Gigabytes radeon 9700 hits 400mhz core clock!!

quote]still doesnt mean they have to manufacture it, even if its ready to go that far... imagine.. if the r300 .15 can do 400+Mhz, and the r350 is "essentially" a shrunken r300, it should be quite a bit higher than 400MHz now, not just the 300MHz as before. everything falls right into place. And theres nothing to stop them from waiting a cycle (releasing a 400Mhz .15 R300 with faster ram in this cycle), then introducing the .13 with even faster ddr2 and core (because of process refinement in that extra cycle) than they originally expected.[[/quote]

It'll be interesting... one thing to note, I doubt the r350 will need the external power supply. Which, in turn, I believe is partially responsible for this very high overclock of the r300. Surely to think approx. 400MHz is not out the question for a .13 r350, but to push that to 500MHz or more I think is questionable without the aid of the external power supply they have currently with the r300. So, I guess the question is... what is limiting high clock speeds of these modern GPUs... power or heat?

 

You can be positive that if ATI indeed release R350 on 0.15 and clockspeeds higher than 400, it will still require an external power supply.

Now, if they release it on 0.13, that's an entirely different story...

 

actually right now i think the limiting factor is CPU's. I think a 0.13u/DDR2 R350 will be overkill for at least another 10-12 months. Just an extension of my earlier posts. ATI will be able to get by with producing an "R325" thats just a higher clocked 0.15u R300 with faster ram due to the sole fact that CPU's dont seem to be scaling enough to take advantage of a DDR2/0.13u part.

(why produce a part that would be very expensive at the current time and wouldnt even be fully utilized due to cpu limitations?? doesnt make sense to me to release R350 until at least a year from now.)
<edit>
this is from a business POV. from a tech standpoint, i'd have to say power
</edit>

 

That's what I meant... perhaps I just didn't word it that well. Off topic... how do you quote a specific person and get it to write in who you're quoting?

 

Unless they use DDR-II, I think they'll still need the external feed, even with a 0.13u ASIC clocked reasonably conservatively. The memory subsystem is a pretty heavy on power compared to current cards; about 1W/chip and 0.5-1W for I/O, per branch. That's ~15W right there. All the discrete, ancillary logic would probably require another 5-10W, so unless they can get the ASIC draw under 20W, underclock/undervolt the chip or go for DDR-II they'll be looking at having to use the 4-pin connector again.

Just my opinion.

MuFu.