http://siliconinvestor.advfn.com/r [...] d=24196847

I have read that the TLB bug is minor and most would never see it, and that a lot of reviewers have could not even cause the bug to happen.

Now these guys claim it is till there, curious, where did they get this information? I have read that the bug causes shutdowns and the old bluescreen of death in windows. Sure a lot of missinformation to sift through.

That article claims that B3 still has the bug.

Not even AMD is that incompetent!

I don't trust that source.

I agree that this seems pretty fishy. Who would even have a B3 stepping CPU to test at this point besides AMD's labs?

I don't trust that source. Mainly because as far as i can remember AMD have said B3 wouldn't be ready until February/March and the fact that their MAIN objective with B3 was to get rid of the TLB bug. I'm sure even Hector wouldn't let them release it without the bug fixed.

They article states that the "Early Silicone" of B3 has not fixed the bug.
All this means is that AMD has not yet fixed the bug.

Considering that this bug is not a new bug, but a long standing bug that AMD has been working on, it's no surprise they have not yet fixed it.

They are simply stating that AMD is still trying to perfect B3.

Seems to me that even with early silicon, a bug that they had to stop ship their flagship server procs on would be the FIRST thing fixed - so I wouldn't put too much stock in this info...

Isn't this the same French website that started the rumor that Intel was pushing back the 45nm Core 2 Quads due to erratum?

Sounds like FUD to me...

I believe Intel probably paid the Author of this article to be negative about AMD.

Highly Suspicious, don't you think?

Well, The point of the early silicon is to test the fix.
AMD has known about the bug for since early 2007, but assumed it was speed releated and that is why the first shipping chips were so slow.

Turns out the error was happening in slower chips than AMD imagined so they needed to finally admit the issue.

The problem now is that they still need to fix the bug that they have had for a year. It's likely not a simple fix, but could be more of a design issue.

More and More it appears to be a minor issue at worst for desktops who could be set to ignore the error and most will rarely or never see the problem.

The big issue is on the server side where AMD is a big player in the virtualization market where the big is most seen. Here they really need to get it fixed.

I would not be surprised if AMD does release a "B3" type revision on the desktop that lacks the fix if they can get enough other goodies into the rev to make it work while.

Why in the world would the bug still be in B3 if the point of revising the chip was to fix it?

Why would Intel spend any money having someone write a negitive article about AMD? AMD supplies this for free and has for the last year.

I make no claims about the accuracy of said article.

However, B3 HAS to be alive and well at this stage of the game if AMD hopes to sell in 1-2 months time. There will be ES of this stepping. Otherwise, you can't do the required in-house testing. The only difference between B3 ES and saleable B3 is, well, they package it for sale. (I am convienently ignoring subtle manufacturing process tweaks here). Either the logic works or it don't.

It is quite possible that the fix for the TLB bug will not appear in B3. As Baron is fond of saying, CPUs are VERY complicated. Even wtih the brightest minds, they might not get it right the first (or second time). That would blow for AMD, but it's not as if there aren't good, solid engineers workign long hours on the problem.

It is also quite possible that this is FUD.

But, it is impossible for the B3 stepping NOT to exist at this point in time.

I disagree. That would just mean that B3 gets delayed. A very familiar thing with AMD.

Thunderman should see this post