Microsoft Releases Patch For Core Processors

[CalvinHobbes]

http://support.microsoft.com/?kbid=936357

More info:
http://www.theinquirer.net/default.aspx?article=40567

 

[Roguestar]

What does it mean reliability?

"We learned that the affected CPUs are the Core 2 Duo E4000/E6000, Core 2 Quad Q6600, Core 2 Xtreme X6800, XC6700 and XC6800."

Does that mean any processors in the E4X00 and E6X00 range?

 

[CalvinHobbes]

That's what it sounds like to me.

 

[gobucks]

any idea if this will improve performance and/or overclockability?

 

[Azimuth40]

The folks over on slashdot have tracked down some info. It seems to have something to do with the TLB (Translation Lookaside Buffer) which is used to map virtual to real addresses. Random events such as hangs, blue screens or the mouse and keyboard stop responding etc. I have had all of the above during my brief stint with Vista Ultimate retail. I have gone to Microsoft, downloaded it after genuine advantage check of course and have run the usual test routines and back on line after messing around for two hours. So far so good. Then again that is what one person said after jumping of the Empire State building as he passed the 50th floor. We shall see

 

[Blain]

Great...
Now, not only do we have to patch our faulty software, but our hardware too. :|
I knew this C2D stuff was too good to be true.

 

[Roguestar]

Blain is healthily skeptical as usual . I'll download this once I'm at home and remember about it, but I've not been having any crashes or hangs (well, none not attributable to BF2142 not being able to f**king alt-tab properly).

 

[coolpurplefan]

Originally posted by: Blain
Great...
Now, not only do we have to patch our faulty software, but our hardware too. :|
I knew this C2D stuff was too good to be true.

Eh, what's kind of funny is I got two single cores to avoid gaming issues with the X2. And now this news comes out. Oh well, hahaha, maybe we'll have to wait for months to see what happens with AMD's new K10.

 

[VirtualLarry]

<div class="FTQUOTE"><begin quote>Originally posted by: coolpurplefan
Eh, what's kind of funny is I got two single cores to avoid gaming issues with the X2. And now this news comes out. Oh well, hahaha, maybe we'll have to wait for months to see what happens with AMD's new K10. </end quote></div>
Was it "studdering" with games? The new MS v4 multi-core patch for XP is supposed to help with that. It avoids ping-ponging threads between CPU cores.

The real issue is somewhat a software defect, and somewhat a hardware design issue. The problem with current dual-core designs is that they are two "cloned" cores on a die. That means two seperate time-stamp counters (high-resolution counters that keep track of accumulated timer ticks), and due to other factors, they aren't syncronized. That means that software that was originally designed for single-core only uses the TSC to compute and compare elapsed time, and if the software thread gets switched between cores that have differing TSC values, it's possible that the TSC appears to go backwards in time to the software. The original design and intent and usage of the TSC never included that possibility. Thus current software bugs out when ping-ponged between two cores with differing TSCs.

The ideal solution would be for all cores to share a single TSC value, that would always appear to move forward, to all software running on any core. But that would require some re-design of the chips, to allow for that.

I've seen the original Unreal Tournament running buggy on a dual-core rig, and it's not pretty. The game's display updates freeze up and then "warp". Smooth movement is impossible, and the game is a mess to play.