RV770 taped out

[GoThr3k]

http://www.fudzilla.com/index.php?option=com_content&task=view&id=5145&Itemid=1

looks promising

[HKPolice]

50% faster? Wow.... put 2 of those on the same pcb and it'll be faster than the upcoming 9800GX2... niceee

[Baleful]

Not being rude, but how is this news? It's the same old crappy generic information from fudzilla. There is nothing in there that is "news".

[MuffinFlavored]

Not being rude, but how is this news? It's the same old crappy generic information from fudzilla. There is nothing in there that is "news".

So Fudzilla is the new Inq?

[Baleful]

I don't have anything against fud, im just saying I don't think this qualifies as news as it's just some VERY VERY generic information.

[Der_KHAN]

we should add a prefix like FUD or INQ to future news threads that use these sources.

maybe then we'd finally get over this endless arguement...

[Jakko]

It's more like rumors.
Early news based on unknown sources information.
But still interesting.

Btw, what does "taped out" mean?

[Ugly n Grey]

50% faster? Wow.... put 2 of those on the same pcb and it'll be faster than the upcoming 9800GX2... niceee

why would the 9800gx2 even be faster than this gens upcoming 3870x2 ? Never mind the next generation of chips.... sorry don't get what you're thinking there

[naokaji]

50% faster? Wow.... put 2 of those on the same pcb and it'll be faster than the upcoming 9800GX2... niceee

doesnt matter, because its the generation after 3870x2 and 9800gx2... nvidia will release something to compete with rv770 by the time it comes out. who will end up with the faster product in 2nd half of 08 still remains to be seen.

[Ugly n Grey]

who will end up with the faster product in 2nd half of 08 still remains to be seen.

Regardless of who builds it, *I* will end up with it :ROTF:

[naokaji]

Regardless of who builds it, *I* will end up with it :ROTF:

i shoudnt try to post when answering a call at work....

[Ace123]

Regardless of who builds it, *I* will end up with it :ROTF:

Haha thats good.

[CERO]

fud....

[Monkeywoman]

q3? thats 9 months from now. member when ppl were saying the R600 was pur pwnage? yah...i'll wait till q3 to come around , thats when i will call the winner.

[GoThr3k]

from tape out to final product is about 6 months for GPU's right?

[xsbb]

we should add a prefix like FUD or INQ to future news threads that use these sources.

maybe then we'd finally get over this endless arguement...

:up: words of wisdom.

[GoThr3k]

:up: words of wisdom.

well next time i will put it in the title...
10 posts about the same subject is boring, can we go back on topic?

[Cooper]

10 posts about the same subject is boring, can we go back on topic?


Yes please

[Zytek_Fan]

we should add a prefix like FUD or INQ to future news threads that use these sources.

maybe then we'd finally get over this endless arguement...

Nice sig pic :D

Another for the BMW Sauber alliance!! :D


The article appears to actually be news for once. 55nm is correct, although I'm not sure if RV770 will arrive in Q2. If anytime in Q2, it will be late...

[[XC] 2long4u]

q3? thats 9 months from now. member when ppl were saying the R600 was pur pwnage? yah...i'll wait till q3 to come around , thats when i will call the winner.

Remember how many times they pushed back the release date of R600? I wonder if this will be the same?

[AliG]

Actually I believe the rv770 is supposed to be based off a 45nm process, unless I'm thinking of the r(v)700

[Trox]

Remember how many times they pushed back the release date of R600? I wonder if this will be the same?

Well, RV670 arrived close to release date.

[AliG]

Remember how many times they pushed back the release date of R600? I wonder if this will be the same?

Don't think so, the r600's delays were due (imo) to most likely first bugs, probably they tried to fix the aa performance but once they couldn't decided to up the shader count to 5 clusters (from 4) of 64 shaders, then the die running extremely hot gave them problems so they tried to get to a 65nm process (I believe using amd's fabs), and once they made that move they decided to convert everything to 65nm and wait for a hard launch (which they did with the r600 at least) on everything, but they had leakage issues so performance just wasn't where it should have been with the r600xtx so that was scrapped and the rv6xx cards had driver issues. In short, basically a bunch of screw ups probably caused by the transition of the merger is why the r600 was delayed and performance suffered, that's probably why k10's performance isn't where it should be, I agree with others that amd kinda gave up with it at 65nm due to various reasons and with k10.5 or k11 we might be seeing the real thing.

This time around, I think they'll be good, the 55nm process is working great for them so if the rv770 is 55nm (and from what I've seen so far the design will be more of a perfected r600 design than a new architecture -which would speed things up a bit-, but that could be wrong) it shouldn't take them near as long as the r600 did as they probably won't need to try to get to smaller nodes etc

[AliG]

Well, RV670 arrived close to release date.

No, the rv670 came something like 2 months early, it was originally planned to be launched with the entire 55nm series but for once the test silicon (I believe a11 or something) worked/had hardly any bugs so they brought it out a head of time to slow down the 8800gt's thunder. Which it did, and then some, look at Nvidia's current market share, the 3870 and 3850 are huge right now (probably due to the g92 shortage since the performance isn't different from the r600's)

[awdrifter]

Any word on how many shaders it'll have? Because the 64 shaders in the RV670 won't be able to compete with the next gen Nvidia stuff.

[AliG]

No one knows for sure right now about the specific details

[LordEC911]

Don't think so, the r600's delays were due (imo) to most likely first bugs, probably they tried to fix the aa performance but once they couldn't decided to up the shader count to 5 clusters (from 4) of 64 shaders, then the die running extremely hot gave them problems so they tried to get to a 65nm process (I believe using amd's fabs), and once they made that move they decided to convert everything to 65nm and wait for a hard launch (which they did with the r600 at least) on everything, but they had leakage issues so performance just wasn't where it should have been with the r600xtx so that was scrapped and the rv6xx cards had driver issues. In short, basically a bunch of screw ups probably caused by the transition of the merger is why the r600 was delayed and performance suffered, that's probably why k10's performance isn't where it should be, I agree with others that amd kinda gave up with it at 65nm due to various reasons and with k10.5 or k11 we might be seeing the real thing.

So much BS in that post...
AA performance is a decent improvement over R580 and is shader based, something that ATi think is needed for their cards.
You really think that AMD can take and redesign an architecture that took roughly 4 years in R&D, R600 was started back when the R300 was released, redesign it, add more trannies and then still have the time to tweak and debug 3 revisions in ~6 months? Wow... Just tweaking/debugging a revision and getting the next revision out takes about 2 months.
Also, AMD has completely different fabs then TSMC. I highly doubt that AMD even thought about trying a 65nm R600 in their fabs. Now they might have tried a spin in TSMC's fabs but that is a completely different situation.
Leakage was thanks to the 80nm PP, so they already knew they had a leakage problem, if anything that is why they thought about pushing it to 65nm.

Other than you comments about the merger, the merger definitely set them back roughly 6-8months from their previous roadmaps, most of that is just wrong.

[Ugly n Grey]

So much BS in that post...


If you have an issue with a post, say what's wrong with it, making one liners is just flame bait :stick: . TY

Regards
UnG

[NH|Delph1]

50% faster than last generation is pretty much what happens with every generation. You should have seen it coming ;)

//Andreas

[RPGWiZaRD]

I think realistic goals would be at least 60~70% boost over dual RV670 if ATI wants to keep up hopes for gaining or challenging NVIDIA's next new gen highend card so 50% claim to me is nothing too impressive at this moment. I think around 20k 3DMark06 stock score I think is to be expected at least by NVIDIA's next gen card if we were to discuss 3DMark scores. ATI has been a bit behind NVIDIA lately so the performance aims for next highend chip has to be high. There's 2 ways to sell cards, by performance or price and the former is the more profitable. I'd rather see ATI getting the performance crown this time than yet again price the products according to the competitor's performance.

[adamsleath]

if only some1 knew wot "rv770" was :lol:

[strange|ife]

this looks like a nice move on ati/amd's part. Card be an NV killer, at least until G100 rolls out later this year.

Will def probably be faster than the GX2

[Unbornchild]

(...)

def probably

(...):p:

[GoThr3k]

http://www.fudzilla.com/index.php?option=com_content&task=view&id=5196&Itemid=1

ES are out (according to Fuad)

[AliG]

If the ES are out it should mean that it will probably be like the rv670's launch are possible come out early, this is good news if true (though no other site has confirmed this yet)

[GoThr3k]

this is good news if true (though no other site has confirmed this yet)

my concern too

[DilTech]

First and foremost guys.. The RV770(yes, I've heard this codename before, and not from fud or theinq) is the performance-mainstream R700. Kind of like the x1950pro was to the x1950xt/x. It's competitor won't be the 9800gx2, it'll be whatever nvidia's next gen part is, plain and simple.

In otherwords, it's challenging the D9P.

If the ES are out it should mean that it will probably be like the rv670's launch are possible come out early, this is good news if true (though no other site has confirmed this yet)

Not quite...

You see, taped out and in the ES phase doesn't mean it's possible to release early. Most recent example has been the G80, which were in the developer phase(as in physically had working cards in the hands of developers) for literally more than a year before they hit stores. There's still a lot more to do before it's ready to launch.

That's assuming that what these sites say is true... The only thing I can currently confirm is that I have had the codename floating around, and the inq and fud aren't where I was told about it first. ;)

[cegras]

The reason why the R600 was :banana::banana::banana::banana: was because the hardware engineers at ATI were too ambitious.

First yields were pathetically low, 20-30%.

And yes, this is real information.

[DShankar]

In these days, taped out means nothing. It could even mean that AMD has concluded on a design, or it could mean that AMD is in production, or it could be that FUD is pulling things out of the air. I believe DilTech mentioned the fallacy of 'taping out' in reference to the R600 lineup and more in an older thread.

[AliG]

The reason why the R600 was :banana::banana::banana::banana: was because the hardware engineers at ATI were too ambitious.

First yields were pathetically low, 20-30%.

And yes, this is real information.
Hmm, is that implying that the original r600 was a stronger performing card, but due to yields they had to simplify/reduce the design?

[cegras]

Don't know / can't say much more, as I freely admit I do not exactly know how the processing structure of the GPU works.

What did happen was that with such a gimped core, the driver team was hard pressed to create things that could effectively utilize it.

And so ends my interesting tidbit of info.