So, do we know anything about RV670 yet?

This is so funny. Look at what Fudzilla posted today:

RV670 has close to R600 performance

But much cheaper claim

Our well informed sources have said that RV670 mainstream chip will be as fast as some high end chips but it will be priced in the mid range. It won’t cost more than $200 per card, but the performance itself might be as good as the R600.

To be realistic we expect that R600 will be a bit faster than the upcoming 55 nanometre chip, but RV670 will certainly be smaller, cheaper to produce and most importantly not as hot as the previous chip.

The chip is still scheduled for Q1 2008.
http://www.fudzilla.com/index.php?option=com_content&task=view&id=2753&Itemid=1

Art of milking indeed. ;) So predictable.
 
FUDzilla:

http://www.fudzilla.com/index.php?option=com_content&task=view&id=2838&Itemid=1

Dual RV670 design kit is out

Why not if two RV630 can work

AMD is preparing a design kit for dual RV670 card and as dual RV630 cards are real, so is the dual RV670.

The story is quite simple as dual RV670 will draw even less power at least theoretically due to to its 55 nanometre design and such a card is possible and very likely. If the story of Quad Crossfire becomes a success ATI will have sub $500 card to fight Nvidia’s Geforce 8800 Ultra or Geforce 8950 how Nvidia likes to call its G92.

After dual RV670 comes R700 and that is ATI’s real high end. The question remains what Nvidia has versus G100..
 
So, until R700/G100, it's possible the top end from both companies will be MCMs?

2x 6 TCP 256-bit (192 SPs and 65nm)

vs.

2x 320 SPs 256-bit (R600 on 55nm)

??

I understand single high-end chips could be released (g90/R680)...but.

Sounds kind of logical if you think how it's been mentioned that both companies (mostly AMD/ATi) will go for a single chip used in MCM for higher-end parts next gen. Perhaps this is a precursor?

While 256-bit may seem like a step backward for both companies, perhaps 384-bit/512-bit is hard to fit on the smaller die, and even more-so, 2400mhz could give similar bandwidth to the parts the single chip is replacing, with a number in-between the 8800gts and 8800gtx. As for the R600, I think it's been shown the extra bandwidth was overkill, and perhaps if the hardware AA fixed in RV670, it could compensate.

hmm...Just a thought. :p
 
Perhaps AMD simply can't do top-end anymore and NV doesn't see any reason to do it because it has no competition? 8)

Heh, exactly how does that work? Wouldnt the next level down become the top-end if both companies drop the current top-end? ;)
 
Heh, exactly how does that work? Wouldnt the next level down become the top-end if both companies drop the current top-end? ;)
No because in that case the next "top-end" would be two-chip boards which would be good in 3DMark but awful in most of "real" games, especially in those without benchmarks 8)
No, really, i'd prefer the real top-end over two-chip monster any time with my eyes closed. But going the way of multichip they'll make more money (and AMD hopefully won't fail for the forth time in a row).
 
November 3D War brings ATI and Nvidia siblings

RV670 took the same receipt as G92 compared to G80: cut the number of R600 units in half and you get RV670. 160 Superscalar shaders consisted out of 32 fatties and 128 regular scalars, blazing high GPU clock... like we said, RV670/G92 are siblings.

Have you had any doubt, DAAMIT is pitching for very same performance that Nvidia is achieving. 10.000 3DMarks is the whole grail, but 9500 should be doable. Performance estimates were based with Kentsfield machine, so this may change if you overclock the CPU at above 3.0 GHz level (QX6800 works at 2.93 GHz).
http://www.theinq.com/?article=42199
 
Perhaps AMD simply can't do top-end anymore and NV doesn't see any reason to do it because it has no competition? 8)

Despite the smiley, that is indeed possible. Not in the sense that ATI couldn't do it, but maybe it just isn't feasible anymore. I think the success (or lack of) of the R700 will lead to a decision about the future of their high-end stuff.
 
I don't know about you guys, but I'm tired of 10"+ 150w+ monsters anyway.

Me too. I wish they would already give us the 8" 75Watt card with three times the performance of the SLI 8800 GTX Ultra.
 
The way things have been going, it is unlikely those 150W monsters are going away.

TDP is the gating factor for performance these days, and as long as 150W products sell, they will find a way to fit enough GPUs/RAM/etc to make it happen.

It's not like going SLI or Xfire really changes the wattage dynamic.
Using 2 80W cards to hit a given performance target is going to bust the 150W power ceiling anyway, and they will be less efficient doing it.
 
Me too. I wish they would already give us the 8" 75Watt card with three times the performance of the SLI 8800 GTX Ultra.

Can they give us that in 10" now? Or you're just engaging in strawmen?
 
Can they give us that in 10" now? Or you're just engaging in strawmen?

Strawmen... I hoped the winkie post icon pointed that out. If you're going to wish, wish big.

Seriously now ... They need to make higher efficiency cards -- working smarter not harder, and all that goes along with doing more with less.
 
I was missing full-length cards ever since they disappeared with ISA and VLB. Too bad the power usage is ridiculous, lol. :devilish:
 
ATi-roadmap.gif


Weird slide, hd2900pro stand in enhusiast segment, and the sucessor hd2950pro in the performance segment.
 
Last edited by a moderator:
Weird slide, hd2900pro stand in enhusiast segment, and the sucessor hd2950pro in the performance segment.
Well, once R680 is out that's the only enthusiast GPU from AMD. Everything else, even if it matches or slightly exceeds R600 will be pushed down - to the Performance segment.
 
Back
Top