Socket 939 non-FX's to be single channel?

http://www.vr-zone.com/#2874

Surely not?

--
Derek

Derek Baker wrote:

http://www.vr-zone.com/#2874

Surely not?


Will it even matter? Even with an athlon the difference between single and
dual chanel ram is basically nothing. On a P4 ram bandwidth is a major
concern, with an athlon it doesn't seem to need mega ram bandwidth to work
well.
--

Stacey

"stacey" wrote in message
...
Derek Baker wrote:

http://www.vr-zone.com/#2874

Surely not?


Will it even matter? Even with an athlon the difference between single
and
dual chanel ram is basically nothing. On a P4 ram bandwidth is a major
concern, with an athlon it doesn't seem to need mega ram bandwidth to work
well.

I guess we'll find out in a couple of months. If you follow the link to The
Inquirer- which doesn't mention memory - and from there to the original
French article, you'll see it does say dual-channel, as we all knew.

--

Derek

On Sun, 11 Jan 2004 06:04:11 -0000, "Derek Baker"
wrote:

http://www.vr-zone.com/#2874

Surely not?

It could be. the current Opteron and Athlon64 FX cores can work in
single channel mode (err, more to the point, a 64-bit channel vs. the
single 128-bit channel they normally use, though the difference
between a 128-bit single channel and two 64-bit channels isn't all
that significant). All they would need to do is force the chips to
only use 64-bit memory access. This could complicate things with
memory installation though, ie some memory sockets on a motherboard
might work with an FX chips but not a regular Athlon64.

AMD needs SOMETHING to differentiate the chips and justify the
significantly higher cost of the FX series. The difference between
the Athlon64 3000+ and 3200+ suggests that cache alone might not cut
it. Higher clock speeds are an option, but everyone knows that
they'll get a higher clocked chip for the same price a few months down
the line. Extra memory bandwidth might be the only thing AMD has left
to differentiate between the chips.

-------------
Tony Hill
hilla underscore 20 at yahoo dot ca

"Tony Hill" wrote in message
.com...
AMD needs SOMETHING to differentiate the chips and justify the
significantly higher cost of the FX series. The difference between
the Athlon64 3000+ and 3200+ suggests that cache alone might not cut
it. Higher clock speeds are an option, but everyone knows that
they'll get a higher clocked chip for the same price a few months down
the line. Extra memory bandwidth might be the only thing AMD has left
to differentiate between the chips.

Or it doesn't need to keep producing the FX series at all. The only reason
for bringing out the FX series originally was to counter any criticism that
Intel's Pentium 4 is dual-channel while Athlon 64 is just single-channel
DDR. If the whole series becomes dual-channel, that marketing faux pas
ceases to be a liability. Even though there doesn't seem to be too much of a
difference in performance between single and dual-channel memory, the gadget
kiddies might object.

Another reason for bringing out the 939 socket might be to get ready for
DDRII memory?

Yousuf Khan

On Mon, 12 Jan 2004 10:56:37 GMT, "Yousuf Khan"
wrote:
"Tony Hill" wrote in message
t.com...
AMD needs SOMETHING to differentiate the chips and justify the
significantly higher cost of the FX series. The difference between
the Athlon64 3000+ and 3200+ suggests that cache alone might not cut
it. Higher clock speeds are an option, but everyone knows that
they'll get a higher clocked chip for the same price a few months down
the line. Extra memory bandwidth might be the only thing AMD has left
to differentiate between the chips.

Or it doesn't need to keep producing the FX series at all. The only reason
for bringing out the FX series originally was to counter any criticism that
Intel's Pentium 4 is dual-channel while Athlon 64 is just single-channel
DDR. If the whole series becomes dual-channel, that marketing faux pas
ceases to be a liability. Even though there doesn't seem to be too much of a
difference in performance between single and dual-channel memory, the gadget
kiddies might object.

The downside to the above theory (making just 128-bit memory channel
Athlon64s) is that OEMs might not like it. Using two DIMMs in a
system is perfectly fine for a mid range or high-end system, but it
doesn't fit in well with the bargain segment. The Athlon64 might not
be in the bargain segment at the moment, but it probably will be in
the not-too-distant future.

Personally I still think that there's a good market for a "Duron64" or
some such thing, basically just an Athlon64 but with only a 64-bit
memory, an 800MT/s HT connection and 256KB of L2 cache. If the
Athlon64s all have 1MB of L2 cache and 128-bit memory bus plus the
1GT/s HT connection they're aiming for, they should have a
sufficiently large performance advantage to justify a the price
difference. Of course, that will probably have to wait until the
chips are shrunk to a 90nm process and it still doesn't solve the
issue of differentiating the Athlon64 and Athlon64 FX (if they decide
to keep it).

Another reason for bringing out the 939 socket might be to get ready for
DDRII memory?

That is a possibility. I haven't heard for certain if DDRII is going
to be supported on the first Socket 939 Athlon64s or not, but it
likely will be before too long.

-------------
Tony Hill
hilla underscore 20 at yahoo dot ca

"Tony Hill" wrote in message
.com...
The downside to the above theory (making just 128-bit memory channel
Athlon64s) is that OEMs might not like it. Using two DIMMs in a
system is perfectly fine for a mid range or high-end system, but it
doesn't fit in well with the bargain segment. The Athlon64 might not
be in the bargain segment at the moment, but it probably will be in
the not-too-distant future.

Well, as it turns out, since dual-channel AMD64's work just fine with a
single channel, this might not be a problem afterall.

Yousuf Khan