+ Reply to Thread
Page 1 of 2 1 2 LastLast
Results 1 to 15 of 17
  1. #1
    Guest

    TLB bug turns AMD's plans upside down big time

    <P>One not so insignificant bug of the K10 architecture seem to be causing more than a few bodies to twist, turn and want to bang their heads over and over again into the closest brick wall. TechReport first reported that the up till now very limited supply of Barcelona processors have been reduced to zero as http//techreport.com/discussions.x/13721AMD has stopped shipping out Barcelonas</A>. The reason is rooted in the Translation Lookaside Buffer (TLB) in the L3 cache of the processor. This errata has forced AMD to cancel all shipments to OEM and channel distributors, only specific customers are getting their orders fulfilled, supercomputer companies TechReport suggests. </P><P>Previous statements, both public and internal, have implied yield and/or frequency issues, but it seems that this TLB bug is the biggest problem of all. AMD revised the Phenom frequencies (read lowered) just before the launch and later even postponed the higher frequency models of Phenom. Once again yield issues and TLB was suggested, but when speaking to Michael Saucier, Desktop Product Marketing Manager, TechReport found out that the http//techreport.com/discussions.x/13724TLB and the lower frequencies are not related</A>. </P>
    <P>This error is found in all B2 revision processors, but is said to be fixed with the B3 revision coming Q1. Large quantities are not expected until late Q1, or even Q2 we've been told, but with the B3 revision, AMD hopes to bump the frequencies, but not even near those that we've previously reported about. Right now there doesn't seem to any kind of solid launch schedule for the coming Phenom models. </P>
    <P>Errata is not uncommon though, and it wasn't that long ago that Microsoft issued a http//www.nordichardware.com/index.php?news=1&amp;action=more&amp;id=6550security patch for Windows that&nbsp;remedied the bug found in the current Intel Core 2 lineup</A>. &nbsp;</P>
    <P>There is a BIOS workaround for the TLB, but it comes with a severe performance penalty of 10-20%, most severe in virtualization environments. The fix simply deactivates the dysfunctional TLB unit but does not disable the L3 cache, which let the processor to continue to work, but at reduced performance. </P>
    <P>Linux users can work around it through a Kernel patch which limits the penalty to around 1%. At first the patch was thought to be limited to selected people, but was later released to the public. With the patch there was a http//techreport.com/discussions.x/13742detailed description of the bug itself</A>;</P>
    <P><EM>"The processor operation to change the accessed or dirty bits of a page translation table entry in the L2 from 0b to 1b may not be atomic. A small window of time exists where other cached operations may cause the stale page translation table entry to be installed in the L3 before the modified copy is returned to the L2. In addition, if a probe for this cache line occurs during this window of time, the processor may not set the accessed or dirty bit and may corrupt data for an unrelated cached operation. The system may experience a machine check event reporting an L3 protocol error has occurred. In this case, the MC4 status register (MSR 0000_0410) will be equal to B2000000_000B0C0F or BA000000_000B0C0F. The MC4 address register (MSR 0000_0412) will be equal to 26h."</EM></P>
    <P>The bug itself causes system hangs during high workloads, like in virtualization systems, but other, more normal, workloads can also trigger the bug and cause the system to hang.&nbsp;You simply have to choose between random hangs&nbsp;or lower performance, you will be able to with an updated version of the OverDrive utility.</P>
    <P>While Opteron has been put on hold to OEMs, Phenom is shipping, even though it contains the very same bug;</P>
    <P><EM>"In fact, AMD knew about the erratum before the Phenom product launch, although its original statements about the issue gave the impression it only affected Phenoms clocked at 2.4GHz or higher."</EM></P>
    <P>If true, someone at AMD has to be sacrificed to the enthusiast gods, because this is just bad. [Some] Current reviews of AMD Phenom processors has to be revised as well as these have northbridges running at higher frequencies than current retail processors. They also lack the performance-reducing BIOS update. </P>
    <P>Last but not least, TechReport has published a short article where they compared performance with and without the TLB patch. Let's just say there's a penalty, but it varies from application to applications, and if Phenom was close to keeping up with Core 2 before, it's looking even less attractive today...</P>
    <P> http//techreport.com/articles.x/13741/1Phenom TLB patch benchmarked</A></P>

  2. #2
    Join Date
    May 2007
    Posts
    414
    Yep it sucks.

    Very low scores. 10% is a minimum penalty.

  3. #3
    Join Date
    Dec 2003
    Posts
    4,510
    Update on the B3 frequencies coming, doesn't look good...

    //Andreas
    "I'm sort of a postmodern vegetarian, I eat meat ironically"
    "You are our Swedish superhero "

    Forget everything, knowledge is your enemy, it's only going to make you second-guess yourself."

  4. #4
    Join Date
    May 2007
    Posts
    414
    Have they lowered the frequencies?

    I don't expect them to get out chips which do factort +2.6GHz by even August.

  5. #5
    Join Date
    Dec 2003
    Posts
    4,510
    2.6GHz should be out Q2, but can still change of course.

    //Andreas
    "I'm sort of a postmodern vegetarian, I eat meat ironically"
    "You are our Swedish superhero "

    Forget everything, knowledge is your enemy, it's only going to make you second-guess yourself."

  6. #6
    Join Date
    May 2007
    Posts
    414
    Yeah but the 9500 goes to 2.7GHz pretty easy anyway so that 2.6GHz will be pointless for anyone who can oc a little on air or has earlier chips.

    Unless it's got more headroom/little quicker/unlocked multi.

    My guess is it'll just be 200MHz HT ref. x 13 simply put. That's a piece of cake on 9500 if they allowed it. The one macci has needs more volts than retail to get 2.7GHz and yet his can do 3.3GHz air.

    Very cool core though compared to any 65nm Kentsfield. Cooler than G0 Q6600 but a little hotter than E6750 stock.

  7. #7
    Join Date
    Dec 2003
    Posts
    4,510
    Well if we're to bring overclocking into it, AMD has pretty much nothing to bring up against C2D

    //Andreas
    "I'm sort of a postmodern vegetarian, I eat meat ironically"
    "You are our Swedish superhero "

    Forget everything, knowledge is your enemy, it's only going to make you second-guess yourself."

  8. #8
    Join Date
    May 2007
    Posts
    414
    Quite obvously C2Q beats Phenom in oc, even more with Penryn.

    C2D loses in perf to Phenom but beats it in oc to make up.

    X2 BE which gets 3500 air performs much worse than C2D at same MHz to compare.

    C2 oc's far the colder you go past zero while AMD chips top out around 0C near 4GHz for X2 and 3.3GHz for X4.

    While C2D/C2Q can go 5.9GHz/5.7GHz below -100C.

    A difference that's always been there since I first saw any westerner do any oc (which incidently was macci). Pentiums oc'd even more than anything new and so did the S939.

  9. #9
    Join Date
    Dec 2003
    Posts
    4,510
    Indeed true, and, at the same time, AMD will never be able to reach Intel's phenomenal yields. Last I heard they were around 95%...

    EDIT: no pun intended :P

    //Andreas
    "I'm sort of a postmodern vegetarian, I eat meat ironically"
    "You are our Swedish superhero "

    Forget everything, knowledge is your enemy, it's only going to make you second-guess yourself."

  10. #10
    Join Date
    May 2007
    Posts
    414
    What do yuo mean by "phenomenal yields" Andreas?

    Are you comparing Penryn to Phenom DC to QC yields?

    I think the deficit would be more than 50% to AMD right now. But not with Intel's first native quad. They will fall quite a lot. Intel had too many problems making a native quad since 2005 at 65nm and 45nm and had to leave it for a much smaller node. Even that only covers the costs mainly rather than the low yields.

  11. #11
    Join Date
    May 2006
    Posts
    608
    Well, putting all speculation aside, I would tell you that if Intel is to be believed, a brisk momentum is to be released into the cpu world, and it hopefully won't take as long as Intel's last great achievement.

    Intel CEO Paul Otellini has stated that Bloomfield is taped out, and this means alot when it comes to the maturity of a design.

    http://www.dailytech.com/article.aspx?newsid=9417

    http://www.dailytech.com/Nehalem+Tap...rticle8927.htm

    Intel's new architecture needs more pins (a new socket) and has an on-die memory controller, as introduced into some older AMD counterparts.

    I honestly hope AMD will have a bold answer.

  12. #12
    Join Date
    Dec 2003
    Posts
    4,510
    Quote Originally Posted by KTE
    What do yuo mean by "phenomenal yields" Andreas?

    Are you comparing Penryn to Phenom DC to QC yields?

    I think the deficit would be more than 50% to AMD right now. But not with Intel's first native quad. They will fall quite a lot. Intel had too many problems making a native quad since 2005 at 65nm and 45nm and had to leave it for a much smaller node. Even that only covers the costs mainly rather than the low yields.
    Intel's overall yield compared to AMD's overall yield.

    Intel's 65nm process was a gold mine, while the 45nm node seems to be working well as well, but there are some architectural difficulties with the QCs at the moment.

    //Andreas
    "I'm sort of a postmodern vegetarian, I eat meat ironically"
    "You are our Swedish superhero "

    Forget everything, knowledge is your enemy, it's only going to make you second-guess yourself."

  13. #13
    Join Date
    May 2007
    Posts
    414
    Well we don't know what the yields for K10h X2 are yet to compare Intel 65nm DC to AMD 65nm DC. They should be far better than k10h X4 if they are separate cores and not fused out quad-cores. These quads are too expensive/pofit damaging/complex to make hence why Intel did not make them but did experiment.

    Penryn QC has bascially the same bug problems as Phenom QC. Obviously speed grades are different though.

    Future designs? All speculation/hype at best yet. I recall January '08 4GHz quads sub 130TW TDP boasts not long ago by some Intel blinded online geeks. Even AMD reps says they have 3GHz on all Phenom chips at the labs and 4GHz with better cooling on some. But that means nothing to me unless you can release a decent and priceerformance worthy chip into retail en masse without platform/support/stability/power/heat problems. So far even 1066 RAM is no go on many MBs with Phenom, let alone CPU speed.

  14. #14
    Join Date
    Dec 2003
    Posts
    4,510
    K10 X2 launch have been moved up till early summer last I heard. Everything is being put into making Barcelona work now.

    //Andreas
    "I'm sort of a postmodern vegetarian, I eat meat ironically"
    "You are our Swedish superhero "

    Forget everything, knowledge is your enemy, it's only going to make you second-guess yourself."

  15. #15
    Join Date
    May 2007
    Posts
    414
    Yep. And there's confusion as to whether it's a separate core or an X4 with two cores disabled (regardless of the pros/cons).

    I think the bug patch tries to make the OS not see the L3 cache. Performance hit is massive though. From 55ns mem latency to 120ns and from 7800 mem read to 4200.
    The bug hasn't appeared yet in my testing. What I thought was the bug turned out to be the AOD utility instead. It will randomly freeze the system (at any frequency). For this reason oc is pretty impossible yet. And the core is very weird (either its the BIOS or CPU, I'm not sure because many BIOS issues we have other boards do not have with same CPU).

    The one time AOD did work the CPU reached 2.9GHz on stock volts. However, it crashed soon after and I don't really know what caused it since 2.75GHz I stability tested for over an hour just fine.
    Too many probems yet for a customer anyway.

Similar Threads

  1. Replies: 0
    Last Post: 09-10-2008, 03:45 AM
  2. Replies: 0
    Last Post: 08-13-2008, 03:34 AM
  3. Replies: 0
    Last Post: 04-27-2007, 02:12 PM
  4. AMD turns to NVIDIA for 4x4 chipset
    By in forum News archive
    Replies: 0
    Last Post: 10-14-2006, 12:09 AM
  5. Replies: 0
    Last Post: 03-02-2006, 05:24 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts