Do Dual-Core CPU's require Dual-Channel Memory?

[Karma Police]

AMD has said that a simple BIOS upgrade on existing socket 939 motherboards will allow a dual-core CPU to be used.

But socket 939 was created after socket 754 to add dual-channel memory, while 754 only has single-channel memory.

Which brings me to this question. Did AMD put the dual-core CPU's in the 939 sockets & not 754 sockets because dual-core chips will need dual-channel, or it it just a coincidence (or something else entirely)?

 

[Karma Police]

Those of you interested in the Turion know the consequenses of the answer to the question above.

 

[jvd]

They have the socket 754 board as thier low end soultion

The 939 is the highend . Since most likely the first dual core athlon 64s will be the athlon 64 fxs they went this route . There is no reason why they need dual channel anymore than the reason why a single core chip needs it . The more bandwidth no matter single , dual , tri , quad the better

 

[mczak]

They have the socket 754 board as thier low end soultion

The 939 is the highend . Since most likely the first dual core athlon 64s will be the athlon 64 fxs they went this route . There is no reason why they need dual channel anymore than the reason why a single core chip needs it . The more bandwidth no matter single , dual , tri , quad the better

That's true, I'm not sure however if it would make sense to do a single-channel dual-core A64 chip, unless it's really clocked low. There is not that much performance hit by using only a single memory channel nowadays on the single-core cpus (especially not the low-clocked versions - a 3000+ socket 754 eats the 3000+ socket 939 for lunch due to its slightly higher clock speed), but a dual-core 2Ghz very likely would take a much larger performance hit (on average - you can always find apps where memory bandwidth doesn't matter).

 

[Karma Police]

That's true, I'm not sure however if it would make sense to do a single-channel dual-core A64 chip, unless it's really clocked low. There is not that much performance hit by using only a single memory channel nowadays on the single-core cpus (especially not the low-clocked versions - a 3000+ socket 754 eats the 3000+ socket 939 for lunch due to its slightly higher clock speed), but a dual-core 2Ghz very likely would take a much larger performance hit (on average - you can always find apps where memory bandwidth doesn't matter).

It makes sense where AMD's mobile unit comes into play. They've announced a dual-core Turion later this year, but no-one is exactly sure what interface it will have. Alot of people thought that since AMD is releasing the dual-core CPU's in 939, then they'd have to release a 939 for the notebooks, too.

But, I guess not. 754 will stay in notebooks, single or dual-core.

 

[mczak]

It makes sense where AMD's mobile unit comes into play. They've announced a dual-core Turion later this year, but no-one is exactly sure what interface it will have. Alot of people thought that since AMD is releasing the dual-core CPU's in 939, then they'd have to release a 939 for the notebooks, too.

This year? Never heard of that. Dual-core Opteron will be mid-year or so, dual-core high-end desktop Q4. AFAIK everything else has to wait til 2006.

But, I guess not. 754 will stay in notebooks, single or dual-core.

Well, 2006 should see a transition from socket754/socket 939 to M2 too, which should support dual-channel ddr2. I suspect the mobiles are going to use the same socket, which would make a lot of sense (ddr2 uses less power, and actually dual-channel ddr1 memory might not be a very good idea for mobiles as it uses more power than a single-channel setup).

 

[Karma Police]

AMD has said they are releasing DTR notebook dual-core CPUs in the 2H '05.

http://www.tomshardware.com/hardnews/20050303_023118.html

 

[Pete]

It's possible that dual-core will need dual-channel in order to avoid being bandwidth-starved. I'm not sure each CPU physically "needs" a single channel, though, as I believe dual (discrete) A64 CPUs still only use the integrated memory controller/HyperTransport services of a single CPU to regulate memory access to both. Of course, that may be an unrelated feature. I'm not sure if quadruple and octuple CPU MBs operate in the same way, with each CPU depending on only one for memory access.

I suspect they limited dual-core to s939 b/c it's their premium platform, though. Dual-core seems unnecy at the low- and mid-range s754 offerings. Dual-core Sempron just sounds wrong on its face.

 

[Karma Police]

So do you think AMD will release the dual-core CPUs only in a 939 format?

 

[mczak]

So do you think AMD will release the dual-core CPUs only in a 939 format?

Would make sense I guess (and later M2 of course). I must have missed the information about the mobile Athlon 64 appearing this year, but since it's DTR only I guess I wouldn't be surprised if it will use socket 939. And it wouldn't surprise me if we don't see the mobile dual-core A64 this year at all neither...