benchmarks showing the amd64 x2 doesnt need all its memory bandwith

[Unknown]

link

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2469

Basicly they used ddr500 and in most instances proved no performance
increase, kind of makes you wonder why the move to m2 unless they are
planing for quad cores

mildly ot but if amd were to go for buffered ram would the be able to keep
socket compatibility with 939 boards

 

[George Macdonald]

link

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2469

Basicly they used ddr500 and in most instances proved no performance
increase, kind of makes you wonder why the move to m2 unless they are
planing for quad cores

I think M2 may be because of a concern that DDR-II is going to eventually
dominate the memory market and that there'll be a price premium for DDR.
BTW I'm not surprised at the lowish performance improvement with the new
"dividers" - it's typical of what we've seen with non-clock locked chipset
memory controllers in the past. DDR500 is, from my POV, something you use
if you want to overclock the hell out of the CPU.

mildly ot but if amd were to go for buffered ram would the be able to keep
socket compatibility with 939 boards

You mean registered DRAM? I'm not sure what changes internally in the
Athlon64 memory controller for s939 but the 2T timing gives essentially the
same timing slack and loss of performance as registering so I don't see the
incentive to do registering for the Athon64.

 

[nobody]

link

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2469

Basicly they used ddr500 and in most instances proved no performance
increase, kind of makes you wonder why the move to m2 unless they are
planing for quad cores

mildly ot but if amd were to go for buffered ram would the be able to keep
socket compatibility with 939 boards

AMD did it even before s939 introduction. It was, and still is,
socket 940, and it can use only buffered, aka registered, memory.
Trust me on that - this is being typed on a dual Opty system.

 

[Ar Q]

link

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2469

Basicly they used ddr500 and in most instances proved no performance
increase, kind of makes you wonder why the move to m2 unless they are
planing for quad cores

mildly ot but if amd were to go for buffered ram would the be able to keep
socket compatibility with 939 boards

Any chance that next-gen PC uses parallel lines microweaved XDR2 on either
main memory or video memory for graphic cards? I saw a lot of posts on Yahoo
Rambus messageboard that many posters are sure that ATI or NVidia will use
XDR2 on the new graphic cards in 2006.

 

[George Macdonald]

Any chance that next-gen PC uses parallel lines microweaved XDR2 on either
main memory or video memory for graphic cards? I saw a lot of posts on Yahoo
Rambus messageboard that many posters are sure that ATI or NVidia will use
XDR2 on the new graphic cards in 2006.

The boards at Yahoo are well known pump 'n' dump venues. The graphics card
could go XDR2 I suppose but for system memory, I believe we'd likely have
to accept the abandonment of a multi-drop memory channels and add-in slot
upgrades.

 

[keith]

AMD did it even before s939 introduction. It was, and still is,
socket 940, and it can use only buffered, aka registered, memory.
Trust me on that - this is being typed on a dual Opty system.

Sure, with the advantage that you can drive more DIMMs *reliably*. ...not
that this board (Tyan S2875S) takes advantage of that feature.