What will replace NV50?

[Megadrive1988]

btw, here's an example of how 'reliable' Inq is

Playstation 3 delayed until 2007

By which time, kids may be older,wiser


By INQUIRER staff: Monday 21 April 2003, 02:06

IT SEEMS Nvidia may well be right and the PlayStation 3 – complete with cells and the like will not produce any return on invesment at all.
The Toshiba-Sony cell tech will now cost an absolute fortune – as Nvidia predicted, and if we see it by 2005, we will all, at the INQ, fall over in surprise.

Sony and its partners – we estimate – will have to invest at least four billion dollars into an early release.

It's the cell version that will take that time to develop, according to Electronics Design Chain.

But, it seems Sony has decided to call what was to be the Playstation 3 the Playstation 4, according to the same article.

Sony don't seem to have the amount of money needed to develop the super design, yet.

http://www.theinquirer.net/?article=9053

so I would not say that NV50 is any more cancelled than the Cell-based PS3 is delayed until 2007 and to be called PS4.

 

[WaltC]

Heh... We all know that eventually the year 3125 A.D. will in fact arrive, but in truth that matters very little to any of us, right?...

Two things rampant in Internet journalism that I loathe:

(1) Hard news stories (supposedly) which are headlined as questions (with a literal question mark at the end of the title.) Examples commonly seen might include news stories entitled thusly, such as, "Will we all get old and die one day?", or "Is nVidia secretly owned by ATi in a backroom plot to annex the US into Canada?" or "Is there a gene for tax cheating and promiscuity?", etc., ad infintum.

Basically, news stories that pose a question in their titles are at best examples of very poor news journalism and are only partially excused even when the question they raise is definitively answered in the story--which almost never happens in such cases, because if the article answers the question definitively then there's no need to pose the question in the first place. Heh... Rather, the correct headline for such a story would be a simple declarative statement summarizing the *answer* to the question. Taking off on the satirical jabs above, if these articles contained definitive answers in the form of meaningful information then the correct titles for them would be, "We are all going to get old and die someday," and "ATi Does Not Own nVidia," and "Specific Ethical Behaviors Cannot Be Empirically or Scientifically Linked to Specific Genes in Specific Human Beings," etc.

This practice is also irrational as what is the purpose in writing a news story entitled with a question which you know you cannot answer?....

(2) The flip side of this is equally appalling, which is news articles headlined or titled with declarative statements of fact which the news story itself never actually demonstrates or proves. An example would be a story entitled, "nVidia cancels nV50," but in reading the story text we discover that this only seems to be the case as nobody knows for sure, and as nVidia hadn't formally announced nV50 as an upcoming product there isn't any factual basis to conclude publicly that nVidia has cancelled it, and the whole story is proven a waste of time when we read that the proclamation on which the article is based originated with unidentified sources that were not speaking for nVidia in any official capacity whatsoever on any subject.

But these are but two sides of the same coin, which is "Idle Speculation" (or, you can leave off the "Idle" as it really makes no difference.) And of course speculation which ruminates on the unknown is the antithesis of verifiable fact, and so these stories belong on editorial pages or in topic forums and should never be passed off as "news." But unfortunately they are and unfortunately many accordingly mistake such articles for "news," which in turn lends these vacuous speculations a gravity which is wholly undeserved.

 

[Ailuros]

Allow me to aid you with a cliff note: Inquirer su**s

 

[Coz]

Allow me to aid you with a cliff note: Inquirer su**s

That's pretty harsh, they're (the Inq) very well respected by the vast majority of hardware news sites who link to Inquirer articles all the time. If they were nothing but BS then I don't think we'd be seeing much linkage to them at all. Yeah, they're not perfect and their speculation is often wrong (but that's common among everyone) but they're more often right. Of course, we never remember the times when they're spot-on we simply remember the times they screw-up.

My biggest beef with the Inq is Fuad Abazovic (or 'Fud' if you prefer) - he comes across as having very little knowledge of what he writes about and fails to convince me that his stories are accurate. For example, this week he posted this story.......

Nforce 4 SLI boards are nowhere found

.....and was promptly inundated with responses by folks who'd seen them on sale (and in many cases bought them). The very next day he put this story up.........

Nforce 4 SLI begins to sell

Whoops!!

 

[Ailuros]

That's pretty harsh, they're (the Inq) very well respected by the vast majority of hardware news sites who link to Inquirer articles all the time.

Harsh? Not a single bit. If I'd own a HW news site and would want my audience to take me a bit more seriously, I wouldn't even deal with it.

If they were nothing but BS then I don't think we'd be seeing much linkage to them at all.

Uhhhhm ok....

Yeah, they're not perfect and their speculation is often wrong (but that's common among everyone) but they're more often right. Of course, we never remember the times when they're spot-on we simply remember the times they screw-up.

The worst thing about it isn't that it's 9 out of 10 not even what I'd call an educated speculation. All they need to do is browse a forum like this one and copy/paste what they'd consider to make more sense. That's actually what they're doing in the end.

The times the site is right about anything, it was either because it was a damn predictable scenario or tons of insider information from multiple sources has leaked out, which means that's it's almost common knowledge amongst more than a few already.

 

[Coz]

All they need to do is browse a forum like this one and copy/paste what they'd consider to make more sense. That's actually what they're doing in the end.

How do you actually *know* that? Have you worked for the Inquirer in the past? If that were true then why have they reported the 'canning' of NV48, NV50 and the non-existence of NV47? Those stories were the exact opposite of what we simple forum-monkeys were speculating about!! I'm not saying those stories are correct but it proves they're not cutting and pasting from forums as you suggest.

Personally, I believe they have well-placed industry contacts and simply report back leaks and rumours from them. That's not perfect but it's still much better information than we get.

 

[Ailuros]

All they need to do is browse a forum like this one and copy/paste what they'd consider to make more sense. That's actually what they're doing in the end.

How do you actually *know* that? Have you worked for the Inquirer in the past? If that were true then why have they reported the 'canning' of NV48, NV50 and the non-existence of NV47? Those stories were the exact opposite of what we simple forum-monkeys were speculating about!! I'm not saying those stories are correct but it proves they're not cutting and pasting from forums as you suggest.

Personally, I believe they have well-placed industry contacts and simply report back leaks and rumours from them. That's not perfect but it's still much better information than we get.

Just read the B3D frontpage newsblurbs of this year's August and October that are related and the connection is easy to make.

To make it easier:

25th August 2004:

Following the Games Convention in Leipzig golem.de is reporting that following the launch of GeForce 6600 NVIDIA will move the NV4x range into the low end segment, replacing the 5200 series with an NV4x based version, and an notebook NV4x part will soon to be introduced on the NVIDIA MXM notebook video board format - a look at the PCI Express Integrators List shows NV43M, indicating that the mobile part coming first is a version of the 6600. However, evidently NVIDIA have signalled the end to the 12 month architectural cycle.

In statement that echoed ATI's previous sentiments, NVIDIA are suggesting that due to chip complexities and lithography cycle times architectural innovation times will be pushed out to about 18 months. Although NVIDIA are suggesting similar things to ATI now, if we look at NVIDIA's NV2x and NV3x generations, they have been on greater than 12 month cycles for the past two generations anyway (taking into account that NV30 was launched in November 2002, which was probably close to its intended introduction). This being the case, given NV40's announcement and wide scale availability, it would suggest that the NV50 series will not be announced until late 2005 with possible wide scale availability in 2006. NVIDIA have already stated to their investors that the low end NV4x parts will last for up to 3 years, indicating that the NV5x range will remain the domain of the high end segments and that NV4x will stay around until "Windows Graphics Foundation" in Micorsofts next geneteration OS, Longhorn.

The release of Longhorn will also be critical to both NVIDIA and ATI's plans - given the timing and the fact that there are not to be any low end NV5x, this may suggest that NV5x is set to be an extended Shader 3.0 architecture, which would indicate that the further Longhorn moves into 2007 the better it would suit NVIDIA's architectural innovation cycle. Presently the expectation is that ATI will introduce their Shader 3.0 part, suggested to be primarily developed by the R300 architectural team, in mid 2005 - ATI may be a little off their 18 month cycle as they chose not to innovate as much this cycle in order to hit the PCI Express transition - and that would also suggest that "R600" based parts would come in the late 2006 / early 2007 period. Regardless of longterm exrapolated timescales, though, its likely that both vendors will attempt to hit as close to the introduction of Longhorn as possible.

15th October 2004:

A report at Digitimes states that TSMC have already taped out around 80 products for their 90nm process and they state that around 40 more will be taped out by the end of the year, with mass production slated for next year. With the main graphics vendors being fabless they will be looking to the likes of TSMC, and others, in order to utilise their latest processes for ever more complicated chips – while 80 products may have already gone through TSMC’s 90nm tape-out its unlikely that they are likely to be anything near as complex logic ASIC’s as required for modern high end desktop graphics.

It is, however, expected that ATI will have based the R500 “Xenon†(XBox 2) graphics part on TSMC’s 90nm process, leaving Microsoft in charge of placing orders on the final chips. If Microsoft wants to ship Xenon in 2005 then the final chip will need to have been developed to leave enough time to begin ramping in order to build inventories for the launch – given that ATI have already stated that some engineers from the Xenon project have been deployed to other projects this suggests that tape-out has already occurred, or is at least imminent. It’s also expected that ATI’s next generation high end desktop graphics chip, R520, will appear in the first half of 2005 and given ATI’s CEO’s comments on the silicon requirements of a Shader Model 3.0, which it is assumed R520 will support, its likely this will also be 90nm based and tape-out should be around this timeframe in order to bring it to market within the first half Q2 ’05.

A couple of conflicting reports for NVIDIA have suggested that they may still be shopping around for their next process beyond 110nm. While IBM or TSMC may normally have been NVIDIA’s first port of call for next generation processes, NVIDIA’s reluctance to use TSMC’s 130nm low-k may be causing them to be a little leery of TSMC’s 90nm node as it is currently only offered with low-k dielectric materials. There have also been persistent reports over capacity and yield issues on IBM’s customer lines which, whilst improving, may well carry over to finer processes which could be another cause for concern. Rumours have suggested that NVIDIA’s next high end chip refresh to “NV47†may boost performance by increasing the number of internal pipelines whilst moving to 110nm, with a larger shift to 90nm later with the NV5x platform – ATI’s continued use of 150nm for R300, R350 and R360 was proof that continuing longer on a tried and tested process can still yield a performance advantage as whilst top end clock-speed may be limited in relation to the newer processes, a greater understanding can allow for larger die sizes with greater parallelism affording more performance per cycle.

Please mark the careful wording of those reports.

I don't care what either any trash tabloid site or potential wannabe speculators have to say; if you can answer me this very simple question I posted on the first side of this thread, than fine:

If you'd ask the INQ for example what they exactly think the "NV50" codename stands for, they'll most likely just lift their shoulders in ignorance.

Finally if you bounce back to page 1 and read carefully enough, you'll see that changes in roadmaps are common place under certain occassions (whereby Microsoft's timetables play obviously a major role). Logically any changes must have occured a long time ago, but it does in no way mean a cancellation of anything.

B3D expected (and it makes perfect sense) another SM3.0 chip from NV for 2005 (insert codename as your heart desires) and in between the lines a WGF2.0 chip shortly before the Longhorn release (insert another codename as your heart desires). I have the let's say "feeling" that the WGF2.0 chip will not carry a "common" or expected codename at all.

 

[Unknown Soldier]

Well I think ATI are most probably wise to use the 90nm technology, but will it fall flat? I think not.

Firstly. since they have been working on the R600 for the XBox2 and this chip uses 90nm, ATI must have quiet a bit of knowledge on how to use that 90nm tech. Coming to the R520, it's wise to use this R600 tech to enhance the R520 because, firstly as we all know from nvidia's NV4x, SM3.0 is ok using 130nm but not efficient enough. With ATI using 90nm, the R520 could actually be a killer card for SM3.0 applications/games. More transistors, maybe better yield speeds and best of all better silicon extractions(less failures). Or so ATI hopes.

All I can say is good for them. They must make the most of it.

This I think has Nvidia worried and that's why they might have cancelled a "NV50". It doesn't mean though that they can't make another "nv50".

I wouldn't be suprised if Nvidia are working on a "NV50" on the 90nm process instead of the reported 110nm process.

US

 

[jvd]

Could it be that sli has effectively killed the nv50 ?

Perhaps two 6800ultras on a board would end up close to the same speed as the nv50 and will be avalible much sooner (dual boards are starting to become avalible in large mounts) that coupled with the modest core enhancments (as the nv50 would be out before wgf and would just be a small step up from sm3.0) made nvidia cancel it instead of sinking more money into it ?

Of course on the flip side they could allways release sli nv50s . But that may kill the market for cheap dual board 90nm nv40s ?

 

[Dave Baumann]

B3D expected (and it makes perfect sense) another SM3.0 chip from NV for 2005 (insert codename as your heart desires) and in between the lines a WGF2.0 chip shortly before the Longhorn release (insert another codename as your heart desires). I have the let's say "feeling" that the WGF2.0 chip will not carry a "common" or expected codename at all.

Ail, its good to hear at least someone is getting those news posts! However one other thing you may want to take note of is the timing in relation to Longhorn. The more Longhorn becomes a 2006 product the less another shader 3 part makes sense late 05 / early 06.

 

[Ailuros]

B3D expected (and it makes perfect sense) another SM3.0 chip from NV for 2005 (insert codename as your heart desires) and in between the lines a WGF2.0 chip shortly before the Longhorn release (insert another codename as your heart desires). I have the let's say "feeling" that the WGF2.0 chip will not carry a "common" or expected codename at all.

Ail, its good to hear at least someone is getting those news posts! However one other thing you may want to take note of is the timing in relation to Longhorn. The more Longhorn becomes a 2006 product the less another shader 3 part makes sense late 05 / early 06.

People are reading the news posts here at B3D, despite what you may think and no I'm not talking about myself, but the general feeling I get on various other boards.

As for 2005 I find it hard to believe that NVIDIA would want to counter ATI's next generation part just with NV40-SLi configs. If they'd have something in the works it could either be just a clock frequency increased NV4x or increased amount of quads. Either way I'd call it a refresh just like I used to call NV25 a "refresh" in relative terms too.

Let's call "NV50" NV50; historically the numbering scheme would indicate a next generation part; it doesn't really take a wizzard to figure out that either IHV would want to release a WGF2.0 part as close as possible to the Longhorn release. Not too early and not too late. None of the two can afford to miss that particular boat.

Finally there are some indications emerging that the forementioned part might have unified units; and that would NOT suggest anything else but a WGF2.0 part. All IMHO of course.

***edit: credit for the speculation in the last paragraph goes of course to Demirug.

http://www.forum-3dcenter.org/vbulletin/showthread.php?t=190859

 

[Jawed]

Maybe NVidia decided that going non-unified for WGF (even though they could, technically) would be an NV30 style disaster and are now busily re-working for unified.

In the run-up to Longhorn, presumably OEMs are going to demand sample WGF 2.0 cards to test (thinking Dell here). Being early with demonstration cards would surely be a big advantage if you want to cut new OEM deals. I suppose that depends on the degree of WGF 2.0 integration into late betas.

On the other hand will the WGF 2.0-specific functionality of Longhorn really be relevant, when it's simply DX9 functionality that's required for the GUI?

Jawed

 

[Xmas]

I wouldn't be suprised if Nvidia are working on a "NV50" on the 90nm process instead of the reported 110nm process.

Which reports do you mean?

 

[Coz]

As for 2005 I find it hard to believe that NVIDIA would want to counter ATI's next generation part just with NV40-SLi configs. If they'd have something in the works it could either be just a clock frequency increased NV4x or increased amount of quads.

I agree. SLi may deliver great performance for the ultra high-end market but you need those new cores and refreshes to stay competitive throughout the entire price range. There's also a compelling economic argument for an NV4x refresh on the 110nm process. The simple truth is that NV need something better to counter R520 (and it's derivatives) and SLi ain't it.

 

[Ailuros]

As for 2005 I find it hard to believe that NVIDIA would want to counter ATI's next generation part just with NV40-SLi configs. If they'd have something in the works it could either be just a clock frequency increased NV4x or increased amount of quads.

I agree. SLi may deliver great performance for the ultra high-end market but you need those new cores and refreshes to stay competitive throughout the entire price range. There's also a compelling economic argument for an NV4x refresh on the 110nm process. The simple truth is that NV need something better to counter R520 (and it's derivatives) and SLi ain't it.

A clock frequency increase would suggest a relatively small transistor count increase, additional quads a much bigger one. Keeping the 222M count for NV40 in mind, makes it sound quite difficult to think of 110nm and higher transistor counts. Power consumption isn't exactly relatively small on NV43's, if one considers that it has only 2 quads (yes clock frequency is at 500MHz).

 

[Unknown Soldier]

Which reports do you mean?

A couple of conflicting reports for NVIDIA have suggested that they may still be shopping around for their next process beyond 110nm. While IBM or TSMC may normally have been NVIDIA’s first port of call for next generation processes, NVIDIA’s reluctance to use TSMC’s 130nm low-k may be causing them to be a little leery of TSMC’s 90nm node as it is currently only offered with low-k dielectric materials. There have also been persistent reports over capacity and yield issues on IBM’s customer lines which, whilst improving, may well carry over to finer processes which could be another cause for concern. Rumours have suggested that NVIDIA’s next high end chip refresh to “NV47†may boost performance by increasing the number of internal pipelines whilst moving to 110nm, with a larger shift to 90nm later with the NV5x platform – ATI’s continued use of 150nm for R300, R350 and R360 was proof that continuing longer on a tried and tested process can still yield a performance advantage as whilst top end clock-speed may be limited in relation to the newer processes, a greater understanding can allow for larger die sizes with greater parallelism affording more performance per cycle.

Seeing as the NV40/5 were 130nm it's easy to see Nvidia working on a 110nm NV50. Of course they've cancelled the NV50 now. And the NV48. Oh and there's no NV47. So say the Inq.

US

 

[Kaotik]

Could it be that say NV50 was an AGP part and NV55 would be the PCI-E version of the same part similar to NV40 and NV45? Cancelling NV50 might then not mean much as they could use NV55 with a bridge chip for AGP.

I fail to see NV45 as "true PEG part" since after all, it's just HSI & NV40 slapped onto one flipchip

 

[2senile]

http://www.digitimes.com/news/a20041220A7039.html

The NV48 will take on ATI’s next-generation high-end graphics chip, the R520, which is also expected to be launched in the second quarter of 2005. The R520, however, will be the industry’s first GPU produced using a 90-nm process at TSMC.

My "bold".

 

[Unknown Soldier]

lol .. so much for the Inq.

US

 

[Unknown Soldier]

BTW. ain't 2Q05 a bit late? I wonder how fast they can push the NV48 and how cool it can keep.

The R520 is expected around 2Q05 too, ain't it?

US