More Nv-30 speculation !!
- Thread starter Ascended Saiyan
- Start date
Careful there.
Some people still vigorously regret doing that.
(In Finland the Elsa Gladiac Ultra appeared at around USD 800+, got mine from Germany for about USD 600... Obviously had extra money in my hands at the time... And the following months looked surprisingly like The Great Memory Price Nose Dive a couple years back, exactly right after I had invested in Mushkin's Rev2 PC133... But hey, at least the Revelator 3D goggles are great for funny dress parties, if for nothing else.)
Some people still vigorously regret doing that.
(In Finland the Elsa Gladiac Ultra appeared at around USD 800+, got mine from Germany for about USD 600... Obviously had extra money in my hands at the time... And the following months looked surprisingly like The Great Memory Price Nose Dive a couple years back, exactly right after I had invested in Mushkin's Rev2 PC133... But hey, at least the Revelator 3D goggles are great for funny dress parties, if for nothing else.)
Another speculation:
http://www.theinquirer.net/?article=5794
At least they admit the IKOS-stupidity...
http://www.theinquirer.net/?article=5794
At least they admit the IKOS-stupidity...
Much as I hate to bump one of these sordid threads...
Does anybody know what happened at nVidia this weekend? Something pretty important and exciting, I feel. I was hoping to be able to get some info on NV30 prior to the launch (a few of you might remember me posting some R300/RV250 info a few months back...) but my dude is being very tight-lipped. They are much more strict about NDAs at nVidia it would seem.
Apparently, this weekend everybody was working virtually 24/7 (very busy, even by the furious standards of late). Something back from the fab?
Taking into account the speculated tape-out period (personally, I believe it was in the first couple of weeks of September), a 1-2 week delay considering possible complications involving the new process and the fact that there have been no reports of early samples thus far, the flurry of activity this weekend may well have been due to the NV30 ASIC, rev. 1 returning from TSMC, right?
Can anybody add to this? Purely speculation at this stage, of course...
MuFu.
Does anybody know what happened at nVidia this weekend? Something pretty important and exciting, I feel. I was hoping to be able to get some info on NV30 prior to the launch (a few of you might remember me posting some R300/RV250 info a few months back...) but my dude is being very tight-lipped. They are much more strict about NDAs at nVidia it would seem.
Apparently, this weekend everybody was working virtually 24/7 (very busy, even by the furious standards of late). Something back from the fab?
Taking into account the speculated tape-out period (personally, I believe it was in the first couple of weeks of September), a 1-2 week delay considering possible complications involving the new process and the fact that there have been no reports of early samples thus far, the flurry of activity this weekend may well have been due to the NV30 ASIC, rev. 1 returning from TSMC, right?
Can anybody add to this? Purely speculation at this stage, of course...
MuFu.
Joe DeFuria
Legend
Yeah, the time line seems about right to get first silicon back from the fab, assuming tape-out in early Sept. Note that by most expectations, the first silicon probably is not in good enough shape to be publically "demoable". (Does anyone expect first silicon ever to be in that shape?)
Aside from the normal "pressure" to determine errata and get fixes in for a respin ASAP, there must be immense pressure with this particular part: because the publicized "launch" is only about 4 weeks away.
There would be a lot of disappointment (from fans and investors) if there is no public showing of working silicon at fall Comdex. It seems to me that if nVidia got first silicon back this past weekend, they are REALLY cutting it close to get a re-spin (assuming one is needed) back from the fab in time to have a public showing at Comdex.
All speculation of course, but it makes for a nice story.
Aside from the normal "pressure" to determine errata and get fixes in for a respin ASAP, there must be immense pressure with this particular part: because the publicized "launch" is only about 4 weeks away.
There would be a lot of disappointment (from fans and investors) if there is no public showing of working silicon at fall Comdex. It seems to me that if nVidia got first silicon back this past weekend, they are REALLY cutting it close to get a re-spin (assuming one is needed) back from the fab in time to have a public showing at Comdex.
All speculation of course, but it makes for a nice story.
Joe DeFuria said:
We might hear some details even before that after the nVidia Iron Developer Conference on October 23 in Japan where whey promise:
You need some fairly okay working silicon before you can annonce 'details' on the NV30. I guess that things will start to leak big time after October 23.
Joe DeFuria said:
To be honest I really don't think they can respin & debug/verify in time.
But, having said that I'm fairly sure the DOOM III demo at E3 was running on A11 R300 silicon. That part was limited it to ~250MHz (mem interface problems) using standard cooling and no doubt had numerous other post-mask fixes, but nevertheless was in good enough shape to run the demo and also show off AGP 8x on VIA's stand. It is possible that nVidia will have demo hardware for Comdex.
MuFu.
P.S. Joe, I posted in your nVnews thread. What's the deal with that anyway?! Way to much winking for my liking...
On both the projects I worked on at my current company, we had demo-able silicon on the first try. While we did some revisions to get yield up and fix some issues with minor blocks, we could have sold them as is.
Both were new processes in new fabs (to us).
So yes, 1st silicon is usually demoable. If its not usually demo-able, you need to fire some verification people as its impacting your schedule way too much. A chip that is completely DOA will put your entire schedule back 4-6 weeks since you can't do post silicon validation on anything.
Both were new processes in new fabs (to us).
So yes, 1st silicon is usually demoable. If its not usually demo-able, you need to fire some verification people as its impacting your schedule way too much. A chip that is completely DOA will put your entire schedule back 4-6 weeks since you can't do post silicon validation on anything.
Joe DeFuria
Legend
Ok, assuming that is true, my "theory" could still hold water, by adding the following:
That nVidia not only wants to have "demoable" silicon for the launch, but silicon that shows a performance advantage over R-300.
In other words, if their silicon is demoable, but either lacks the Mhz, or has some crucial (to marketing) feature not working as it should, nVidia may not WANT to demo such a part publically. That's what I really mean by "publically demo." (Demo that part so that it is shown in a favorable light compared to the competition.)
It would be different if R-300 didn't exist, and all nVidia had to do was demo something faster than GeForce4 or Radeon 8500.
Yup I agree with that.
Regardless of NVIDIA choosing to climb a bigger hill with .13um, people will still be rightfully disapointed if NV30 is slower then R300 since NVIDIA has had more time to engineer the product.
If NV30 isn't a obvious step above R300 expect NVIDIA's marketing team to be burning the midnight oil...
Regardless of NVIDIA choosing to climb a bigger hill with .13um, people will still be rightfully disapointed if NV30 is slower then R300 since NVIDIA has had more time to engineer the product.
If NV30 isn't a obvious step above R300 expect NVIDIA's marketing team to be burning the midnight oil...
Leadteak will release the PR on its next NV30 card (WinFast A330 + 128Mo DDR (?)) at the end of november, and release time is...to be determined
http://www.3dchips-fr.com/News/news.php?date=2002-10-16&newsnumber=5
Which is why you'd do more pre-silicon verification and risk reduction on the front end.
Its not like there's two guys in a back room putting it all together and hoping it works right without testing anything.
The fact that it was running on the IKOS means that there's likely no LOGICAL/FUNCTIONAL errors that aren't known by now.
The analog section is likely the big scary part for them, but even that might not be a show stopper for demos. They have a digital breakout to circumvent the RAMDACs if they come back busted (Or must, since they every product to date has had the option for external ramdacs for the secondary monitor).
I'm not saying its definately demo-able, don't get me wrong. But I don't think its obviously not demo-able either.
Its not like there's two guys in a back room putting it all together and hoping it works right without testing anything.
The fact that it was running on the IKOS means that there's likely no LOGICAL/FUNCTIONAL errors that aren't known by now.
The analog section is likely the big scary part for them, but even that might not be a show stopper for demos. They have a digital breakout to circumvent the RAMDACs if they come back busted (Or must, since they every product to date has had the option for external ramdacs for the secondary monitor).
I'm not saying its definately demo-able, don't get me wrong. But I don't think its obviously not demo-able either.
Similar threads
