Nvidia GT300 yields are under 2%

[ownage]

Has this been posted yet?

GT300 nude shot.



GT300-A1 version
GT300
700 MHz Drawing clock (MHz) 700 MHz
Processor clock (MHz) 1600 MHz * 512 = 819200
281.6GB/s GDDR5 <225W 281.6GB / s GDDR5 <225W

GT200b GT200b
648 MHz Drawing clock (MHz) 648 MHz
Processor clock (MHz) 1476 MHz * 240 = 354240
159.0GB/s GDDR3 <204W 159.0GB / s GDDR3 <204W

http://translate.google.com/translat...late.google.nl

[Tim]

Whoa. Okay that's a surprise if it's GT300.

I've seen that marbe top before in pictures, I believe it was the first real shots of G80. (Just grasping at straws here lol)

[AMDDeathstar]

After yesterdays spanking of its stock price it would have suprise me more if wasn't any leaks.

Good to see some news.

[jaredpace]

Whoa. Okay that's a surprise if it's GT300.

I've seen that marbe top before in pictures, I believe it was the first real shots of G80. (Just grasping at straws here lol)

Maybe you saw it here?

http://www.xtremesystems.org/Forums/...6&postcount=58

[003]

Maybe you saw it here?

http://www.xtremesystems.org/Forums/...6&postcount=58

So is this picture fake or real?

[jaredpace]

its real.... really a picture of a 576mm chip, but not a 452mm chip

[Oberon]

Looks like it's a fake.

[perkam]

Considering how many duds coming out of those GT 300 wafers, I'm surprised we haven't seen any sooner.

I don't know why Nvidia continues to use the Behemoth strategy that landed them in trouble with the GTX 280 over and over again.

Perkam

[003]

its real.... really a picture of a 576mm chip, but not a 452mm chip

Is it the G300 or not?

[perkam]

Is it the G300 or not?

Perkam

[jaredpace]

oh sorry, not it's not a gt300, it's a picture of a gt200 posted on XS a year ago. I figured you would have gotten that by reading the thread and seeing g200 A2 printed on the top of that IHS that was removed from that die... and then noticed that die was the same as the one in the picture....
and then that die photo used in a chiphell post

....

Whatever, no it's not GT300.

[saaya]

Has this been posted yet?

GT300 nude shot.



GT300-A1 version
GT300
700 MHz Drawing clock (MHz) 700 MHz
Processor clock (MHz) 1600 MHz * 512 = 819200
281.6GB/s GDDR5 <225W 281.6GB / s GDDR5 <225W

GT200b GT200b
648 MHz Drawing clock (MHz) 648 MHz
Processor clock (MHz) 1476 MHz * 240 = 354240
159.0GB/s GDDR3 <204W 159.0GB / s GDDR3 <204W

http://translate.google.com/translat...late.google.nl

rofl such an obvious fake...
man chiphell is such a joke, they got great access to ES hardware but keep posting fakes, what gives?

[jaredpace]

man chiphell is such a joke, they got great access to ES hardware but keep posting fakes

+1, They talked non-stop for the last 3 months about RV870 being "native dual-core"....really had me going too.

[grndzro]

Charlie said specifically that 7 chips out of the risk wafers booted. He must have a couple sources that work at TSMC. Otherwise Nvidia would have sued him in a heartbeat. On that note He would have been sued many times over for his articles on Nvidia if they were false.

The fact is he hasn't been sued or armtwisted because Nvidia has nothing on him. He even borrowed an electron microscope to prove the defects with the G80 and G90 series. Sure if you are an Nvidia fan you won't like what he posts.....but please for the love of god show proof when bashing the man.

On the GT300 fiasco. it is a completely new architecture down to the cache. Even making it on 65nm would be terribly difficult given the complexity of it. add to that 40nm process.....yea 7 chips sounds about right.

They were hot wafers, it was the 1st batch, and given the yield it looks like the CEO saw the writing on the wall just looking at the design of the chip and exercised his stock option even b4 TSMC knew anything about GT300.

GT200 yields were awful when the process started and it cost Nvidia a lot of money because the yields practically killed their profit margin. And now U have the GT300...a much more complex chip than the GT200 on an even newer process, This is a huge mistake for Nvidia and it's about time for people to quit the kool aid.

Flame me bash me IDC, when the chip is pushed back to allow a redesign just take a step back and look at Nvidia's actions over the last couple years since the release of the Radeon 4000 series, you just might see Nvidia as I do.....a company that needed a reality check. I don't want Nvidia to crash and burn, I want both companies to have a staggered release scheduele that gives the performance crown to the newest products. it's good for innovation.

[Cybercat]

Charlie said specifically that 7 chips out of the risk wafers booted. He must have a couple sources that work at TSMC. Otherwise Nvidia would have sued him in a heartbeat. On that note He would have been sued many times over for his articles on Nvidia if they were false.

The fact is he hasn't been sued or armtwisted because Nvidia has nothing on him. He even borrowed an electron microscope to prove the defects with the G80 and G90 series. Sure if you are an Nvidia fan you won't like what he posts.....but please for the love of god show proof when bashing the man.

On the GT300 fiasco. it is a completely new architecture down to the cache. Even making it on 65nm would be terribly difficult given the complexity of it. add to that 40nm process.....yea 7 chips sounds about right.

They were hot wafers, it was the 1st batch, and given the yield it looks like the CEO saw the writing on the wall just looking at the design of the chip and exercised his stock option even b4 TSMC knew anything about GT300.

GT200 yields were awful when the process started and it cost Nvidia a lot of money because the yields practically killed their profit margin. And now U have the GT300...a much more complex chip than the GT200 on an even newer process, This is a huge mistake for Nvidia and it's about time for people to quit the kool aid.

Flame me bash me IDC, when the chip is pushed back to allow a redesign just take a step back and look at Nvidia's actions over the last couple years since the release of the Radeon 4000 series, you just might see Nvidia as I do.....a company that needed a reality check. I don't want Nvidia to crash and burn, I want both companies to have a staggered release scheduele that gives the performance crown to the newest products. it's good for innovation.

What's Charlie's proof? Do detractors of his articles need proof that he clearly has an agenda of his own? All you need to do is look through the history of his news articles. The man has nothing good to say about NVIDIA.

Just because it sounds likely that NVIDIA might be having problems with their new design doesn't make it so. A new architecture isn't enough to give them issues, NV40, G80, many new architectures released without a hitch (irrespective of die size). If there's a problem it's with the process. Regardless, do we have any reason to think it will affect the launch late this year? All I'm hearing from critics is conjecture.

And from I've seen there was never a major yield issue with the GT200. It was just a lot more expensive to make than its competition, but hey, at least it was faster.

[LordEC911]

What's Charlie's proof? Do detractors of his articles need proof that he clearly has an agenda of his own? All you need to do is look through the history of his news articles. The man has nothing good to say about NVIDIA.

Just because it sounds likely that NVIDIA might be having problems with their new design doesn't make it so. A new architecture isn't enough to give them issues, NV40, G80, many new architectures released without a hitch (irrespective of die size). If there's a problem it's with the process. Regardless, do we have any reason to think it will affect the launch late this year? All I'm hearing from critics is conjecture.

And from I've seen there was never a major yield issue with the GT200. It was just a lot more expensive to make than its competition, but hey, at least it was faster.

The fact that Charlie is NOT the source is why some of us are so sure of this news. I heard it from an informant/source on another site who gave me the correct codenames to Evergreen back at the beginning of August, which is when all the sites were mixing up the codenames. He has also had very good info in the past and is the one that heard G300 is larger than G200 and he even knew what information Charlie was going to post a day before he posted it.

Yields are the main factor in the cost per die. G200 did have some problems with yields, which is expected when you rely on salavaging, G200b had even more problems needing an extra respin.

New architecture on a new process with an ungodly large die size is more than enough to correlate with the supposed yields.

[Luka_Aveiro]

Charlie Demerjian "news" are just as real and true as 3 dollar bills (limp bizkit first 2 albums rule)

[AffenJack]

He has also had very good info in the past and is the one that heard G300 is larger than G200 and he even knew what information Charlie was going to post a day before he posted it.

So your Sourve is believing G300 is larger than G200 or did you mean G200b?
I'm sure it's not larger than G200. It's maybe in the middle of both, maybe a bit smaller than G200b.

[Cybercat]

The fact that Charlie is NOT the source is why some of us are so sure of this news. I heard it from an informant/source on another site who gave me the correct codenames to Evergreen back at the beginning of August, which is when all the sites were mixing up the codenames. He has also had very good info in the past and is the one that heard G300 is larger than G200 and he even knew what information Charlie was going to post a day before he posted it.

Yields are the main factor in the cost per die. G200 did have some problems with yields, which is expected when you rely on salavaging, G200b had even more problems needing an extra respin.

New architecture on a new process with an ungodly large die size is more than enough to correlate with the supposed yields.

And there are plenty of other times when he was wrong...or his source was wrong, or whatever. He's biased, it's obvious, and that's really all there is to it. The only conclusion one can come to about him or what he writes is that you need even more salt than usual with what you read of his.

But some people are coming to other conclusions, like the GT300 being a disaster, the next FX series, NVIDIA's HD 2900, and all this nonsense. People calling large dies a huge mistake are forgetting that those large dies, more often than not, accomplished the goals they set out to. How about people take a "wait-and-see" approach instead, rather than calling Charlie's information fact?

[Calmatory]

And there are plenty of other times when he was wrong...or his source was wrong, or whatever. He's biased, it's obvious, and that's really all there is to it. The only conclusion one can come to about him or what he writes is that you need even more salt than usual with what you read of his.

But some people are coming to other conclusions, like the GT300 being a disaster, the next FX series, NVIDIA's HD 2900, and all this nonsense. People calling large dies a huge mistake are forgetting that those large dies, more often than not, accomplished the goals they set out to. How about people take a "wait-and-see" approach instead, rather than calling Charlie's information fact?

Charlie has nothing to do with the rumour. He is not the source.

OTOH I suggest people to take "wait-and-see" approach instead, rather than calling this rumour bogus.

[Cybercat]

Charlie has nothing to do with the rumour. He is not the source.

Hence why I said:

And there are plenty of other times when he was wrong...or his source was wrong, or whatever.

[003]

So your Sourve is believing G300 is larger than G200 or did you mean G200b?
I'm sure it's not larger than G200. It's maybe in the middle of both, maybe a bit smaller than G200b.

G300 is reported to be in between GT200 and GT200b.

BTW,
There seems to be a lot of unspoken confusion being tossed around with the proper spelling of G300. It was leaked a while back that nvidia was dropping the T from the codename, because T stands for Tesla and apparently they have something slightly different in mind for Tesla this time around.

So the correct codename is G300, not GT300. GT200 is still correct for the GTX280.

[LordEC911]

Well, I'm just putting it out there that the info comes from a very good source that has been very accurate in the past and that it is NOT from Charlie. Take it at face value but don't discount it just because Charlie heard the same info and misinterpreted some of it, though the general gist of his article is spot on.

Edit- And yes, the most recent info on G300 from that source is that it is "as big as G200", 65nm.

[Chumbucket843]

Well, I'm just putting it out there that the info comes from a very good source that has been very accurate in the past and that it is NOT from Charlie. Take it at face value but don't discount it just because Charlie heard the same info and misinterpreted some of it, though the general gist of his article is spot on.

Edit- And yes, the most recent info on G300 from that source is that it is "as big as G200", 65nm.

well if you say charlie is right then you should believe there are 104 wafers per die which would be smaller than gt200a. i dont see how he can misinterpret a number.

[Russian]

Charlie has nothing to do with the rumour. He is not the source.

OTOH I suggest people to take "wait-and-see" approach instead, rather than calling this rumour bogus.

Yes, but he is perpetuating the rumor. no? giving it any sort of validity merely validates the crap that he posts on a daily basis.