NVIDIA GF100 & Friends speculation
- Thread starter Arty
- Start date
Yield, power and clocks?
Metal spins are used to fix functional bugs. I can't call a chip which needs a metal spin to work as a "working" chip. If by production quality you mean yields, power and clocks, then they need a silicon respin to fix.
thatdude90210
Regular
Maybe they are seeing less performance than they were expecting from a 512 shaders monster, just saying...
Well they obviously had working A1, too, they did demo if it even if ran barely.
Gamestopper bugfree and bugfree are different things, too, are they not?
I mean, as long as these are engineering samples and prototypes, a bug or two won't kill you even if they would probhit you from putting them to retail channel?
http://ic.tweakimg.net/ext/i/imagelarge/1267547763.jpeg
http://ic.tweakimg.net/ext/i/imagelarge/1267547764.jpeg
http://ic.tweakimg.net/ext/i/imagelarge/1267547765.jpeg
pictures of the board and cooler
its got all 12 chips its got 512 shaders, or are they able to cut off shader units and not effect the bus width?
http://translate.google.com/transla...e-gevoelige-plaat-gevangen.html&sl=auto&tl=en
pictures are from here
http://ic.tweakimg.net/ext/i/imagelarge/1267547764.jpeg
http://ic.tweakimg.net/ext/i/imagelarge/1267547765.jpeg
pictures of the board and cooler
its got all 12 chips its got 512 shaders, or are they able to cut off shader units and not effect the bus width?
http://translate.google.com/transla...e-gevoelige-plaat-gevangen.html&sl=auto&tl=en
pictures are from here
Last edited by a moderator:
Yes but isn't there a new chip hecto something or other that comes out before NI
metal spins on R600 raised clocks by a lot. A11 ran @500, A12 a bit above that and A13 was able to reach the desired clocks. A11 needs the XTX cooler to stay cool @500Mhz.
Tweakers also mentions the reference cooler i "quite pricey"
I am sure silent_guy and aaronspink will disagree.
Me? Gotta study VLSI more before I can say either way.
it can raise clocks if the functional bugs are causing instability of the chips at needed clocks. That's what my understanding of what slient_guy was saying.
Last I checked both ATi and NV tie memory controllers to RBEs/ROP partitions, respectively. TMUs and ALUs are generally separated from ROPs/memory controllers.
For the usual hairsplitting A3 didn't go into production in December. It was sent to the fab in early December.
As far as metal spins concerns yes it would be wiser to ask someone like silent_guy what they can or cannot do.
thatdude90210
Regular
Maybe S/A was too optimistic with their prediction of 8000-10000 cards. Tweaker.net says 5000 worldwide at launch.
Nothing substantial most likely. If they truly have a problem with very few dies reaching full 16SMs operational, then it's very likely that they truly found out after going into production as rumored. And yes they'll obviously go for a re spin for that one.
What if the A3 was a "lets try it if it can help before full respin" and the results were almost same as A2.
Why would they bother blacking out the PCI-e interface and power plugs? Do they have a "Super PCI-e" interface with 3x the checkmarks that is under NDA?
FenderBender
Newcomer
He could just be going on transistor count
I think he is. G200 was the biggest chip by transistor count at the time (65nm, 1.4B transistors at 570 mm² in 2008), but not the biggest in terms of die area (which would probably be the 90nm dual core Itanium 2 at 596 mm² in 2006).
GF100 at 2.9B transistors and unknown size beats the 2.2B HD5800 and also the 8-core Xeon, at 2.3B.
Similar threads
- Deleted member 2197
- Architecture and Products
