Well then coming back from CeBit, Dave, what's your take on the situation? From what Inq mentions and what you've heard yourself? C'mon, we plebians are going batty out here!DaveBaumann said:
Inquirer: We Have NV40 Sample...
- Thread starter mreman4k
- Start date
DaveBaumann said:
That's the problem, Dave, in this case...
I was commenting only on what was written, as it was written, with no regard for the personality doing the writing. Basically, there are so many unanswered questions the article brings to mind--questions which could have been answered easily, I think--that I have no confidence that the author was actually looking at what he thought he was looking at. If he was, the problem is that he failed to communicate the fact.As well, I think it is because of this particular trait of erroneous and/or incomplete writing that people have developed any "general INQ skepticism" they may have. Such skepticism comes as a result of the writing, rather than as comment on the individuals doing the writing, which seems to me to be irrelevant...
IE, if Hemingway had written a "stinky" book, the fact that Hemingway was the author wouldn't change the fact that it stinks...
In defense of the Inq., lately many of their rumours do have alot of truth in them. For instance, all the rumours they launched about the P4 Prescott were true in fact. So, I wouldn't dismiss their rumours as easily as people tend to do every time. Because of lately many have come true.
Guden Oden said:
Actually, I can clearly recall all of the posts I read (well, not all of them, of course) that were made by so-called "insiders" in the last few weeks of nVidia's "countdown" relative to the nv10 launch, who swore up & down they were running prototype nV10s (yes, that far back), and who swore that the pipeline organization for them was 8x2...
And as you point out this happened again with nV30, nVidia's "next-gen" gpu post the nV10-25 gpu family. In the case of nV30 it was compounded by the fact that nVidia also officially misrepresented the gpu as 8x1, and I think is still doing that, as unbelievable as that may be.
So anyone who *isn't* skeptical of the current "16 pixel pipeline" claims, is, well, a dollar short somewhere, I think...
Such skepticism seems entirely warranted and appropriate.I notice in the Inquirer pieces to date a distinct absence of any acknowledgement of how the pixel-pipeline organization has oft been misrepresented in the past prior to the introduction of new nVidia gpus, not to mention that nVidia has intentionally misrepresented its nV3x hardware *officially* since it shipped in terms of this specification, and that calls the INQ's technical and professional competence into question, I think. If they don't know that despite what nVidia states to the contrary, nVidia never made or sold an 8x1 nV3x gpu, I think questions as to their technical competence are beyond dispute.
I think skepticism about a genuine 16-pixel-pipeline claim for nV40 is the only thing that is approriate at this time. This of course does not mean that I or anyone else should think that a 16-pixel-pipeline gpu is a technical impossibility for either ATi or nVidia. It simply means that so far there is no *credible evidence* to support the idea that either ATi or nVidia have any imminent plans to produce and market such a gpu, either for R4x0 or nV4x. Considering, especially, nVidia's yield problems relative to producing and marketing a 4x2 .13 micron gpu, problems consistent for well over a year now, problems not solved by switching FABs for nV3x production nor by the subsequent revision tweaks this entailed, that is why I find the notion of a genuine 16-pixel-pipe .13 micron nv40--not impossible--but highly unlikely. It seems to me that in order to accomplish this nVidia would have to have completely scrapped nV3x, and that nV4x would have to be a clean, new architecture from the ground up, at the very least. I am, at best, dubious that anything approaching this has happened. Time will tell, but as of now I've seen nothing that would cause me to actually think nV40 is 16x1.
WaltC said:Guden Oden said:
Actually, I can clearly recall all of the posts I read (well, not all of them, of course) that were made by so-called "insiders" in the last few weeks of nVidia's "countdown" relative to the nv10 launch, who swore up & down they were running prototype nV10s (yes, that far back), and who swore that the pipeline organization for them was 8x2...
And as you point out this happened again with nV30, nVidia's "next-gen" gpu post the nV10-25 gpu family. In the case of nV30 it was compounded by the fact that nVidia also officially misrepresented the gpu as 8x1, and I think is still doing that, as unbelievable as that may be.
So anyone who *isn't* skeptical of the current "16 pixel pipeline" claims, is, well, a dollar short somewhere, I think...
I notice in the Inquirer pieces to date a distinct absence of any acknowledgement of how the pixel-pipeline organization has oft been misrepresented in the past prior to the introduction of new nVidia gpus, not to mention that nVidia has intentionally misrepresented its nV3x hardware *officially* since it shipped in terms of this specification, and that calls the INQ's technical and professional competence into question, I think. If they don't know that despite what nVidia states to the contrary, nVidia never made or sold an 8x1 nV3x gpu, I think questions as to their technical competence are beyond dispute.
I think skepticism about a genuine 16-pixel-pipeline claim for nV40 is the only thing that is approriate at this time. This of course does not mean that I or anyone else should think that a 16-pixel-pipeline gpu is a technical impossibility for either ATi or nVidia. It simply means that so far there is no *credible evidence* to support the idea that either ATi or nVidia have any imminent plans to produce and market such a gpu, either for R4x0 or nV4x. Considering, especially, nVidia's yield problems relative to producing and marketing a 4x2 .13 micron gpu, problems consistent for well over a year now, problems not solved by switching FABs for nV3x production nor by the subsequent revision tweaks this entailed, that is why I find the notion of a genuine 16-pixel-pipe .13 micron nv40--not impossible--but highly unlikely. It seems to me that in order to accomplish this nVidia would have to have completely scrapped nV3x, and that nV4x would have to be a clean, new architecture from the ground up, at the very least. I am, at best, dubious that anything approaching this has happened. Time will tell, but as of now I've seen nothing that would cause me to actually think nV40 is 16x1.
Fanboy!!!
sonix666 said:
Look, the Inquirer is a rumor mill...
That's what it is, and when I say that I don't say that critically because I *like* the Inquirer and I admire the spirit of what they attempt to do, and I have corresponded a few times with individuals staffing the site who seem friendly, reasonable, and chordial. So, I have to separate my personal feelings for the people there from an objective examination of whether what they say is likely or not. The two are entirely separate, to me.The truth is, as AlphaWolf states, that when you are all over the map with your rumors a few of them will inevitably hit the mark...
That doesn't speak to any overall credibility--remember that what they run with is usually unattributed, unverifiable rumor--and nothing more. For instance, INQ has stated in separate rumor pieces its written that nV40 is 175M transistors, in other nV40 "anonymous insider source" rumors they've run, it is 205-210M transistors because it is a "16 pixel pipeline" gpu. Gee, you think maybe one or the other transistor-count rumor might turn out to be close to the truth? I do--but then I might have guessed much the same myself, for a number of reasons... Still, though, it remains to be seen whether either transistor-count rumor printed to date will be accurate... That's the point to ponder: the idea is not to "trash" the Inquirer, but simply to put it into perspective and appreciate it for what it is as opposed to what we might prefer it to be.Geeforcer said:
Heh...
You had to ask me that, didn't you?Truth is, I've spent more time thinking about what I consider it likely isn't than I have in considering what it probably is...so maybe I should break this down as follows...
What I think it probably is: 8x1/2, with greatly improved DX9 functionality. (Without a serious streamlining of nV3x architecture, though, I have difficulty imagining them able to support 475MHz at .13 microns at 8x1, though, considering their complete inability to support 8x1 in nV3x at that clockspeed at the same .13 microns, but I find that more likely than....)
What I hope it could possibly be, but find highly unlikely: 16x1 with greatly improved DX9 capability (I find current clockspeed rumors of 475MHz @ .13 microns to make this particularly improbable, but the clockspeed rumors could simply be wrong.)
What I hope to heck it is NOT: 4x4, or something kinky like that, with adequate DX9 hardware functionality of a degree to allow a continued lip-service to DX9-grade functionality.
I am content, though, to just wait and find out...
