left Apple and went to AMD. Not everyone is motivated by money, he may be motivated by the challenge or by doing something novel. When something more novel comes along, he follows it. Be interesting to see where he lands next...AFAIK he's never been involved in a startup, maybe that's his next challenge.

      1. mi1400

        Re: Mystery?

        Not a mystery... Perhaps they read and gave serious thought to my last post...

        AMD ... Intel's intentionally brewed Anticompetitiveness Manipulation Decoy.

        An excerpt from my Dec 2, 2007 post...

        I am personally an AMD Fanatic but AMD's lately Apple like reasoning have made me so disappointed, like ... Intel-like-Quadsicore approach is not an "Angelic" approach (recently Intel has spoken out on record that it was near impossible for AMD to solder two X2 like Intel beacuse of integrated HyperTransport etc) ... then AMD's stance that Phenom is also not an Angelic-way to empower notebook. I usyually say AMD has developed taste for eating bread crums from the floor. and perhaps Intel-AMD's 1999 secret agrement is hint/factor of why SSE4a lags SSE4.1 hence keeping AMD walking leaning-head behind Intel and still thrusting ancient 4000+ X2 bases notebooks on Enthusiasts, this also confirmed their TDP charm is broken and was a myth. Intel has silently opened a new front in notebooks and AMD is still not getting out of Angel-ic/Apple-ic phobia. Rememebr when AMD had upper hand in Opetron Vs Xeon they put a patetion (Opteron Xeon Duel) on thier site for fans to sign in convincing Intel to show up with its Xeon to get humiliated infront of whole world and the Flash video on did just that, now when the tables are turned Intel has not posted such patetion thing to call for it fans join the AMD-humiliation competition. I think we should abandon fanfare and become wise-customers thinking from brain not heart.


        Does anyone remember the AMD's MegaHertzigovinia (MegaHz-igovinia) pun to Intel too ... anyone share an article on secret enslavement agreement where by AMD is bound to keep its SSE throttled behind Intel's SSE or was it related to floating point calculation throttling !?! ... anyway it is the truw face of AMD and just an Intel's Decoy to keep anti-competitive monitoring courts shut up .. saying look we have shared the patents staisfying FRAND clauses and see AMD where camera focusing n corner of court AMD eating something from floor with diaper all filled with pee and feases .. looking toward Intel with all the gratitude to help AMD live no matter in this condition.

        1. Anonymous Coward
          Anonymous Coward

          Re: Mystery?

          Amanfrommars1 is posting under a new ID?

        2. Zmodem

          Re: Mystery?

          AMD are competitive, the AM3+ just needs a new bridge with pci-e 3.0 and faster ram

          core speed is a different thing, compared to game benchmarks that give you a score based on all speeds

          download fractal generator that uses cpu only

          download settings

          the fractal :: took 3m 47.61s on a athlon II x250 2core

          there are plenty of render settings etc, to mess about with

          AMD were alot faster then intel from the mid 90s, until intel sandy bridge, all the loosers still brought intel for 3x the cost to get less work done

      2. E 2

        Re: Mystery?

        It would be novel indeed if AMD sold (in volume) a CPU that was competitive in the datacentre!

        1. Zmodem

          Re: Mystery?

          they would be if AMD used ddr4, mounting server images to ramdisks or using virtual machines, need faster ram

          and pci-e 3.0 is better for thunderbolt/ethernet links

  2. PleebSmash

    Zen: good?

    Skylake looks terrible. Barely any IPC increase for desktop users, and no mainstream core increase. Overclocking is better than the messed up Haswell and Broadwell... but even some Sandy Bridge users won't need the upgrade. On mobile battery life and graphics performance is improved somewhat, but GT3 and GT4 parts are hard or impossible to find yet.

    AMD has hinted at a 40% instructions per clock increase for Zen. They could actually deliver that since the Bulldozer cores were so bad compared to Intel. They already deliver good price/performance in most segments, especially with low-cost APUs. Maybe Zen will make 2016 the year of AMD.

    As if it wasn't already hard enough for AMD, their potential savior/flop may be scheduled for release in Oct. 2016, a full 13 months from now and long after the huge back-to-school season. Zen will be competing against Kaby Lake rather than Skylake, for what that's worth.

    1. BinkyTheMagicPaperclip Silver badge

      Re: Zen: good?

      Nope, people are not comparing like with like. All the complaints so far are comparing Skylake with Haswell-E. The DMI link between the processor and chipset bandwidth has doubled, and the chips are finally shipping with TSX that works (unless you were running Haswell-EX previously).

      40% isn't enough for AMD to be competitive, and they're not serious about their HSA architecture, because it's not included in their higher end processors.

      I agree, by the time Zen comes out both Skylake-E (fast, lots of cores) and Kaby Lake (who knows!) will be out, and it'll still be nowhere.

      1. PleebSmash

        Re: Zen: good?

        Zen will include 8 core desktop chips. Not the 4-module, "8 core" clustered multithreading chips from Bulldozer, but 8 hopefully physical cores with simultaneous multithreading, like Intel. People still bought Bulldozer chips even with AMD's failure to compete with Intel.

        AMD APUs for laptops have been competitive with Intel, because they offer decent graphics performance at a low cost. Skylake may be bridging the graphics gap with AMD/Carrizo, but Skylake GT3 and GT4 parts are either missing or expensive. If AMD can put out a new mobile APU with a big boost in IPC and a graphics boost, it could win against Intel in any machine without a dedicated graphics card. The ultimate win would be getting Apple to use its chips instead of Core i5 and Core i7 in Macbook Air and Pro, since it would be an acknowledgment that they tick all the boxes: cost, low power/heat, high CPU performance, high GPU performance. It would help counteract this too.

  3. Anonymous Coward

    Evidence that MSFT is really buying AMD?

    Wasn't the whole point of Zen to be a sort of architecture convergence? Proper integration of CPU with the ATI GPUs AMD bought... and with the CPU being either an x64 or high-end ARM part with packages pin compatible regardless of CPU architecture. Presumably it was that broad strategy which made Jim "k7k8A4A5" Keller the *perfect* choice for running the show.

    If the Wintel Corp., really is sniffing about AMD, it'll presumably be in a desperate bid to crush the ARM in the server/desktop aspect... obviously abhorrent to both the Win' and the 'tel... Hence JK's been completely wasting his time these last few years.

    A tenner on AMD's high end ARM chips... and AMD... both being gone (the way of Nokia) within two years.

  4. Ragequit

    Well that blows..

    I can only hope that means the design for the Zen is effectively finished and the only thing holding back a shipping product is getting their 14nm Fab ramped.

    @AC - AMD already announced it has abandoned the ARM play in favor of going all in with the Zen core for desktop, mobile, and potentially server as well.

    That said I hope him leaving isn't premature and in reaction to that project being scuttled.

    1. PleebSmash
    2. h4rm0ny

      Re: Well that blows..

      With the timescales that CPUs work to, the architecture part will be finished now (will have been for a little while) and the next parts are to turn it into a real and usable piece of hardware - not a small task. Jim Keller may be leaving because he's too expensive to just keep around with AMD's limited funds now that his stage is done, he may be leaving because he wants to find a new challenge elsewhere or simply have some time off (I imagine heading up a new CPU architecture is a pretty exhausting job). In either case, he will have delivered them all the architecture for Zen by this point. I would have expected him to stay on and continue working on successor chips personally, but then Jim Keller has never really stuck around that long at one company - check his work history. Hopefully if Zen is a success he will go back to AMD to do more for them.

  5. Anonymous Coward
    Anonymous Coward

    AMD please give us something to rival Intel on performance per watt and make it cheaper.

  6. El_Fev

    AMD's Problem

    Is that their CPU's are always compared against CPU@s that are a couple of process shrinks ahead of them. One wonders if both CPUS were produced on 14nm at the same time, what the results would be.

    1. BinkyTheMagicPaperclip Silver badge

      Re: AMD's Problem

      Intel would still win. Check out Sandy Bridge vs AMD FX benchmarks.

    2. Anonymous Coward
      Anonymous Coward

      Re: AMD's Problem

      AMD has many challenges but the days of substantial gains from a reduced node size are long gone for everyone. A node shrink can lower power consumption a bit and increase density, but the increased density makes cooling more difficult.

      AMD's latest Carrizo chips are on 28Nm even though they were planned to be on 20Nm but none of the Fabs were able to get decent 20Nm yields. Despite staying on 28Nm AMD was able to make substantial improvements in power reduction and IPC, quite comparable to Intel dropping a node size. Any process below 32Nm yields very small performance gains. Zen will be on 16Nm FinFET which is a good step forward on many fronts.

      As far as Jim Keller leaving, he did this before the K8 was released also. He's an interesting guy who gets bored easily unless he has a new challenge. After Zen comes Zen plus an evolution of Zen that is already in full development. Thus AMD is set for several years so it's no surprise that Jim has left for another challenge. He may return to AMD some day in the future. One never knows for sure. Either way Zen is a great new architecture core and it will deliver excellent performance.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like