RV560/570 Gemini roadmap
- Thread starter kemosabe
- Start date
Fodder said:
Is it a mockup? Afaik that was the Asus X1600 X2 (although they did mirror the picture). Asus supposedly also has functional models of the X1800 X2 and X1900 X2, but only the X1600 X2 would go into MP (from what I've heard). And the X2 cards would be able to run in "4GPU" setup.
Here's another pic of the Asus X1600 X2
The picture you see is of Asus' Dual 1600XT setup..
I think it is due launch somewhere in March as a prelude to Gemini.
Besides wanting to know if it works like SLI (i.e. Quad VPU) I'm more interested in what looks like a Carbon Heatsink to cool the two 1600's.
They can be passively cooled in single slot so I wonder about about this setup..
I think it is due launch somewhere in March as a prelude to Gemini.
Besides wanting to know if it works like SLI (i.e. Quad VPU) I'm more interested in what looks like a Carbon Heatsink to cool the two 1600's.
They can be passively cooled in single slot so I wonder about about this setup..
I'd expect quad CF to be a reasonable expectation for the near future (under the monkey see monkey do perspective which works both ways).
If the heatsink should be more than the board would actually need even with two chips, it could indicate a very potent overclocker.
http://www.gecube.com/press-release-detail.php?id=60847
And since Gemini was on an official ATI roadmap doc, one assumes they won't be the last.
And since Gemini was on an official ATI roadmap doc, one assumes they won't be the last.
The inq claims we are going to see plenty of price cuts. A x1600 pro for 109 is a steal IMO:
http://www.theinquirer.net/?article=30137
Does this kind of pricing bring up the possibilities of quad XF? Also do you guys have any experience with XF on two x1600s? For 250 we could get an x1800 xt.
http://www.theinquirer.net/?article=30137
Does this kind of pricing bring up the possibilities of quad XF? Also do you guys have any experience with XF on two x1600s? For 250 we could get an x1800 xt.
http://www.theinquirer.net/?article=30217
Could 256-bit be the difference between RV560 and RV570?
At 80nm, how big is that thing? NV has moved the bar for modern 256-bit down to 196mm2. . .but RV530 is only 150mm2 at 90nm.
Could 256-bit be the difference between RV560 and RV570?
At 80nm, how big is that thing? NV has moved the bar for modern 256-bit down to 196mm2. . .but RV530 is only 150mm2 at 90nm.
I think Dave already suggested that the last number will not be a 1 -> 2.kemosabe said:Hmm, at the very bottom of this page our Chinese comrades are claiming RV570/RV560 will be 12-1-3-? and 8-1-2-?, respectively, with both sporting a 128-bit bus. Any buyers?
RV570 12-1-3-2
RV560 8-1-2-2
I'm guessing RV570 will be ~260 million transistors (on 80nm) which I'm guessing is enough for people to confuse it as a 256bit chip.
You just blew my mind. (Well, you and HKEPC, upon rereading.)kemosabe said:
Fudo thinks and Xbit announced that RV570 is 256b, so maybe it is and RV560 is 128b. And maybe, as Dave suggested, RV570 is straight outta TSMC and RV560 is outta UMC.
Actually, didn't ATI use UMC back in the RV250 days? Have they used them since then for any GPU? If not, would it be reasonable to drop an ALU per TMU for higher yields (thinking like Jawed, here) in getting reacquainted with them?
Pete said:
It was RV280 (9200) that UMC fabbed back then. Since these would also be relatively high-volume parts, getting UMC's fab capacity back on board wouldn't be a luxury. Obviously the design team has conceded that 8-1-2-2 would be far more balanced an approach, now that we've all seen where RV530 falls short. And RV570 could shape up to be a formidable part for the mid-range segment even with a 128-bit bus, assuming that ATI can depart from its recent history of being too late to the party. Unless that roadmap has changed, volume production in July/August is not going to change their situation at all.
Last edited by a moderator: