It seems so, yes.
But there's also winter launch, we could hope for multi-chip cards, R680 or what have they.
So, do we know anything about RV670 yet?
- Thread starter nicolasb
- Start date
But there's also winter launch, we could hope for multi-chip cards, R680 or what have they.It seems so, yes.
Im sure if they really really wanted to, they could of came out in November with something thats worth 600$ and blows the ultra away, but it would seem that isnt the plan at AMD....atm. No ...lets make awesome cards that will work very well within a certain low cost bracket . Thats kewl, but with that plan, theres the risk of loosing more ground to NV in the enthusiast market.
I know...patience is a virtue....Im sure there will soon be something from AMD/ATI that will cater to us enthusiasts alot more...which seems to be r680 Q1 08 or this summer with r700, but it would of been nice of AMD/ATI to make sure that its not one uped again by Nvidia this November 15th with R670
It's right there in the text. Re-pipelining and library changes are just that, changes to the logic, which can affect the final transistor count needed to build the affected blocks. "Oh, hey, the library we use for this twelvety bit adder now saves us 200M transistors compared to the version we had to use on the last process node", etc.
It probably wasn't a matter of room. Afaik TSMC's maximum reticle size is around 21mm*21mm, and i think G80 may have hit the limit. R600 had 50mm² more to go, and i adding hardware AA resolves and even UVD wouldnt even come close to the limit.
If that's the kind of thing you say when you're not bashing ATI, what sort of thing do you say when you are bashing them?
Never say never again...
Here's an interesting slide: http://www.tweaktown.com/news/8444/amd_s_q4_graphics_line_up_revealed/index.html
Interesting matchup of competing products. Wonder when it was drafted...?
Here's an interesting slide: http://www.tweaktown.com/news/8444/amd_s_q4_graphics_line_up_revealed/index.html
Interesting matchup of competing products. Wonder when it was drafted...?
If that's the kind of thing you say when you're not bashing ATI, what sort of thing do you say when you are bashing them?
Then it would contain rantings and not facts.
A few things:
R680 is not due out this quarter.
There is no $299 RV670 product (unless it's dual GPU with only 512MB frame buffer total - 2x256MB).
No way in hell is AMD going to try to sell a graphics card with a $799 MSRP
It was clearly drafted at a time when whoever drafted it assumed that an 8800GT would be outperformed by a 320MB GTS. If it's an accurate reflection of ATI's thinking at the time, this doesn't bode well for how RV670 performance will compare to G92: roughly on a par with 320MB GTS, and comfortably below R600. That would put it a long way behind 8800GT. But I'm inclined not to believe its authenticity, on the grounds that I'm a nasty, horrible, cynical person.
The world largest and most successful computer company, Apple, comes to mind.
With a market cap worth more than Intel, IBM, Dell etc.
Which is NOT due to their PC sales
iPod, iTunes, and iPhone (as well as the iRon lung and other various iTems).The industry standard is 25x25.
Which is NOT due to their PC sales
Still over 2 million sold computers the last quarter is nothing to look down upon
But yeah, if the power charateristics are right and the price is right, I can see this 55nm card be used with a lot of OEMs.
Yeah, I misunderstood that. I understand why the transistors were able to be cut (as you clearly wrote), but I thought the changes to the library were possible because of the shrink, not just an architectural change that could've been done on the same node (such as the changes to the pipeline.)
Thas interesting...R680 at such a high price.
I wouldnt be interested unless its a single chip monster........i would shun it if its a dual chip implementation, especially at the quoted price.
jimmyjames123
Regular
Very brilliant strategy from NV. I think they knew that details of the 8800 GT would be leaked early to ATI/AMD (like name, memory bus width, etc) before actual performance numbers were leaked. So they may have been able to confuse ATI/AMD about the real high performance of the GT card. For instance, how many people would have expected a "GT" to handily outperform a "GTS" that uses a wider memory bus?
http://www.tcmagazine.com/comments.php?id=16657&catid=2
HD3870: 775MHz GPU, 1200MHz Mem
HD3850: 668MHz GPU, 828MHz Mem
HD3870: 775MHz GPU, 1200MHz Mem
HD3850: 668MHz GPU, 828MHz Mem
I hope for ATI's sake....the R680 is a monster chip with 6 texture quads and 6x(16x5) shader processors on a 512 bit external memory interface.
A MCM based around the RV670 is viable and is being touted as the R680 in various forums. But the margins will be really low on these and i do not think it can scale in performance.
A monster R680 would still beat the G80 ultra by no more than 10% but i think it would be relatively easy for ATI to design one on 55nm process.
That way it will give them breathing space if G9x series based on a beefed up G80 tips up.
A MCM based around the RV670 is viable and is being touted as the R680 in various forums. But the margins will be really low on these and i do not think it can scale in performance.
A monster R680 would still beat the G80 ultra by no more than 10% but i think it would be relatively easy for ATI to design one on 55nm process.
That way it will give them breathing space if G9x series based on a beefed up G80 tips up.
