Phenom TLB patch benchmarked

We have been following the story of a chip-level problem in AMD’s quad-core Opteron and Phenom processors all week. This bug—CPU makers prefer to call them errata—can cause system hangs in specific, rare circumstances. This sort of obscure problem is not really uncommon in microprocessors, but CPU makers are often able to fix them on the fly with little impact to the end user. This particular erratum is especially unfortunate because the fix for it involves sacrificing a substantial amount of performance.

This week’s developments have included the revelation that this bug affects all “Barcelona” quad-core Opterons, leading to a “stop ship” order on quad-core Opterons to most customers. The erratum also affects all speed grades of Phenom processors, which are still shipping to PC makers and resellers. AMD admitted the presence of the erratum prior to the Phenom’s public introduction, but the firm’s initial statements gave the impression that the erratum affected only virtualization, which is a server-class application and an uncommon use for a desktop CPU. In truth, the erratum can cause instability with desktop-style usage patterns, as well, and systems with Phenom 9500 and 9600 processors will have to be patched and suffer the accompanying performance penalty.

One thing we haven’t known is exactly how that performance penalty would look—until today. We can now offer you some preliminary benchmarks that demonstrate the impact of the BIOS-based workaround for the problem.

Err-what-um?

I don’t wish to re-hash too much of what we’ve already covered this week, but we should recap briefly the nature of the erratum. The problem involves the chip’s translation lookaside buffer (TLB) and L3 cache. AMD has provided a technical description of the problem as part of its documentation for a unsupported patch for the Linux kernel that alleviates the problem with only a minor performance hit. The specific circumstances that can lead to the data corruption and system hang are most likely to occur during periods of high utilization of all four CPU cores. Technically, AMD refers to the problem as errata number 298, but the problem has become more widely known as the TLB erratum.


Those are the basic outlines of the problem, but we should address something else before we move on. Some folks seem to be confused about the likelihood that the erratum could affect a system’s stability. We don’t really have any good way of quantifying that at present, but we can offer a few nuggets of wisdom. AMD says the problem is very rare, and we believe them. CPU makers do a tremendous amount of qualification testing before releasing a product, precisely because they want to avoid show-stopper problems. The Barcelona Opterons were rock solid when we conducted the testing for our review of those chips. We did run into some stability problems with our early Phenom test systems, but we’d trace those issues back to a pre-production Asus motherboard. The production version of the Asus M3A32-MVP Deluxe that we’ve since tested was much more stable, even without the erratum workaround applied. We used an MSI motherboard in testing for this article, and not once did the system lock or crash during hours of testing without the TLB patch applied.

The TLB erratum is a big deal largely because of the standards CPU makers have established for themselves, in which utter stability is a guarantee, not an optional feature. Even if the likelihood of a crash is extremely rare, of course anything less than 100% stability is unacceptable. AMD has helped establish those expectations, and of course, the industry is correct to expect CPU makers to live up to them. We would be intrigued to see some handicapping of the exact odds of a TLB erratum-induced system hang occurring during the life of an average PC, but such considerations probably aren’t going to fly for most customers. Even if the erratum occurs very infrequently, no one wants the possibility of a system crash at the worst possible moment hanging over his head, and no one wants a “broken” chip. That’s surely why AMD has directed motherboard makers to enable the workaround by default in their BIOSes, with no option to disable it, even though it slows performance.

AMD has taken the unusual step of pledging to release a version of its Overdrive tweaking utility that will allow users to disable the workaround, however, which says something about the severity of its performance impact.

As for that impact, it’s tough to estimate entirely. We’ve heard estimates of 10% and “10 to 20%.” We do know that the BIOS-based workaround for the TLB erratum disables some problematic logic in the CPU, but does not disable the L3 cache entirely.

The test

With that said, we can move on to the test conditions. Several key things made this test possible. One was the fact that MSI was able to supply us with a BIOS for its K9A2 Platinum motherboard that includes the TLB erratum workaround. Thus, we tested with an earlier revision of the MSI board’s BIOS (version VP.0B7) and with the newer, patched BIOS (version 1.21). Per AMD’s guidance on this issue, MSI apparently did not include a menu option to disable the workaround. In fact, the BIOS doesn’t look to offer any cosmetic indicator that the workaround is in place.

Another key to making this test possible was the help of the fine folks at NCIX, who supplied us with a production Phenom 9600 processor for use in testing. Thanks to them for their assistance.

I’ll spare you the giant table here, but we generally tested using the configurations described in the testing methods section of our original Phenom review. The notable exceptions, of course, are the production Phenom 9600 and the MSI motherboard.

That original Phenom review overstated the Phenom 9600’s performance for two reasons. The most obvious, of course, is the fact that we didn’t have the TLB erratum patch applied. The other is that the north bridge clock on our Phenom engineering sample was running at 2.0GHz. AMD told us that was the correct north bridge speed, but our experience with production Phenom 9600 chips has proven otherwise; the correct clock is 1.8GHz. The north bridge clock is critical to performance in this CPU architecture because the L3 cache runs at the speed of this clock.

As a result, we have included scores for several Phenom CPUs in the graphs on the following pages. The ones marked “Phenom 9600 – TLB patch” and “Phenom 9600 – No TLB patch” come from the production Phenom 9600 and the MSI motherboard. The scores marked “Phenom ES” come from our Phenom engineering sample and the pre-production Asus motherboard. The “Phenom ES 2.3GHz” is what we mistakenly represented as a Phenom 9600 in our original review. With all of these results present, you should be able to see the impact of both the lower north bridge clock and of the TLB erratum patch.

The warm-up: Memory subsystem performance

Since the TLB erratum workaround is likely to affect the performance of the memory subsystem most directly, we’ll start by looking at some synthetic memory benchmark results. Note that performance results for these tests don’t translate directly into real-world application performance. These more focused benchmarks are instead designed to stress the memory subsystem, including CPU caches and main memory.

This first test lets us whip out a fancy-looking graph. Bear with us as we show off briefly.

I hesitated to include these results because, well, the graph is really hard to read. The problem? All three of the Phenom CPU configs running at 2.3GHz overlap almost entirely. I do think that’s enlightening, though, because it shows us that L2 and L3 cache bandwidth don’t appear to be affected by the TLB workaround. We can isolate the larger 1GB block size in this test, though, and see a more tangible difference.

Here, the TLB patch’s impact begins to be apparent. It’s even more apparent in Sandra’s Stream-like test of main memory bandwidth. Have a look.

Ow. Memory bandwidth drops from roughly 5.4 GB/s without the TLB patch to 3.65 GB/s with it.

Of course, bandwidth is only part of the story. Let’s look at memory access latencies, as well.

The TLB patch exacts an access latency penalty of 40 nanoseconds at our sample block size. We can look at the latency picture more fully with a terrifying-but-useful 3D graph.

In these graphs, yellow represents L1 cache, light orange is L2 cache, and dark orange is main memory. What you’re seeing here is memory access latencies at various block and step sizes, in a way that exposes latency for the various stages in the memory hierarchy.

These latency graphs demonstrate the same basic principle that the bandwidth graphs did. The TLB erratum workaround exacts its performance penalty by slowing main memory performance. The actual impact on memory performance is much greater than the 10% number we’ve seen floating around, but the patch’s affect on application performance will depend on whether the app’s working data set can fit into the CPU’s on-chip caches. The more often an application must reach into main memory, the greater the impact of the erratum workaround will be.

The real test: Application performance

Now for the real test: actual applications one might run on a Phenom processor. We’ve used a number of elements of our usual CPU testing suite. If you’re not familiar with them, we introduced them more properly in our original Phenom review. We’ll dispense with the pleasantries here and give you the results.

As we predicted, the TLB patch’s influence on performance varies from one application to the next. Everything depends on how that application uses memory. The largest impact, by far, is in the Firefox test, which really takes a hit when the patch is present. Web browsing isn’t exactly an uncommon activity, either. In other tests, like the Sandra multimedia benchmark that uses SIMD instructions to generate a picture of a Mandelbrot fractal, the patch barely slows performance at all.

Conclusions

So the BIOS-based workaround for the TLB erratum can have quite an effect on performance. How close were the estimates we’ve heard of a 10% performance drop? Let’s summarize our results and consider the percentage differences.

No TLB patch TLB patch Difference
Sandra cache and memory bandwidth 6527 5932 9.6%
Sandra memory bandwidth – FPU 5403 3650 38.7%
Sandra memory bandwidth – ALU 5401 3648 38.7%
CPU-Z memory access latency 59 99 50.6%
WorldBench – Microsoft Office 2003 SP-1 369 399 7.8%
WorldBench – Adobe Photoshop CS2 521 595 13.3%
WorldBench – Firefox 298 536 57.1%
WorldBench – Microsoft Windows Media Encoder 9.0 248 272 9.2%
WorldBench – WinZip 10 305 321 5.1%
picCOLOR overall score 9.74 7.21 29.9%
Valve Source engine particle simulation benchmark 62 55 12.0%
Valve VRAD map build time 182 191 4.8%
SiSoft Sandra Multimedia Integer x16 130697 130648 0.04%
SiSoft Sandra Multimedia Floating Point x8 169434 169373 0.04%
Total average difference 19.8%
Average difference without memory subsystem tests 13.9%

Across every test we ran, the difference between the Phenom 9600 with and without the TLB patch averages out to 19.8%. However, if we rule out the synthetic memory tests and consider only the application tests, that difference drops to 13.9%.

The most troubling results here are the applications where we see large performance drops with the TLB erratum workaround active, including the Firefox web browser and the picCOLOR image analysis tool. If one happens to spend a lot of time running an application whose memory access patterns don’t mix well with the TLB patch, the result could prove frustrating. The BIOS-based workaround for the TLB erratum may achieve its intended result—system stability—but it comes at a pretty steep price in terms of performance.

For the average retail PC consumer, this price might not be unacceptable. After seeing the Firefox test results, I spent some time browsing the web with our Phenom-based test system, and it didn’t feel noticeably sluggish to me compared to most modern PCs. Then again, I doubt whether the average sort of consumer is likely to purchase a system with a quad-core processor. One wonders where that leaves AMD and the PC makers currently shipping Phenom-based PCs. I’m not sure a recall is in order, but a discount certainly might be. And folks need to know what they’re getting into when purchasing a Phenom 9500 or 9600-based computer this holiday season. Caveat emptor, indeed.

In fact, a credible source indicated to us that at least some of the few high-volume customers who are still accepting Barcelona Opterons with the erratum are receiving “substantial” discounts for taking the chips. One would hope consumers would get the same consideration. The trouble is, I doubt AMD would have shipped Phenom processors in this state were it not feeling intense financial pressure.

AMD’s other major concern here should be for its reputation. The company really pulled a no-no by representing Phenom performance to the press (and thus to consumers) without fully explaining the TLB erratum and its performance ramifications at the time of the product’s introduction.

As we’ve reported elsewhere, AMD does plan to fix the TLB erratum with a new revision of its quad-core chip due some time in mid-to-late Q1 of 2008. Once the new revision is available, the Phenom 9500 and 9600 will be replaced by the 9550 and 9650, with the -50 suffix denoting the updated silicon and higher performance. Most users will want to wait until those new Phenom models are available before paying full price for a Phenom processor or a system based on one.

Comments closed
    • pluscard
    • 12 years ago

    Actually, Kyle Bennett, editor in chief of HardOCP is challenging review sites to produce the TLB issue on a desktop machine.

    He’s claiming they’re using the issue to gain hits a the expense of AMD and the consumer.

    He’s overclocked a phenom to 3ghz and ran it 14hrs without fault.

    • Damage
    • 12 years ago

    I have made a correction to the text of this article. Previously, it read:


    AMD admitted the presence of the erratum prior to the Phenom’s public introduction, but the firm’s initial statements gave many the impression that the erratum affected only the 2.4GHz version of the Phenom, which it had decided to delay. In truth, all current Phenom variants have the bug, and systems with those processors will have to be patched and suffer the accompanying performance penalty.

    Now, it reads:

    —-
    AMD admitted the presence of the erratum prior to the Phenom’s public introduction, but the firm’s initial statements gave the impression that the erratum affected only virtualization, which is a server-class application and an uncommon use for a desktop CPU. In truth, the erratum can cause instability with desktop-style usage patterns, as well, and systems with Phenom 9500 and 9600 processors will have to be patched and suffer the accompanying performance penalty.
    —-

    Please see here for more information:

    §[< https://techreport.com/discussions.x/13764<]§

    • Lazier_Said
    • 12 years ago

    I remember those garbage Prescott space heaters pretty well.

    But I don’t seem to remember forum fanboys pushing to “buy the overheating, underperforming POS anyway because we need to support the company in their time of need.”

    • Iketh
    • 12 years ago

    I need a little help… apparently the lock-up i wrote about a few posts down is in fact my 7950gt becoming flaky, as it has occurred a second time now in window’s desktop….

    so my phenom 9500 is running folding@home 24/7 flawlessly, overclocked to 2.5ghz… i had already sent an email to MWave requesting a full refund on the phenom and they surprisingly obliged despite their policy clearly stating no returns on boxed retail processors (cudos MWave!)… i have 7 days to get the processor back to MWave, and I don’t know what to do now…. replace the phenom with a 6400+ ($169 tigerdirect) or keep the phenom, any advice will help me decide (i make use of all 4 cores regularly by capturing full game videos with fraps and rendering the movies, among others)

    what would u do in my position?

      • Flying Fox
      • 12 years ago

      You could drop by the forums and we can discuss this. 😉

    • muyuubyou
    • 12 years ago

    The “good news” for AMD is that they don’t have that many flawed Phenoms to sell. They will surely have to price them lower, and apparently the problem doesn’t reproduce with 3 cores, so their upcoming 3-cores from this stepping will work okay.

    It will surely cost them some, but not as much as many would think. AMD is better now, in real terms, than when they had the performance lead but no big OEMs were selling their chips. People are overreacting.

    • rivieracadman
    • 12 years ago

    Lets sum this up, because from the comments there must be a huge supply of fanbois/grls on this forum:

    1. Phenoms 1st in system cache and memory bandwidth

    2. Phemon (No TLB) 9600 memory access beats all Intel chips

    3. Phemon (No TLB) 9600 and Phemon ES (2.3 and 2.6) beat Q6600 in Worldbench Office

    4. Phemon ES (2.6) beats X2 6400+ (3.2Ghz) in Worldbench Adobe

    5. Phemon ES (2.6) beats Q6600 and X2 5600+ in Worldbench Firefox

    6. Phemon ES (2.6) beats Q6600 and FX 74 in Worldbench Windows Media Encoder

    7. Phemon ES (2.6) leads AMD’s pack in Worldbench Winzip

    8. Phemon ES (2.6) leads AMD’s desktop processors in picCOLOR

    9. Phemon ES (2.6) beats all other AMD Processors in Valve Source engine particle simulation

    10. Phemon ES (2.6) beats all other AMD Processors in Valve VRAD map

    11. Phemon ES (2.6) beats all other AMD Processors in Sisoft Sandra Multimedia Int x16

    12. Phemon ES (2.6) beats all other AMD Processors in Sisoft Sandra Multimedia float pt x8

    SO … disapointments aside, the 2.6Ghz Phemon beats the Q6600 in four of the 12 benchmarks. In almost all the other tests it either defeats or comes very near defeating the entire Athlon X2 and FX processors currently available.

    In most of the tests where the Phemon loses to the Q6600 it is very close. So close in fact it would be impossible to tell the diffrence between the two in real life useage.

    Serveral sites including Legit Reviews where able to overclock the Phemons to over 2.7Ghz on stock voltages, and 3Ghz ( a little hot ) but that’s with the stock cooler too.

    §[< http://www.legitreviews.com/article/597/12/< ]§ §[<http://www.xtremesystems.org/forums/showthread.php?t=166792<]§ So what's the problem? It's a better chip. It's faster then the chips their currently making. It can handle more threads. It has around the same overclocking range as existing Athlon x2s, and we'll be seeing updated and quicker versions in a few months along with Black and FX editions. Why all the bad press? These guys are obvious contenders their just having a real slow start. I don't plan on moving our Opteron farm to Intel for a 10% or even 30% performance diffrence. We're talking real world viable answers to upgrading hundreds of thousands of dollars in existing hardware. Why would you want to spend a couple hundred dollars more to replace a healthy AM2 rig for 10% or even 20%? I just don't get it? Maybe I'm just getting old.

      • provoko
      • 12 years ago

      Good job. Definitely Phenoms are still a great upgrade for AM2. However, for other motherboards (new computers), Core 2 would be the better value. I wish they made a socket 939, I would still buy one now.

      The Opteron farms using Socket F-class definitely benefit from the Barcelona Opterons until now, the TLB hurts virtualization the most. So I’m sure companies wanting to upgrade will have to either wait for the revision, take into account a 10-20% performance drop, or move to Intel.

      • flip-mode
      • 12 years ago

      One problem with your hypothesis is its dependence on unavailable chips and overclocked chips.

      Unavailable chips don’t count, plain and simple, no explaination needed.

      Performance and frequencies obtained by overclocking are only marginally useful. Overclocking results vary, period; overclocking is more art than science — it depends on so many variables (other system components, skill or experience, the chip itself, clean power, ambient temperature) that it isn’t a straightforward metric by which to judge a CPU. It’s almost certain that all Core2 based chips, whether quad or dual, will OC to 3.0 stable. K10 is a different story, and the bug only complicates matters. A K10 at 3.0 will have massive power consumption compared to Core2. Any K10 could be hit by the TLB bug. Every K10 will have less IPC than Core2. K10s can’t clock as high as K8 and are therefore sometimes beaten by K8.

      So your post oversimplifies the situation.

      • TakkiM
      • 12 years ago

      quote:

      Why would you want to spend a couple hundred dollars more to replace a healthy AM2 rig for 10% or even 20%? I just don’t get it?

      You don’t get it because you didn’t factor-in the power consumptions in your plan! The 45nm Penryn derivative chips consume far lesser power than your beloved “Phemon” (its called Phenom BTW!!) and far outperforms them!

      You referring to real-world?!? then get real!! one of the largest cost for a data center is the operating expense for power consumption! For a lot of companies, time is money! the faster you get things done the better! with faster chips, things get done faster… its as simple as that!!

        • provoko
        • 12 years ago

        I see you haven’t read the barcelona opteron review.

          • melvz90
          • 12 years ago

          your point being on memory bandwidth benchmarks… its a common knowledge that the K8/K10 had always the edge on memory bandwidth benchmarks

            • provoko
            • 12 years ago

            My point was on wattage regarding TakkiM’s comment above.

    • pluscard
    • 12 years ago

    Just to lend some perspective on this Intel/AMD thing:

    -Internationally, Intel currently has 3 separate unfair trade practice investigations going on. In Japan (JFTC), Intel didn’t refute the charges and agreed to comply in the future, paying a small fine. The EU charged Intel as well, and Intel has asked for more time to answer. The answer is now due in Jan’08. Korea has yet to file their charges, but is collaborating with the EU and Japan and exchanging data.

    -Intel has a history of rigging benchmarks as was discussed here in the last month or so. In one example, Intel rigged their compilers to detect an AMD processor, not utilize it’s features, and return a worse result. The issue was exposed when a simple patch turned off the “AMD check”, resulting in a much improved result for the AMD machine.

    -Despite Intel outspending on research by a factor of ten, AMD has been the driver of most technical innovations in the last 8 years. 64bit support, integrated memory controllers, hypertransport connect system, native dual and quad core, power management, etc. INTC typically competes thru process and larger cache (brute force). The core2duo had some elegant design features, notably the shared cache, but most would agree the C2D was a response to AMD trouncing the P4 with it’s Athlon 64.

    While I’m the first to agree that AMD’s execution has been anything but spotless, harping on AMD’s mistakes, while ignoring the overwhelming evidence that Intel will us any tactic, legal or illegal to maintain it’s monopoly is foolish.

    More importantly, the end result of this media frenzy may be a return to an “Intel Only” market. I’m sure Intel will wait a respectable period of time before raising prices across the board, and throttling back it’s R&D efforts, just to keep up appearances.

    Plus

      • amdsroadkill
      • 12 years ago

      I will not get too arguemenative because AMD is not worth beating up now is it.

      But the license AMD has for making X86 chips is an Intel license Intel made this chip. That is a fact we canb all look up. About AMD? It has produced a broken chip. That is the topic here.

      Intel has a very good chip out now and it is not broken.
      Likely, because of the current disaster AMD has with its broken and buggy core, Intel will gain huge market share in the server area. This is the most profitable sales area in the businesss. AMD is already losing money as a company. This wreck will just add to the problems AMD has with its current junk bond credit rating. it is good to see that even the AMD fans see the same.

    • pluscard
    • 12 years ago

    Is it just me, or does it seem that TR has been “piling on” to AMD’s issues lately?

      • bfellow
      • 12 years ago

      I don’t “recall” AMD ever doing a major recall of chips then make the existing ones slower. That’s why everyone is discussing this historical precedence.

      • Entroper
      • 12 years ago

      It seems to me that issues have been “piling on” to AMD lately. Don’t shoot the messenger. 🙂

      • flip-mode
      • 12 years ago

      They’re just making a living by reporting the truth. TR’s coverage of these issues has been linked all over the place. I wouldn’t be surprised if coverage of these issues has resulted in something near the highest unique page hit count TR has ever seen.

      • provoko
      • 12 years ago

      It’s you. TR is just doing their job, a great one too.

    • Lord.Blue
    • 12 years ago

    Just make sure you are getting the B3 stepping on these chips and you’ll be fine. That is, when they release the B3 stepping.

    • sigher
    • 12 years ago

    “Most users will want to wait until those new Phenom models are available before paying full price for a Phenom processor or a system based on one.”

    See that’s the advantage from the fallout between techreport and AMD, that they now dare to say such truths.
    Here’s hoping they have a fallout with intel and other companies too 🙂

    • sigher
    • 12 years ago

    The word errata does not really mean bug as such, the word refers to the listing of (known) bugs, so it’s more the description of the bugs than the bugs itself as it were.

    Also when speaking about errata and the stability you might have mentioned that EVERY CPU has a list of bugs, even after several updates to the steppings, including the intel ones, like the math bug core2 has that when patched for greatly reduces its math speed, although generally a patch in applications is only needed for scientific/medical applications.
    As it was put in the article it made it seem like only the phenom has an errata and that all other CPU’s are rock stable and have no bugs that could affect it in such ways (reliability)

    When I read the errata list of both intel and AMD CPU’s I’m shocked by the bugs I must say, some are quite bad and have no workarounds, it’s better to not know and just have your computer have some issue/crash with some program once every 2 months or so without knowing why I then think :/

    Not that this TLB bug isn’t a rather extreme one performance-wise, requiring special mention.

    • sigher
    • 12 years ago

    The ‘extreme use’ of RAM by FF is only extreme in the eyes of the beholder and dependant on settings and customer expectations.
    If you want it to precache pages you are likely to go to and keep things in RAM to be speedy then yeah it uses a lot of RAM, but is that extreme? you have RAM for the exact purpose of speeding up and extending the use of applications.
    And if you wish you can disable a lot of the functions that require the RAM and have lots of pointlessly unused RAM, and a slower browsing experience.

    What I find extreme is the use of RAM by such things as mousedrivers, that could easily do with 1/100th of it and are always loaded.
    Or ‘shell extensions’ that are useless often, and even when useful are so only 0.001% of the time you use the computer.

    • Daveburt714
    • 12 years ago

    Thanks for the article Scott.. You answered alot of questions I had! Even though the news was bleek, and not what I had hoped for, I can still see some potential there…

    I know it’s totally foolish to buy a flawed chip, when you could get a Q6600 that would OC like crazy, but I still find myself wanting one…:)

    If they would release an Unlocked Black Edition in the next few weeks for less than $300 I’d buy one, and a new 790FX Mobo to go with it….

    If I could get it to run at 3/3/3/2.6, even with the patch it would still be fairly competitive, and if the patch turned out to be irrellevent to my apps, so I could safely turn it off….. ???

    I’m feeling like the only person in the world who actually wants an Unlocked Phenom…. But I don’t care… (hehe!)

      • Voldenuit
      • 12 years ago

      l[

    • Chaos-Storm
    • 12 years ago

    Amd would have been better off not mentioning any of this, and not issuing a patch. I doubt anyone running a desktop system would have noticed. There would have been a lot less complaining, that’s for sure.

    I have a Phenom 9500, and I don’t think i’m ever gonna enable this patch. Chances are the errata will never manifest itself anyways….

      • BoBzeBuilder
      • 12 years ago

      When you sell your chip to the mass public, you need to make sure that it simply works. Its good that they mentioned it, and that they found a workaround, because if they didn’t, someone else could have discovered it and that damages the companies credibility among its costumers.
      AMD is just darn unlucky nowdays I guess.

      • credo
      • 12 years ago

      No, thats something Intel would have done. Besides, what happens when a year later some white paper leaks out to the press proving AMD knew about the issue far before its release? I think that would have been even worse.

      Well, worse if they weren’t financially kaput right now… Let’s just say this proves AMDs ethics.

        • Flying Fox
        • 12 years ago

        You give AMD too much credit. They are out to make money. If they can hide it do you seriously think they will not do it? The fact that they let the guy writing up the erratum “on vacation” while knowing about the bug before launch, plus not pro-actively notifying Barcelona customers (it was launched a while back after all), did not give me confidence that they are handling this properly themselves.

          • credo
          • 12 years ago

          No, I dont think they would have shipped the chips unless they had to. That probably would have been the proverbial straw as it were. Fact of the matter is, they could have just kept shipping the chips under the radar like you said and acted like they didnt know anything was wrong. but they didnt.

          financially speaking, this was probably the worst move they could have made….

            • mushroom
            • 12 years ago

            AMD would not acted like they didnt know of this problem, because should that happen, the public would feel that this problem caught AMD off guard, and might want a recall!!.. this would be the worst move both financially and reputation…
            In downplaying the issue with an errata on launch, it avoids both 😉
            a sly and cunning move…
            AMD is a business, i dont think it is any better morally or ethically than other company…

            • Flying Fox
            • 12 years ago

            q[

      • CingKrab
      • 12 years ago

      Intel made that mistake 14 years ago with the FDIV bug, and I doubt they’ll repeat it. AMD would be foolish if they made the same mistake.

    • 5150
    • 12 years ago

    Shouldn’t this warrant a recall, rather than a “patch that degrades performance.” They advertised this chip, it doesn’t perform as advertised, send me a working chip. If I bought a car and three of the four disc brakes worked, I’d expect them to replace the fourth brake with a disc brake, not a drum brake.

      • cass
      • 12 years ago

      Well, it supposedly doesn’t hurt the linux systems but 1% using the kernel patch.

      Maybe there will be a windows patch to do the same and avoid the bios hack.?

      Your are not really in danger because your processor is 18% slower than you thought, so the car analogy isn’t quite fair. I am watching the prices to see what happens though.

      • eitje
      • 12 years ago

      if you bought a car and got gas mileage 18% lower than advertised, would you expect a recall? 🙂

        • cass
        • 12 years ago

        I wouldn’t be happy about it but, this has been the reality since the 1970’s cafe standards were introduced and the epa estimated gas MPG numbers were put in car windows. The reasoning was that the test was run with white gas and at a certain rpm after warmup and the car makers installed an “epa program” in the ecu of their cars to run very lean and low power at that rpm range. Now when me and you bought it, we never got within 10% of the estimated fuel mileage, and it was common to be off by more thatn 30%. In spite of this I don’t ever remember a fuel mileage recall.

        The car gas mileage works the same way as processor performance.. you can’t pay any attention to the claims or the ratings, you have to read some impartial and thorough reviews to get an idea.

        I know you weren’t addressing me, but this is still how it works and I did see your smiley. 🙂

          • Entroper
          • 12 years ago

          That’s funny, since I regularly exceed the EPA estimated mileage on my own car (both before and after the 2008 EPA revisions).

        • 5150
        • 12 years ago

        If they told me I would get 25 MPG driving under the same specs they tested it at and I only got 20.5 miles, then yes, I would expect a recall, or my money back.

          • d2brothe
          • 12 years ago

          Well…you’d be SOL…

            • 5150
            • 12 years ago

            Sure as hell wouldn’t buy their brand again.

    • Iketh
    • 12 years ago

    I’ve had the Phenom 9500 for just over a week now on an M2N-E mobo without the patch. I run folding@home 24/7 except while gaming. Believe it or not, while reading this article, my computer froze for the first time. I stared blankly at the screen for 2 minutes, somewhat intrigued, and watched 3 garbled lines appear within 2 minutes. I couldn’t believe I was witnessing the bug first-hand while reading this article.

    Granted, that was 1 crash in a week’s worth of f@h and gaming with the likes of FSX (which stresses all 4 cores). The fact that it wasn’t a complete freeze since graphical anomalies very slowly began covering the screen, showing that 2 cores are in a perpetual state of fighting one another for L3 real-estate… so what if it occurs during a write command to the harddisk?

    I don’t want the patch, i want a refund from MWave and a new 65nm 5600+. Better yet, I’ll pair this mobo with a BE-2xxx and give to my brother for Christmas, and buy myself a QX6600+mobo.

      • Flying Fox
      • 12 years ago

      This does not necessarily mean you are hitting that particular bug.

        • Iketh
        • 12 years ago

        lol ok… i havent had a lock-up browsing a website in years

          • Iketh
          • 12 years ago

          better yet, when was the last time YOUR computer locked up browsing the web on a stock system, or even after u’ve settled on a stable overclock? or when ur running prime95 or f@h on a stock system even without proper case ventilation?

          and then the fact this system was built in Jan. 07 with an Athlon x2 4200+ and now wants to lock up with a phenom…

          come on, get real

            • StashTheVampede
            • 12 years ago

            Your description isn’t describing what will happen to your machine if you encounter this bug — your machine isn’t totally dead. The # of end users that affects is SMALL. How do we know this? Look at the vendors that are affected by this issue: Cray.

            Cray tends to build some pretty massive supercomputing clusters that do A LOT of low level work to squeeze out all the juice possible from the machine. With the dozens of reviews/reviews published on the net, how many had frequent lockups associated with all four cores maxed out?

            • Iketh
            • 12 years ago

            so someone please enlighten me… which component can be a culprit of what i described? mind you, as i already stated, this system never locked up before with an athlon x2 4200+

            • StashTheVampede
            • 12 years ago

            There could be … handful of other things, in your box, that could have gotten flaky or gone bad BEYOND just dropping in a while new processor. The chipset could need a few updates, the video, etc.

            Besides, did your machine perform a hard lock as described? If yours didn’t, then the chances of you being affected are very low.

            • Iketh
            • 12 years ago

            it was hard locked as it can be, other than the very slow appearing garbled lines on screen, about one a minute… it reminded me of someone ripping out a piece of hardware while the computer is running (don’t ask how i know this)…

            definitely possible my 7950gt is becoming flaky, and if so i apologize to the phenom… but how damn coincidental it occurs when reading this article while running a phenom 9500…

            i sure hope it wasnt the bug, because i emailed MWave from my online receipt, then went to their website to read their policy and it clearly states no returns on “boxed retail processors”… i can see their response already….

      • redpriest
      • 12 years ago

      If you had the bug you would have had a complete lockup and a machine check, not the weird behavior you are describing.

        • Iketh
        • 12 years ago

        what’s a machine check?

      • pluscard
      • 12 years ago

      It’s my understanding the “race condition” is rare, and typically only seen when using virtualization. I doubt what you saw was the bug.

    • tinokun
    • 12 years ago

    The percentage differences in page 4 don’t seem to match the scores (the bigger the difference, the more obvious this is).

    • provoko
    • 12 years ago

    That damn L3 cache! When I first saw this L3 cache, more than a year ago, I wasn’t too impressed, I figured if the Core 2s had a shared L2 these Phenoms (Barcelona at the time) should use the same method, no point in going into another cache. Now it turns out the L3 is a major problem in performance AND stability!

    Good news is that the L3 cache is Barcelona’s problem. If AMD can solve the erratum through a revision, they can also speed up the L3 cache’s performance with that same revision. With an efficient L3 cache, the Phenoms may actually show equal performance to the Core 2s.

    By the way, is this errata even necessary? From reading the article, it’s only a rare stability issue. How rare is it? If this cpu were tested with ORTHOS without erratum, when would that instability occur?

    • Kulith
    • 12 years ago

    You freakin people. You don’t buy crappy products from a crappy manufacturer to keep the competition alive. If you do that your an idiot, and you have no idea what the definition of competition even means.

    AMD came out with a horrible cpu, the phenom. And it sux. We all agree AMD took the wrong road somewhere and are currently taking a direct flight to extinction. You DO NOT help a company that is making bad products, If you do you are encouraging the existence of crappy cpus to remain on the market.

    You pay for a cpu from a company that knows what they are doing. This way competitors (AMD) will have to learn to get their act together and finally learn how to make cpus before they go bankrupt.

    Dont you guys even know what competition is? The good products survive, the shitty ones go to a landfill and then the drawingboard and hopefully end up as good products. Simple as that.

      • Franken13
      • 12 years ago

      Don’t you remember the the intel p4 spaceheaters – hundreds of millions of them were sold to an uninformed public who thought they wee wondeful -they keep old peoples feet warm on those long winter nights.
      They sucked wind and made intel billions upon billions

        • Kulith
        • 12 years ago

        No I dont actually. Feel free to enlighten me if you will.

        Im not biased against AMD at all. If amd comes out with a better cpu than intel and the benchmarks support it, then by all means ill support it as well.

        I doubt those people bought the intel because they wanted to keep competition alive. Keeping low quality products in production is what ruins competition.

        If you dont buy AMD crappy cpus now, you are telling them “Wake up and make something worth buying” Thats what competition is all about.

          • credo
          • 12 years ago

          AMD was doing just fine. IMHO they would never have released a product w/ bugs before Intel’s shoddy business practices put them into the (deep)red. Thats the difference between intel and amd i guess. And thats why there are plenty of ppl out there, besides yourself, that understand the whole picture. Good thing the world of technology isn’t in your charge.

            • sigher
            • 12 years ago

            Go download the errata list of the old AMD CPU’s, what do you mean ‘they never would have released a product with bugs”? every CPU has bugs.

            And he’s right about competition, in theory, the only issue being that it costs billions and many years to start a competitive CPU firm so there will not be new competition to take over if AMD dies, meaning that if it does intel will be a monopoly, and I think we all know how great monopolies are huh.
            Of course a drop in sales could make the shareholders force new management I guess, then you’d basically have a new AMD and competition did work.

            Also the phenom aren’t that expensive really (the MSRP that is, the retailers overprice it a bit currently, for some reason) so if you don’t like the intel platform and can do without the most extreme speed it is a doable CPU.

      • Krogoth
      • 12 years ago

      It is bit of a stretch to called the Phenom terrible.

      It is not that great and didn’t deliver to expectations of enthusiasts.

      Guess what? The majority of the market doesn’t care about percentage differences in benchmarks. The total cost of the platform is what matters. As long a Phenom-based platform costs less then its competition. It will do fine. Quite frankly, any modern CPU is bloody overkill for mainstream user.

        • bfellow
        • 12 years ago

        The problem is Phenom 9600 chip is on average about same price as Intel’s current quad core Q6600 and performs 13.5% slower without the errata and now probably around 27.4% slower with it.

          • Flying Fox
          • 12 years ago

          Which means eventually market forces will make them adjust prices to more reasonable levels to be “competitive”.

      • credo
      • 12 years ago

      You’re an idiot if you dont. What do you think pushed Intel to its current standard? AMD. If AMD was gone then you would be paying a lot more money than the cost of a 9500 for far less of a product right about now.

      Besides, AMD wouldnt be in this spot if Intel didnt pull its BS tactics of selling chips at (estimated) cost. Sure it was great business, hitting AMD right when they purchased ATI, but it was still BS.

        • Kulith
        • 12 years ago

        Dude, I wholeheartedly support competition. Competition is the greatest thing that ever happened in this world.

        But when people say “you should buy the Phenom to keep the competition alive” I just get frustrated. That is BS and has nothing to do with competition.

      • lyc
      • 12 years ago

      lol @ “your an idiot”

        • gerbilspy
        • 12 years ago

        Great catch! I noticed that and had the same reaction! 🙂

    • totoro
    • 12 years ago

    Incidentally, who are all you people?
    I recognize like three names.

      • jobodaho
      • 12 years ago

      the fan boys that come out swinging after every article. It’s almost laughable at how fired up they get.

        • provoko
        • 12 years ago

        Haha, I see more haters. By the way, who’s totoro and jobodagi? Haha.

        • Flying Fox
        • 12 years ago

        Yeah, they all seem to come out of the woodwork all of a sudden. 🙂

          • d2brothe
          • 12 years ago

          Lol…this is terrible…its the only time in my life I’ve felt the urge to say …”don’t be hattin”….somehow I feel dirty 😛

          Never seen so many opinionated idiots in one place before.

      • BoBzeBuilder
      • 12 years ago

      I’ve been here for quiet a while now. but I don’t recognize YOU. welcome fanboy.

        • totoro
        • 12 years ago

        I’ve been here longer than you have.
        Get off my lawn!

    • Snake
    • 12 years ago

    I am going to open up another part of this discussion with this question:

    Is this AMD’s architecture (finally) coming back to bite them in the rear?

    In other words, AMD hyped the benefit of having their memory controllers integrated into each and every CPU. While this policy can increase theoretical bandwidth and performance – due to the memory controller being optimized to the CPU’s needs – did this architecture lead to this TLB problem due to the complexity of the design process?

    By having the memory controller on each CPU the memory controller must therefore, by definition, be re-created on each die design. Did this decision on architecture cause AMD’s eventual problems, by:

    a) a new major CPU die redesign oftentimes requiring a new motherboard design in order to handle the new memory controller (see part b)

    b) not allowing a refinement of design due to the fact that said memory controller got recreated in each new CPU design (while Intel, having a separate Northbridge, could refine and correct this chip under separate design, and stick with the same chip for a large range of CPU’s, thereby making the motherboard manufacturer’s job “somewhat” easier).

    AMD’s AM2 socket was introduced to reduce the number of sockets and motherboard designs that AMD’s processors required for each redesign. But has AMD’s decision to integrate the CPU and memory controller finally shot them in the foot – because of L3 cache and constant redesigns instead of refinements and reusage of pre-existing proven support parts?

      • provoko
      • 12 years ago

      It think they did shoot them selves in the ass with the L3 cache. It was a good idea, but they would have probably been better off with a MUCH larger shared L2 cahce.

      • Flying Fox
      • 12 years ago

      I am not sure if the TLB is part of the memory controller? It should be more around the cache stuff?

    • steelcity_ballin
    • 12 years ago

    Ug terrible. Does AMD have a QA department anymore? Why release it if you know it’s bad and further steep your name in mud? I was a huge AMD fan for the longest time and JUST as you think the entire “lol only hackers have amd / intel is better” stigma was wearing off, they’re back at it.

    /me takes a trip down barton 2500+ lane.

      • Flying Fox
      • 12 years ago

      As has been repeated so many times, this bug is pretty rare. It is being blown out of proportion. They do have testing, pretty extensive at that. Damage even acknowledged that this is now being touted as such a big deal because of the CPU makers setting the bar high in the first place.

      But then the performance hit of the update is a big enough downer already.

        • sigher
        • 12 years ago

        As I understand it this bug’s issues increases with clockspeed, meaning that they can’t make parts faster than 2.6GHz while this bug is in place, not a big deal eh? think again, overclocking potential matters, as does selling faster parts.

          • poulpy
          • 12 years ago

          Of course selling higher clocked parts is important but that’s irrelevant to the fact they do proper testing, the rareness of the issue nor that it has been blown out of proportion.

          I disagree on the o/c part, it’s interesting to see how fast the architecture can go and a selling point for enthusiasts albeit global market wise o/c-bility is a pretty insignificant factor really.

            • sigher
            • 12 years ago

            Well let’s be realistic, OC’ing isn’t something done by a few ‘enthusiasts’ anymore, EVERY motherboard and EVERY manufacturer now has support for OC’ing and it’s part of all their marketing.
            I’d say that well over half the people buying CPU’s now expect it to be overclockable, that’s counting people buying retail made boxes, if you just count buyers of individual CPU’s it might be as high as 90%, or more in my personal estimates.
            Even chipset makers release windows utilities to make OC’ing more approachable.

            • mboza
            • 12 years ago

            Do all dell m/bs have overclocking options? Tyan? Supermicro? How many retail boxed cpus are sold, against the number of systems sold by Dell, HP, etc?

            Sure, there might be many more enthusiasts than there once was, but there are many many more non-enthusiast PC users who do not know what a front side bus is, let alone how to make it go faster.

            And if all these proocessors are so good all the time, why not just bin them higher?

          • Flying Fox
          • 12 years ago

          Race condition bugs usually show more easily with a change in clock speeds, in this case upwards. Does not mean slower clock will never see the problem, just harder to reproduce. You just need the right star alignment and other factors to trigger it.

    • kilkennycat
    • 12 years ago

    One more big snag ‘down the road’ regarding the BIOS fix…………

    When the TLB problem is finally fixed and the 9×50 processors are readily available, if one is contemplating purchasing a motherboard then a wait for a motherboard with a BIOS compatible with the 9×50 versions would be very wise.( i.e a motherboard with the TLB BIOS ‘fix’ removed )…

    The MB vendors have reportedly (and stupidly) not given the user the option in the BIOS settings to disable the fix. Why not? Just another non-default BIOS setting. I thoroughly distrust post-sale promises by MB vendors with regard to BIOS fixes. And there is always the risk of ‘bricking’ a motherboard with a user-initiated BIOS update.

    It seems that AMD should stop ALL Phenom shipments right now to salvage their tattered reputation. No doubt bank and investor pressure is too great and AMD’s spineless management has caved in. What can you expect from a company that was beating Intel at its own game until they indulged in the financial catastrophe of acquiring ATi. The huge debt load has played its part in diverting focus from AMD’s core business and no doubt put pressure on AMD to rush Phenom and Barcelona to market and cut corners in fully testing the K10 design.

    Sad, considering that I had jumped for the first time from Intel to AMD (X2) when I built my most recent personal high-end machine… the most stable PC that I have ever possessed. Now that I am contemplating a new personal high-end system-build, it seems sadly inevitable that I will have to jump back to Intel (Penryn).

      • b4b2
      • 12 years ago

      “AMD claims it has handed off the BIOS workaround to motherboard makers for implementation, and Saucier told us the company’s guidance to partners included an enable/disable option in the BIOS. AMD also has plans for an update to its Overdrive overclocking utility for Windows that will allow users to toggle the erratum fix on and off.”

      Quoted from earlier TR article on the Phenom on Dec. 3

      Sounds like it shouldn’t be too big of an issue.

        • bfellow
        • 12 years ago

        TR already quoted that Saucier was wrong according to AMD. It won’t be disabled in anyway.

    • Jeff Grant
    • 12 years ago

    Does the patch affect overclocks either up or down in any way? Can the chip be overclocked to compensate for any performance loss?

    • Jeff Grant
    • 12 years ago

    Does the patch affect overclocks either up or down? Can the chip be overclocked to make up any differences in performance?

    • Krogoth
    • 12 years ago

    The professional-level stuff is what really matters.

    The hit isn’t as bad, but it degrades the Phenom below Conroe-performance. That is not good.

    Mainstream market on the other hand doesn’t care as long as Phenom are dirt-cheap. OEM builders will choose it over its competition. The bottom line cost of the platform is what matters.

    The initial revisions of K10 are taking a page from Intel. They AMD’s own Willamette.

    • Dposcorp
    • 12 years ago

    Wow.
    I will still buy one if the price is right, if for no other reason then to try and keep the competition alive.

    Who else has been able to really compete with Intel the last 5-10 years when it comes to CPUs on the desktop?

    Transmeta?
    Via?
    Motorola?
    IBM?
    Sun?

    If somehow AMD does fail, and I do not think they will, is there anyone that can step up and provide a viable option with competitive performance on the desktop?

    • Convert
    • 12 years ago

    Well, TR seems to be really on the ball. Great job guys.

    • Spotpuff
    • 12 years ago

    That latency increase really sucks… AMD get your act together. 40ns increase in latency? Sheesh.

      • sigher
      • 12 years ago

      I wonder if, if I were CEO, of AMD I would have recalled the whole batch and cut my losses, it would be pricey and AMD is in debt already, but still..
      You could always rename them to ‘consumer value CPU CV01’ and sell them for $50-$100 for the cheap computers to get some money back I guess, disable some cores even maybe.

    • d2brothe
    • 12 years ago

    All this article really tells me is that if you have a first gen phenom, you shouldn’t enable the fix. I can’t see that its worth it for a desktop processor. Seems your system is much more likely to crash due to bad ram, bad psu, dead hd, or even more likely…windows…than it would be due to this bug, thus there seems to be no reason to incur this penalty. Also, I might point out…this fixes extreme effect on firefox might be due to firefox’s extreme use of system ram. I really don’t think this is as bad as it seems…If I had an AM2 system, I’d certainly be willing to upgrade to a phenom with a moderate discount.

    • cass
    • 12 years ago

    I would like to know how AMD ever started down this path. I mean somebody had to either lie or just not have any clue at AMD. It should have been obvious without the TLB erratum during early simulations that the Phelop (rhyme with flop) was not going to be competitive with the q6600 in any way. I mean I feel sure they had some type software simulation ability (well maybe).

    I am not cutting them the same benefit of a doubt TR is. I don’t see any way they should not have know they were inferior to start with… Intel had their parts out for over a year now and they new full well what the mark was in terms of IPC and ghz. Yet even with a large target painted on the wall at point blank range, they somehow aimed in the other direction. That just screams bad management, and a total disconnect with reality. Especially when coupled with the decision to ship the bad parts while furnishing modded ES procs for reviews and setting up nonrepresentative test systems for evaluation.

    Yeah Phelop, Phlop, or dog call it by any name. It ain’t worth what they are currently asking. I’ll wait.

      • lethal
      • 12 years ago

      I’m sure they knew what they were aiming to…. its just that you can’t do much to a concrete wall when all you got is a water gun =P.

      • sigher
      • 12 years ago

      Your attempt at making a joke from the name isn’t quite working I fear, you knew that too while editing and before release, and yet you released it 🙂

    • lordT
    • 12 years ago

    A 57.1% performance drop in Firefox. Youch!!

      • flip-mode
      • 12 years ago

      Both large and meaningless.

        • moose17145
        • 12 years ago

        lol i was thinking the same thing. Even running almost 60% slower i would imagine its still running so fast you can’t tell the difference. Meaning its probably still running faster than on my 3.2GHz P4 rig.

          • sigher
          • 12 years ago

          Interesting to see the effect of this issue on various apps yeah, you can learn a lot from it, perhaps AMD should and learn how they can make a kickass CPU by finding out what exactly matters in the end. (or intel can learn even more what they can tweak)

    • axeman
    • 12 years ago

    Ouch. Ouch..OUch OUch Ouch Ouch.

    • emorgoch
    • 12 years ago

    Quick Question: With the new round of testing on the MSI board with production BIOS, and the retail Phenom chip, did you guys ever encounter the TLB bug when the TLB patch was not applied? If so, were you able to reproduce it consistantly?

      • sroylance
      • 12 years ago

      He covered that in the review. The barcelonas were stable under testing. While the phenom test rig was not stable, he hangs that to a pre-production motherboard and not the TLB errata. Considering the bug was not found during AMD’s pre-release testing (which is probably quite exhaustive) we can assume that triggering this bug is very rare in usual server/desktop workloads.

    • spuppy
    • 12 years ago

    I don’t know if it’s just me, but I think a big part of this story is that AMD sent processors with overclocked northbridges to reviewers in an attempt to inflate their launch benchmark scores!

    So they launched a broken CPU, but at least that is fixable. It’s entirely inexcusable to outright LIE to the hardware enthusiast community.

    The coverage of this issue has been spectacular, Scott. Thanks!

      • Voldenuit
      • 12 years ago

      Yeah, imagine if the ES units sent to reviewers were doctored, how badly the Lake Tahoe systems must be fudged. Anyone done any gumshoe work to look at published figures?

        • spuppy
        • 12 years ago

        I don’t think anyone who went to that thing actually published benchmarks, did they?

    • flip-mode
    • 12 years ago

    It’s not as bad as I expected. It basically takes a chip I wouldn’t buy and turns it into a chip I wouldn’t buy.

    Interestingly, after filtering out the benches that I didn’t consider to be worth a damn and then obtaining a new average, that average (12.46%) was very close to the original average.

    The only bench that *REELY* matters to a person like myself is the Window Media Encoder but that’s still essentially a 10% drop and not that far from the bench suite average either.

    So the question is, what’s 13% lamer than lame? AMD is up the creek until the get either a new architecture or a vastly improved mfg. process.

    Thanks for the article TR.

      • shank15217
      • 12 years ago

      it is a significant revision of the k8, i m not sure what you mean by a new architecture. One of the biggest issues with these new processors is that they have just too many architectural changes in one family. They increased ipc, 4 cores per die and a very complex cache hierarchy in one generation. Many of their ipc improvements could have been added to the k8 family without introducing the other improvements but they jumped the gun. This is unlike amd’s previous efforts which were constant incremental performance improvements. The k10 is a very good architecture, it will live up to it’s potential no doubt, but the competition isn’t standing still so even if they get it all together they will just be keeping up.

        • flip-mode
        • 12 years ago

        I mean K11.

    • gratuitous
    • 12 years ago
      • indeego
      • 12 years ago

      /[

        • provoko
        • 12 years ago

        Haha. I was going to put it in red.

      • tfp
      • 12 years ago

      Hey your post is gone, does that mean you are too.

      hahaha

        • eitje
        • 12 years ago

        as of December 11th at 1am, his post is back.

          • tfp
          • 12 years ago

          haha

          • JustAnEngineer
          • 12 years ago

          Is this anything like that disappearing-reappearing ink from /[

    • just brew it!
    • 12 years ago

    That’s very disappointing. They really need to get that B3 stepping out the door ASAP!

    • alex666
    • 12 years ago

    A single question to AMD management: WHAT WERE YOU THINKING?!

    • Voldenuit
    • 12 years ago

    I’ve thought very hard about this, and there is one way for geeks to help keep AMD alive:

    *[

    • Pax-UX
    • 12 years ago

    Come on AMD don’t dye on us! Get that silicon out the door!

    • Bensam123
    • 12 years ago

    Wow, if I bought a AMD chip I’d get a x2. Screw the phenom.

    • srg86
    • 12 years ago

    My current understanding is that clock-for-clock, the Phenom is on average 10% slower than Core 2 Quad 65nm and 14% slower than upcoming 45nm (please correct me is I’m wrong).

    With the TLB patch enabled, that’s now 23.9% (65nm) or 27.9 (45nm).

    That is defiantly crippling. What this does also do is turn a relatively rare bug that you may not affect you (though I’d rather not be in that position) into something that causes problems for almost everyone with the hit in performance. In case of Opterons, at least for the moment, you can’t even buy the things!

    I simply can’t recommend the K10 based chips until the B3 stepping comes out, and then they will have Penryn to account for. Not what AMD needs at all! Once the B3 is out, Phenom, as AMD intended it would have been, will be about 1 year late.

    This is very sad, we need the competition, even die hard intel fanboys need the competition.

    • fpsduck
    • 12 years ago

    Thanks Damage for fast update TLB patch report.

    The situation for DAAMIT is really worse with this patch.
    I think they gonna release more patch
    or fix this bug in the next revision.

    • Maks
    • 12 years ago

    Holy cow! This is nothing less than a castration! Ouch!

    Well, I guess AMD has to give away this crappy Rev. B2-Steppings for as much as a celeron costs.

    • indeego
    • 12 years ago

    Phenom = boring. Let’s move ong{<.<}g

      • shank15217
      • 12 years ago

      move on to what? i think everybody on the board already understands your preference in platforms, so maybe you should say something insightful.

        • tfp
        • 12 years ago

        I don’t think so, the point is there is no need to beat a dead horse…

        It is good to know the impact of the bios fix but I have never seen so many write ups on one topic here at TR. Must be part of the new format/new TR.

          • totoro
          • 12 years ago

          What does that mean, “the new TR?”
          This is why people come here, because Damage and Co. do a thorough job.
          This is a new CPU family and should be the focus for awhile.
          They did the same thing with Conroe.

            • tfp
            • 12 years ago

            I don’t remember them covering anything to this extent before. It’s all a part of the New TR, more opinion pieces, more news pieces, ect. More has changed over the last year then the layout.

            Its not a problem at all, just saying its a difference from a year or two ago.

            I don’t know why you are so defensive its not even your site.

            • Flying Fox
            • 12 years ago

            If by “more” you mean you seem to see them mentioned often, then it is in fact due to the new layout where news and review articles are mixed (seems that blogs are not in the “featured rotation” anymore).

            It is not like TR chose to report so many bad things about AMD, they are just reporting. And the increased in reports is actually a response to our request for “more coverage”. And quite frankly due to the effects of Web 2.0, you have to do more in order to get more coverage or someone else would be doing it. Of course, if there is no problem with the launch there won’t be much to report, right?

        • indeego
        • 12 years ago

        Actually I was what would have been considered an AMD fanboi until 2007. I had never even considered buying an Intel proc/board until then. I even lambasted Intel for letting Anand release benches a full 4 months prior to release of core2duo.

        But then I saw it in action, and I saw it’s execution, and no response from AMD and particularly bad fumbles. If AMD had simply said: Intel, you won on performance, but we’re going to go back with our bread+butter approach of *[value]8, I would have stayed AMD.

        I haven’t looked back since. AMD has only themselves to blameg{<.<}g

      • provoko
      • 12 years ago

      Yea, boring,, that’s why this article got 76 comments in less than a day… haha.

        • indeego
        • 12 years ago

        The article was written expressely to bring lots of viewers in. Reminds me of talk shows: talk about something that is bound to get a lot of peopel riled up, and watch the ad revenue roll in. Rush limbaugh? Dr. Laura? Tom Leykis? All “personalities” that incite rage and fist raising in their listeners.

        TR is now editorializing along with their tech articles. (The line is thin, in fact you can refresh the front page and get either an opinion, a news article, or a benchmark.) I hope the tech articles don’t suffer as a resultg{<.<}g

          • provoko
          • 12 years ago

          Um okay… So you wouldn’t want anyone to bench an erratum to see it’s performance hit?

            • indeego
            • 12 years ago

            Not for a failed platform, no. Move on. Let the dog lick its woundsg{<.<}g

    • achaycock
    • 12 years ago

    I am absolutely horrified with AMDs latest track record. It’s reminiscent of 3DFX starting to go down the drain.

    I have always had a soft spot for AMD. It was the K6-2 that allowed me to have Pentium II like performance (close enough anyway) on a budget. The AMD Duron gave me a huge performance boost after that on an incredibly tight budget and since then, the Athlon, Athlon XP and Athlon64 have all proven to be credible, low cost parts that have offered me more than enough performance.

    The fact is though, this does not make me an AMD ‘fanboi’. I have always been able to acknowledge Intels strengths, it was simply a case that push come to shove, AMD always utilised my existing infrastructure in such a way that they were the lower cost alternative.

    Phenom, however, is too little too late. In fact worse than that, I really cannot recommend it at all for any reason. It is too late, it’s performance is woeful and every time I read about the Phenom, I keep going to different sites to work out which Intel board will form the basis of my next upgrade.

    I sincerely hope AMD are simply experiencing a bad patch. They have some incredibly talented people and we truly need decent competition from both camps to benefit as consumers.

    • DaveJB
    • 12 years ago

    Yowp. I had a feeling this was gonna be bad, but now the memory controller is so slow that the chip looks like a quad-core Athlon XP!

      • shank15217
      • 12 years ago

      a quad core athlon xp would be faster across the board. Infact a quad core athlon 64 at 3.2 ghz woudl actually be fairly competitive on the low end quad core front.

    • DrDillyBar
    • 12 years ago

    Holy Crap. More of an impact then I was letting myself believe.

    • davidedney123
    • 12 years ago

    I get the impression that launching these CPUs in their current broken state was a marketing/sales led decision rather than an engineering led one. Kind of like the Pentium 4. When will the weasels learn that they need to leave the difficult decisions to people who think in more than two dimensions?

    Dave

    • A_Pickle
    • 12 years ago

    You know…

    …I’ll be honest. I used to be something of an Intel fanboy, largely because I hated to think that I’d associate with what I perceived to be a congealing mass that was AMD fanboys in the geek community.

    I was ecstatic when the Core 2 Duo came out and smashed computing records to pieces, it was like saying, “Hah!” to all those AMD fanboys….

    …but I’ve changed quite a bit since then. This article was utterly… and excruciatingly… painful to read…

    • kurt-o
    • 12 years ago

    How about running some benches with the L3 cache disabled? I think those numbers would be interesting to see.

      • Jeff Grant
      • 12 years ago

      Does the patch affect overclocks either up or down? Can the chip be overclocked to compensate for any peformance loss?

      • provoko
      • 12 years ago

      That would be interesting, but how do you do that?

    • Archer
    • 12 years ago

    B3 stepping in March with a fix for this issue.

    That’s a LONG wait.

      • Peldor
      • 12 years ago

      What’s another $600 million among friends?

        • flip-mode
        • 12 years ago

        maybe only 500mill since they’re still selling them to some customers

    • Gungir
    • 12 years ago

    AMD reminds me of a cartoon character repeatedly whacking themselves on the head with a frying pan and sinking a little closer to the ground every time. It would be almost funny if it weren’t such dismal news for them. I really want to support them over Intel, but it’s getting harder every time news like this comes around.

    • moose17145
    • 12 years ago

    All i can think to say is “OUCH!!!”

    • Archer
    • 12 years ago

    AMD, now circling the drain.

    • Firestarter
    • 12 years ago

    This is really dismal, I don’t see how anyone would want a first-generation Phenom now.

      • mortifiedPenguin
      • 12 years ago

      perhaps somebody requiring quad-core performance (rendering or something) and has a preexisting am2 system and needs it now without paying 1000+ for a full system upgrade? not saying that it’s going to be particularly common, but it could happen.

      note: am2+ IS backwards compatable with am2 right? please correct me if i am wrong.

        • crazybus
        • 12 years ago

        You mean $350 upgrade? A quad core Intel and motherboard can be had for that.

    • BoBzeBuilder
    • 12 years ago

    WTH? I thought the patch was supposed to boost performance? what a mess.

    • TakkiM
    • 12 years ago

    Its nice to see AMD documenting their blunders!

    Aside from a dismal launch… they now have to worry about their quality…

      • UberGerbil
      • 12 years ago

      Intel documents their blunders too. And yes, it’s nice of them both (unlike Sun, who requires you to sign an NDA).

Pin It on Pinterest

Share This