NV40 Test samples, pics inside

Depends. Tantalum versions go up to 100s of uFs and can thus be used for supply line filtering. Can't remember if their downside was leakage current or high ESR. Anyway, the bigger tantalum ones are - well - bigger...

 

About the number of caps

I seem to remember some previous leaked PR images for some card in the NV3x line that only had a few but the final production cards had more than in the leaked pics.

Anyone else remember this? Or am I hallucinating about remembering things I saw?

 

Better to find out now than after you've bought one. Just consider it as part of your pre-purchase research.

Besides you might be pleasantly surprised if R420 provides compariable or better performance without any of the NV40 power/heat issues.

 

Right about now the boys at NVIDIA are probably none to happy about not developing a Low-K solution.

 

After their experiences with NV30 and new processes, maybe they decided to go for a known, established process. A repeat of NV30 process problems and subsequent delays and redesign would have been disasterous for Nvidia, so they decided not to gamble on unproven processes.

ATI with their technique of trialling new processes on simpler, mid/low range chips in advance of using that process on the next generation high-end seems to be working well for them.

 

That's my expectation (mostly) now, hence my lack of enthusiasm and comments about the performance coming at a cost not worth paying. We'll find out soon enough anyway.

More than likely, and that would be fine had NV not bad mouthed the process in public. They should have simply said that they didn't want to risk the process this time around, and leave it at that.

 

I don't know how the CEO of a technology company can get away with saying that a technology being employed by IBM NASA and your chief competitor as well as a 1/2 dozen companies is 'dangerous' and not be called on it.

 

IBM use Low K through as manufactured by Dow think the call it SiLK. When it comes to power consumption, heat, and cross talk Low K is better.

 

Low K is a form of SOI isn't it.. I don't believe NVIDIA is using this or they wouldn't need two power connectors if the process allowed 50% less powe consumption.. Let me read thes articles.

 

SOI used to be Sapphire grown on as a dielectric but looks like others are coming out. I never herd of Silk. We will see how good it is when ICs come out using it.

http://www.digitimes.com/NewsShow/MailHome.asp?datePublish=2004/4/12&pages=A5&seq=20
As for this, its funny that a metal is used as a dielectric. Any metal can migrate and form pits in the silicon (electron traps). It should be faster to deposet alumina-aluminium oxide (Al2O3) then growing oxide-nitride-oxide (ONO)( known as nitrogen glass developed by IBM). The nitrogen glass don't migrate like metals do so its much better but nitrogen glass is very fragial and alumina-aluminium oxide is not. For ppl that like to overclock there mem will not like this new Samsung mem because the bump in heat and volts/amp will speed the migration process and basicly short the mem chip.

 

I believe low-k and SOI are two separate things.
http://www-306.ibm.com/chips/technology/makechip/breakthroughs/soi.html
http://www-306.ibm.com/chips/technology/makechip/breakthroughs/lowk.html

There's actually a pretty good (though simplistic) overview of the silicon making process on that website, though it might be a bit dated.

http://www-306.ibm.com/chips/technology/makechip/?source=leftnav

 

SOI is Sapphire grown as a dielectric and Low-K is Black Diamond ( I think its a carbon based/Graphite) grown as a dielectric. I think Black diamond is better dielectric and is not as fragil but will take more time to grow.

 

Hmmm caps...

I remember once on a pager I rotated a cap by 90 degrees..which fixed are pager to pass the old FCC specs and we could ship that product


Anyways as said before caps can be used for lots of things. Yea you normally don't use caps like these for impedence matching but none of us really knows what the requirements are, what voltage management system they have, ect. Too many varibles. Heck it may only need two caps, just NV felt safe to add a two extra. So the ammount of caps really does not mean jack.

 

Never would have thought they make tantalum caps with packaging other than the common cylindrical electrolyte caps, but looks like you're right : http://www.elna-america.com/PDF/SK.PDF

 

SOI is commonly known as Semiconductor on Insulator. It is a process where by some method is used to grow an insulating layer into the silicon crystal. The transistors are then manufactured so that they abut with this insulatig layer. The primary benefit is a reduction in transistor leakage. This reduction in leakage can either be used to increase the speed of the process or the power of the process. Semiconductor on Sapphire (SOS) is one form of SOI.

Low-K is generally refering to an inter and intra-layer dialectric with a dialectric constant better than SiO. There are multible materials on the market for Low-K work each with their own trade-offs between dialectric constant and manufacturability. In general, the better the dialectric constant, the more fragile the material hence the higher risk involved.
Of course the best semiconductor dialectric material that has been demonstrated is air-bridge which effectively uses air as the dialectric constant.

Aaron Spink
speaking for myself inc.

 

Everything I've seen uses SOI and Silicon on Insulator interchangeably.

 

Must have been a brainfart.