Quote Originally Posted by Metroid View Post
They can disable or set the frequencies. That is how they split up the main chip into many fragments which is usually called market segments, that means the same chip can make many derived products from it like 4850, 4870 and 4870x2.

Metroid.
Quote Originally Posted by madmossy View Post
What will be interesting to see is just how much of a difference the GDDR5 will have over the GDDR3. If both GPU's were at the same clock speed just how much of a difference will the extra 900mhz on the member make (2ghz vs 3.6-3.8ghz effective).
Quote Originally Posted by Metroid View Post
The 4850 uses lower bandwidth due to 256 bit-wide + GDDR3 and frequency, so that should imply on a lower performance than 4870. The 4850 probably could not have shaders being disabled due to performance issues and competition. The best bet was to keep its design and performance being controlled by its bandwidth and frequency. A 4850 overclocked with a GDDR5 would be interesting.

Metroid.
This is what i mean - if the 4850 and 4870 are using the same core, with the same number of SP's, is the memory bandwidth of the GDDR5 alone going to increase performance of the 4870 much above the 4850? Isn't that like comparing a 512mb current card to a 1gb or 2gb version (i believe there are 1 or 2 kicking around), whenever you see benches on those huge memory cards there is virtually no performance gain with the extra bandwidth, even at high resolutions.

Quote Originally Posted by CBONE View Post
Sounds like the 4000 series are going to be the new golden boys of GPGPU applications. It would be nice if AMD hit the ground running with this release.
Hmm, nothing sounds like anything to me personally so far, not until the launch day reviews! No harm in some speculation i guess.