AMD M2.... new info (I think)

[K404]

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2476&p=1

And to summarise...

M2: 940-pin, not compatable with 939 or existing 940pin procs. DDR2 memory. 1207 is still around a year away.

Release: Q2 2006

Rumoured to have an M2 4000+ as a 2.6 GHz 512K L2 cache. To me, sounds like the DDR2 version of what we`re using with 200MHz added.

More power draw...exceeding 100W for ALL chips. (who said heat wouldnt change??)

Theres more, but cant be bothered picking it apart right now

Kenny

P.S...Please dont shoot the messenger

 

[Alanore]

TBH it looks kind of like Socket M2 is a stop gap measure until AMD can gets its new FAB up and in volume production.

There going to be an increase in the TDP so that the FX-59 is possible.

After the new fab is in production AMD has a good chance to flood the market with once again cheap CPUs, so it wont really matter about the power consumption as Intel has no significant peroformance advantage with its 'back to the future' Pentium 3 archiecture.

After Intel strike back with their low power dual core/Quad core. AMD will move over to the 65nm process. Possible a quicker migration than from 130nm to 90nm as they still have another FAB thats capable of meeting todays requirements thanks to its yeild rate.

Possible in the time when the two FABs are running together companies like DELL may consider taking up AMD processors espically thanks to the AMD legal case against Intel, even if AMD loss they have won becuase intel will be in fear they will repeat the law case with new evidence, And the fact that AMD can suply the demaned. Its likely that they will at best use Laptop and Server chips.

On the major front AMD are not likely to overhaul their processors soon, they have the winning forumula, so only tweeks are needed to keep up with the 'M'. It looks as though Intel will take the advantage when it moves to the 65nm process. Although chances are when AMD get around to the 65nm process they can do a better job of it, thanks to their advanced FABing techniques.

I also look forward to Intels second attempt at Hyper threading. TBH the Pentium 4 was a total f**k up, apart from something that contained the most predicatble branch predictions, like media encoding. Its cache hits ratio made it so bad that it costs performance. If anything they way programs were compiled was its saving graces, as well as the symthetic benchmarks. Although hort pipeline processors really dontbenifit much from HT.