Something everyone missed about the GTX460....

[Sokar]

http://www.pcper.com/article.php?aid=954

It has come to my attention that apparently these chips are A1 silicon... How often does something of this level of complexity ship on A1 silicon? Especially since NV is not particularly known for their prowess with this node as seen with GF100....

And then it hit me. Why no "full part"? Why the super conservative clock speeds? Why go the extra mile to bin every single chip individually to get the absolute lowest voltage possible on every chip when really thermals are not a problem with that card so you could save the money and just set them all at the high voltage anyways.

Really i'm wondering how long it takes to test and validate a chip like this because I doubt they had time to do it... So they just tuned everything down and hope that these things hold up until they have proper testing and a respin with good enough yields.

 

[Voo]

You know, sometimes the only thing one can do is to admire the creativity of some people *gets some popcorn*

Nice troll.. but I don't think we'll get more than 5pages, just not the best topic.. maybe add a few singled out benchmarks and some out of context quotes?

 

[Shilohen]

nVidia starts with A0 I think, so this should be the 2nd spin, no? I may be wrong of course, I easily mix up those.

I really don't mind them however, I have two EVGA coming in my way and I'll most likely use the step up option if a better, fully enabled, chip gets released.

 

[Idontcare]

You know, sometimes the only thing one can do is to admire the creativity of some people *gets some popcorn*

Voo we are so on the same wavelength here, I read the OP and lol'ed with exactly your sentiment...

 

[happy medium]

Voo we are so on the same wavelength here, I read the OP and lol'ed with exactly your sentiment...

Well let me in on what your thinking, I don't get it. :hmm: Pm or whatever.

On topic:
I said this at launch but most people didn't comment. I thought A1 silicon was the so called "hot lot" and usually not ready for production.

 

[Grooveriding]

http://www.pcper.com/article.php?aid=954

It has come to my attention that apparently these chips are A1 silicon... How often does something of this level of complexity ship on A1 silicon? Especially since NV is not particularly known for their prowess with this node as seen with GF100....

And then it hit me. Why no "full part"? Why the super conservative clock speeds? Why go the extra mile to bin every single chip individually to get the absolute lowest voltage possible on every chip when really thermals are not a problem with that card so you could save the money and just set them all at the high voltage anyways.

Really i'm wondering how long it takes to test and validate a chip like this because I doubt they had time to do it... So they just tuned everything down and hope that these things hold up until they have proper testing and a respin with good enough yields.

Thanks for joining the forum to start us off with this gem

http://forums.anandtech.com/showthread.php?t=2096434&highlight=

and now this excellent topic sure to infuse the rabid fanboys with tales of gtx 460s soon to come fully unlocked and supercharged and faster than gtx 580 sli with a die tweak.

 

[Leyawiin]

/snip

Charlie? AMD employee? Major stockholder?

 

[tviceman]

Thanks for joining the forum to start us off with this gem

http://forums.anandtech.com/showthread.php?t=2096434&highlight=

and now this excellent topic sure to infuse the rabid fanboys with tales of gtx 460s soon to come fully unlocked and supercharged and faster than gtx 580 sli with a die tweak.

LOL!

In reality though, if a 384 shader @ 750-775 mhz is the reference design (when released) it'll likely be faster than a 5870, which will mean Nvidia will have either tied or recaptured the performance per square mm^2 space that ATI fans have been bragging about for a long, long time.

 

[bryanW1995]

3870 was a1 silicon. amd made a very big deal out of it at the time b/c they were at the time getting their asses smoked. looks like both sides can pull a rabit out of the hat every now and then.

 

[Idontcare]

Well let me in on what your thinking, I don't get it. :hmm: Pm or whatever.

On topic:
I said this at launch but most people didn't comment. I thought A1 silicon was the so called "hot lot" and usually not ready for production.

"hot lot" is a manufacturing procedure designed to reduce the cycle-time for a batch of wafers (called a "lot" in fab-speak)...has nothing to do with the stepping, although it is common for the first lot on a new stepping to be put on "hot lot" priority so the design guys get their info back asap regarding whether the stepping is good to go.

A1 stepping just means it is the first respin...whatever came back from the fab with the A0 stepping was deemed to be not production-worthy and thus a respin to stepping A1 was performed. (and probably hot-lotted for the first few lots as well)

 

[Rifter]

forgive my lack of knowlege on this subjet but why does this even matter at all? I would assume most new products/chips go through a few minor redesigns before getting to the retail shelves do they not?

 

[bryanW1995]

now that I think about it, the amd guy's comment on 3870 was that they were using "first silicon". don't know if that was a1 or a0. funny, that card also took a slow/hot card and trimmed it down while giving similar performance. too bad tsmc is holding both sides back, I'll be interested to see what happens when amd switches to gf.

 

[ronnn]

I think he means nvidia is at the start of a very nice run of ever improving chips. Not sure of his reasoning, but I would think the 460 is the start of something good.

 

[GaiaHunter]

LOL!

In reality though, if a 384 shader @ 750-775 mhz is the reference design (when released) it'll likely be faster than a 5870, which will mean Nvidia will have either tied or recaptured the performance per square mm^2 space that ATI fans have been bragging about for a long, long time.

1 year after NVIDIA might tie in performance per mm^2.

Cool.

If the rumors hold any truth, NVIDIA achieved that just in time to face SI.

 

[lsv]

You know, sometimes the only thing one can do is to admire the creativity of some people *gets some popcorn*

Nice troll.. but I don't think we'll get more than 5pages, just not the best topic.. maybe add a few singled out benchmarks and some out of context quotes?

Hey VoO is that you?

 

[tviceman]

1 year after NVIDIA might tie in performance per mm^2.

Cool.

If the rumors hold any truth, NVIDIA achieved that just in time to face SI.

Actually it's been way longer than 1 year. Ever since the gtx200 series vs. the 4800 series, Nvidia has had significantly larger dies and only slightly better performance.