NVIDIA says Fermi yields are fine, promises 512-core version

Quote:

---

DigiTimes had a chat with NVIDIA general manager of MCP business Drew Henry. When speaking about Fermi, Henry said rumors about yields being lower than 20 percent are completely untrue. He claims they currently have everything under control, and says they have a chance to launch a 512-core Fermi in the future.

Q: What are your comments on the rumors that Nvidia has blocked some graphics cores with problems on their GeForce 480/470 chips causing these chips to have less than 512 cores due to Taiwan Semiconductor Manufacturing Company's (TSMC's) low yields?

A: Nvidia does not comment on unannounced products; however, we have a chance to launch a graphics chip with 512 cores in the future.

TSMC's yields for its 40nm process has met our expectations and market rumors about the yields being lower than 20% are completely untrue. We currently have everything under control.

---

DigiTimes
via DVH

512 core versions? Wonder what their power draw is going to be... This is good news though. Both the good yields and a 512 core version coming will mean a tougher competition for ATI, which may force the prices down, which is exactly what we as the consumers want :)

If I already had a 480 I'd be kinda upset with this news of a 512 core version when that was supposed to be the original spec.

Quote:

---

we have a chance to launch a graphics chip with 512 cores in the future.

---

Translation: Not soon.

Quote:

---

market rumors about the yields being lower than 20% are completely untrue

---

Latest rumor was that yields are between 20% and 30%.
But they won't dispute that, it seems...

I'm of a mind to believe him.

Newegg just got a whole bunch of GTX 470s in stock, though the GTX 480s are still a bit scarce..

Things certainly look to be improving though.

So expect a triple slot cooler for 512core version?

Quote:

---

Originally Posted by Carfax

I'm of a mind to believe him.

Newegg just got a whole bunch of GTX 470s in stock, though the GTX 480s are still a bit scarce..

Things certainly look to be improving though.

---

I was thinking exactly the same, right now i can pick 6 different brands of 470GTX from 4 different stores starting from 350$ but i can only find single 480GTX and for 510$ , the 470 yields looks promising and it is only 40$ more than the cheapest HD5850 at the egg

Quote:

---

Originally Posted by Face

DigiTimes
via DVH

---

Uhmm... no... If yields were fine GTX 480 would have 512 cores already. :rolleyes:

Quote:

---

Originally Posted by kemo

I was thinking exactly the same, right now i can pick 6 different brands of 470GTX from 4 different stores starting from 350$ but i can only find single 480GTX and for 510$ , the 470 yields looks promising and it is only 40$ more than the cheapest HD5850 at the egg

---

Or... very few GPUs bin to 480 specs so they sell them as 470s. ;)

If the yields are fine then that's good. I'll just take his words at face value ( >=20% ) and wait until a long term pattern of availability emerges. A 512 card would be nice.

I'd also like to see them fill out the rest of the product range with fermi products. Not just a dual card to compete with 5970 but midrange and lowend products too. Competition across the range would be ideal, for me as a consumer. :D

The architecture itself is designed with 512 cores. Saying there is a chance for a 512 core version is like saying the sky is blue...

I'll believe an NV PR drone just slightly less than I'll believe a biased reporter like Charlie. Charlie has been known to spew bull:banana::banana::banana::banana:, but Nvidia PR has no problems with outright lies (such as the reason why PhysX won't work in mixed GPU systems: there is 'intricate communications' between the PhysX card and the Gfx card, despite the fact that any problem running mixed GPU PhysX has been shown to be artificial.)

Quote:

---

Originally Posted by kemo

I was thinking exactly the same, right now i can pick 6 different brands of 470GTX from 4 different stores starting from 350$ but i can only find single 480GTX and for 510$ , the 470 yields looks promising and it is only 40$ more than the cheapest HD5850 at the egg

---

TD has had 480s in stock for like the last weeks straight....at least they're reported to be in stock at the site :shrug:

I just ordered one....I hope they're stock level is accurate :rolleyes:

they say they have a chance of releasing one. it doesnt mean they will.

this was 2 questions we would all like to know, with answers that mean almost nothing. lets move along now

I don't get it. What would they be lying about? He said "rumors about the yields being lower than 20% are completely untrue"

That could mean yields are currently at 21% or 101% (sarcasm on that last point)....

Oh, come on. Which idiot thinks there will never be a 512 core Fermi? The GF100 architecture was designed for 512 cores. It will eventually have 512 cores. Or do you guys think Nvidia is going to abandon the Fermi architecture altogether because Charlie "told you so" that it was "unmanufacturable"?

I'm assuming that by above 20% yields they mean above 20% are usable, not above 20% can use all 512 cores.

I love it how people keep blowing heat issues out of their asses. Unlocking all of the die won't do a thing to the cooling needed if yields are improving and they can do a better job cherry picking.

wasnt the original yield number supposed to be based on higher clocks, the cores from the 480 or better and being lower voltage, so that 20% could well be true but they lowered the cores, lowered the clocks and upped the voltage so if u do any of those it would change yields.

not trying to be combative but when u change what u will use it greatly changes what your yields are.

I recall seeing something similar with the 7900GTX, that is all...:p:

No, no, the yields are NOT under 20%. They're under 21%. GET IT RIGHT! :p:

Quote:

---

Originally Posted by annihilat0r

Oh, come on. Which idiot thinks there will never be a 512 core Fermi? The GF100 architecture was designed for 512 cores. It will eventually have 512 cores. Or do you guys think Nvidia is going to abandon the Fermi architecture altogether because Charlie "told you so" that it was "unmanufacturable"?

---

No Nv will probably release a couple of token cards...

Quote:

---

Originally Posted by zanzabar

wasnt the original yield number supposed to be based on higher clocks, the cores from the 480 or better and being lower voltage, so that 20% could well be true but they lowered the cores, lowered the clocks and upped the voltage so if u do any of those it would change yields.

not trying to be combative but when u change what u will use it greatly changes what your yields are.

---

Bingo

Quote:

---

Originally Posted by stinkas

No Nv will probably release a couple of token cards...

---

More then likely samples for reviews. That's part of the reason why I mentioned the 7900 GTX. Or can we include the 8800 Ultra?

Quote:

---

Originally Posted by Eastcoasthandle

More then likely samples for reviews. That's part of the reason why I mentioned the 7900 GTX. Or can we include the 8800 Ultra?

---

there have been several cards from both ATi and nvidia that were the fastest of their time but virtually impossible to buy. the 512sp version would only offer ~6% more performance over a 480 at the same clocks. that's something to keep in mind. clockspeeds and power consumption are more important. these are the areas that can and need the most improvement.

wernt clockspeeds pretty good? faster than other nvidia cards

and if you take an 5870 and max it out at stock volts, and do the same with a 480, i think the 480 has alot more potential, and can help reduce alot of that perf/watt suckage people complain about.

woot for one click, double posting, lol

Quote:

---

Originally Posted by Chumbucket843

there have been several cards from both ATi and nvidia that were the fastest of their time but virtually impossible to buy. the 512sp version would only offer ~6% more performance over a 480 at the same clocks. that's something to keep in mind. clockspeeds and power consumption are more important. these are the areas that can and need the most improvement.

---

Hmm, no they weren't. Not in the context that you are replying to my post. The history regarding the 7900 GTX and 8800 Ultra is apparent.

Quote:

---

Originally Posted by Manicdan

wernt clockspeeds pretty good? faster than other nvidia cards

and if you take an 5870 and max it out at stock volts, and do the same with a 480, i think the 480 has alot more potential, and can help reduce alot of that perf/watt suckage people complain about.

---

But does it need to be on water?
Water is a pain in the @ss and expensive...

Drew Henry just tales truisms (Down with SKYMTL and annihilat0r) ... 'better said 768 CC on 28 nm ...

Nice to see Face again on XS ;)

Quote:

---

Originally Posted by SKYMTL

The architecture itself is designed with 512 cores. Saying there is a chance for a 512 core version is like saying the sky is blue...

---

Quote:

---

Originally Posted by zanzabar

wasnt the original yield number supposed to be based on higher clocks, the cores from the 480 or better and being lower voltage, so that 20% could well be true but they lowered the cores, lowered the clocks and upped the voltage so if u do any of those it would change yields.

not trying to be combative but when u change what u will use it greatly changes what your yields are.

---

Quote:

---

Originally Posted by Eastcoasthandle

More then likely samples for reviews. That's part of the reason why I mentioned the 7900 GTX. Or can we include the 8800 Ultra?

---

^^This.

Quote:

---

Originally Posted by Dainas

I love it how people keep blowing heat issues out of their asses. Unlocking all of the die won't do a thing to the cooling needed if yields are improving and they can do a better job cherry picking.

---

in order to improve yield they would need to tweak the process not wait out of luck.... + enabling the disabled core would raise the power draw.... duh.....

Quote:

---

Originally Posted by Sn0wm@n

in order to improve yield they would need to tweak the process not wait out of luck....

---

What? No. A process will increase in yields "by itself" with time.

Yields so great they couldn't release the 512 version right off the bat!!!

Right Nvidia... right.... :yepp:

Quote:

---

Originally Posted by Sly Fox

Yields so great they couldn't release the 512 version right off the bat!!!

Right Nvidia... right.... :yepp:

---

Yeah just like how fast the 470 and 480 came out.

I am sure nvidia is on top of their shat

Quote:

---

Originally Posted by annihilat0r

What? No. A process will increase in yields "by itself" with time.

---

its both but its hard to say which is more important. tsmc gives clients the tools to design a chip and then they manufacture it. being fabless doesnt mean that the process side is a free lunch either.

vias are being problematic. apparently ATi found a way around it.
http://www.semiconductor.net/article...ia_Defects.php

Quote:

---

Originally Posted by HuffPCair

Yeah just like how fast the 470 and 480 came out.

I am sure nvidia is on top of their shat

---

I think the problem with 512 at this point is not only yields but additionally power requirements.

I.E. The 512 although using only 6 shaders, will use significanly more power than a 480(just look at the difference between the 470 and 480), additionally to justify the another product it will need higher clocks than the 480 itself, as a 6 percent difference hardly needs a new product.

512 shaders + 750mhz clock with the current process means a card that needs a waterblock and 2 8 pin power.

Quote:

---

Originally Posted by annihilat0r

What? No. A process will increase in yields "by itself" with time.

---

the yield are due in part to complexity + if the process helps... so its a yes

Quote:

---

Originally Posted by tajoh111

I think the problem with 512 at this point is not only yields but additionally power requirements.

I.E. The 512 although using only 6 shaders, will use significanly more power than a 480(just look at the difference between the 470 and 480), additionally to justify the another product it will need higher clocks than the 480 itself, as a 6 percent difference hardly needs a new product.

512 shaders + 750mhz clock with the current process means a card that needs a waterblock and 2 8 pin power.

---

Yeah I agree. Am waiting to see how much BS they throw out in a few months.

Quote:

---

Originally Posted by Sly Fox

Yields so great they couldn't release the 512 version right off the bat!!!

Right Nvidia... right.... :yepp:

---

Dont forget they need to save something for the refresh.

Im shure the yields werent good enough for 512 at launch. But another reason for not launching a 512 version could be the refresh that at some point wil come.

Quote:

---

Originally Posted by annihilat0r

What? No. A process will increase in yields "by itself" with time.

---

I really don't think so.

If TSMC works to improve yields, yields will improve. Or if Nvidia tweaks the chip it might get better yields.

AMD improved yields by doubling the number of vias, and making some other tweaks to the design of Cypress (and of course the smaller die also naturally gets higher yields).

I don't see how it would magically improve by itself.

Is nVidia going to do a B1 respin? Kyle from [H] does not seem to think so, at least not anytime soon and I believe Charlie said the same thing. So, B1 respin in time for Christmas 2010 season :D

Just sticking to nVidias own schedule for Fermi here. Remember Fermi was supposed to be here "very soon" back in September 2009. "Very soon" in nVidia lingo equates out to 6-8 months in real world lingo :rofl:

Does anyone even know what this 'yield' number represents? Is it the percentage of chips that can be used as 480/470, or the percentage that can be used for 480/470/460?

Quote:

---

Originally Posted by Zonker

Does anyone even know what this 'yield' number represents? Is it the percentage of chips that can be used as 480/470, or the percentage that can be used for 480/470/460?

---

that's the thing, they could easily shift the goalposts just to say they have higher yields.

Quote:

---

Originally Posted by LiquidReactor

Is nVidia going to do a B1 respin? Kyle from [H] does not seem to think so, at least not anytime soon and I believe Charlie said the same thing. So, B1 respin in time for Christmas 2010 season :D

Just sticking to nVidias own schedule for Fermi here. Remember Fermi was supposed to be here "very soon" back in September 2009. "Very soon" in nVidia lingo equates out to 6-8 months in real world lingo :rofl:

---

that would take 6 months... if they stayed on schedule. its best to spend more time on what they will be doing on 28nm.

Quote:

---

Originally Posted by Sn0wm@n

in order to improve yield they would need to tweak the process not wait out of luck.... + enabling the disabled core would raise the power draw.... duh.....

---

duh yourself

The thing is even ignoring tweaking refinements, not all cores are the same, even on the same wafer there can be big differences in TDP. Some dead, barely usable, some good, some very good and some excellent. Nvidia had the 480 configured so they could get numbers were not phantoms cards, this full 512 version looks like it could be a phantom. Creme of the creme these are the ultra rare perfect examples.

Quote:

---

Originally Posted by annihilat0r

Oh, come on. Which idiot thinks there will never be a 512 core Fermi? The GF100 architecture was designed for 512 cores. It will eventually have 512 cores. Or do you guys think Nvidia is going to abandon the Fermi architecture altogether because Charlie "told you so" that it was "unmanufacturable"?

---

To improve yields, you don't only need a new metal layer scheme, but a complete new scheme. Not just going from A4 to A5, but from A4 to B1, this costs a lot of time. So in my opinion, Nvidia better starts designing the upgrade from Fermi, because fixing it takes too much time.

Quote:

---

Originally Posted by Apokalipse

I really don't think so.

If TSMC works to improve yields, yields will improve. Or if Nvidia tweaks the chip it might get better yields.

AMD improved yields by doubling the number of vias, and making some other tweaks to the design of Cypress (and of course the smaller die also naturally gets higher yields).

I don't see how it would magically improve by itself.

---

As the process matures, small tweaks are possible that will slightly increase yields, not much but a little.

Quote:

---

Originally Posted by LiquidReactor

Is nVidia going to do a B1 respin? Kyle from [H] does not seem to think so, at least not anytime soon and I believe Charlie said the same thing. So, B1 respin in time for Christmas 2010 season :D

Just sticking to nVidias own schedule for Fermi here. Remember Fermi was supposed to be here "very soon" back in September 2009. "Very soon" in nVidia lingo equates out to 6-8 months in real world lingo :rofl:

---

Bx silicon seemed to be likely but that rumor changed from possibly seeing it in Q3, to Fermi II in Q4.

Quote:

---

Originally Posted by Zonker

Does anyone even know what this 'yield' number represents? Is it the percentage of chips that can be used as 480/470, or the percentage that can be used for 480/470/460?

---

Yield, in this context, means total number of usable dice.
Those are then split into different bins, GTX480/470/460 depending on clocks/voltage/TDP/defects.

Quote:

---

Originally Posted by LordEC911

As the process matures, small tweaks are possible that will slightly increase yields, not much but a little.

---

In other words, it doesn't happen by itself. It happens because tweaks are made as they use it and learn more about it.

Quote:

---

Originally Posted by Zonker

Does anyone even know what this 'yield' number represents? Is it the percentage of chips that can be used as 480/470, or the percentage that can be used for 480/470/460?

---

i think its the number of salvageable chips....

Quote:

---

Originally Posted by Dainas

duh yourself

The thing is even ignoring tweaking refinements, not all cores are the same, even on the same wafer there can be big differences in TDP. Some dead, barely usable, some good, some very good and some excellent. Nvidia had the 480 configured so they could get numbers were not phantoms cards, this full 512 version looks like it could be a phantom. Creme of the creme these are the ultra rare perfect examples.

---

it all depend on how good the process and the chip was in the begining...


like say if fermi didnt have room for failure included in the design and the transistor problem at tsmc arise how can they fix it right in such a short time????


there are loads of variable to look at before even thinking your comment is correct.... not all dies are the same... sure... but then again if the chip is made to have some sort of bufffer in case something goes wrong on the production side then they can improve it over time... exactly what ati did ....

Quote:

---

Originally Posted by SKYMTL

The architecture itself is designed with 512 cores. Saying there is a chance for a 512 core version is like saying the sky is blue...

---

exactly... pure marketing speak, the art of saying something without saying anything at all :D

Quote:

---

A: Nvidia does not comment on unannounced products; however, we have a chance to launch a graphics chip with 512 cores in the future.

---

so IF there would be an upcoming product with 512cores, then he wouldnt comment on it... as nvidia doesnt comment on unannounced products... so he basically said they do NOT plan a 512core card at this point in time... which DOES confirm that yields are not exactly great... they wouldnt need a ton of 512core cards, this would be an ultra which is very limited in numbers... but they dont even have those few thousand gpus it seems...

Quote:

---

TSMC's yields for its 40nm process has met our expectations and market rumors about the yields being lower than 20% are completely untrue.

---

if it has met nvidias expectations how come they continously ripped on tsmc in the past and pushed them to admit issues with their 40nm process and talk about why their yields have problems and apologize to nvidia? repeatedly!
who said yields are lower than 20%? the last i heard was 20-40%, so he pretty much confirmed that yields are 20% or a bit higher... which isnt exactly great by any means... its acceptable for the price points they target, but not good... if they would depend on gf100 alone they would be in serious trouble.

pretty amateurish statements imo, he unconsciously and inententionally confirmed what he tried to deny :lol:
1. no plans for a 512core card anytime soon
2. yields are around 30% = not very good

i wonder how they define yields though... since the 460 with 384sps will be based on gf100, that means they will be able to use even more chips that they would have had to throw away as garbage otherwise... 384sps means 4 disabled blocks, plus 3 disabled memory controllers... and a much higher vcore and tdp range per sps block... so their combined yields for 480, 470 and 460 should be way higher than those of the 480 and 470 alone...

so maybe thats where the 20-40% yield rumors come from?
20% for 480+470 and 40% for 480+470+460? that would make sense and explain why digitimes talked about such a wide range of 20-40 instead of saying ~30%...

Quote:

---

Originally Posted by Bobsama

No, no, the yields are NOT under 20%. They're under 21%. GET IT RIGHT! :p:

---

Yeah, it's funny how they haven't denied that, isn't it?

Quote:

---

Originally Posted by Sly Fox

Yields so great they couldn't release the 512 version right off the bat!!!

Right Nvidia... right.... :yepp:

---

It's Nvidia, what did you expect? :p:

Quote:

---

Originally Posted by saaya

i wonder how they define yields though... since the 460 with 384sps will be based on gf100, that means they will be able to use even more chips that they would have had to throw away as garbage otherwise... 384sps means 4 disabled blocks, plus 3 disabled memory controllers... and a much higher vcore and tdp range per sps block... so their combined yields for 480, 470 and 460 should be way higher than those of the 480 and 470 alone...

so maybe thats where the 20-40% yield rumors come from?
20% for 480+470 and 40% for 480+470+460? that would make sense and explain why digitimes talked about such a wide range of 20-40 instead of saying ~30%...

---

Yeah, that makes sense if they're talking about GF100 in particular.

It was a poorly worded question really. It would interest me if the following questions were answered:

"Nvidia, what percentage of fermi chips coming out of the tsmc fab can be binned as 470 and what percentage can be binned as 480?"
"Also, are you at this moment holding back any top binned silicon that will be release as geforce 512?"

But I guess such questions would never be answered through official channels. Lucky we have charlie:)

Quote:

---

Originally Posted by Zonker

It was a poorly worded question really. It would interest me if the following questions were answered:

"Nvidia, what percentage of fermi chips coming out of the tsmc fab can be binned as 470 and what percentage can be binned as 480?"
"Also, are you at this moment holding back any top binned silicon that will be release as geforce 512?"

But I guess such questions would never be answered through official channels. Lucky we have charlie:)

---

If they did that, AMD would know pretty much exactly what to drop their prices to so that Nvidia doesn't make a profit off the cards, assuming AMD has headroom in their pricing.

I just got my gtx 480 today, supposedly the 2nd or 3rd batch's that etailers are getting. (bought it April 23rd) It's a EVGA model, and the serial number ends 000000314

kind of makes you wonder just how many cards they had at launch.

how about this for a good analogy of the chance nvida makes the 512 ultra within 4 months of coming out with fermi.....



i would have a better chance of having a 3some with these 2 :D

Man, those TWO look really great, can i join ? :p:

Quote:

---

Originally Posted by LordEC911

As the process matures, small tweaks are possible that will slightly increase yields, not much but a little.
.

---

these are intel's yields. there are no dates for obvious reasons but i would guess that decent manufacturability is right around when they start on the next node. that's why they can make 700mm2 tukwila on 65nm.
http://images.anandtech.com/reviews/...mprovement.jpg

Quote:

---

Originally Posted by saaya

exactly... pure marketing speak, the art of saying something without saying anything at all :D

so IF there would be an upcoming product with 512cores, then he wouldnt comment on it... as nvidia doesnt comment on unannounced products... so he basically said they do NOT plan a 512core card at this point in time... which DOES confirm that yields are not exactly great..

[...]

pretty amateurish statements imo, he unconsciously and inententionally confirmed what he tried to deny :lol:
1. no plans for a 512core card anytime soon
2. yields are around 30% = not very good

---

You might be on to something there... in any case it's worth considering.
Your title should probably read "Xtreme BS detector" instead :)

Dont need a BS detector when someone at nV speaks these days lol

Quote:

---

Originally Posted by STEvil

Dont need a BS detector when someone at nV speaks these days lol

---

Yeah, but you might need a BS protector:
http://lh5.ggpht.com/_BEz2tA7YEWg/SR...Q/s800/665.jpg

Quote:

---

Originally Posted by stinkas

But does it need to be on water?
Water is a pain in the @ss and expensive...

---

You're joking right? @ water cooling

Doesn't have to be but if you want to get higher stable overclocks I'd go water over the stock coolers any day. Plus going water is quieter...if you are a stickler for fan noise.

IMO watercooling isn't a pain. Its actually pretty simple. Expensive...I could agree with that if you're going to water cool a triple sli set up., and you are going for the top shelf parts. Water cooling one card on one loop...wouldnt be that expensive.

Quote:

---

Originally Posted by skycrane

how about this for a good analogy of the chance nvida makes the 512 ultra within 4 months of coming out with fermi.....



i would have a better chance of having a 3some with these 2 :D

---

Sry I had to comment on the threesome thing...I would never say never when it comes to threesomes...those two look like college chicks which up yur chances of a threesome with them by like 20%. Add the fact they dont seem shy about kissing each other in public...another 20%. If they are drinkers...u have an 70% chance unless u have a puss filled hump on your back, a lazy eye and a peg leg...even then slip em some X and u still would have a shot.

In regards to the 512 coming out in 4 mnths...yea I have to agree wit ya.

about watercooling... if you get the right gear its not a pain at all...
buy your vga with the waterblock installed, use those click vents to connect the tubing, you install them on both ends of the tubing, they are water tight, then you connect the blocks and radiator and tubing by just clicking it together.
if you have a good res/pump combo bleeding is a piece of cake too nowadays... all in all id say with the right equippment and a good guide or some experience you can set up watercooling (cpu+gpu) within 1 hour...
if your experienced 30mins might even be possible...

abit late but...


http://i42.tinypic.com/2e6g3dt.jpg

Quote:

---

Originally Posted by JkS

abit late but...


http://i42.tinypic.com/2e6g3dt.jpg

---

hah, i love that guy... what ever happened to him?
imo he should get his own late night show or something! :lol:

Well, I have no idea what is going on in their business, but there are 2 yields - what they get from TSMC (die to wafer yield) and what the net yield is out of their factory (manufacturing yield). Charlie was indicating that yields were <20% and they were claiming that TSMC yields were >20%. Both can be true.

If TSMC is at 30% (die to wafer yeild) but half of the cards don't make it through all of the back-end testing or there are manufacturability issues (manufacturing yield), you would see a net yield of <20%.

As well, when they said yields were "within expectations", that would be perfectly true if their expectation was only 30% yields, right?

Theoretically, everyone could be telling the truth.

Quote:

---

Originally Posted by saaya

hah, i love that guy... what ever happened to him?
imo he should get his own late night show or something! :lol:

---

I think he lives in a different country now with his family. He probably made a deal with the US:P