i7 has a TLB BUG

I'm sure that Intel won't release a new batch of Core i7 like AMD did with Phenom.

Why not? Intel generally releases multiple steppings of every architecture over time.
We've seen that with both the 65 nm and 45 nm Core2s.
So aside from the tick-tock scheme, they generally also release a new stepping roughly every 6 months.
(AMD does roughly the same, it also took AMD about 6 months to release the new Phenom stepping, it was just obscured a bit because the launch of the first stepping was already delayed by 3 months).
 
There are actually two separate issues, 1) the SPEC CLARIFICATION AAJ1 and 2) the errata on Core i7 that relate to the TLB.

Correct, people seem to confuse the two, thinking that programmers have to work around some kind of TLB issue in the Core i7.
The 'heads up' to programmers is the newly defined TLB behaviour that was introduced last year on the Core2.
This document lists only 3 bugs related to the TLB:
http://download.intel.com/design/processor/specupdt/320836.pdf

For two of them there are workarounds, either in the OS (AAJ54) or via a microcode update (AAJ42). The third (AAJ1) is very insignificant (when a TLB error occurs, the overflow bit in one of the status registers is undefined. Which it generally was anyway, just because you don't know what code was executed earlier... In this case the difference is that even if you did know what the status was earlier, you still can't rely on it being the same after the TLB error occured.... But you were never interested in that bit anyway, at that point, so it is of no consequence).
 
The improved steppings usually increase yields also, so why wouldnt they release a newer chip down the line.
 
When I said a new batch I meant a new stepping with a fix in the silicon that might give a better performance, not just a plain new stepping.
 
When I said a new batch I meant a new stepping with a fix in the silicon that might give a better performance, not just a plain new stepping.

dude come on, The fix might cause an undocument performance hit, but who cares. It's not like there were hundreds of reviews out that praised its performance and then it suddenly sucked ass when released into the wild. Core i7 is what it is. Those who have it got exactly what they were promised. Intel could have increased performance by drinking more koolaid, jumping up and down on their heads, and throwing mud at each other. It doesn't matter since we're getting what we're promised. Hell you might as well bitch about a Celeron that could have been Core 2 if Intel hadn't "fixed" it before it left the plant.
 
Back
Top