A-Data will be announcing DDR3 at Computex

[Wolf2000me]

Hey,

While DDR2 is barely becoming mainstream the memory manufacturers are thinking ahead, eventhough DDR3 is not spotted on roadmaps until 2008.

http://www.adata.com.tw/adata_en/ada...hp?news_id=152

Nothing about latencies though. I wonder why

[skool h8r]

hah, latencies, i expect 12-7-7-4 to be 'good' latencies for DDR3.

[Gizmo]

DRAM technology still has to contend with certain realities of physics. Although the data transfer speeds have increased significantly since the days of EDO, even the best DRAM still only achieves a latency of about 45 nS. Increasing the clock speed simply means that you have to waste more clocks waiting.

[madcatmk3]

Originally Posted by gizmo

DRAM technology still has to contend with certain realities of physics. Although the data transfer speeds have increased significantly since the days of EDO, even the best DRAM still only achieves a latency of about 45 nS. Increasing the clock speed simply means that you have to waste more clocks waiting.

I have a feeling some time in my life time, probly fairly soon we will see the acceptable limit on silicone. For the time being I hope for GDDR4 or GDDR5, that will make much more of differance.

[MONKEYMAN]

GDDR is different from normal DDR, they are made and operate in different ways.

[aghastpumpkin]

Why do we need this? We already know DDR2 gives barely any extra power to a new athlon am2 machine.

[Wolf2000me]

Originally Posted by aghastpumpkin

Why do we need this? We already know DDR2 gives barely any extra power to a new athlon am2 machine.

It's probably some PR move as DDR3 doesn't show up on any roadmap till 2008

By that time we will be using a bit more cpu cores than we do now which will require a lot more memory bandwidth.

[Áedán]

Originally Posted by madcatmk3

I have a feeling some time in my life time, probly fairly soon we will see the acceptable limit on silicone.

IBM's managed to get silicon switching at 60GHz, which is normally territory reserved for gallium arsenide. Gallium arsenide is expensive and difficult to work with, but suits high frequencies (>250GHz) well.

[madcatmk3]

Originally Posted by Áedán

IBM's managed to get silicon switching at 60GHz, which is normally territory reserved for gallium arsenide. Gallium arsenide is expensive and difficult to work with, but suits high frequencies (>250GHz) well.

When theres profit theres a way, while currently we can keep going a while when they hit a road block is when we make something better.