- Friday, 10 July 2009
- INQ Mobile
- RSS
Drawing on my fine command of the English language, I said nothing - Robert Benchley
Intel denies Core i7 TLB bug
IT SEEMS INTEL’S new Core i7 chip may be a bit soggy on the translation lookaside buffer (TLB) front, with the firm hinting at a bug in an official Specification Update.
A TLB, found in pretty much all current desktop and server processors, is a CPU cache used by memory management hardware to improve the speed of virtual address translation.
Chipzilla’s spec update notes "In rare instances, improper TLB invalidation may result in unpredictable system behavior, such as system hangs or incorrect data". The document goes on to say "Developers of operating systems should take this documentation into account when designing TLB invalidation algorithms. For the processors affected, Intel has provided a recommended update to system and BIOS vendors to incorporate into their BIOS to resolve this issue."
A TLB flaw could have serious implications for Intel’s street cred, but the firm insists there is nothing to get excited about. "This is a spec clarification that points back to a programmer's application note written by Intel in April 2007. The same spec clarification is in our Penryn docs pointing to same app note, which advises on programming techniques to avoid any issue", an Intel spokesman told the INQ.
The Intel spinner went on to say that, as far as the spec clarification was concerned, the problem didn’t affect Core i7 or any Nehalem processor. "That was for Core 2 Duo and that has been fixed via BIOS updates long ago. Any outstanding software programmer developer manual changes are shown in all product spec updates so it was listed", he noted.
The INQ was also assured that if one were to look through the various Core i7 updates in the errata part of the spec update, there would be some mention of TLB (unrelated to the previous mention), but that anything significant had been fixed before Core i7’s launch. µ
L'Inq
Intel’s official Specification Update (page 37 AAJ1 Clarification)
Share this:
Business Intelligence Software
Customer Relationship Management (CRM)
Enterprise Accounting Software
Enterprise Resource Planning (ERP)
Enterprise Systems Management
If you read the document that supposedly is from last year and supposedly contains Core 2 Duo errata, it clearly has on every page:

"Intel® Core™ i7 Processor
Specification Update, November 2008"

Nice try Intel, but unfortunately some people can read.
Has anyone else noticed the two phrases in the document #320834 page #7:

* Supports Intel 64 Architecture

* Optimized for 32-bit applications runnig on advanced 32-bit operating systems

Two thoughts come to mind:

1) My impression of AMD's approach to the x86 arch has been design for AMD64 and support backward compatibility w/ x86-32, while Intel designs for x86-32 w/ support for iAMD64.

2) Does Intel still not want to "fully" support iAMD64 in an optimal fashion, in an effort to avoid undermining the good-ship Itanic?

Me wants more 64-bit benchmarks...

HB
Let's see...
AMD has a TLB bug in Barcelona that causes system hangs under rare, but recreatable circumstances.
AMD comes clean and takes a massive hit in the market and on Wall Street, still talked about today - "The infamous TLB bug" "AMD Barcelona mucho massive suckage" etc.

Intel has a Core i7 TLB bug that causes system hangs under rare, but recreatable conditions.
Intel does not come clean, looks the other way, takes no hit in the market and no hit on Wall Street.

So.... either AMD's PR dept. is incredibly dumb, or Intel's business ethics are incredibly impaired.

Not sure of the answer here, but after reading about the infamous Intel/Microsoft/Screw HP/Screw The Customer "Vista Capable" fiasco, I'm leaning towards Intel's market ethics being in need of a revamp.
Admitting a TLB bug like AMD did with barcelona is clearly the wrong way.
AMD only got bad publicity by being honest intel doens't want the same thing.
So essentially its the same sort of issue AMD had with Barcelona? Funny, one company does poorly based on such a thing but I daresay people will still buy the i7 regardless.

Mind you, Barcelona lack of performance probably didn't help matters much.
Has our world ever been an honest man's world? Clearly not, and honesty DOES NOT pay, especially in business. However, as long as AMD stays in the game, I will be buying their CPUs. Because those cute bastards from Intel sorely need competition...
I agree with brainkilla in one thing - I will buy AMD as long as they will be in market. But I personaly prefer people who play game by rules. That marketing step wasn't wrong step. That way AMD said that they will play this hard for them game honestly to the end. And many users like me heard it right - AMD will not hide the truth! I can forgive many things but I hate lie!
***
BTW, seven out of top 10 supercomputers build on AMD CPUs. Two - on IBM PowerCell CPUs (?) and one - on Intel Xeon. No, Intel CPUs are not bad, they are just expensive! So my home supercomputer will be on AMD too ;)