AMD delays faster Phenoms, juggles B3 allocations

We reported earlier today that AMD was denying problems with the upcoming B3 stepping of its quad-core processors, the new rev of the chip expected to squash the TLB erratum and, perhaps more importantly, to cast off the performance hit cause by the erratum workaround. AMD maintains that some B3-step chips are still on track for introduction this quarter, but it has delayed the introduction of faster Phenom 9700 (2.4GHz) and 9900 (2.6GHz) processors into the second quarter of 2008. Here’s AMD’s public statement on the matter:

AMD will now introduce the AMD Phenom 9700 and 9900 models in Q2 2008. This decision was based on OEM input on how AMD should prioritize its next two waves of AMD Phenom processor models. Based on these customer inputs, AMD will continue to prioritize volume-based Phenom products, including the AMD Phenom triple-core processor introduction for consumer and commercial markets this quarter, and now a new energy–efficient 65W AMD Phenom 9000e series processor in this quarter (instead of Q2). AMD Phenom 9700 and 9900 models will immediately follow next quarter.

So AMD will move ahead with triple-core and low-power versions of the Phenom in Q1, but not with higher Phenom speed grades.

An AMD spokesman we questioned insisted this decision was not the result of a delay in B3 silicon, but the result of a reshuffling of priorities for chip allocations. He characterized the ramp of B3 silicon as still "on track," beginning in Q1 and extending into Q2, with higher volumes coming later in the process. Rather than lead with higher speed grades of Phenom desktop processors, AMD will ship the first B3 chips as quad-core Opterons.

Since Opterons command higher prices than their desktop equivalents, this move could make good financial sense for AMD. That’s especially true because the TLB problem is apparently most likely to manifest in server-class workloads involving virtualization, and the firm has curtailed quad-core Opteron shipments to all but a select group of high-volume customers who are likely getting a discount on B2 chips affected by the TLB erratum.

In reality, though, AMD’s change of plans may be prompted by two issues that aren’t entirely related. Technically, the TLB erratum affects all speed grades of AMD’s pre-B3 quad-core processors, regardless of clock frequency. AMD might be skittish about introducing a higher-priced, higher-performance Phenom that isn’t based on B3 silicon because the erratum workaround can sometimes cause substantial performance reductions, and AMD has directed its PC OEM and motherboard partners to enable the workaround by default. The other, more directly related issue is AMD’s ability to deliver chips capable of running at higher clock frequencies in sufficient volumes to make the launch of a product at 2.4GHz or above feasible. The firm is likely counting on newer batches of chips to deliver the improved clock speed tolerance needed to make higher speed grades possible in volume.

One question raised by the delay of the Phenom 9700 and 9900 is the fate of the B3-rev Phenom processors at 2.2GHz and 2.3GHz, originally scheduled for release in Q1 as Phenom models 9550 and 9650. When pressed on this question, AMD would no longer commit to a Q1 release date for these products. The most likely outcome, in our view, is that these products will be delayed into the second quarter alongside the higher Phenom speed grades. In fact, AMD confirms that its triple-core and energy-efficient Phenoms will initially be based on the B2 stepping.

Comments closed
    • pluscard
    • 12 years ago

    AMD makes a statement:

    “Our OEM customers are placing more priority on our energy-efficient and triple-core processors, where the volumes and ability to differentiate products take precedence over the more prestigious, but much lower volume, higher performance quad-core products,” AMD concluded.

    §[<http://www.semiconductor.net/article/CA6522563.html<]§ AMD is clearly going after the big OEM's first.

      • Flying Fox
      • 12 years ago

      As long as they can make money and survive I don’t really care much.

    • Krogoth
    • 12 years ago

    I suppose that on the bright side, it means that current batch of Conroe-based units will have a longer product lifespan.

    It is because Intel has very little reason to push out Nehalem dynasty into the fray if the upcoming mainstream Penyrn-units can outrun K10s by a healthily margin.

    • BIGIRON
    • 12 years ago

    I will always be an AMD fan, I have been since I bought a 486 66mhz cpu in the way back days. AMD will get back on course here shortly they have had times like this before and done just fine. I look at the situation alot like I do sports, sometimes my teams win and sometimes they dont. A good season turns into a dynasty then a few bad years come and you just realize thats they way things are.

    Who would have ever thought AMD would challenge Intel in the first place but they did and they beat them for quite a few years. Now its AMDs turn to take their lumps but I can bet you it wont be forever and some of you who are jumping ship will be back.

    The TLB error isnt that big of a deal it rarely effects anyone, if some of you are so worried about errors occuring why are you running vista? As my mom has always said “this to shall pass” AMD will be fine and it might even join with IBM…how delicious would that be?

      • Heiwashin
      • 12 years ago

      Guess they need to hire a new coach for the next few seasons then.

      • indeego
      • 12 years ago

      Bad teams get new ownersg{<.<}g

    • AMDguy
    • 12 years ago

    AMD is conceding the high end quad core market to Intel, which makes sense. To make money, AMD needs to differentiate its products from its competitor. In this case it means going for the low speed low power OEM market. People really shouldn’t be surprised by this move as AMD has been forced into it before.

    In terms of their tech, AMD is struggling to produce quads with adequate performance. When they try to boost the speed of the K10 memory controller above 2GHz it draws too much power and gets too hot. So if AMD can undervolt their 2.4GHz processors and sell them as energy efficient models, then that’s a way of turning their lemons into lemonade.

    As others have noted, AMD tried to do too much at one time when they did Barcelona: 65nm, quad core, K10, new power saving circuits, and new MC. As with the ATI merger, it will take time for AMD to digest Barcelona and really get it running.

    I’m still an AMD fan, but I know they’ll be limping along for a while longer. Hopefully they’ve directed all their resources at making 45nm a hit.

    • pluscard
    • 12 years ago

    Just a note – when you click on a persons name, the # of posts appears not to be accurate. I clicked on my own name and it shows one post only.

      • eitje
      • 12 years ago

      that’s your number of posts in the forum, not the jazz system which runs the front board.

      god knows what your number would be if the jazz comments were included. 😉 (just kidding!)

    • Fighterpilot
    • 12 years ago

    The NVidia/AMD combo used to be de riguer here at TR when I joined…they seemed a solid pair.
    Its a shame Intel didn’t pick up ATI.They would have had a much better run over the last year or so with Intel’s recources behind them.
    Add to that NVidia buying AMD and there would have been two powerhouse graphics/processor companies instead of the below par combination we see now with AMD/ATI.

      • blastdoor
      • 12 years ago

      Here’s the main argument for why nobody can take over AMD and continue to make x86 processors:

      §[<http://overclockers.com/tips01276/<]§

        • pluscard
        • 12 years ago

        That cross license is re-negotiated from time to time by the two parties. AMD first released x86-64 (originally called amd64) in 2003, and gave that to Intel. That code is now a part of all C2D cpus.

        In anycase, with billions of cash on hand, and some major support from the OEM’s, I don’t see AMD interested in being taken over. They welcome cash infusions from minor players, such as the recent $600+M from the arabs. (AMD’s average selling price is still 1/2 that of Intels, making them the more profitable choice for the OEM’s).

        It’s not clear to me that ATI was going to survive to deliver the 3870 without AMD’s investment. Remember, AMD outlayed $5.4B in cash for the deal.

        AMD’s debt will likely be erased by the antitrust settlement next year, and if they are able to continue to gain share, they should be profitable once they hit a $2B/quarter run rate.

          • newwind
          • 12 years ago

          I am not here to debate the ability of *AMD or *INTL microchips. pluscard you seem to not have correct stock/company information on *AMD. They sold 11% of the company for that 622 million cash infusion and details of the convertible bonds will not be known for at least six more months. On Dec 12th they filed a 8-K for Material Impairments. This is due to the fact that they will have to write down assets associated with ATI as they are not actually worth what they paid. That is there words not mine. The cash on hand is now under 900 million. *AMD does not have billions in cash. They are burning cash rather quickly and as someone that actually is financially routing for *AMD I really do hope they can turn it around.

          I am also not sure any settlement would wipe out the 7.3 billion in debt they currently hold. I am not a lawyer though so I cannot honestly say what sort of judgment is possible or reasonable.

            • pluscard
            • 12 years ago

            On Sept 29, AMD had $1.5B in cash and short term investments per yahoo finance. We know they received the $622M in cash mid quarter. Their cash and short term investments at the begin of Q3 was just 100k more than at the end, so it’s a reasonable bet they still have around $2B in cash, although we’ll find out for sure on Jan 17.

            The material impairment does not affect cash – it’s the difference between the marketvalue of ATI at the time of purchase and today.

            Despite all the bad press, AMD has continued to gain market share, not only since 2003 when the Opteron was introduced, but also thru 2006, and 2007, despite Intels success with the C2D line.

            It may be what’s very important to enthusiasts may have little value to the big OEM’s and the mainstream market.

            • green
            • 12 years ago

            i personally don’t think it’s a reasonable bet they’d have $2B in cash
            that would assume that all divisions of amd made an operating profit
            considering the events of the whole tlb debacle i’d think the computing systems division would be at a loss
            ce, gfx, etc probably did fine but from memory the computing division has been dragging things down for a while
            i would bet the same would be true for q4-07
            but i could be wrong. q4 usually has higher sales due to the holiday period. so jan 17 it is.

    • Stranger
    • 12 years ago

    I’ve heard that the erratum is limited to virtualization can anyone confirm or deni?

    §[<http://www.lostcircuits.com/cpu/amd_phenom/<]§ "What does all of this have to do with Erratum 298 or the bug it describes? The answer is very simple: Only in situations where hardware virtualization is used and there is heavy load on the CPU can there be a race condition where the wrong TLB data may be written to the L3 cache before being updated in the L2 cache." if thats the case I'm just going to get a cheap quad core and screw virtualization.

    • cegras
    • 12 years ago

    3850 is good. Have a 256 version. It ‘jumps’ a bit when moving in a big area – probably to load textures, but performance is great. Can’t believe I actually lived without AA.

      • Corrado
      • 12 years ago

      I’m picking up a 512mb one in the next week or two. Glad everyone seems to agree they are a really good deal.

    • kilkennycat
    • 12 years ago

    AMD has totally loused up with the Barcelona/Phenom introduction. Not only with the TLB problem itself , but with their directives to the motherboard manufacturers… enable the TLB fix by default AND not allow it to have a bypass switch in the BIOS user-settings. So, how long are we going to wait for motherboards to be shipped with BIOS versions that remove this kludge ?? How long to wait before wholesalers and retailers turn over their stocks of boards with the kludge BIOS? User-updates of a motherboard BIOS is still a risky exercise – especially since it would probably have been completely unnecessary if a user-accesible bypass BIOS setting was available. Of course, assuming that the board manufacturer actually gets around to issuing a BIOS with the TLB-kludge removed….

    Compounding this lousy situation is the abysmal compatibility of Phenom quads with existing AM2 motherboards, some of which will end up never having quad-compatible updates… regardless of the TLB problem…, either because of some physical design considerations (e.g: weak power-regulators) or fading BIOS support. See:-

    §[<http://www.tomshardware.com/2007/12/26/phenom_motherboards/<]§ The ATi diversion of focus and resources has been and expensive and a potentially fatal exercise for AMD. Dave Orton and other past-ATi principals must be laughing all the way to the bank. AMD overpaid at least $1billion for an under-performing company.... and borrowed heavily to do so....

      • Sargent Duck
      • 12 years ago

      I agree with almost everything you said, except the ATI part. That’s the only part of AMD right now that’s actually performing. The 38xx cards are good, and ATI’s other divisions are all doing really good, like tv tuners, xbox 360 graphics cards, multimedia chips and so on.

      Of course, I think AMD shouldn’t have bought ATI at all, and spent $5 billion in other things, like building new fabs and R&D, but eh, what do I know?

        • indeego
        • 12 years ago

        The ATI buy was a good buy but they did it at the wrong time.

        (I think nvidia and AMD a much more natural fit, but instead I see nvidia buy AMD since Nvidia is a much better run company.)

        ATI has lost some share, AMD paid a hefty price for them, and revenue hasn’t caught up well over a year after purchase. ATI isn’t getting wins on the high end of the line where they really need the high-end profit margins. We may still yet see nvidia acquire AMD, eventually nvidia and intel are going to butt heads more than either can affordg{<...<}g Have I mentioned I abhor ATI's driversg{

    • indeego
    • 12 years ago

    The phenom name is tainted. Move on, AMDg{<.<}g

      • Lurkmore
      • 12 years ago

      Next one will be called the Ph’uh’nom processor..

    • Sargent Duck
    • 12 years ago

    I realize this is just pure out and out dreaming on my part, but if they somehow put 939 pins on one of those Phenom processors, I’d gladly take an early stepping (B2?) off their hands. As I’m sure many others would.

      • DrDillyBar
      • 12 years ago

      I thought the same thing about Core2 S478.

        • tfp
        • 12 years ago

        Heck I still think the same thing. I would even take a laptop CPU core duo or core2.

        • slot_one
        • 12 years ago

        Where’s my Super Socket 7 Phenom, god DAAMIT? 😛

      • SS4
      • 12 years ago

      Yeah thats definatly dreaming 😛

      Proof is i have one of the first lga 775 board and it wont support c2d, only P4… so yah, even if there were to make a 939 pin ur board would still not support it : /

      • Flying Fox
      • 12 years ago

      How about an Intel-made chipset for those X2’s? Now that’s a dream. 😆

        • derFunkenstein
        • 12 years ago

        Step 1: AMD CPU
        Step 2: Intel chipset
        Step 3: ???
        Step 4: PROFIT!

    • orthogonal
    • 12 years ago

    This is just more anecdotal evidence that says AMD’s 65nm process is a bust. I’ve said it before and I’ll say it again, the whole TLB errata issue was purposely overblown by AMD so they could buy themselves some time to tweak a little more out of the process so that power issues would hopefully yield higher binned parts in greater quantity. The excuse that customers want low clocked/low powered parts is just smoke and mirrors and excuses. Of course there’s a market for higher performing parts, Intel completely owns it, and now AMD is tacitly conceding that market to them since they know they would not be competitive on power in higher clocked parts (140W TDP for the 9900, ouch).

    Frankly, focusing on the low-end (high volume) is foolish. They don’t yet have the cost model or the economy of scales to generate profit in the low end, they need to raise ASP’s and get premium priced products in the market. There’s a direct correlation to 2003-04 when their midrange parts started at $300+ with their profitability. When you have to sell your “best” stuff in the bargain bin, you’re never going to do well.

      • Flying Fox
      • 12 years ago

      What else can they do? They just cannot put any high end parts out. Some revenue is better than no revenue, no?

    • My Johnson
    • 12 years ago

    Jeez, they still can’t ship their next chip? Heads need to roll.

    • zimpdagreene
    • 12 years ago

    Well I have been a Intel man for about two years anyway. I use to be AMD all the way. But when Intel had the price war to try to get back at AMD, I went with the cheaper price. So now the new core had came out and Intel on top again. So if times are right again AMD will hit top again. They just need some smarter engineers to come up with something fast and reliable that you don’t have to buy another motherboard to make it work.
    O and me as CEO. I would take a pay cut also.

    • Chairman_Now
    • 12 years ago

    #3: I’m sorry to have to say this, but better that you get clued in: you’re an idiot.

    Yeah, I guess AMD not doing what they said they’d do is all my fault, and I guess I should be just be fine and buy Intel.

    Thanks for the constructive comment.

    • Damage
    • 12 years ago

    I’ve added a single sentence to the end of the story to reflect the news about tri-core and EE parts being based on the B2 stepping.

      • ssidbroadcast
      • 12 years ago

      I wonder if a 3-core Phenom will still be vulnerable to TLB erratum.

        • tfp
        • 12 years ago

        I would guess yes but the chances are even lower because its only 3 cores hitting the buffer instead of 4.

    • stix
    • 12 years ago

    Pooh baby comes to mind. Buy early get screwed.

    • Chairman_Now
    • 12 years ago

    This is it. I’m turning in my AMD fanboi badge and gun. I’ve been screwed with too many times:

    – No upgrade for the QuadFX, and I have three of them. Really, it’d have been better to just tell the truth rather than lie about it, AMD. Hey Suzy Pruitt, where the hell is my non-Phenom uprade? Why is ASUS saying that there won’t be one, ever?

    – No widespread BIOS compatibility for Phenoms, so my AM2 socket machine is end-of-roaded.

    – Phenom just generally sucks, it seems.

    Now this. Forget it. Intel for me for quite a while. Bye, AMD, it was fun.

      • SHOES
      • 12 years ago

      yep I believe this is happening more and more AMD better pull a rabbit out of their hat but as long as AMD realizes the problem they are creating it should inspire them to create a better product…. We can only hope.

      • pluscard
      • 12 years ago

      First post ever?

        • Damage
        • 12 years ago

        Just to serve notice, pluscard: You’re on thin ice here due to your past behavior in AMD-related discussions. I don’t know whether you’re somehow employed by a “viral” PR firm or not, but ultimately, it doesn’t really matter. If you pollute this thread with propaganda, you will be banned. If you again misrepresent our reporting willfully, you will be banned. Consider yourself warned.

          • pluscard
          • 12 years ago

          Of course, if I was bashing AMD and ATI’s products, my posts would be welcomed, right?

          Do what you have to do.

            • Flying Fox
            • 12 years ago

            It depends if your bashing makes sense. If you post and say the X2 is worse than the Pentium D, then you will get the same ribbing from us too. 😉

          • RambodasCordas
          • 12 years ago

          Well I think he is absolutely right, and about the Viral PR, I think this site and their authors have been a great contributions (from the bad outdated SB600 to the B3 issue) don’t you think?
          And why not some article about Intel errata list and TLB bug too?

          Intel SB issues:
          §[<http://www.anandtech.com/showdoc.aspx?i=3067&p=1<]§ §[<http://www.anandtech.com/storage/showdoc.aspx?i=3167<]§ Intel TLB bug: §[<http://www.heise.de/ct/08/02/018/<]§ Originally Posted by HP Intel has provided a microcode update as a critical software fix. This Intel microcode update addresses an improper Translation Lookaside Buffer (TLB) invalidation that may result in unpredictable system behavior such as system hangs or incorrect data.

            • Kurotetsu
            • 12 years ago

            “This site does not suck AMD’s wang on a non-stop basis, so it must be a propaganda machine for the eeebil Intel!”

            Laughable stereotypes like you and pluscard really make my day. Keep it up.

            • RambodasCordas
            • 12 years ago

            “Laughable stereotypes like you and pluscard really make my day. Keep it up.”

            Funny. Besides this site who gives so much for the TLB bug?
            Besides Intel TLB is much worst since it may lead to data corruption. So why no body talks about it?

            I think guys like you (Intel paid puppets) need to learn a lesson or two from AMD. While AMD is clear with the issues Intel does not:

            Flaws In Intel Processors Quietly Patched
            §[<http://hardware.slashdot.org/article.pl?sid=07/06/26/2152246<]§ And has you can see here Intel initial CPU revisions had issues that Intel asked Microsoft to fix: §[<http://blogs.msdn.com/ptaylor/archive/2007/07/16/windows-update-xp-and-vista-for-intel-multi-core-processors.aspx<]§ Your post was very laughable to me, too bad there is very little content in it, maybe because of some brain problems you might have?

            • Flying Fox
            • 12 years ago

            TR sort of brought this story into the limelight. They should be the ones doing the follow-ups. To not follow-up will be irresponsible IMO.

            Do you know the fact that you guys are popping up just to defend this makes things look even worse? This is the effect of hysteria. AMD has not encountered this before so they handled this poorly. Intel learned their lesson with the FDIV bug (both in terms of internal testing and PR) so they have handled this more quietly ever since.

            The issue and follow-ups need to be reported, but no need to generate all these responses. Doing so just pop the item up on search engines and other news aggregation sites, thereby just adding fuel to the fire.

            I for one think the bug is no big deal. The fix however is a big deal since it has noticeable performance impact.

            • Kurotetsu
            • 12 years ago

            So the fact that Intel was intelligent enough to see the problem, develop a solution, and implement said solution in such a way that the end-user doesn’t even notice is something they should be vilified for? What’s the point of blatantly announcing a flaw to the world when it, and its fix, apparently had very little to no impact on end-user performance? As you said; nobody talks about it.

            The same could have been said of AMD’s problem, except they decided to MAKE it a big issue by releasing a performance degrading patch that, according to many AMD cultists such as yourself, was never needed to begin with. I would think such stupidity would make AMD more worthy of ire than Intel is, but whatever.

            Oh, I should probably mention, as far as your ‘Intel paid puppets’ comment is concerned, that I’m using an AMD system right now. Which, what a surprise, makes your comment as devoid of merit as your post.

            • pluscard
            • 12 years ago

            They released a patch – likely the performance penalty wasn’t their main focus. AMD has to do it better than Intel, and with the patch, the part is flawless.

            If you scan the various review sites, it’s more than the “amd cultists” that believe it’s not necessary for desktop systems, given the inability of anyone to reproduce the error, and considering the performance penalty.

            The q6600 is a very good chip at $275, as is the e6600 at $225. Since there is competition, we also have the Phenom quad 9500 for $189, or the unlocked black edition 9600 for $239, which according to the newegg reviews overclocks to 3ghz on air.

            Regardless of which you choose, it’s a good time to build a computer.

            • TheTechReporter
            • 12 years ago

            The difference is that Intel was smart enough to _not_ make headlines with their TLB patch. AMD really can’t afford to make this TLB issue so public when they are already struggling against Intel.

            Oh, and to #29, this is _not_ just a virtualization issue (there is actually a Tech Report article about this), but the key is that it’s _very_ difficult for _any_ end-user to reproduce, enthusiast or not.

            • green
            • 12 years ago

            probably because it’s seriousness was challenged by things like this:
            §[<http://blogs.zdnet.com/Ou/?p=559<]§ so "apparently" it's like relying on the pentium fdiv behavior before it was disclosed that it was borked then having it suddenly fixed in later chips and your software ends up breaking (sorry i couldn't think up a better analogy)

        • willyolio
        • 12 years ago

        last post ever?

          • BobbinThreadbare
          • 12 years ago

          I laughed.

            • flip-mode
            • 12 years ago

            It made me laugh when I saw that you laughed.

        • jobodaho
        • 12 years ago

        Says the guy with only one post…

      • HighTech4US
      • 12 years ago

      I have already done the same. I upgraded from an Athlon X2 4200+ to an Intel Q6600 2.4ghz safely overclocked to 3.0ghz in August 2007. At that time I no longer believed the L I E S that were coming from Hector and company.

      Also, at that time I was ridiculed for going with the Intel Q6600 as AMDs quadcore was coming very soon and would kick ass and take names.

      As history now shows my move was the right one as AMD still has no answer to even Intels lowest quadcore and won’t for some time to come.

        • Chairman_Now
        • 12 years ago

        Funny, same thing here. When it was time to add a fourth “quad”, I went with the same CPU – and it runs my custom software at the same rate as the QuadFX at 3.0ghz. And that’s with a bandwidth penalty, as I use both gigabit connections on the QuadFX boxes, whereas there’s only one on the Intel Quad box that I built.

        Moreover, the Intel box uses a lot less AC, and produces a lot less heat. And it cost a lot less than the QuadFX boxes did, or even still do.

        I’ll be adding 8-12 more cores here during the year (or two or three more machines). They won’t be AMD.

        AMD’s Suzy Pruitt (not sure what her role there is) said in a press release that they wouldn’t be abandoning the QuadFX folks. Nothing announced since. Won’t hold my breath.

      • leor
      • 12 years ago

      this is actually good news if you have quad fx, 1027 opterons will work fine in those boards

Pin It on Pinterest

Share This