nVidia Q2 financial results

[nitromullet]

Is this the project you are talking about?

http://www.engadget.com/2010/08/11/darpa-enlists-nvidia-to-build-exascale-supercomputer-thats-100/

http://www.easyir.com/easyir/custom...rsion=live&releasejsp=release_157&prid=610136
That.

They are probably referring to the one in golem's link...

DARPA is handing out $25 million to NVIDIA in order to develop "high-performance GPU computing systems." Specifically the Defense Department's research and development arm is aiming to address a so-called "crisis in computing," and if all goes well, the four-year project will eventually yield a "new class of exascale supercomputers which will be 1,000-times more powerful than today's fastest supercomputers." That's a pretty lofty goal, but NVIDIA will be aided by Cray, Oak Ridge National Laboratory and a half-dozen US universities along the way.

 

[brybir]

they probably wrote down their gf100 chips, figured they would take the hit now and get it over with. next quarter should be much more positive for nvidia with gtx 460 sales ramping up strongly and fermi II getting much closer to reality.

They cannot book revenue shipments for Fermi II until they actually start "selling" them to AIB vendors. It is going to be a while before revenue shipments of Fermi II start to impact Nvidia's bottom line, even more so if ATI maintains the current schedule of development and release they are on now.

And we don't really know how well the 460 is selling right now, we just speculate that it is selling well because it has positive momentum right now in the enthusiast realm. That does not necessarily translate into profits for Nvidia.

Much of the future speculation is wait and see.

 

[bryanW1995]

sorry, you're right, just like 5xxx had "positive momentum".

gtx 460 is the only decent midrange or better card right now from a price/performance perspective. there have been some deals on 470 and 5850, but those 2 are generally too much $$. nvidia has a rabid fanbase that hasn't had much choice for the past year; are you trying to say that this pent up demand for a decent nvidia card in combination with the gtx 460 isn't going to lead nvidia to a "more positive" 3rd quarter?

 

[akugami]

http://news.cnet.com/8301-13924_3-20013543-64.html

nVidia's CEO Huang basically stating that their CPU strategy is to invest in ARM.

And importantly for those who swore up and down Intel was going to get a beat down from nVidia courtesy of the FTC, Huang also states the recent FTC decisions/sanctions aren't really related to the lawsuit between Intel and nVidia which is a contract dispute.

Personally I've always felt that nVidia was moving away from its reliance on x86 and the PC for most of its revenue and moving to markets such as ARM via its Tegra line. It only makes sense to diversify so you don't have all your eggs in one basket so to speak. nVidia will always remain a GPU company though.

Decisions like these are going to make or break nVidia in the next few years. If its Tegra line is successful then nVidia is going to remain very healthy but if it fails and with the dwindling prospects in the PC video card markets, it's going to look like rough times for nVidia.

 

[Pantalaimon]

Their gross margin took quite a nosedive from the usual 45+% to only 16,6%

 

[SlowSpyder]

sorry, you're right, just like 5xxx had "positive momentum".

gtx 460 is the only decent midrange or better card right now from a price/performance perspective. there have been some deals on 470 and 5850, but those 2 are generally too much $$. nvidia has a rabid fanbase that hasn't had much choice for the past year; are you trying to say that this pent up demand for a decent nvidia card in combination with the gtx 460 isn't going to lead nvidia to a "more positive" 3rd quarter?

In another thread there was some discussion about the 460 being a good part for Nvidia's busines, or not. No doubt it's a great card for the consumer, near 5850 performance for less money. But I think the possible problem from a business perspective is that the cards are selling for not that much money compared to the 5850 and 5870 while using a similar sized GPU (actually the Nvidia part is a bit larger even). So who knows how much Nvidia is or isn't making on the parts.

 

[tviceman]

But I think the possible problem from a business perspective is that the cards are selling for not that much money compared to the 5850 and 5870 while using a similar sized GPU (actually the Nvidia part is a bit larger even).

"The GF104 die is measured at 13.7x24.2mm, i.e. 331.54mm2. Note that some publications incorrectly mentioned die size being 366mm2 - thus, bigger than AMD's current high-end chip. That is not correct, if only by a small margin [332 vs. 337 mm2]. Direct competitor: AMD Cypress, 337mm2."

The chip is actually smaller than Cypress. Don't believe everything Charlie writes. Yeah, yeah I know it's BSON but this is actually a well written, nonbiased informative article.

http://www.brightsideofnews.com/news/2010/8/9/nvidia-fermi-geforce-die-sizes-exposed.aspx

 

[SlowSpyder]

"The GF104 die is measured at 13.7x24.2mm, i.e. 331.54mm2. Note that some publications incorrectly mentioned die size being 366mm2 - thus, bigger than AMD's current high-end chip. That is not correct, if only by a small margin [332 vs. 337 mm2]. Direct competitor: AMD Cypress, 337mm2."

The chip is actually smaller than Cypress. Don't believe everything Charlie writes. Yeah, yeah I know it's BSON but this is actually a well written, nonbiased informative article.

http://www.brightsideofnews.com/news/2010/8/9/nvidia-fermi-geforce-die-sizes-exposed.aspx

Last I remember reading, it was larger. So I guess we're not sure < shrug >. But even if they are what you say, I'm guessing AMD is getting a lot higher average prices for their very similar sized GPU than Nvidia in this case. Of course, once Nvidia has a fully functional part, they will obviously get more for those GPU's than the current 460 parts.

 

[tviceman]

Last I remember reading, it was larger. So I guess we're not sure < shrug >. But even if they are what you say, I'm guessing AMD is getting a lot higher average prices for their very similar sized GPU than Nvidia in this case. Of course, once Nvidia has a fully functional part, they will obviously get more for those GPU's than the current 460 parts.

Nvidia did take a big step to cut some of their costs regarding all fermi related-products by securing more wafer orders than AMD (and thus getting a discount). I think more than anything this is why we have not seen price cuts from the 5850/5870.

Regardless, I have no doubt that AMD averages a higher profit per sale with cypress than Nvidia with GF104, and even with a 384 shader part coming soon, the average selling price of GF104 will still be lower so Nvidia is more reliant on volume.

 

[tincart]

"The GF104 die is measured at 13.7x24.2mm, i.e. 331.54mm2. Note that some publications incorrectly mentioned die size being 366mm2 - thus, bigger than AMD's current high-end chip. That is not correct, if only by a small margin [332 vs. 337 mm2]. Direct competitor: AMD Cypress, 337mm2."

The chip is actually smaller than Cypress. Don't believe everything Charlie writes. Yeah, yeah I know it's BSON but this is actually a well written, nonbiased informative article.

http://www.brightsideofnews.com/news/2010/8/9/nvidia-fermi-geforce-die-sizes-exposed.aspx

Who cares what Charlie says. Nordic Hardware gave separate "confirmation".

Charlie also suggested he had access to the hard numbers for x-ray measurement which would give accurate results.

The BSON numbers seem to be coming from an Xtreme Systems forum thread. If I had to put it to numbers, SemiAccurate gets a Full Of Shit rating of 6/10 and BSON gets a FOS rating of 7.5/10 so I'll take one full of shit die size claim over the other.

 

[bryanW1995]

regardless, the die sizes are similar. nvidia has lower costs but amd has significantly higher asp's. right now amd is still in the driver's seat, but nvidia is at least on the bus finally.

I think that gtx 460 is nvidia's 3870. We all know what came after that, and the fermi architecture looks to have an even stronger starting point. however, amd isn't exactly resting on its laurels like nvidia did with g92 and its 14 rereleased, rebadged cards. hopefully SI and fermi II are able to reignite a new graphics war for us all.

 

[HurleyBird]

nvidia has lower costs

I don't suppose you can back that wild claim with any kinds of facts?

 

[SlowSpyder]

I don't suppose you can back that wild claim with any kinds of facts?

Yea, I've heard that mentioned in other threads, but who knows if it's true or not. The GPU's are of similar size, AMD is rumored to have very solid yields, both the 4601GB and the 5870/5850 use a 256bit connection to DDR5 memory (though the 768MB 460 may be a simpler PCB with it's 192bit connection). Both use similar power, have non-spectacular coolers. So really, who knows.

 

[HurleyBird]

Well, regardless of size I'm sure AMD has lower costs right now as they have a mature product with good yields. Seeing as we only have harvested GF104 products, yields can't be too great. Who knows who will have lower costs once GF104 becomes more mature though.

 

[golem]

Well, regardless of size I'm sure AMD has lower costs right now as they have a mature product with good yields. Seeing as we only have harvested GF104 products, yields can't be too great. Who knows who will have lower costs once GF104 becomes more mature though.

Isn't using harvested chips a plus for cost (use parts that you would otherwise throw away) but a negative for ASP (using parts you could possible use for a full product in a cut down product)?

 

[HurleyBird]

Isn't using harvested chips a plus for cost (use parts that you would otherwise throw away) but a negative for ASP (using parts you could possible use for a full product in a cut down product)?

Yes, but the fact that they are ONLY using harvested chips is not a good sign as far as yields are concerned.

 

[golem]

Yes, but the fact that they are ONLY using harvested chips is not a good sign as far as yields are concerned.

ah got it. But in that case, there is a silver lining. Unless they can't get fully functioning chips then they could be stockpiling those chips for another product soon to go up against SI. Which would only do good things for prices (for consumers at least).

 

[SlowSpyder]

Isn't using harvested chips a plus for cost (use parts that you would otherwise throw away) but a negative for ASP (using parts you could possible use for a full product in a cut down product)?

Obviously selling those chips that don't make the cut for something is better than throwing them away. They could be selling thousands of chips for $20 vs. throwing away thousands of chips ($20 pulled out of my ass). But, who knows if they are really making much on them, maybe not making anything but just taking less of a loss vs. throwing them away.

As mentioned above, harvesting them for something is better than throwing them away for nothing, but when your top part is a harvested chip it probably is saying something about yields.

And I remember reading something on how selling harvested chips can have a negative effect long term, if the company is going to keep that product line alive. Say the GF104 matures nicely and they are able to get huge yields of fully working parts. They may have to 'artificially' cripple some to keep the GTX460 alive, parts that they could have sold at higher price points possibly.

I do see this part improving for Nvidia, a fully functional GF104 will probably fit between a 5850 and 5870, Nvidia will charge more for it obviously and get a healthier profit.

 

[tviceman]

I don't suppose you can back that wild claim with any kinds of facts?

It's not really a wild claim. Nvidia has secured much more of the 40nm process than AMD at TSMC - hence part of AMD's supply constraint - and therefore Nvidia gets discounts on wafers. Even super fan #1 Charlie has alluded to this. So if yields are the same for GF104 and cypress, Nvidia is paying less per chip.

 

[evolucion8]

It's not really a wild claim. Nvidia has secured much more of the 40nm process than AMD at TSMC - hence part of AMD's supply constraint - and therefore Nvidia gets discounts on wafers. Even super fan #1 Charlie has alluded to this. So if yields are the same for GF104 and cypress, Nvidia is paying less per chip.

But AFAIK the GF104 yields are far from being the same as Cypress.

 

[TemjinGold]

Exactly. Nvidia may or may not be paying less per wafer (seeing as TSMC is at capacity, it isn't a given that the larger order is getting a discount) but that doesn't necessarily mean Nvidia is paying less per chip. You are assuming that GF104 is at the same or close to the same yield as Cypress for this to be true.

 

[brybir]

sorry, you're right, just like 5xxx had "positive momentum".

gtx 460 is the only decent midrange or better card right now from a price/performance perspective. there have been some deals on 470 and 5850, but those 2 are generally too much $$. nvidia has a rabid fanbase that hasn't had much choice for the past year; are you trying to say that this pent up demand for a decent nvidia card in combination with the gtx 460 isn't going to lead nvidia to a "more positive" 3rd quarter?

One product in Nvidia's product pipeline will not make or break the company, nor will one particular part lead them either up or down. Besides, we have very little information about what it actually costs to make, gross margins, etc. You cannot ever assume just because something sells well (and we do not know that) that it is making money for Nvidia or any other company. Profits come from the margin between selling price and cost of making the chip. If Nvidia is making $10 per chip and sells 1 million of them, it makes $10 million dollars that year in profit from that card...not that much for a company that does $4 billion dollars in business and whos historical profits are in the hundreds of millions of dollars range. Change that number to $30 a chip or higher and it starts to make more of a difference, but unless we know that number, all we can do is guess.

Besides, it could end up having a short run. If the "rumors" hold true and ATI is implementing a large price drop to clear inventories before SI launches in November, it could see some fierce competition, squeezing margins.

And in any event, the bread and butter of these companies are the big buyers like HP and Dell who buy millions of cards. The enthusiast market may hold higher margins, but it is these lower cost and high volume parts that allow companies to amortize their R & D and production costs over a product lifetime.

Bottom line is that we have no idea what impact 460 will have on Nvidia's bottom line. While you can generally expect it "should" be having a positive impact on Nvidia's quarter, it is far from certain.

 

[brybir]

regardless, the die sizes are similar. nvidia has lower costs but amd has significantly higher asp's. right now amd is still in the driver's seat, but nvidia is at least on the bus finally.

I think that gtx 460 is nvidia's 3870. We all know what came after that, and the fermi architecture looks to have an even stronger starting point. however, amd isn't exactly resting on its laurels like nvidia did with g92 and its 14 rereleased, rebadged cards. hopefully SI and fermi II are able to reignite a new graphics war for us all.

The thing you are missing here is that while margins are very important to know in regard to whether a product is profitable, you have to look at Nvidia's ability to spread its R & D costs of its product family out over multiple chips.

Somewhere is a number ($$$$$$$$$ most likely) that represents the R & D to develop and product the Fermi chips that are out. Ideally, you can then spread this cost out over lots of chips at different price points. If Nvidia spent $100 million developing Fermi and sold 100 million chips, the R & D cost per chip would be $1, making the rest of the margin profit. But, if they only sell 10 million chips, R & D eats away $10 per chip and so on.

So we cant say that Nvidia's costs are higher or lower until, as I stated previously, we know the margins on the chips, and the rest of the costs that went into production, R & D etc.

I would also say that it is unlikely that Nvidia has lower costs on a similar sized die, given how behind they were in releasing Fermi. As has been previously reported, ATI spent quite a bit more time working on process technology issues than Nvidia (although I dont know this to be a fact, it was just talked about a lot on this forum) and has quite a bit more time on the process node in actual product before Fermi started to even sample. Usually this should translate into ATI having better control of defects, production issues etc...leading to lower production costs. But again, that is just my speculation given the normal tendencies in semiconductor manufacturing.

 

[Janooo]

Unsold product sitting in a warehouse has a value assigned to it, which is reported on the "assets" part of your balance sheet. If that product value declines for some reason then the company needs to "write it down."

So, for example, if NV had 1 million G92 chips in a warehouse somewhere valued at $20/ea with nobody willing to buy any more at that price it would be reasonable to write down that inventory from a value of say $20,000,000 to $1,000,000. Totally fictitious example, but you get the point.

nVidia has some product sitting around somewhere (the inventory) and it is recorded in their books at a certain value.

However, nVidia had to decrease the book value of that inventory in order to make it accurately reflect the value they could get from selling it by $70 million (according to Janney Capital analyst Nick Aberle). This is generally recorded as an expense at the end of the financial quarter.

So nV has some stock of something that they now expect to sell for $70 million less than they had planned previously.

Well, in this case the write off is down to zero. NV does not consider the inventory to be sell-able.

Ross Seymore – Deutsche Bank Securities
Just on the inventory charge side of things and the gross margin is the gross margin benefitting at all in the third quarter because of written off inventory?
David White
No. We – you know we wrote off what we anticipated would not be sellable, right, and so it would be inconsistent to assume one quarter that you are going to write it off, the next quarter you are going to sell it. So, we wrote off what we believe was an excess and so our guidance doesn’t include any of that selling through.

http://seekingalpha.com/article/220...-08-01-10-earnings-call-transcript?part=qanda

 

[postmortemIA]

NV is going to be in shoes that AMD and many many other hardware companies have been for a while. Real problem is intel bundling GPUs onboard with their CPUs. Like Microsoft does IE with windows. That gives them huge market share. Most people want cheap computer, Nvidia has almost nothing to offer against $3 GPU from intel. I am surprised that they managed to stay profitable for this long.