NVIDIA Reports Financial Results for Second Quarter Fiscal 2011

[btarunr]

NVIDIA reported revenue of $811.2 million for the second quarter of fiscal 2011 ended Aug. 1, 2010, down 19.0 percent from the prior quarter and up 4.5 percent from $776.5 million from the same period a year earlier.

On a GAAP basis, the company recorded a net loss of $141.0 million, or $0.25 per share, compared with net income of $137.6 million, or $0.23 per diluted share, in the previous quarter and a net loss of $105.3 million, or $0.19 per share, in the same period a year earlier. GAAP gross margin was 16.6 percent compared with 45.6 percent in the previous quarter and 20.2 percent in the same period a year earlier.



Results were impacted by a large inventory write-down and a charge related to a weak die/packaging material set.

The inventory write-down was a consequence of weakened demand for consumer graphics processing units (GPUs) as higher memory prices and economic weakness in Europe and China led to a greater-than-expected shift to lower-priced GPUs and PCs with integrated graphics.

The weak die/packaging material set was used in certain versions of previous generation MCP (chipset) and GPU products shipped before July 2008 and used in notebook configurations. The charge, of $193.9 million, includes additional remediation costs, as well as the estimated costs of a pending settlement of a class action lawsuit consolidated in the District Court for the Northern District of California in April 2009 related to this same matter. The settlement is subject to certain approvals, including final approval by the court. Excluding this die/packaging material charge and the associated tax impact, non-GAAP net income was $20.1 million, or $0.03 per diluted share.

"Rapidly changing market conditions made for a challenging quarter," said Jen-Hsun Huang, NVIDIA's CEO and president. "We delivered excellent results in Quadro professional graphics, Tesla GPU computing, and our Tegra system-on-a-chip business. But our GeForce consumer business fell significantly short of expectations amid weak PC demand in Europe and China. Although demand among end-users remains uncertain, we expect to drive revenue and grow market share with new products that are gaining momentum in each of our businesses."

Outlook
The outlook for the third quarter of fiscal 2011 is as follows:

• Revenue is expected to be up 3 to 5 percent from the second quarter.
• GAAP gross margin is expected to increase to 46.5 to 47.5 percent.
• GAAP operating expenses are expected to be approximately $300 million.
• GAAP tax rate of 17 to 19 percent.

Second Quarter Fiscal 2011 Highlights:

• NVIDIA launched and shipped the GeForce GTX 460, bringing the gaming benefits of the Fermi architecture to lower price points.
• NVIDIA launched and shipped a range of Quadro products based on the Fermi architecture for workstation professionals.
• NVIDIA extended its reach in supercomputing as IBM started offering products based on Tesla; these also use Fermi-generation GPUs.

For more information, please visit this page.

View at TechPowerUp Main Site

 

[JATownes]

One word...OUCH!!!

 

[WarEagleAU]

Well now that the GF100 chips have shipped, the next quarter should be better. Plus ATI had a huge run there for a minute too.

 

[bear jesus]

The thing that worrys me is the size of the gf104 die, as being bigger than the 5870's i would assume it would have an effect on their profit margin of what i would assume will be their most popular card (mid range cards are normally the big sellers thus where a lot of the money is for ati and nvidia)

oh and of corse the ati 6xxx cards coming out this year and price drops for the 5xxx cards may be anoher pain for nvidia, hopefully their gf108 cores will give some awesome performace for the price and get them some more sales (im not saying current sales are bad but im sure nvidia thinks they could be better)

 

[Benetanegia]

The thing that worrys me is the size of the gf104 die, as being bigger than the 5870's...

It's actually a few mm smaller than Cypress and remember that the GTX460 is a very crippled part, not only on clusters but on the clocks too. This is obvious when most stock cards with stock voltage reach 25-30% overclock and a slight voltage increase allows 40% OCs easily and with little effect on temperatures. There's even rumors of some partners preparing some 1 Ghz GTX460 variants, so the potential is obviously there. A full GF104 die running at 750-800 Mhz would be just as fast as a HD5870 or very close and if it was run higher than 850Mhz it would actually beat it, though it wouldn't have as great OC potential as the GTX460 does, since I assume they would only OC to 950-1000 Mhz.

Die size is a very bad metric for how much a card costs to make and it's profits anyway. There's many other factors. First of all, the chip accounts for about 10% of the price of a high-end card ($25-50), so even if the there's a significant die difference that doesn't mean a card would cost a lot more. It all depends on the bussiness model and relations wth partners etc. For example if chip A costs $20 and chip B costs $30, that means B costs 50% more (it's 50% bigger too), but the final prodcts may actually end up costing $300 and $310 respectively, not a big deal at all.

And second, does Nvidia pay the same as AMD for waffers? Certainly not. Nvidia is 40% of TSMCs bussiness, Ati is about 15%. Ati is going to leave in favor of GlobalFoundries, Nvidia stays. Believe me, Nvidia does not pay nearly as much to TSMC as AMD does, they need them to stay at all cost.

 

[wahdangun]

yup, not really surprised after all they make dumb move, and make fermi late for 6 month!!!! and after their 3 bigs AIBs are failing except for EVGA.

 

[PCpraiser100]

Pwned.

 

[wahdangun]

It's actually a few mm smaller than Cypress and remember that the GTX460 is a very crippled part, not only on clusters but on the clocks too. This is obvious when most stock cards with stock voltage reach 25-30% overclock and a slight voltage increase allows 40% OCs easily and with little effect on temperatures. There's even rumors of some partners preparing some 1 Ghz GTX460 variants, so the potential is obviously there. A full GF104 die running at 750-800 Mhz would be just as fast as a HD5870 or very close and if it was run higher than 850Mhz it would actually beat it, though it wouldn't have as great OC potential as the GTX460 does, since I assume they would only OC to 950-1000 Mhz.

Die size is a very bad metric for how much a card costs to make and it's profits anyway. There's many other factors. First of all, the chip accounts for about 10% of the price of a high-end card ($25-50), so even if the there's a significant die difference that doesn't mean a card would cost a lot more. It all depends on the bussiness model and relations wth partners etc. For example if chip A costs $20 and chip B costs $30, that means B costs 50% more (it's 50% bigger too), but the final prodcts may actually end up costing $300 and $310 respectively, not a big deal at all.

And second, does Nvidia pay the same as AMD for waffers? Certainly not. Nvidia is 40% of TSMCs bussiness, Ati is about 15%. Ati is going to leave in favor of GlobalFoundries, Nvidia stays. Believe me, Nvidia does not pay nearly as much to TSMC as AMD does, they need them to stay at all cost.

can you give me link about that thing i BOLD?

its not make sense at all, after all ati right NOW have bigger market share than NVDIA, and have ship more card than nvdia


EDIT: and btw ati is the first partner that use their 40nm process (HD4770) and its take months for nvdia to catch up.

 

[CS]

GF104 is larger then Cypress but it is a salvaged part,so the cost might not be as high,yields on it should very good as well,as transistor density is lower.I don't believe NV is paying less then ATi per wafer.Either way Nvidia needs to sell 460 at low margins,they still have no low end chips that compete with ATi.

 

[Benetanegia]

can you give me link about that thing i BOLD?

its not make sense at all, after all ati right NOW have bigger market share than NVDIA, and have ship more card than nvdia


EDIT: and btw ati is the first partner that use their 40nm process (HD4770) and its take months for nvdia to catch up.

I don't have a link just something I remember reading a while back 3-4 months ago. The same article also mentioned that 40nm was still a very small portion of TSMC's revenues. Most of the market was still 55 nm and even 65nm was far more important than 40nm. Remember that moving to smaller process is not a direct advantage for foundries, it's actually an inconvenience they have to deal with if they want to compete. If they could stay on one node forever they would, plain and simple. AMD supporting 40 nm first does not mean as much as "how much money you spend in our waffers".

I dont think what I said is that difficult to believe anyway:

1- Ati only uses TSMC mainly for consumer graphics, since it's core logic, professional GPU and HPC markets are really small, whereas Nvidia uses it for GeForce, Quadro, Tesla, Ion and Tegra products all of which vastly outsell AMD competing products except the GeForce in this Quarter. We don't know what will happen in the next quarter, but we do know that in the past Nvidia sold almost twice as much and it's back then when the deals were made*. In regards to the future Ati would certainly have the power to reverse the situation if it weren't because they are leaving. But as I said, deals where made months ago and back then Nvidia had a far greater power to pressure TSMC. They still have, because it is the only big customer that's still 100% loyal, yet.

2- Nvidia's chips were bigger, which means that for the same ammount of cards sold Nvidia ordered many more waffers. Fact of the matter is that the ammount of waffers ordered were probably 3-4 times more for Nvidia. You just cannot talk about a chip being bigger and the disadvantage that it means, without taking into consideration that it also means a small advantage. It's bussiness, after all.

* To clarify, Ati winning this quarter does not and cannot affect previous and current deals. Only future deals can be affected and those are spoiled by the fact AMD is going to GF, no matter what good deals TSMC may offer.

GF104 is larger then Cypress

Like I said it's actually a little bit smaller. Read the link I provided.

 

[CS]

They report no source,Nordic hardware had a de-lidded GF104 die and it was measured larger then Cypress...

 

[Benetanegia]

They report no source,Nordic hardware had a de-lidded GF104 die and it was measured larger then Cypress...

They do report a source : Nvidia.

 

[CS]

They do report a source : Nvidia.

Did you actually read the article ?

However, when Fermi came onboard, nVidia turned into "Howdy, we don’t disclose die sizes of our chips… Sorry man!"

 

[wahdangun]

I don't have a link just something I remember reading a while back 3-4 months ago. The same article also mentioned that 40nm was still a very small portion of TSMC's revenues. Most of the market was still 55 nm and even 65nm was far more important than 40nm. Remember that moving to smaller process is not a direct advantage for foundries, it's actually an inconvenience they have to deal with if they want to compete. If they could stay on one node forever they would, plain and simple. AMD supporting 40 nm first does not mean as much as "how much money you spend in our waffers".

I dont think what I said is that difficult to believe anyway:

1- Ati only uses TSMC mainly for consumer graphics, since it's core logic, professional GPU and HPC markets are really small, whereas Nvidia uses it for GeForce, Quadro, Tesla, Ion and Tegra products all of which vastly outsell AMD competing products except the GeForce in this Quarter. We don't know what will happen in the next quarter, but we do know that in the past Nvidia sold almost twice as much and it's back then when the deals were made*. In regards to the future Ati would certainly have the power to reverse the situation if it weren't because they are leaving. But as I said, deals where made months ago and back then Nvidia had a far greater power to pressure TSMC. They still have, because it is the only big customer that's still 100% loyal, yet.

2- Nvidia's chips were bigger, which means that for the same ammount of cards sold Nvidia ordered many more waffers. Fact of the matter is that the ammount of waffers ordered were probably 3-4 times more for Nvidia. You just cannot talk about a chip being bigger and the disadvantage that it means, without taking into consideration that it also means a small advantage. It's bussiness, after all.

* To clarify, Ati winning this quarter does not and cannot affect previous and current deals. Only future deals can be affected and those are spoiled by the fact AMD is going to GF, no matter what good deals TSMC may offer.



Like I said it's actually a little bit smaller. Read the link I provided.

sorry i can't believe you without any actual prof or any link, to back it up.

1. you forget one thing, ati is using TSMC for all of its chipset, and they have firepro, firegl line up,

and btw its really affect current deal, because it must increase their capacity in TSMC because its HD 5XXXX selling like hot cakes and right now nvdia don't have any answer for low end DX 11 part where the bulk sales are.

 

[Benetanegia]

Did you actually read the article ?

Yeah I did read the article (and other commens in the past), did you? The reason they are saying all that is because, someone from Nvidia, once again, told them the die sizes.

What I have not seen is that article in Nordic HW where it was measured, all I found is this:

http://www.nordichardware.com/news/71-graphics/40710-geforce-gtx-460-core-revealed.html

Nothing is demostrated, they don't even mention what their measurements are. Not even a ruler or hell even a coin so that we could compare... BSN on the other hand is very specific: 13.7x24.2mm.

sorry i can't believe you without any actual prof or any link, to back it up.

1. you forget one thing, ati is using TSMC for all of its chipset, and they have firepro, firegl line up,

and btw its really affect current deal, because it must increase their capacity in TSMC because its HD 5XXXX selling like hot cakes and right now nvdia don't have any answer for low end DX 11 part where the bulk sales are.

Like the article I linked said Nvidia is completely dominating 40nm waffers. How many GPUs have been sold this quarter is meaningless. Nvidia didn't have products, hence they didn't sell. Now they have prodcts and they have most of TSMCs capacity. That is what matters. TSMC is going to increase capacity because of both, but mainly Nvidia, obviously, since it's Nvidia who's taking most of the orders currently.

 

[CS]

Yeah I did read the article (and other commens in the past), did you?

What makes you ask that ? Clearly I have red everything...

The reason they are saying all that is because, someone from Nvidia, once again, told them the die sizes.

Nowhere in the article do they mention that.. You just assume so , or wish so ?

What I have not seen is that article in Nordic HW where it was measured, all I found is this:

http://www.nordichardware.com/news/71-graphics/40710-geforce-gtx-460-core-revealed.html

Nothing is demostrated, they don't even mention what their measurements are. Not even a ruler or hell even a coin so that we could compare...

But you believe an article without source ?

At least they have a die...


Semiaccurate reports the same die size...

 

[Benetanegia]

What makes you ask that ? Clearly I have red everything...



Nowhere in the article do they mention that.. You just assume so , or wish so ?

In the forums there were some comments which suggested that was the case.

But you believe an article without source ?

At least they have a die...

Semiaccurate reports the same die size...

No, I believe an article which doesn't reveal a source, but that claims real measurements, above one that does not provide any measurements and pretends that a picture is a proof. I could take a picture of a dog and claim that it is a cat, and you don't even need to be a genious if you know what I mean.

Especially, because they are claiming the same size as Semiaccurate I know it's not correct. Semiaccurate... please...

 

[CS]

In the forums there were some comments which suggested that was the case.

... ....so FUD over everything ?

No, I believe an article which doesn't reveal a source, but that claims real measurements, above one that does not provide any measurements and pretends that a picture is a proof. I could take a picture of a dog and claim that it is a cat, and you don't even need to be a genious if you know what I mean.

What the hell ?!? I would know it's a cat !

The picture is clearly of a GTX460 768mb.

Especially, because they are claiming the same size as Semiaccurate I know it's not correct. Semiaccurate... please...

Please,what ? Semiaccurate never had good info ?

 

[Benetanegia]

The picture is clearly of a GTX460 768mb.

You could have said, no I don't know what you mean. Just because they pisted a picture of a GF104 die, does not make their claim any more accurate.

Please,what ? Semiaccurate never had good info ?

Semiaccurate is well known to fake and inflate Nvidia numbers (when lower is better of course, because GTX480 is only 5% faster than 5870... ). Always. i.e they claimed:

Fermi: more than 600 mm^2 -> truth 529 mm^2
GT200b: more than 500 mm^2 -> 470mm^2
G92b: 270 mm^2 -> 231 mm^2

Of course after real numbers are known, they always make a false "I told you" article where they claim to have said the correct numbers. Charlie Demerjian will soon adopt the name Winston Smith, will start working on the Minitrue (Ministry of Truth) and SA may very well start writing their articles in Newspeak.

... ....so FUD over everything ?

Actually SA is FUD and that NH "article" looks like FUD, since it was obviously created with the only idea of saying SA numbers were true without giving any proof, because... "look it's a cat!". Show me a single Charlie Demerjian article where there's a single proof instead of a link to another SA or INQ article. I'm just asking for one...

 

[bear jesus]

Really would not the ram chips be something that could be used for a relative size compairison of the g460's core, im assuming that the gddr5 chips on there are of a standard size thus make it very easy to confirm the size of the actual die?

i would work it out myself but its 3am and im drunk lol, time for some much needed sleep

 

[Benetanegia]

Really would not the ram chips be something that could be used for a relative size compairison of the g460's core, im assuming that the gddr5 chips on there are of a standard size thus make it very easy to confirm the size of the actual die?

i would work it out myself but its 3am and im drunk lol, time for some much needed sleep

No, the only way to measure that is on the real chip with a caliper, and better if it's digital so that user's eyesight is taken out of the ecuation, although you still need a person that actually knows how to use it precisely, which is not common and certainly these people in NH, SA etc are not trained.

Any other method fails miserably and sadly is the other methods like using a ruler or comparative methods (like the one you suggested) which are used by these people. For example a ruler is NOT accurate at all to measure this kind of things and most people don't even know how to properly use a ruler or simply lack the experience or the eyesight (me included). The main problem is from where exactly in the scale you start counting? And where do you end? Are you actually chosing the exact place you wanted to choose, in other words is your eyesight accurate enough? Are you at a 90º angle above both marks? Etc, etc, etc...

The marks are usually 0.5-1 mm thick so you can actually miss the real measurement by as much as 1 mm and that actually happens a lot (subjectiveness also plays a big role in this case, think Charlie measuring). Most people that I have seen in my life, measure by putting the object inscribed between the two lines, which is a fatal error and results in one extra mm. And that's a lot guys. i.e 18x18 = 324, but 19x19 = 361. <<-- BTW isn't the similarity between those two results and the ones at hand with GF104, hmmm... somewhat... telling?

 

[AndreiD]

Benetanegia, I have no idea what you're talking about half of the time.
Your key strokes reveal the green rage you have inside.

And yes, the NH die measurements are correct, or I presumed they just measured the die with their fingers? Jesus.

If I throw a GTX460 in the ocean I bet you'll jump in after it so you can save it and caress it with your hands and love it for eternity.

PS: I like that people are actually giving you links to articles, aka real proof, and all you're getting as counterarguments are just words you made up, which you think are solid arguments just because you spewed them on this forum.

 

[Benetanegia]

Benetanegia, I have no idea what you're talking about half of the time.
Your key strokes reveal the green rage you have inside.

And yes, the NH die measurements are correct, or I presumed they just measured the die with their fingers? Jesus.

If I throw a GTX460 in the ocean I bet you'll jump in after it so you can save it and caress it with your hands and love it for eternity.

PS: I like that people are actually giving you links to articles, aka real proof, and all you're getting as counterarguments are just words you made up, which you think are solid arguments just because you spewed them on this forum.

The only one who has posted a link it's me. I didn't even got a link to NH and had to find out and post it myself, so stop trolling and no matter how many accounts you create you are not fooling anyone. And BTW since you care to brag about how everyone is posting proofs, would you care to demostrate that NH "measurements" are correct, especially since those measrements don't really exist since NH never mentioned any numbers?

EDIT: Now that I had time I searched for a link to support my claims about TSMC. Not the article that I first read, but this should suffice. I never claimed my numbers about TSMC were true anyway, I only claimed to have read them btw. What the actual especific numbers are do not change the fact that my point was valid.

http://www.digitimes.com/NewsShow/MailHome.asp?datePublish=2010/7/1&pages=PD&seq=207

AMD may see tight chip supply from TSMC in 2H10

Latest news
Monica Chen, Taipei; Joseph Tsai, DIGITIMES [Thursday 1 July 2010]

AMD's chip supply from Taiwan Semiconductor Manufacturing Company (TSMC) may face shortages in the second half as Nvidia has already placed a large amount of orders to TSMC in March and April and may squeeze AMD's order out of the already fully-loaded capacity, according to sources from graphics players.

Since Nvidia is set to launch its new GeForce GTX 460 GPU later in July and GF106 and GF108 in August and September, respectively, the company has already pre-booked a large volume of capacity from TSMC.

Since TSMC is likely to give its major clients, Nvidia and Qualcomm, supply priority, the sources believe AMD may not be able to share much of TSMC's capacity in the second half of 2010.

Nvidia plans to launch two versions of the GeForce GTX 460 graphics card with prices ranging between US$230-250 and targeting the ATI Radeon HD 5830. In August, Nvidia will launch the GeForce GTS 455/450 priced at US$129-179 and in September, the GF100-based GPU will be launched with a price below US$100 to replace the GeForce GT 240 and targeting AMD's ATI Radeon HD 5600/5500 series GPUs.

 

[AndreiD]

I'm not a troll, I just recently made an account after visiting this site for years.
I was talking about the GF104 die size btw. Because of the sarcasm that didn't transcribe.
GF104 sizes:
http://techreport.com/articles.x/19242/2 - 320mm^2
http://www.semiaccurate.com/2010/07/21/gf104gtx460-has-huge-die/ - 367mm^2
Some other sources say: 331mm^2 (lab501 forums)
Nordic Hardware: 366mm^2

And Jesus, people gave you this link:
http://www.nordichardware.com/news/71-graphics/40710-geforce-gtx-460-core-revealed.html
It leads you directly to the Nordic Hardware news article, or was it so hard to open the link?

And I presume someone who has that kind of machinery lying around is just going to measure the die size with a ruler?
Let me bold this out for you GTX460 die size > HD5870 die size (which means ATI makes more dies / waffer and in the end makes more money)

 

[Benetanegia]

And Jesus, people gave you this link:
http://www.nordichardware.com/news/71-graphics/40710-geforce-gtx-460-core-revealed.html
It leads you directly to the Nordic Hardware news article, or was it so hard to open the link?
http://www.nordichardware.com/image.../Grafik/GF104/fullimages/GTX460-768GD5_06.jpg
And I presume someone who has that kind of machinery lying around is just going to measure the die size with a ruler?
Let me bold this out for you GTX460 die size > HD5870 die size (which means ATI makes more dies / waffer and in the end makes more money)

Apparently you can't read!

It was me who posted that link! Jeeeesus!

me it's me: Benetanegia

<<<<<< this one.

Anyway...

So out from 4 different sources (3 actually since SA and NH is the same in this case), we have to believe the one that says it's bigger. Funny, who is biased??. 320 mm was said by at least TechReport, Xbitlabs and Anandtech and BSN and Fudzilla said 330-340.

By cathegories:

Press:

TedhReport/Anand/Xbit >>>>>>>> Nordic Hardware

Non-press/blogs:

BSN/Fudzilla >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Semiaccurate