View Full Version : 2xNV40
Ok, I am going to run with the small text until I hit a wall and/or someone with credibility PM's me and says I'm an idiot... although I don't believe anything yet, it's far more fun this way. A dual NV40 board would explain...
~The oddity of of 130m=>175m transistors, yet 4x2/8x0=>16x1/32x0 (+ enhancements).
~Why a dev I spoke to saw a 8x1/16x0 NV40GL ES and managed to get this re-confirmed with nV recently despite a spate of new, reliable information suggesting otherwise.
~The ludicrous 3-rail draw of 175W.
~The leaks last year that suggested NV41 was NV40+HIS (i.e. they could both be 8x1/16x0).
~The fact that the Chaintech roadmap from January says NV40 will debut in about a month's time but now we have learnt that it will hit shelves in late April/early May at the earliest.
~In conjunction with the above point, all the confusion regarding "old" and "new" NV40s in the face of consistent news regarding steppings etc.
~The unusually paranoid security at CeBIT.
Lastly... why I heard some lunatic rantings about nV debuting 3Dfx tech this year in a major way (something like SLI, they said). It was slated for Q4 but you'd think that if they had something like this on slow-burn they'd implement it early to try and save face in the light of a 12-pipe R420. V5 6000 was so close to being launched, after all...
Be nice. :D
Oh, and there's this from Veg-eX [source of the recent NV40 numbers & screenshots]:
"Well that's all... at least all I can say now, there is something else but I risk too much and I have a promise, maybe later in this month, but I have said everything I know, hope you like it, and again excuse my bad english.
Just posted this on B3D but I figure the reaction here will be far more entertaining.
(pb) :bash: :spank: :spam: :fu: :kar: :screwy: :clap: :drooling:
Rock on,
MuFu.
Geforce4ti4200
03-21-04, 12:08 AM
I somehow doubt nv40 is sli nor dual core (bam)
Me too, but every time there's a new product generation coming people talk about multiple-ASIC implementations. Why should now be any different? :D
If you think about it, in this case it's actually a little more likely than usual.
But it does tie together all the 3Dfx rumors in a more plausible way. There was a thread running at Rage3D for a while with a person that supposedly had an inside contact. The inside contact insisted a 3Dfx product would be released this year and I found that very unlikely especially since they said it would be separate from an nvidia product. Your speculation kind of resolves a lot of that I think...if there is anything actually to all the 3Dfx rumors.
jimmyjames123
03-21-04, 12:31 AM
Did you guys see this pic posted by PaulS at Beyond3D forums ( http://www.beyond3d.com/forum/viewtopic.php?t=10894&start=40 )?
Just received this from a top secret source. Not sure what revision this is, but it seems legit. Surprised they left out DVI and AGP though
NOTE: this is actually a Voodoo 5 5500 pic, photoshopped FX-style, but it still looks cool :D
Thats the spirit. If we're gonna speculate, why not do it with a twist? :D
I guess this weekend will be called 'crazy NV40 rumor weekend' from now on. :o :clap:
jimmyjames123
03-21-04, 12:38 AM
LOL, that is so true. You have to admit though, it's been quite a fun, exciting, and interesting weekend even if it is all just speculation at this point. :D
Spiritwalker
03-21-04, 01:42 AM
It in a way does make sense. Would explain how they could get the supposed true 16x1 on so few transistors. Maybe there really is more use and a need for the bridge chip than agp to pcie...
zakelwe
03-21-04, 02:30 AM
But then people will be saying nvidia copied XGI ! :)
I'm gonna plump for 4 cores, just like the Rampage, 4 4x2 so that is 16x2 or maybe even 16x8 when nvidia's marketing people go completely banana's .:D
Regards
Andy
mmm.. very interesting the 2xNv40 possibility , thanks for sharing Mufu.. :)
it seems weird to me that Nvidia with heavy security ,allowed the INQ to see their "NV40" behind the scenes. without any NDA.. dont you think?
i have now the THeory that what the INQ saw ,was something Nvidia wanted people without NDA to see. so probably the info of the INQ is part of the True. not all.. meaning what the INQ saw was probably an NV41 or the singlechip version of the NV40 or NV40 non-ultra. :)
some possible scenarios..
1)Nv40 ultra is a 175 mtrans chip ,all what the INQ saw.
(the most likely situation)
2)THe Nv40 of the INQ was the NV41.. or the NV40 non ultra
3)the Nv40 of the INQ was the half of the NV40.. that should explain the 2 conectors.. ?
4) none of above.. hehe
NVidia is allowing mis-information to pass to the INQ and to FOrums and to little DEv ,on purpose just to see how ATI is going to react. ;)
from all those the dual chip is the most impressive one ,seems wishfull thinking, but in technology nothing is impossible.. already the volari Duo its a dual chip and it was made by a small company,and 3dfx invented the tech. :)
but if we speculate that the card is a dual one ,it is possible that each chip can do 16x1,32x0 and that they together can do ->32 x1 ,64x0 ? sounds very unrealistic , but it will be very nice to have a card with such power.. :)
btw.. found this..
john carmack.. Note that this doesn't mean that technical directors at the film studios will have to learn a new language -- there will be translators that will go from existing languages. Instead of sending their RIB code to the renderfarm, you will send it to a program that decomposes it for hardware acceleration. They will return image files just like everyone is used to. Multi chip -and- multi card solutions are also coming, meaning that you will be able to fit more frame rendering power in a single tower case than Pixar's entire rendering farm. Next year.
http://www.tech-report.com/etc/2002q3/nextgen-gpus/index.x?pg=3
that was told in june27 2002
i dont know when it will happen in -gaming cards- made for gamers by ATI or Nvidia ,but sooner or later i think ,it will,since once you reach directx9 and beyond (something like directx10) real life quality will be possible ,so there will be no need for too much graphics enhancements in video cards ,for a very long time ,so the focus will jump more to tweaks and performance. and its far cheapier mixing 2 gpus already made than designing a new one. the idea of parallel Rendering ,sharing the work between more than one hardware is not new ,as we already know and SOny is already following that philosophy (multiple cpus) a very ambitious one in the PS3.
:smoking2:
Just posted this on B3D but I figure the reaction here will be far more entertaining.
(pb) :bash: :spank: :spam: :fu: :kar: :screwy: :clap: :drooling:
Rock on,
MuFu.
Nice of you to pop by for your entertainment ;)
bloodbob
03-21-04, 05:16 AM
I heard it was dual board too Dual Vacuum cleaner :P LOL
ATI and Nvidia do multichip solutions but not for the mainstream I think it will stay that way for a long time to come.
I also got this from beyond3d...I think it was mufu's post there
http://www.nvnews.net/vbulletin/attachment.php?attachmentid=4965&stc=1
bkswaney
03-21-04, 07:37 AM
That would be a wild shocker for the world there.
My god 2 8pipe cards running at 500Mhz with Nvidia's drivers
and 1.2Ghz GDDR3. :scarey: :super: (dan) :naughty:
I guess it's possible. :)
ATI would be dog meat with no way to counter
to there next card. A new 420 Maxx. :D
That would be a wild shocker for the world there.
My god 2 8pipe cards running at 500Mhz with Nvidia's drivers
and 1.2Ghz GDDR3. :scarey: :super: (dan) :naughty:
I guess it's possible. :)
ATI would be dog meat with no way to counter
to there next card. A new 420 Maxx. :D
I don't quite think ATI are panicking yet. We should all wait till the reviews comparing each other pop out. Till then it's all rumours etc. Certainly R20 isn't a slouch as it brings more than improvements over the old r3XX line. As for the Maxx thing..they have 4 way GPU boards used in professional sims (SimFusion) reviewed by Darkcrow. They could do a two GPU card...not quite the same as a dual core mind u. I would think ATI could also do a dual core GPU as well if indeed what Nv40 is....if!
Muya is right - people definitely shouldn't think ATi are out of the race here. If the multi-chip solution turns out to be true, it's because it's all nVidia could do to top what ATi has coming with the R420. If it's still just a single chip solution, then you have even more reason to think they'll be competitive (and possibly more).
I've certainly not seen or heard anything which makes me think ATi are to be discounted this generation.
I highly doubt they would do two 8x1 chips and call it 16x1. Look at how badly the Volari tanked doing the same thing.
Muya is right - people definitely shouldn't think ATi are out of the race here. If the multi-chip solution turns out to be true, it's because it's all nVidia could do to top what ATi has coming with the R420. If it's still just a single chip solution, then you have even more reason to think they'll be competitive (and possibly more).
I've certainly not seen or heard anything which makes me think ATi are to be discounted this generation.
Well problem is we haven't much yet except CMR4 numbers. Wish some would leak a few more details. Bith NV40 and ATI are promising certainly to be a performers.
I highly doubt they would do two 8x1 chips and call it 16x1. Look at how badly the Volari tanked doing the same thing.
To be fair, nV aren't XGI. They have a history of making high performance parts and will have carried on exploring the possibility of multiple-ASIC solutions when they absorbed 3Dfx.
If the Volari Duo hadn't ever been released this speculation would be even less credible, IMHO.
Toaster
03-21-04, 04:08 PM
NV40 is a single chip. And those who are thinking it will trample all over the r420... well it won't. The r420 will most likely be ( slightly ) faster.
zakelwe
03-21-04, 04:42 PM
We need an r420 leak, the lack of information on this part is fairly interesting in itself.
Regards
Andy
jimmyjames123
03-21-04, 04:44 PM
NV40 is a single chip. And those who are thinking it will trample all over the r420... well it won't. The r420 will most likely be ( slightly ) faster.
Care to expand on this? How in the world could you possibly know this? And when you say "faster", faster in what situations?
We need an r420 leak, the lack of information on this part is fairly interesting in itself.
We know it has 16 pipelines with 12 enabled now.
Demirug
03-21-04, 04:52 PM
We know it has 16 pipelines with 12 enabled now.
The question is only on which level they the 4 piplines are deactivated.
DIE oder Design?
Exactly. Very curious about that!
vBulletin® v3.7.1, Copyright ©2000-2012, Jelsoft Enterprises Ltd.