Nvidia GT300 yields are under 2%

[ahmad]

Well I am not surprised. Nvidia was the one saying "we will produce monolithic unbelievably huge and fast chips because AMD's low cost smaller GPUs are a waste of time". Maybe its time to reconsider.



Frankly this is the only piece of info that we have so for now it will have to suffice as debate material.

Getting a bit tired of this, the 'I know more but I'm not telling'. No offense to you buddy but c'mon, what possible use is there for posts like that with no substance except for, 'So... not true'.

QFT

Getting quite tired of that as well. As far as I can tell BenchZowner is just an nvidia fanboy and a troll. No disrespect to you BenchZowner but trying to prove to others that you have more information than everyone else without producing anything doesn't help your credibility.

[zerazax]

When Jen Hsun opened his Can of Whoop Ass he apparently spilled it in his lap.

Definitely lol'd

[570091D]

Hey guys, heads up, I got a preview look at Charlie Demerjian's next article:

this

i find this hard to believe considering that 3 months ago we were hearing about gt300 yeilds in the 20-25% range. i think charlie needs a girlfriend.....

[003]

i think charlie needs a girlfriend.....

actually, more like a boyfriend! hell, just a friend!

[K404]

300mm wafers.... 416 samples.....

Which genius fancies figuring out the die-size?

Assuming that the sample count is right......

[zerazax]

300mm wafers.... 416 samples.....

Which genius fancies figuring out the die-size?

Assuming that the sample count is right......

Wafers being circular makes it a bit hard

[highoctane]

300mm wafers.... 416 samples.....

Which genius fancies figuring out the die-size?

Assuming that the sample count is right......

Yeah sample number seems a bit high!!

[LordEC911]

Ummm... lots of mis-info and assumptions being made in this thread.

#1- Charlie is NOT the source. A different source leaked some info and a very good informant even posted up the numbers Charlie was going to quote yesterday. This informant quoted similar troubles but slightly higher yields, still under 10%.

#2- G300 never was getting 20-25% yields, that was a compeltely made up rumor based on RV740 yields.
G300 was either just taped out or not even tapedout 3 months ago...

#3- G300 is supposedly around G200 size, even though it is on 40nm vs 65nm.

[K404]

Wafers being circular makes it a bit hard

I know thats why I didnt have a shot myself

There must be a formula for stuff like that seeing as its industry-standard

[AMDDeathstar]

300mm wafers.... 416 samples.....

Which genius fancies figuring out the die-size?

Assuming that the sample count is right......

It's 104 per wafer
rumors say it 23x23 529 mm2
If my rough math is right that would be around 3.2 billion transitors

[K404]

104 is MILES off (welll....x4) the 416 in the original post.

If you feel inclined, can ya work out the rough transistor count or die size for 416 samples?

We can say "this is probably BS" with maths and science instead of just "Its BS coz its FUD"

[LowRun]

Getting a bit tired of this, the 'I know more but I'm not telling'. No offense to you buddy but c'mon, what possible use is there for posts like that with no substance except for, 'So... not true'.

QFT, especially when the guy has been so wrong in the past

[Serra]

I really hope this is not true, because if it is... wow. Boatload of fail.

[Mechromancer]

Charlie: Nvidia's GT300 yields are less than 2%



Nvidia Fanboy: No, it’s not true! It’s impossible.



Charlie: Search your feelings… you know it to be true…



Nvidia Fanboy: NO!

Charlie: Even if GT300 yields are 100%, R8XX will blow it out of the water (Says the AMD fanboy)

I'm making light of fanboyism. Fanboys get in denial mode and act like nothing is wrong. Makes me laugh.

[Sadasius]

^^^ LMFAO

[safan80]

Demerjian is the Uwe Boll of tech reporters.

[Jamesrt2004]

I believe theres some truth to this.. maybe not the stupid less then 2% but i can believe that they weren't what they where expecting, and now it's been twisted and turned into this..... :/

[Andrew LB]

Can you point out any examples where he is wrong? He was right about the number of shaders in Cypress, so can you provide 19 articles where he is wrong, to prove your correct percentage?

I've been reading Charlie's stuff for about a decade now and while he has made some very good predictions, the fact remains that he's one of those guys who typically throws out a whole wide range of predictions in order to hedge his bet.

He predicted on Tuesday, 5 May 2009 that RV870 would have 1200 stream processors. In November, 2008 he predicted 2000, and I've also seen him throw out the 960sp number as well for the chip.

Anyone who throws out three nicely spaced out numbers is bound to have one of them land pretty close to the mark.


What really gives me a laugh is how we revisit this same topic every single product cycle, and more importantly how Charlie is still trying to convince the online community how "accurate" he always is.

Oh well, at least the guy seems happy.

[cegras]

From what I've read, Charlie's articles usually tend to be right.

[zerazax]

104 is MILES off (welll....x4) the 416 in the original post.

If you feel inclined, can ya work out the rough transistor count or die size for 416 samples?

We can say "this is probably BS" with maths and science instead of just "Its BS coz its FUD"

There were 4 wafers.. hence 4 x 104 = 416

[Chumbucket843]

From what I've read, Charlie's articles usually tend to be right.

charlie has a profile on XS?

[Andrew LB]

From what I've read, Charlie's articles usually tend to be right.

If you google "Charlie Demerjian RV870 (R800, RV8xx, etc) shaders (stream processors, SP's, etc)" you'll find that he has claimed 1200, 960, 2000, 1400, and 1600 over the past year.

He did the same with every GPU/CPU since I got into PC tech over a decade ago.

Here is one incident where Charlie's little 'shell game' didn't quite work out for him.

http://www.theinquirer.net/inquirer/...l-ceo-confirms

Yes... that is Charlie in the bunny suit.

[Newblar]

Consider the source though.

semiaccurate.com .. semi accurate?

[Katanai]

Demerjian is the Uwe Boll of tech reporters.

Couldn't have said it better myself...

[SocketMan]

He predicted on Tuesday, 5 May 2009 that RV870 would have 1200 stream processors.

He called it "RV870" ?
Can you put a link for that article please,I can only find this one here:

http://www.semiaccurate.com/2009/09/...-shader-count/


" It is 1600 for Cypress, 800 for Juniper, going down from there."