- Wednesday, 25 November 2009
- INQ Mobile
- RSS
AMD now stands for Avnet Monopolises De'market - Top AMD reseller
AMD delays Phenom 2.4 GHz due to TLB errata
AMD WILL ONLY LAUNCH the Phenom 9500 and 9600. Even though the channel already got its hands on the Phenom 9700 (2.4 GHz) part, it will have to be pulled off from the shelves.
In a weird deja-vu, it turns out that the company found an errata in the TLB (Transition Lookaside Buffer), just like Intel did earlier this year with complete Core marchitecture. However, unlike Intel, that has a micro-code update function in all of its CPUs, AMD is forced to delay the introduction of the part.
This comes as a huge hit to AMD, at the time that evrything was looking somewhat better. On paper and in practice, its Chipsets Series 7 are world's finest at the moment - PCIe Gen2 implementation is near-perfect, old Athlon 64 X2 parts got a new life on them (insane HyperTransport overclocking options), while Radeon HD 3850 and 3870 are once again sparking a huge war in world of 3D graphics.
However, it turns out that the processor division of AMD botched this time around. We asked AMD to provide us with official statement on this huge problem, and we got a reply back from the firm:
"The Translation Lookaside Buffer (TLB) errata is an L3 protocol issue causing a system hang when running certain client workload applications independent of platform. AMD is immediately introducing an updated BIOS which will correct the TLB errata".
This problem was found during speed-binning the B2 revision processors, and this was the cause for the Phenom FX 3.0 GHz delay. It turns out that some CPUs running at 2.4 GHz or above in some benchmarking combinations, while all four cores are running at 100% load, can cause a system freeze.
While benchmark software vendors had some problems detecting how new memory controller work - both SiSoft Sandra XII and Everest 4.20 were returning single-channel results, regardless of memory controller being used in SC/DC, dual-72-bit or 144-bit- we are still shocked to learn of this situation, that will end up with dire consequences for AMD's management.
AMD already issued a fix to all of its motherboard/system partners, so if you already own a 790FX motherboard or plan to buy a Phenom system, make sure to update the BIOS. 9500 (2.2 GHz) and 9600 (2.3 GHz) parts are unaffected by the errata. Some 9500/9600 parts may even be overclocked to 2.6, 2.8, 2.9, 3.0 GHz and they will have no problems whatsoever, while some will have this error. In any case, you should update the BIOS of your shiny new Phenom system is clocked higher than 2.4 GHz.
According to sources close to the company, 2.4, 2.6, 2.8, 3.0 and faster parts will appear on the market when the B3 revision comes to life. B3 revision is key CPU stepping for AMD's future, since it is considered a speed-bin of B1 and B2. With this errata fixed, Phenoms will have no problems competing against Intel in clock-per-clock action, but it will bring serious disappointment to AMD fanboys and market itself. They are probably blubbing already.
Now, we wonder did AMD learn a lesson here and will the company put micro-code update function in its future cores such as Bulldozer, Fusion and so on, or will the Texas part of the company still consist out of 9 to 5vers who will go home at 5PM and not fix problems that are causing this company to bleed. ยต
Share this:
ShareBusiness Intelligence Software
Customer Relationship Management (CRM)
Enterprise Accounting Software
Enterprise Resource Planning (ERP)
Enterprise Systems Management
TLB is "translation lookaside buffer", not "transition...".
Looking at the way things are going, it should be amd.ati.com, not the other way around.

I mean, 7-series chipset emerged thanks to ATI, HD3800 is also their child. All while AMD is struggling to produce a 65nm part above 2.4 (er, 2.3)Ghz frequency. I used to REALLY like AMD. Used to...
Triumph of business over engineering. I bet they can all point to perfect paperwork. 

AMD is an engineering company not a business. The business part needs only to support the engineering company not nine to five it.
re: "or will the Texas part of the company still consist out of 9 to 5vers who will go home at 5PM and not fix problems that are causing this company to bleed"

If they wanted to they could employ a night shift team for this sort of thing. In fact why not the whole hog and employ round the clock development and testing? 

Do this for 2 years, it will increase costs but should buy them some tech advantage. It won't be cheap, but it's an option.
Honest question:
Do you know what you are talking about ? 

For using a Microcode update, you have to integrate it into the BIOS, so it could be loaded during system start up.

Now you praise Intel for having a MicroCode patch (obviously, loaded via a bios update) and then you criticize AMD for working on a BIOS update (obviously for a Microcode patch).

Please do a proper read up, before you are stating such nonsense ...

cheers

Alex

P.S: 
you might start to read here:
http://www.securiteam.com/securityreviews/5FP0M1PDFO.html 

Is it just me or is it difficult to understand why 2.4Ghz has a big problem and needs to wait for a new revision and 2.3Ghz is just OK?
I just found a video on YouTube and do not know if anyone else is aware of it:

http://www.youtube.com/watch?v=i_q_9a4zDps

An AMD tech demonstrates the Phenom 2.2Ghz CPU and at the end of the vid he takes it to 3.0Ghz. I don't know if this was a handpicked part, but it does show the lowest graded quad Phenom being able to reach 3.0Ghz if not for any specified amount of time.
You obviously haven't noticed, but AMD have had the ability to do microcode updates since the K7. As it turned out, they broke the microcode update functionality in the later K7's, but fixed it again for the K8's.

The thing about microcode updates is that they only allow fixing a small subset of features. Just like Intel can't fix all the Core 2 bugs with microcode updates, AMD can't fix all the K8/Barcelona bugs with microcode updates. The only difference is that one of AMD's bugs is a showstopper ...
"This problem was found during speed-binning the B2 revision processors, and this was the cause for the Phenom FX 3.0 GHz delay." seems to imply that before they found this errata, they had no idea what was causing the problem with 3.0ghz parts. 

So does that mean there is a chance higher clocked parts will be released sooner? I mean, if it's just a bios update to fix the problem, what's the hold up?

Infact, if just updating the bios fixes the issue, why did they need to pull the 2.4ghz part from stores? Did I miss something?
http://www.securiteam.com/securityreviews/5FP0M1PDFO.html
This sounds like a case of a driver transister for a block of memory/registers being too weak (hard to be more specific several thousand miles away from their test lab though)... AMD would not be alone in being caught by this particular class of gotcha. ;)

The good thing is that AMD have spotted this problem *before* they started shipping the faster parts... Anyone remember Intel's FDIV bug ?

If my guess is right AMD should not find this one too hard to fix. In the meantime there is a chance that you might be the lucky owner of a part that was blessed by the Process Gods.
Can you pls post the link to the errata - would love to read it myself
While I do not dismiss that this is an issue, it's not clear to me that this is the cause for the launch delays and lower bins at launch.

Is this issue specific to desktop (Phenom) only? If it is a general issue to K10, why are there no 2.0-2.4GHz K10's available right now? Theoretically they should be able to pump out 2.2GHz and 2.3GHz parts out on Barcy with no problems, no?

Call me a skeptic but this seems like a convenient excuse and I think there are issue(s) beyond this TLB errata limiting the clockspeeds of the Phenom at release.

It also seems quite late in the game to be finding this out, as I believe even this trusted rag mentioned that there were Phenom samples (internal) that were clocking much higher (May 7,2007):

"We understand that people in Austin were 'dancing in the aisles'. When asked if that was because of B0, we were told that people are very happy, very very happy, but he had 'never heard of such a thing'."

"So what do you end up with? A massive gain in frequency. How massive? Almost 500MHz. Instead of the much touted launch parts, look for five SKUs at launch, AM2 quads at 2.6GHz, 2.7GHz and 2.9GHz, a dual at 2.7GHz and a quad on socket F at 2.8GHz."

In the words of Bob Eucker... juuuusssst a little outside...and I suppose now we are to believe the massive shift in bins is due to this TLB errata?

Either AMD knew about this problem much earlier then they have let on and they are now using it as an excuse to cover up other issues, or their validation process leaves a lot to be desired.
If the decode of an instruction includes a vector that doesn't work, you substitute a new vector. This isn't rocket science and Intel wasn't the first to enable fixes.


This couldn't be a problem with an unvectored decode, because that would have turned up months ago.
If this really is a speed-bin problem, why can some 2,3 GHz parts overclock to 3,0GHz without any problems? Why cant AMD just launch some of theese as a 2,4GHz part?
the author does not know anything, truly dumb nutshell inside.
AMD has sucessufully succumbed to automationitis. Each successive chip design beyond k7 gets added delays. All due to a 'don't rock the boat' policy. 
Translation Layoff Beaurocracy.
Anyone daring to stick their head above the mountain of position stuffers is quickly silenced.
To me this is just a big fat lie. The real reason is probably due to low yields.
I'do not not think they add a feature on there cpu, intel since the dawn of time has always sell patched cpu they just found a more economical way of patchin their stuff AMD want to give the better quality produc like they say good stuff sell by himself.
And I would not be happy anyway to buy a cpu 300$+ that need once in a while a patch to work, And every one take that feature it is a great thing, Dam it it don't work take it back and give me somting that work.