So, do we know anything about RV670 yet?

Well, unless they also standardize the amount of memory for each number as well as the stock clocks, this will just be as confusing. I can just see all the abuse manufacturers are going to try pulling... 3600, 3650, 3650.5, 3655, 3655+....

And with the practice of cards that have more memory getting slower ram, what will get a higher ranking? Or will it be inconsistent and even MORE confusing? Ugh
 
I not like this new naming scheme without le/gt/gto/pro/xl/xt/xtx/xtpe.
I challenge you to pronounce "HD3800XTX in CrossFireX running on QuadFX, supporting DX10" :D
Regardess of the product names of the above - if I spend that much on comptech, I probably could throw in one more "X" to prefix my wife. :LOL:
 
Dual FP16 bilerps - that's enough for me! :LOL:
There's certainly enough cache bandwidth to do that if they wanted to.
But in this mole-hunting-slides about R650 AMD claimed, that the more tex-power is not reached through more TMUs, but through some optimizations in the AF-algorithm. ;)
Sounds like they're doing NVidia style TMUs, and actually making some use of the 32 address units and insane cache bandwidth.
 
Btw, not that its any great shock at this point, but AMD confirmed RV670 is 55nm in the call today. I think that's the first official confirmation.
 
But in this mole-hunting-slides about R650 AMD claimed, that the more tex-power is not reached through more TMUs, but through some optimizations in the AF-algorithm. ;)

So do you think AMD-RV670 will win against Nvidia-G92 since that was part of R600 weakness?
 
Back
Top