Official 7900 reviews thread

The word lateness doesn´t really fit in that context, since all those follow-up 80nm parts were never intended to be introduced just now. Actually the process target needed for RV560/RV570 has only been qualified by TSMC last month. RV505 was already taped out some time ago, recently RV560/RV570 and RV535 have followed, so they are perfectly on schedule.

According to ATi´s own roadmaps, they would´ve required a ("cut-down") R520 anyway to fill that gap, because there is no other chip that would´ve made sense in that context.

 

OK there are mixed messages about these two around, and you're implying they're both 80nm.

Is RV560 90nm or 80nm?

Is RV570 90nm or 80nm?

What is the difference between these two?

Jawed

 

RV570 & RV560 are 80nm according to ATI's roadmap. The difference between them I'm guessing is their configuration for different price points. (>$200 and <$200)

edit

RV570/RV560 Details

 

There .. she is fixed now boss.

 

What strikes me as odd is that all four cores (RV560, RV570, RV535 and RV505) have March as their samplingdate. But yet RV535 and RV505 will go into MP in May, while RV560/RV570 will go into MP in the Jul/Aug timeframe which is two months later. So I wouldn't be too suprised to see ATi trying to push RV560 and RV570 out the door sooner. But then again, ATi is reported to have some serious R520 volume left so we might see X1800GTO for another couple of months with it's price dropping to $199 as well as some nice X1800XT deals.

I posted some performance indications in the other topic. Going by those percentages I'd say RV570: 12-1-3-1 and RV560: 8-1-3-1 and this seems in line with ATi talking about future X1900 derivatives (plural) staying in the lead in last months financial webcast. Or maybe they even meant R590.

 

I'm now thinking that last number won't be 1.

 

My mind is still boggling at a 5-die line-up. Damn. With four of them being x-x-3-x parts, I wonder if that reinforces my theory of the dark shader quad per shader unit.

I have to admit my eyes glazed over with that list of SKUs and performances and I haven't taken them seriously.

I can take 12-1-3-1 and 8-1-3-1 more seriously though.

I suppose there's a chance that laptop parts might be getting some priority ahead of all these... erm...

Jawed

 

The problem with IHVs is that when they write (roadmaps) or say (investor/analyst conferences) something about their own future products being in the sampling phase, it could mean quite a lot, technically.

You should think of it as the timeframe they require to ensure their design is (high volume) production ready. We also shouldn´t forget that all these quoted months are estimates. Generally speaking, 3-5 months from S to P sounds about right, but that heavily depends upon the possible scenario that they might need more respins...(that could easily add another 2 months, until finally reaching P).

RV505/506 were taped out a lot earlier than RV560/RV570, so this quoted roadmap doesn´t seem to be that accurate, either.

PS: RV505 should be fabbed at TSMC, while RV516 is done at UMC.

 

No one expects the Inquisition!

DC, no, I didn't mean to imply NV had problems with 90nm similar to ATI's. I am curious to know if they had problems, period, which is what I think Jawed is after: simple curiosity. I don't think an earlier 90nm G71 would've cannibalized G70 sales as much as fed NV's margins. (Calling G71 R480 is a bit of a stretch, tho, considering R480 used the same process as R420 and R423 and didn't bench differently, whereas G71 moved to a new process, lost some transistors, and gained some "blenders." But I can see and so concede your point.)

trini, again, I'll agree that G71's lateness is irrelevant WRT this thread and sub-forum, but I don't think it's an uninteresting or unworthy topic in general. If there was a delay moving to 90nm, did it impact G80? Is it still on for July rather than, say, November? Was it intended for 90nm or 80nm? Heck, did it affect RSX?

True, ATI needed 90nm (to make their die size financially feasible?), whereas NV simply would have used 90nm to boost margins further. But both are motivated to improve margins as much as possible as soon as possible, which is all the incentive they need.

Sorry if I inferred anything untoward. That was just my immediate reaction to (explanation for) your post, though I later granted that you considered it irrelevant to this thread or this forum. Still, I thought your dismissal of his question a bit harsh, given the generally inquiring atmosphere of this site. He opened a door, and you slammed it shut. I think you caught his finger in doing so, thus this huuuge digression (from the thread topic).

OK, so let's toss NV's behavior pre-NV40 (though they did move to 110nm first, with NV43). It's true that since NV30 they've been more conservative with their high end, but it also seems that we've seen more new high end cards than process changes. IOW, 90nm took a while to arrive--3 yrs, actually. With every GPU up to NV30 (arguably to NV40, as tho NV30 was late, it was really the architecture that lagged), I believe NV pushed the process envelope harder than anyone else, and they reaped the benefits. I'm not sure why they wouldn't want to get back on track. (Intel's certainly not wasting any time, and the 65nm Conroe looks set to kick AMD's 90nm [read: fat] ass.)

nForce or something?

No, but we're discussing this after months of sky-high GTX-512 prices and knowing that ATI wanted to have a 90nm R520 out around the same time as G70(!). NV's been riding pretty comfortably, but to think they'd be roughly a year behind ATI in moving to 90nm seems a bit much. I don't believe NV's rude awakening with NV30 means they're now the turtle in the process race, but I'm slow to embrace the obvious.

 

Sure sounds like good news...

 

7600GT is already beating the 6800GS in all benchies and 7600GT also up to 60% faster than x1600xt in most benchies. Seriously, I don't know what your source is smoking....
________
Tigerblood

 

Did you actually take the time to read the rest of the thread? *cough* something about a mix-up with numbers *cough* And they are "performance indications", nothing more nothing less. There's no one who says RV560 is going up against 7600GT. I have a feeling RV570 will do just fine. And fyi, everything else from the same source was spot on even about the G71 being nothing more than a die-shrink/speedbump of G70 three months before the launch when everyone was still speculating about a possible 32 pipes card.

 

Hugs all around!

Deity, no, but I'd hate to picture a 110nm R580 or even R520, let alone think of one at 600+MHz. I don't think they had an option given what appears to be their long-term plan of R5x0 leading into R600. But this is me assuming Xenos is similar in some respects to R5x0, just closer to R600.

I think this is the horse dizietsma spotted when he took that wrong turn. So, no, no, a thousand times, no.

Anyway, I agree with the rest of your post. Well, except for that abrasive bit at the end, but it's cool.

Back on topic, I noticed this gem in THG's conclusion:

I'll attribute that and forgetting AVIVO to lack of time/sleep, but only 'cause trini's abrasiveness made me smile. Next time, I get snarky.

 

Dunno, haven't looked at the date. But isn't that a sort of "official" statement, even if late?

Though who cares about 7800GTX512 now anyway?

EDIT: you don't happen to have a link? I'd just like to check if I got something wrong there, now I ain't sure.

 

Im glad someone finally said something in a review about the shimmering problem. Props to [H] for finally stepping up. If others would do the same, perhaps NV would actually do something about it. Although, I doubt it. As they would lose a good chunk of FPS doing it. Just nice to see something "official" about it. Ive been saying the same thing they said since I got my 2405FPW the first week it came out, and then shortly after a GTX.

 

It wasnt meant to be a limited (read exotic) part, atleast that was the impression I got reading the first batch of reviews, point being Nvidia got away with it with no good reason.

 

Here are my 7900 pages...there have already been a ton of retail cards announced, I have specs and links for all those as well...

http://www.gpureview.com/GeForce-7900-GTX-card-383.html
http://www.gpureview.com/GeForce-7900-GT-card-384.html

 

CoD2, BF2 and FEAR proved them woefully wrong

But those games weren't around when 7800GTX released and when X1800XT should have released, so at that moment NVidia would have looked safe.

Jawed

 

It is one thing to look at benchmark data and another to look at sales data. In my experience NVIDIA are doing very well against the X1800 and X1900 with the 7800 series - I don't see this trend changing with the advent of the 7900 cards.