ATI Radeon HD 4000 Series discussion

[perkam]

We still dont know if the shaders have their own clock domain. If that's true your calculations are wrong. Also, 3870 is bottlenecked by its 16TMUs, you don't know how 32TMUs can change things. My bet is 14k minimum. And a 3870 scoring 10k? With what CPU?

You dont get it

If you get 10k stock currently with the 3870, I'm predicting you'll get 12.5 - 13k stock with a 4870.
If you get 11k stock currently with a 3870, I'm predicting you'll get 12.75 - 13.25k stock with a 4870.
If you get 12k stock currently with a 3870, I'm predicting you'll get 13 - 13.5k stock with a 4870.

As I said, 25-30% is realistic. Any more than that is wishful thinking due to the reasons provided in my last post here:

But whatever it is, it gives you the intuition of knowing that if a 320-shader, 256-bit, 16/16 TMU/ROP card (HD 3850/3870) only manages to get twice the 3d06 points compared to a 120-shader, 128-bit, 8/4 TMU/ROP card (HD 3650) on the same architecture, a 160 shader increase with no change to the 256-bit bus and no change to ROPs will not have a performance effect greater than 30% at best (my 12.5k estimate was a 25% increase over the current 3870s stock 10k score).

And you can't discount the reasoning there. If the 3650 was 256-bit the score would be around 6.5k, which would make the 3870 score with almost 300% the shaders (and double the TMUs AND triple the ROPS !!) of only 10k completely incompetent, so you can see only 50% more shaders will not have a huge effect. It's called diminishing returns, look it up.

I have my reasoning for my estimated score, I would like to see yours.

Perkam