MSI's AM2 Mobo Pic's

[cx323]

SLI marchitecture, six S-ATA drives, has 4 DDR 2 dim slots, two PCIe 16X slots, two PCIe 4X slots, two PCI and MSI's special CNR like slot.

http://gathering.tweakers.net/forum/list_messages/1108517

[bluegreenshxt]

oh gawd, already?

Now I feel bad for having just bought so much stuff :(...ah well I ain't upgrading for a couple of years now :p

[stlouis1]

its looks like socket 939 still with a couple extra rows of pins in the socket, is the only difference ddr2 support?

[LazyAzN]

no NB fan eh.?

[BAM!]

i don't see any major advancements shown yet other than support of DDR2. gotta sit this one our for a while before i make the decision to switch. i think 939 will still be popular for quite a while.

[stlouis1]

ya, doesnt dd2 have high latencies?

[BAM!]

yes it does, but who knows.... every socket switch AMD did seem to have some sort of advancemet such as higher clock speed support and memory controller and hyper transport improvement. i just gotta wait it out to see the results people get before i even think about switching.

[stlouis1]

maybe they'll come up with a netburst knock off for they're memory controller to compensate for the latency ;) i guess amd has something going though, they have too, the extra bandwidth would probly help with the HT or something. if not, DDR2 latencies will drop eventually i guess

[Skorp]

Latency isnt really that important for DDR-2 if you push the MHz high enough. You only need to get DDR2-800 running at latencies of 4-4-4-10, where it will have identical latency to DDR-400 at 2-2-2-5, and as you can see here (http://www.anandtech.com/memory/showdoc.aspx?i=2306&p=9), there is RAM out there which can run faster than DDR2-800 and use timings of 4-2-3-8. This is actually lower in absolute terms than the fastest DDR-400 (because latency numbers represent RAM clock cycles). That was a year ago, and DDR-2 production / speeds / latencies / cost will all improve in one great big lump when AMD launches AM2.

AM2 itself mainly brings about DDR2 support, but i've heard fron l'INQ and other places that AMD plan to put PCI-E controllers on the crossbar switch, nestling next to the memory controller. Tripped. Out.

[stlouis1]

so does this mean we're going back to single channel memory with the "faster" ddr2

[Skorp]

uh, say what?

DDR-2 is dual-channel, mate, and while HT provides more than enough bandwith as it is, they're upping the speed of that too.

Dual-channel DDR-400 gives 6.4GB/sec. Overclock that a bit, max ends up being 8GB/sec. Dual'ed DDR2-1066 would give 17GB/sec. Mmm, tasty megahurts, eh?

[therealwesty]

its looks like socket 939 still with a couple extra rows of pins in the socket, is the only difference ddr2 support?

I am pretty sure Socket-AM2 uses the same pin configuration as Socket-940. Without counting the pins it certainly looks like that is the case. Socket-AM2 and Socket-940 won't be compatible of course, even if the CPUs fit into the other socket.

I've read a few places that AM2 would have a 940 pin-count. I threw together this quick comparision image. I put Socket-939 on there so you can see where the pin in missing. Both Socket-AM2 and Socket-940 have the pin that 939 doesn't. So if AM2 is 940 pins it has to be the same as Socket-940.

http://home.cogeco.ca/~cwest7/PCStats/Socket-AM2.jpg

<i>Yeah the image is large, I didn't want to lose the detail when we are looking at wee holes in a piece of plastic.</i>

[Skorp]

Look at the picture closely. Dont look at the pins, look at the lack of them. Look at the pattern of the blank spots. AM2 isnt the same as 940 or 939; all three use seperate patterns.

[yahooadam]

the blancking spots are at the same height

but in different positions horizontally

[LazyAzN]

Does switching to m2 mean higher chance of breaking the pins? Since the pin slots look smaller than 939 and 940.

[stlouis1]

its the one blank (lower right) on the m2 socket thats different, might mean the motherboard pcb's wont have to change much, would make the transition easier and more cost effective, meaning better for the consumers

[yahooadam]

actually no

for eg, the left blank, on M2 is 7 pins away from the edge, on 940 its 7 pins away
The next blank along on 940 is 2 columns to the right of the first blank, on m2 its in line

etc etc

same number of blanks as 940, but their in different places

and no, i doubt M2 will be any more easy to break the pins on
even then ive never had a problem, its a ZIF socket (zero insersion force) so u should never be able to break it, unless u do somthing wrong

[therealwesty]

Ah I can see what you mean with the difference now. Well it will make it one little bit hard for the newbs to screw up if the processor just plain won't fit. I am somewhat glad to see that.

[yahooadam]

yeh, u dont want a proc to fit in a mobo it shouldnt

they do look remarkably similar, if u dont count pins ;)

[Skorp]

For those who havent seen it yet, there have been pictures of Socket F (http://tweakers.net/nieuws/39753) flying around for some time now, and theres probably another topic with them in already. I'm just plugging them here for convenience, next to the AM2 pictures. This socket will have 1207 pins, support the new Opterons for multiprocessor systems, use FB-DIMMS, and use the land grid array system used by Intel on LGA775.

http://skorp.ecwhost.com/1207.jpg

The desktop socket will probably go LGA sooner or later, but its unlikely that we will see FBDIMMs reach our systems, as the main benefit is allowing more RAM sockets per motherboard while making PCB design simpler. Thats also discussed in another topic (http://forum.pcstats.com/showthread.php?t=35162).

[yahooadam]

fbdimms will be a nice improvement

as it fixes issues with interfearence and such

i recon it should catch on for home users too

well aslong as an uber high speed ram thing doesnt come and steal its light