View Full Version : R423 will have 16 pipes(16x1) and will own nv40
nutball
03-24-04, 07:08 AM
Alright, 8, 16, schmixteen, schmirtytwo.
How about an architecture that is capable of being configured to be any of 8x2, 16x1 and 32x0???
Has that been ruled out?
Demirug
03-24-04, 08:06 AM
Alright, 8, 16, schmixteen, schmirtytwo.
How about an architecture that is capable of being configured to be any of 8x2, 16x1 and 32x0???
Has that been ruled out?
Everthing is possible but in any case you need one base architecture.
Wenzula
03-24-04, 11:50 AM
@nv40:
“a -wider- R300?” – The R420 seems to be a tweaked r3xx with for added pipes.
“when the R423 will be released?!” – the r423 will come after the r420, I have seen a roadmap. On this the chip is supposed to be out in Q3.
Because of the pipe-configuration, i will try to contact my sources. Maybe he can confirm, wheter it is 8x2 or 16x1.
Carbon Unit
03-24-04, 02:51 PM
I wouldnt waste my money in the NV40, it will be blown away by the NV45 due out about 3 months after the NV40's release (foshiz) owned:
zakelwe
03-24-04, 03:54 PM
tongue in cheek mode
You know, I've almost got bored of r420 and nv40 already, they are just using a brute force approach to past architectures
Bring on nv50 and R500 .. brand new architectures .. no pipelines . massive cache .. grid arrays .. " just in time " pixel production , Hamiltonian matrices.
What are your thoughts ?
Regards
Andy
Wenzula
03-24-04, 04:10 PM
Well, i personally think nvidia did the first step with nv30 and will do the second one with nv40.
ati still has to do the step. they are still use a "conservative" architecture, compared to nvidia. As far as i know. At first the r400 was intended to use a completly new architecture, but the performance was to slow. So ati decide to skip the chip. Instead they will bring us the r420 which is nothing more than a tweaked r3xx core. The really revolutoinary step will come with the r500.
p.s.: nv40 will be real 16 pipes!!!
Well, i personally think nvidia did the first step with nv30 and will do the second one with nv40.
ati still has to do the step. they are still use a "conservative" architecture, compared to nvidia. As far as i know. At first the r400 was intended to use a completly new architecture, but the performance was to slow. So ati decide to skip the chip. Instead they will bring us the r420 which is nothing more than a tweaked r3xx core. The really revolutoinary step will come with the r500.
p.s.: nv40 will be real 16 pipes!!!
How did the nv30 have anything to do with "the first step"? What is "the first step" anyways?
einstein_314
03-25-04, 02:03 AM
Well, i personally think nvidia did the first step with nv30 and will do the second one with nv40.
ati still has to do the step. they are still use a "conservative" architecture, compared to nvidia. As far as i know. At first the r400 was intended to use a completly new architecture, but the performance was to slow. So ati decide to skip the chip. Instead they will bring us the r420 which is nothing more than a tweaked r3xx core. The really revolutoinary step will come with the r500.
p.s.: nv40 will be real 16 pipes!!!
Hey Wenzula, how do you know it will be a real 16 pipes?
Evildeus
03-25-04, 02:52 AM
Perhaps because he was at Cebit?
Geforce4ti4200
03-25-04, 03:03 AM
Perhaps because he was at Cebit?
yes, but would he be under NDA also nvidia would just tell everyone its 16 pipes. for a few months after the nv30 was out, everyone thought it was 8x1 but then someone found the truth, it was 4x2!
I think it won't have 16 pipes as we know just be able to process 16 pixels at a go....which ever way works its fine as long as the output is equivalent to 16 individual pipes each processing a pixel.
Evildeus
03-25-04, 07:28 AM
yes, but would he be under NDA also nvidia would just tell everyone its 16 pipes. for a few months after the nv30 was out, everyone thought it was 8x1 but then someone found the truth, it was 4x2!
Yeah, i know.
PS: Perhaps he is breaking his NDA. But i thought that Nv was less secretive this time. And the partners can talk also, you know ;)
he wouldn't be under NDA unless he was told directly by them (NVIDIA), he heard of someone who heard it someone at cebit, that doesn't apply to him...
Wenzula
03-25-04, 11:06 AM
Well let's say i have my sources... :-)
Wenzula
03-25-04, 01:09 PM
Hmm, iu have heard that the r420 will be launched aline with the nv40. So you can guess which date it will be :-)
gordon151
03-25-04, 06:08 PM
Well let's say i have my sources... :-)
We can only hope this isn't your source here (http://www.amazon.com/exec/obidos/tg/detail/-/B00001ZWV7/ref=ase_indrasnet/104-5997639-1932765?v=glance&s=toys) :p.
Hey Muya, do you mean it (ab)uses the vertex shaders to pump out extra pixels? And if it does this, what are the limitations ( besides the lack of filtering ) of doing it this way?
Hey Muya, do you mean it (ab)uses the vertex shaders to pump out extra pixels? And if it does this, what are the limitations ( besides the lack of filtering ) of doing it this way?
Dude, I am just basing my speculation on whats being typed around the forums over the net :D I am not exactly sre but, (ab)use of VS been discounted for the most part by various ppl at b3d. So i dunno.
Demirug
03-26-04, 02:44 AM
Dude, I am just basing my speculation on whats being typed around the forums over the net :D I am not exactly sre but, (ab)use of VS been discounted for the most part by various ppl at b3d. So i dunno.
They abuse of the VS is not possible with NV40 because the vertexshader can only use pointsampling as texturfilter. The only thing that you can do is using the VS for Z/stencil passes without colors. But in this special case you are not shader limited. If a IHV want he can generate 5 Z/stencil pixels per clock per pipe. The number of ROPs are make the limit. And even if you have ROPs you still need bandwith to write all this information in the memory.
Geforce4ti4200
03-26-04, 05:02 AM
the rumor mill has slowed down lately.......no more rumors comming in before final release of both cards? anyway I think even the r420 will own the nv40 unless somehow the nv40 has 16 real pipes which some say yes, others say 8x2
yeah, i think those with some beta boards or whatver have been told to pipe it before the launch or at least artificially with hold info till next week so that the appetite for nv40 info grows even more.
Evildeus
03-26-04, 07:42 AM
At least JC have one NV40 board running in his dev PC
incidentally, he also mentioned that he had just recently "gotten an nv40 card and plugged it into [his] dev system" and that "it rocks". It was a spontaneous aside, but boy, you can't beat that kind of endorsement if you're nvidia. (At least, I'd like to continue believing that it was an entirely spontaneous comment.. )
http://www.beyond3d.com/forum/viewtopic.php?t=11091
GlowStick
03-26-04, 11:41 AM
At least JC have one NV40 board running in his dev PC
http://www.beyond3d.com/forum/viewtopic.php?t=11091
Yeah, it is confirmed now that all major devs have nv40 samples, but whats more intresting about that article is, he says 'doom3' is close :D :D :D
Yeah, it is confirmed now that all major devs have nv40 samples, but whats more intresting about that article is, he says 'doom3' is close :D :D :D
why can't leak specs!!! NV will hardly have a go at them for doing so ;)
Wenzula
03-26-04, 02:26 PM
@gordon151: I never base myself on amazon:)
Well, my guess is 16x1 and 32x0 for nv40 with around 450 - 500 Mhz coreclock and 600 MHz gddr3. Maybe John Carmack can tell us more:)
vBulletin® v3.7.1, Copyright ©2000-2012, Jelsoft Enterprises Ltd.