AMD to introduce Socket M2

[Yousuf Khan]

It's a 940-pin socket, but incompatible with existing 940-pin Opteron
socket. It supports DDR2 memory, and it should be here by the beginning
of 2006.

The Tech Report - AMD to standardize on Socket M2
http://techreport.com/onearticle.x/8927

Yousuf Khan

 

[Rob Stow]

It's a 940-pin socket, but incompatible with existing 940-pin Opteron
socket. It supports DDR2 memory, and it should be here by the beginning
of 2006.

The Tech Report - AMD to standardize on Socket M2
http://techreport.com/onearticle.x/8927

Yousuf Khan

Have you been able to find anything solid about whether the
Socket M2 chips will support DDR *and* DDR2 or whether DDR
support will be dropped ?

I would expect it to support both, as AMD initially hinted at
last year, but every single socket M2 reference motherboard I
have read about so far has been DDR2 only. It is quite possible
that I am hearing about the same DDR2 motherboard over and over
again, but at the same time I haven't read even a rumour about
testing Socket M2 with DDR.

 

[George Macdonald]

It's a 940-pin socket, but incompatible with existing 940-pin Opteron
socket. It supports DDR2 memory, and it should be here by the beginning
of 2006.

The Tech Report - AMD to standardize on Socket M2
http://techreport.com/onearticle.x/8927

Socket M2 for "desktop", i.e. single socket systems. The originally
expected 1207-pin Socket M2 will be Socket F for multi-socket systems
according to http://www.dvhardware.net/article4872.html & Digitimes "free"
click-thru. Also here
http://www.xbitlabs.com/news/cpu/display/20050510043057.html with pics of
S1 & M2.

I hope we're not going to go to yet another heatsink retention mechanism
for M2.

 

[nobody]

Socket M2 for "desktop", i.e. single socket systems. The originally
expected 1207-pin Socket M2 will be Socket F for multi-socket systems
according to http://www.dvhardware.net/article4872.html & Digitimes "free"
click-thru. Also here
http://www.xbitlabs.com/news/cpu/display/20050510043057.html with pics of
S1 & M2.

I hope we're not going to go to yet another heatsink retention mechanism
for M2.

Anyway to upgrade to the new CPU you'll have to replace your
motherboard and memory, so don't cry about the heatsink. Besides, if
you buy the CPU retail, the new heatsink will be included.
NNN

 

[Rob Stow]

So S1 and M2 are going to end the days when you could plug an AMD
"mobile" CPU into a desktop motherboard.

That's too bad - I loved my Athlon XP-M in my old Socket A
desktop, and I've been reading that the Turion is turning out to
be a nice chip for Socket 754 desktops.

 

[YKhan]

Have you been able to find anything solid about whether the
Socket M2 chips will support DDR *and* DDR2 or whether DDR
support will be dropped ?

Well, somebody here pointed out that DDR2 is a superset of the DDR1
specs, so it's likely that DDR1 could be supported by the same memory
controller. It's simply possible that most mobo makers aren't bothering
to put the support in. They expect the initial mobos to be performance
mobos, so most people will be equipping with DDR2. The DDR2-800 has
likely reached performance points now where it's higher per clock
latency is now acceptable in absolute terms.

Yousuf Khan

 

[YKhan]

Yeah, but the Socket S1 638-pin is likely what is necessary for AMD to
achieve even better power thermal characteristics on laptops.

Yousuf Khan

 

[Rob Stow]

Well, somebody here pointed out that DDR2 is a superset of the DDR1
specs, so it's likely that DDR1 could be supported by the same memory
controller. It's simply possible that most mobo makers aren't bothering
to put the support in. They expect the initial mobos to be performance
mobos, so most people will be equipping with DDR2. The DDR2-800 has
likely reached performance points now where it's higher per clock
latency is now acceptable in absolute terms.

Thanks. That give me more reason to hope that DDR backward
compatibility actually gets implemented in the socket F and
socket M2 chips.

I know people with rooms full of servers who are already getting
anxious about the prospect of not being able to carry their large
investments in ECC Reg DDR forward with them when they eventually
migrate to the next generation of servers. Given the likelihood
that they will end up replacing servers rather than upgrading
them, I'm not sure how legitimate their anxiety is, but it is
there nonetheless.

 

[Kai Harrekilde-Petersen]

Thanks. That give me more reason to hope that DDR backward
compatibility actually gets implemented in the socket F and socket M2
chips.

As I recall, the IO drivers are very similar - SSTL_2 and SSTL_1.8.
Mixing SDR and DDR IO drivers are a much bigger pain in the a**.

Cheers,


Kai

 

[George Macdonald]

So S1 and M2 are going to end the days when you could plug an AMD
"mobile" CPU into a desktop motherboard.

That's too bad - I loved my Athlon XP-M in my old Socket A
desktop, and I've been reading that the Turion is turning out to
be a nice chip for Socket 754 desktops.

Could be that S1 is targeted at the thin 'n' light market. They still make
Athlon64 for DTR and "mobile" as well as Turion so a Socket M2 for low
powered Athlon64 in a notebook is not impossible.

 

[George Macdonald]

Anyway to upgrade to the new CPU you'll have to replace your
motherboard and memory, so don't cry about the heatsink. Besides, if
you buy the CPU retail, the new heatsink will be included.
NNN

No it's not about reuse of an old heatsink - it's the delay for all the
heatsink mfrs to build to the new standard. K8 heatsinks took a while to
proliferate beyond a couple of Rube-jobs and are still nowhere near as
abundant as socket 370/A and P4 heatsinks - it took a while to have a
decent set of good, reasonably-priced, low-noise choices... and the one I
want is often out of stock. I hate that feeling, that if the original
fails it's nigh impossible to find an emergency supply.

 

[Grumble]

No it's not about reuse of an old heatsink - it's the delay for all the
heatsink mfrs to build to the new standard. K8 heatsinks took a while to
proliferate beyond a couple of Rube-jobs and are still nowhere near as
abundant as socket 370/A and P4 heatsinks - it took a while to have a
decent set of good, reasonably-priced, low-noise choices...

Huh... Georges?

I don't think heatsinks produce any sound.

:-þ

 

[Felger Carbon]

Huh... Georges?

I don't think heatsinks produce any sound.

HSs don't, but HSFs are more commonly sold, and HSFs certainly produce
acoustic sounds/noises. By "more commonly" I mean 99.997%. ;-)

 

[Yousuf Khan]

Could be that S1 is targeted at the thin 'n' light market. They still make
Athlon64 for DTR and "mobile" as well as Turion so a Socket M2 for low
powered Athlon64 in a notebook is not impossible.

Yes indeed, and I'm pretty sure the smaller number of pins will allow it
to be fit into tinier little palmtops as well as notebooks.

Yousuf Khan

 

[epaton]

perhaps someone could enlighten me why this is a good idea, if i recall
there were some benchmarks out showing that going up from 400 to 500mhz on
the fastest dual core amd64 only gave a small boost to one specific test,
basicly proving that amds arn't bus limited.

now since i assume they will need more pins and a new socket when they
integrate the pciexpress controller its just more confusion.

 

[keith]

perhaps someone could enlighten me why this is a good idea, if i recall
there were some benchmarks out showing that going up from 400 to 500mhz on
the fastest dual core amd64 only gave a small boost to one specific test,
basicly proving that amds arn't bus limited.

Maybe if you included some hint of what you're talking about...

now since i assume they will need more pins and a new socket when they
integrate the pciexpress controller its just more confusion.

Confusion? Why would an end-user care about pins? Conversely, if you
care about pins, perhaps you'll take a few minutes to figure out what's
going on. In short, there are *many* reasons to change sockets. It's not
done "just because".