I am sure Dave B will disagree with me on this one...
- Thread starter JoshMST
- Start date
trinibwoy said:
Mmm. Maybe a cigar is just a cigar this time. But I suspect Josh has at least somewhat better sources than that, and the evidence so far (i.e. his articles) suggests those sources are "NV camp" more often than not ("NV camp", please note, is more general than just NV itself).
Hey jb,
Ack! I just re-read my response to you, and it sounded somewhat more cold and unfriendly than I was trying to get across. My apologies, I didn't mean to sound short or angry.
You are very well correct in that if NVIDIA truly was the source of the initial rumor, then it could have filtered down to people that we have considered very reliable in the past. So, if that is the case, then the leakage and non-working pipelines would be the first two things to go, but I think that test yields so far are the culprit for the most part.
Ack! I just re-read my response to you, and it sounded somewhat more cold and unfriendly than I was trying to get across. My apologies, I didn't mean to sound short or angry.
You are very well correct in that if NVIDIA truly was the source of the initial rumor, then it could have filtered down to people that we have considered very reliable in the past. So, if that is the case, then the leakage and non-working pipelines would be the first two things to go, but I think that test yields so far are the culprit for the most part.
Why is everyone assuming that the G70 will be available in significant quantities for another few months? It's not like NVidia hasn't done a "paper" launch before. Even if they've got working silicon it doesn't mean they're in any posistion to actually mass produce the stuff immediately.
Now, if there's been news of a ramp up in production that I missed that's a different matter...
Now, if there's been news of a ramp up in production that I missed that's a different matter...
Eolirin said:
The only rumours regarding G70 availability so far have been that several AIB's have started receiving shipments. Gainward is also counting down to something in 11 days on its website. I don't think there was such feel good sentiment about availability weeks before the NV40 launch (although I don't remember much of what was going on at that time).
kemosabe said:
Could an additional reason be long term positioning? NV has been totally quiet for a year themselves. With Longhorn possibly coming out in 2006 this could be the last big splash before then minus a possible Spring refresh.
Could it be both parties are positioning their products to clear out inventories suffeciently while ensure just enough time, but not too much, to fit in this new gen + refresh, without causing major timing pains with Longhorn?
Megadrive1988
Veteran
hmmm... If R520 really only has 16 pixel pipelines, maybe one of the ways that ATI could make up for that is an equal amount of vertex shader pipelines. 16 + 16, thus 32 pipelines, but not 32 *pixel* pipelines. we are still on a seperate vertex | pixel pipeline configuration here, with R520.
or we have 16 super EXTREME pixel pipelines (lol) with like 3 or 4 ALUs per pipe :O
or we have 16 super EXTREME pixel pipelines (lol) with like 3 or 4 ALUs per pipe :O
Megadrive1988
Veteran
ondaedg said:
It is my understanding that both G70 and RSX are probably refreshes of NV40, and thus, NV4x architectures.
not unlike NV2A Xbox GPU and NV25 GeForce 4 Ti being refreshes of NV20 GeForce 3, and thus, were both NV2x architectures. although it might not be quite that simple anymore.
perhaps G70 and more likely, RSX, have some elements of Nvidia's true next-gen part, a completely new architecture (NV5x) which will not be seen until 2006, and is no doubt geared around Longhorn and WGF 2.0 and will have SM4.0 and will compete against ATI's R600.
look at NV20 GeForce 3. it was ready to go onto the market by fall 2000 but was held up until the spring of 2001. the NV2x architecture was in use through all of 2001 and 2002, until the NV30 GeForce FX arrived in, what, spring 2003? plus or minus a month or so. The NV2x architecture was the basis of GeForce 3, Xbox, GeForce 4 Ti, all the way through NV28 - GeForce 4 Ti 4800, if I am not mistaken.
the introduction of new architectures and refreshes have slowed down since then. GPU architectures have been stretched out. so, for these reasons, I kind of doubt that G70 GeForce 7800 and even PS3's RSX are fully based on a completely new NV5x architecture, it is my belief that they (G70 and RSX) are NV4X architectures, maybe with a sprinkling of NV5x
how closely RSX is related to G70 is something us outsiders (not in the know) will have to wait to find out
okies, these are my 2 cents
i recall similar rumors about the R420 shortly before it tapped out, and it had no yeilds problems what so ever.
the X700 had major problems tho, they went on low transistor counts under high clocks, and it didnt worked well for them, heck it didnt worked at all. ive seen only 3 retail X700xt`s cards untill they recalled the whole thing, and brought to the X800xl/ns to the table.
yes i do think that ATi had some problems with yeilds at start, this is the first time they havent tested the water first with mid/low end cards. and tapping out a 350mil~ card on a new proccess is risky, but all the rumors from/ablout TSMC suggest that the proccess works very well, and that they will be able to supply decent yeilds.
and imo comparing the R520 to NV30 is just blasphemy, GF5800 had many problems besides the GPU(like how they tried to "solve" the memory bandwidth problem).
now about all this dellays...
IMHO ATi did had problems with yeilds which they fixed. but they turened up with much less headroom then they excpected/wanted.
ATi lost the last round, yes not a very big loss but a sufficent one, they CANNOT afford to lose again.
since ATi has much less headroom to manuver then they would have liked, they are taking a risk(dont know big is it, and ill never will) by letting NV present thier product line first. after the G70 will be officialy out ATi can tune their cards just right so they`ll beat the G70 and allow the maximum yields as posible.
hopefully the R520 will be just what you`ll need to beat the G70, it will not blowup your minds off, and present amazing speeds and preformance.
i agree with some of the articles which say that there is no way ATi or TSMC are ready for such cards under 90nm, but ATi must take these stepps now, because the cannot afford to redesign their cards again, G70 nothing more then a pipmed NV40, a pipmed R4xx wont provide any competition nowdays. and a new design on 110 is kinda pointless imo.
im sorry for any grammar and spelling misstakes, i dont have the time to use a spell checker now.
i recall similar rumors about the R420 shortly before it tapped out, and it had no yeilds problems what so ever.
the X700 had major problems tho, they went on low transistor counts under high clocks, and it didnt worked well for them, heck it didnt worked at all. ive seen only 3 retail X700xt`s cards untill they recalled the whole thing, and brought to the X800xl/ns to the table.
yes i do think that ATi had some problems with yeilds at start, this is the first time they havent tested the water first with mid/low end cards. and tapping out a 350mil~ card on a new proccess is risky, but all the rumors from/ablout TSMC suggest that the proccess works very well, and that they will be able to supply decent yeilds.
and imo comparing the R520 to NV30 is just blasphemy, GF5800 had many problems besides the GPU(like how they tried to "solve" the memory bandwidth problem).
now about all this dellays...
IMHO ATi did had problems with yeilds which they fixed. but they turened up with much less headroom then they excpected/wanted.
ATi lost the last round, yes not a very big loss but a sufficent one, they CANNOT afford to lose again.
since ATi has much less headroom to manuver then they would have liked, they are taking a risk(dont know big is it, and ill never will
) by letting NV present thier product line first. after the G70 will be officialy out ATi can tune their cards just right so they`ll beat the G70 and allow the maximum yields as posible.hopefully the R520 will be just what you`ll need to beat the G70, it will not blowup your minds off, and present amazing speeds and preformance.
i agree with some of the articles which say that there is no way ATi or TSMC are ready for such cards under 90nm, but ATi must take these stepps now, because the cannot afford to redesign their cards again, G70 nothing more then a pipmed NV40, a pipmed R4xx wont provide any competition nowdays. and a new design on 110 is kinda pointless imo.
im sorry for any grammar and spelling misstakes, i dont have the time to use a spell checker now.
Bouncing Zabaglione Bros.
Legend
DOGMA1138 said:
The problem is "just enough" may not be enough to get people to upgrade, or get people to move away from Nvidia if Nvidia also brings new things to the table (such as finally providing gamma-corrected AA).
Do you want to know my immediate reaction to the story of R520 being 16 or 24 pipes with the refresh being 32 pipes? "Oh well, I guess I'll hold off on my upgrade until there's a 32 pipe monster to really make it worth spending the money".
So by only *just* beating Nvidia, ATI muddies the waters and allows for lots of uncertainty, or for early adopters like me to put off buying while I wait for the true powerhouse they could make if they wanted to (and will do a few months down the line). Being just a litte better than the opposition is not much of a reason to upgrade in favour of R520 unless you have to upgrade now for some reason.
Instead ATI should release a product that is a decisive winner, one that is so good there is no ambiguity of choice. Just as with R300 and NV30, it was completely obvious to all but the hardcore fans that R300 was miles better for gaming, so too should ATI strive to make another decisive product that is so good it virtually makes the choice for the customer.
Bouncing Zabaglione Bros. said:
The thing is, by the sounds of it releasing the "decisive winner" is going to mean making a lot less money. I suspect the big debate within ATI right now is exactly where the balance point lies between performance and fab costs, and I don't think a solid answer will be possible until nVidia announces.
Bouncing Zabaglione Bros.
Legend
Charmaka said:
Exactly. Development of these very complex chips and technology has become a balance between eeking out as much profit from your R&D as possible, verses your ability to get new customers and keep market share with the latest mindblowing tech.
ATI went ahead with R300 while Nvidia were complacent. Then Nvidia spent a lot of money on developing NV40 and selling it very cheaply. ATI needs to fight back this round with another R300 type product, as otherwise they will be seen as having faded away in the face of Nvidia's comeback. This may cost ATI more in the short term than eeking out the same older tech, but it should be counterbalanced by the higher number of chips per wafer at .90 (assuming ATI can get reasonable yields).
The smaller process should be an advantage to ATI that can allow it to produce better products - so they that's what they should do to claim back market share and revenue. The alternative is to get into a pricewar (which hurts everyone) or continue to be a marginal choice over the competition for some reasons only.
ATI has traded off the reputation of R300 long enough. They need to bring out another product with some real bang to it. Waiting for R600 is just too long. Now is the time when they need to consolidate their reputation in advance of R600 and it's supposedly revolutionary technology. When R600 arrives, ATI needs to have a reputation that can be trusted for this new tech, and not have R600 seen as just another in the long line of a slow decline since R300.
I want all those 32 pipes dammit!
Similar threads
