More info regarding Phenom TLB issues

[taltamir]

Originally posted by: Zstream
The bios update removes the potential of hitting this area in the CPU which causes the bug. So, with the bios update it is 10% slower then normal. So learn your tech and come back to post.

According to AMD it is up to 10% slower once the crashing part of the chip is disabled...
Some people though claim it is 10-20% in reality. But I cannot be sure about their credability...

But if AMD admits to it being 10% slower I would expect it to really be even slower then that...

 

[SX2012]

You guys are blowing this way out of proportion

-If your going out of your way to buy a amd's long awaited quad core chip at launch and build it yourself your probably also aware of something called the internet, and early adopters should always know this kind of stuff can happen with early revisions. When i bought mine I already knew of this Errata BS but didnt give a shit

-AMD acknowledged a bug thats hard even to reproduce in a lab and that doesn't effect probably 95+ percent of the customers who bought the chip but now we all think we cant play Crysis anymore now

-They pulled most of the product line in response but still allowed us to get our hands on atleast something instead of making us wait for the next stepping

-Every processor ever made has imperfections like this more or less severe

-It seems like people are posting like they received broken CPU's

however AMD should have caught a potentially severe bug like this before mass production through Quality Assurance especially if they were oblivious while fabbing the opterons, i would fire them all right now and start from scratch, probably with people from aerospace or automotive, they never launch bad products HAH! Come to think of it, in the grand scheme of things this Errata is so minor i should format the internet and go to sleep.

 

[bfdd]

Really not that big a deal? Crashing during Photoshop benching? It's in EVERY cpu and the only way to fix it is to decrease performan by 10-20% that's not a big deal at all...

 

[taltamir]

nope. its completely normal for a cpu to crash while running photoshop etc... If you don't want it to crash a 10-20% decrease in performance is perfectly acceptable... Even though it wasn't acceptable in say, every other cpu ever released by AMD or Intel.

 

[harpoon84]

Well, since he already bought the Phenom, he has to justify the purchase one way or another, right?

Denial is always the first step.

 

[taltamir]

I wonder if he will install the bios patch... or will he do as AMD suggested that enthusiasts could run it in the faster, unpatched mode, if they are willing to risk occasional crashes... I think they tried putting a nice spin on that though... I guess if you are willing to OC why wouldn't you be willing to run an unstable chip at stock...

Actually, with an FX board he might be able to overclock it once the defective part of the chip is disabled.

 

[harpoon84]

Originally posted by: taltamir
I wonder if he will install the bios patch... or will he do as AMD suggested that enthusiasts could run it in the faster, unpatched mode, if they are willing to risk occasional crashes... I think they tried putting a nice spin on that though... I guess if you are willing to OC why wouldn't you be willing to run an unstable chip at stock...

Actually, with an FX board he might be able to overclock it once the defective part of the chip is disabled.

http://techreport.com/articles.x/13741/3

I'd advise performance junkies to avoid the 'patch'... unless they are having SERIOUS stability problems. Then again, why would a performance junkie be running a Phenom anyway? LOL

Personally, I'd take the occasional crash over such a significant performance penalty.

Of course, the best course of action would be to avoid Phenom altogether until it is fixed.

 

[Viditor]

Originally posted by: harpoon84

Originally posted by: taltamir
I wonder if he will install the bios patch... or will he do as AMD suggested that enthusiasts could run it in the faster, unpatched mode, if they are willing to risk occasional crashes... I think they tried putting a nice spin on that though... I guess if you are willing to OC why wouldn't you be willing to run an unstable chip at stock...

Actually, with an FX board he might be able to overclock it once the defective part of the chip is disabled.

http://techreport.com/articles.x/13741/3

I'd advise performance junkies to avoid the 'patch'... unless they are having SERIOUS stability problems. Then again, why would a performance junkie be running a Phenom anyway? LOL

Personally, I'd take the occasional crash over such a significant performance penalty.

Of course, the best course of action would be to avoid Phenom altogether until it is fixed.

I agree...at least that's what I'm doing.

 

[AmberClad]

Originally posted by: ViRGE
It's a hardware flaw. They can work around it with microcode at a sizable performance penalty, but fixing it requires a new spin of the processor.

I suppose that answers my other question, which would have been -- how has this whole issue grown to such proportions? If I recall, Core 2 has had it's own various errata, including one with the TLB buffer, but they apparently quietly patched those and everyone went about their business like nothing ever happened ...

 

[bradley]

Originally posted by: AmberClad

Originally posted by: ViRGE
It's a hardware flaw. They can work around it with microcode at a sizable performance penalty, but fixing it requires a new spin of the processor.

I suppose that answers my other question, which would have been -- how has this whole issue grown to such proportions? If I recall, Core 2 has had it's own various errata, including one with the TLB buffer, but they apparently quietly patched those and everyone went about their business like nothing ever happened ...

Pretty much. Disscusion of Intel's TLB bug was fairly limited on these forums. Although, the responses are in sharp contrast to the ones currently found here.

http://forums.anandtech.com/me...AR_FORUMVIEWTMP=Linear

 

[v8envy]

There's a world of difference between the Intel errata, where Intel did not stage a Tahoe press event with carefully selected apps to show the expected performance of their gear knowing full well a 0-57% performance hit (depending on the app) will land in the customer's laps a week later and this.

Yes, hardware and software can and does have bugs. The Intel fix, while kept under wraps did not bring their CPU performance down to the same level as the Netburst product. AMD's errata brings the Phenom performance in line (or below) that of existing X2s, never mind core quads they are ostensibly competing with.

That said, you could just pretend the Tahoe event was showcasing the B3 stepping processors coming in Q1 of next year, like techreport said.

I'm very glad there are plenty of AMD supporters willing to buy the current generation product. More power to you guys, please keep buying so cheapskates like me will have a CPU choice in the future.

 

[harpoon84]

Originally posted by: bradley

Originally posted by: AmberClad

Originally posted by: ViRGE
It's a hardware flaw. They can work around it with microcode at a sizable performance penalty, but fixing it requires a new spin of the processor.

I suppose that answers my other question, which would have been -- how has this whole issue grown to such proportions? If I recall, Core 2 has had it's own various errata, including one with the TLB buffer, but they apparently quietly patched those and everyone went about their business like nothing ever happened ...

Pretty much. Disscusion of Intel's TLB bug was fairly limited on these forums. Although, the responses are in sharp contrast to the ones currently found here.

http://forums.anandtech.com/me...AR_FORUMVIEWTMP=Linear

The difference is Intel's microcode update didn't effect performance by 14%... and they didn't claim it to only affect a certain clockspeed and use it as an excuse of why a higher end SKU was pulled at the last minute.

Come on, open your eyes for a bit, the situation is FAR different. AMD had bad press from the moment it used the TLB as an excuse for not launching the 2.4GHz Phenom, and the whole Lake Tahoe PR exercise just put everyone on edge as it seemed something was not 'quite right' about the Phenom launch... and they were right.

 

[myocardia]

Originally posted by: AmberClad
I suppose that answers my other question, which would have been -- how has this whole issue grown to such proportions? If I recall, Core 2 has had it's own various errata, including one with the TLB buffer, but they apparently quietly patched those and everyone went about their business like nothing ever happened ...

I would think most of the reason it has gotten so much attention was that Intel went to work immediately on getting rid of their TLB bug, while AMD has known about theirs for what 3 or 4 months now, and it will be roughly 2 more months, before the first Phenoms/Barcelonas ship, that don't have the bug.

 

[AmberClad]

Originally posted by: myocardia
I would think most of the reason it has gotten so much attention was that Intel went to work immediately on getting rid of their TLB bug, while AMD has known about theirs for what 3 or 4 months now, and it will be roughly 2 more months, before the first Phenoms/Barcelonas ship, that don't have the bug.

That, yes, and I think it also helps to have semi-competent PR people who don't around uttering comments like this :

AMD spokesman Phil Hughes told us the TLB issue has been designated errata number 298. When questioned about when AMD would update its technical documentation to include the erratum, Saucier said the person responsible for the updates is "on vacation," although he expects an update "by the end of the year."

Ideally, you would want your PR guys to give the impression that everything's under control, not that your people are off on vacation while the house is burning down.

 

[bradley]

Seemed the Intel bug displayed itself enough to make a microcode fix necessary. This also included Xeon processors, making it more than incidental. According to Intel, applying the microcode doesn't cause a speed hit, but I've never actually seen this claim tested. Obviously both AMD and Intel have microcode upgrade capability, and only AMD's has ever forced a recall. Guess it was the luck of the draw.

 

[bradley]

Originally posted by: myocardia


I would think most of the reason it has gotten so much attention was that Intel went to work immediately on getting rid of their TLB bug, while AMD has known about theirs for what 3 or 4 months now, and it will be roughly 2 more months, before the first Phenoms/Barcelonas ship, that don't have the bug.

No offense, but how can we be certain of whether Intel was aware of the bug? And if so, what does it say about Intel to have taken several months to discover it; I believe the first revision was in July 2007.

No question this was handled very poorly by AMD, and perception is a large part of reality. But at least buyers have prior knowledge as now to avoid both Phenom and Barcelona. I certainly remember scratching my head wondering about my C2D instabilities, until after applying the microcode.

 

[taltamir]

Handling it poorly is a very important deal. AMD made it perfectly clear that it knew, and was hiding it. They got cought with their pants down. If intel knew then it kept the illusion of innocence...

Most people think there is a difference between having an accident and doing something on purpose. And anyone who did NOT buy a phenom, like me, isn't mad at the performance decrease (why should I? it actually justifies my choice to get a X2 6400 as a drop in replacement). No, we are mad at their handling it, at the deliberate lies and misdirections.

 

[bradley]

OK. Although, these conversations remind me of the technological equivalent of sporting event fanaticism. It's crazy. For instance, I'm a fan of both the Mets and Yankees, but my love for baseball overrides either. Same holds true for Intel vs. AMD. I'm more of a fan of technological advancement in a fair marketplace. I understand both of these concepts will be called into question by NY baseball and CPU fans alike.

Seems Intel is mostly given the benefit of a doubt, and AMD is held to an elevated standard. I'm a pragmatist and try to envision an even-handed standard, of course, the realist in me sees this as a social impossibility. And I certainly don't want to stop anyone from rooting for or against their favorites. I'm actually more surprised at how many people apply these patches without ever reading the description.

Similar bug, problem, and workaround, different outcome. That's the bottom line for me.

 

[taltamir]

Except what you see here is AMD fans feeling hurt and betrayed.. For the past four years I vehemently told everyone that there is absolutely no reason to ever buy intel.

And then they go and do this to me.

 

[myocardia]

Originally posted by: bradley
OK. Although, these conversations remind me of the technological equivalent of sporting event fanaticism. It's crazy. For instance, I'm a fan of both the Mets and Yankees, but my love for baseball overrides either. Same holds true for Intel vs. AMD. I'm more of a fan of technological advancement in a fair marketplace. I understand both of these concepts will be called into question by NY baseball and CPU fans alike.

Yet the last two people to post in this thread (besides you) did so using an AMD CPU. Or is the new definition of Intel fanboy one who owns more AMD CPU's than Intel CPU's?

 

[taltamir]

Not besides me (edit: it seems it wasn't referring to me after all, oh well), I ALSO own an AMD cpu, in fact I own many. I have a X2 6400 black edition in my main comp. an athlonXP 2500 barton in my second comp. I have a X2 3800+ sitting on the table besides me, as well as an athlon XP 2500+ barton and a 2400+ ALSO sitting on the table... I need to sell them on ebay when I get the chance...

My last intel CPU was my Pentium3 550mhz bought in 2003. The next CPU I use, either for me or for my family, will be an intel though... unless AMD makes some magical comeback.

 

[bradley]

I mostly have no problem with either company. I've built both amazing systems surrounding AMD and Intel chips, and truly plodding, kludge-ridden, buggy pieces of dung beetle food. We live in the most advanced technological age in the history of the world, and everyone's too busy complaining to even take notice. The world isn't going to come to an end owning a system 10 percent slower than the average.

 

[bradley]

Originally posted by: myocardia

Yet the last two people to post in this thread (besides you) did so using an AMD CPU. Or is the new definition of Intel fanboy one who owns more AMD CPU's than Intel CPU's?

I'm not sure what you mean or whom you're referring to with this statement?

 

[myocardia]

Originally posted by: bradley

Originally posted by: myocardia

Yet the last two people to post in this thread (besides you) did so using an AMD CPU. Or is the new definition of Intel fanboy one who owns more AMD CPU's than Intel CPU's?

I'm not sure what you mean or whom you're referring to with this statement?

On a forum such as this, the person writing a post is always referring to the person they quoted.

 

[AmberClad]

Originally posted by: bradley
Seems Intel is mostly given the benefit of a doubt, and AMD is held to an elevated standard.

I certainly don't think it's always been that way. If anything, it used to be the opposite, with many people rooting for the underdog, as well as calling shens on the early preview benchmarks of Conroe.

But since Conroe, the "new Intel" has seemingly tried to be very open about their new and upcoming products. Things like seeding almost every hardware review site on the web with ES QX9770 chips...even when the chips aren't 100% ready for primetime (I got a good laugh out of the "QX9770 - Too Hot for TV" headline). Their higher ups, especially Gelsinger, have been pretty willing to grant interviews. I think that's scored them some points with the media/reviewers, and that may have filtered down to the enthusiasts.

AMD on the other hand, seems to have lost touch a bit with their enthusiast support base. From reading Anand's blog entries, getting details out of them about their upcoming plans has been pretty tough. They've been focusing on top tier sales (Dell, HP, and Acer, for example), and you hear smaller channel partners and people like ASUS's president complaining about the inadequate supply of AMD/ATI chips and feeling screwed over.