Nvidia threatens partners at CeBIT

[BenSkywalker]

DUDE? There are no holiday numbers!!!

My apologies, made you look bad by editing my post while you were posting more inaccuracies.

 

[Schmide]

My apologies, made you look bad by editing my post while you were posting more inaccuracies.

I'm not inaccurate? The Q4 and FY 2010 (July-Sept) is reported to the SEC end of January. The Q1 FY 2011 (Oct-Jan) is reported of April.

Edit: I listed the FY dates wrong but the information is sill correct.

 

[Schmide]

I paid that post every bit of attention it deserved. You didn't even glance at the links I provided-

Maybe you should pay more attention?

Over the Christmas holiday quarter nV outsold ATi 2:1. Enthusiasts that hold their breath from benchmarks are a miniscule portion of the overall market.

nV is currently in FY 2011. I have no idea where you get your financial news from, but I provided the links in that thread, read them.

Meh- I'll repost them and quote them, don't want misinformation being spread

http://phx.corporate-ir.net/phoenix.zhtml?c=116466&p=irol-newsArticle&ID=1392142&highlight=

At the very least you're being grossly misleading by claiming Holiday numbers!!! Regardless, no holiday sales were reported in this financial statement!!! Those will come in April-May.

 

[BenSkywalker]

The nV report I linked to is SEC File/Film 000-23985 10613402 and was filed with the SEC on Feb 17th as per mandated by their public standing.

If you would like here is the link to the actual Form8K online documentation- http://www.sec.gov/Archives/edgar/data/1045810/000104581010000002/0001045810-10-000002.txt

The link is going to appear messed up as SEC links always are which is why I didn't link to the SEC directly in the first place. Honestly, I think you are a bit over your head in this discussion, when I stated explicitly what numbers were reported I meant exactly what I stated

At the very least you're being grossly misleading by claiming Holiday numbers!!! Regardless, no holiday sales were reported in this financial statement!!!

Read the link I provided. The dates covered by the report are explicitly spelled out, also now have links directly to the SEC for you too. The time covered is exactly what I stated it was and is spelled out. Misreporting on those reports means jail time for the executives. Trust me, they aren't going to risk going to jail in anticipation of you coming out of nowhere with this very confusing line of thought you currently have.

 

[tviceman]

People aren't really looking at it this thoroughly on that front. GTX480 512SPs, GTX470 448SPs, GTX 260 384SPs, GTX 250 320SPs- not saying it is how they will go, but using nV's setup it is possible that they only yield 3% of their chips as 480s and it is possible they could hit close to a 95% useable yield rate. This is one of the issues I have with believing a lot of the talk about how they are going to stop production. No, it isn't ideal to use your largest die for 4 seperate chips, but it certainly is within their ability to do so if they decided to. Not saying they will go that route, not even saying they are absolutely going to launch, but when I look at all of the 'nightmare' scenarios I am reading about, seems like it makes a lot more sense to sell off ~1/2 good chips for ~$200 then to throw them away. Think 5830- somehow nV can't do the same thing when they by design make their chips with that in mind?

If the yields at the highest end aren't really that great, I fully expect nvidia to stock pile as many "bad" chips that still operate with 320sp or even 250sp. That wasn't what I was referring to though. I am referring to nvidia making a chip specific for the HPC market, and then whittling out (instead of just disabling) some unnecessary HPC functions like double precision floating point that most average Joe consumers won't use.

I regress though as I completely admit to knowing very little about GPU chip design, process, and manufacturing. I am just looking forward to more market competition.

 

[BenSkywalker]

I am referring to nvidia making a chip specific for the HPC market, and then whittling out (instead of just disabling) some unnecessary HPC functions like double precision floating point that most average Joe consumers won't use.

Not trying to pick on your arguments, just giving you a rough idea of what that would entail. Now speaking strictly from the perspective of yields on consumer parts you are absolutely correct, they would almost certainly improve if they took the approach you are talking about. Where it gets tricky is how viable would that market be for them. Right now they can take any chip that doesn't yield a proper Tesla part and sell it off as a 470, or 460 or whatever else. If they are running dedicated fabs just for the Tesla parts not only do the lose the capabilities of selling off 'bad' chips as lesser models they also need to have a seperate fab process altogether for them. This costs in the low millions of dollars. From a business perspective, what they are likely looking at is the overall cost of running a seperate line entirely for Tesla seperate from GeForce compared to how much their yields are going to be hurt by the additional Tesla hardware. From all the information we have available, the DP support and ECC are the only real things that they would be likely to drop, so how many yields end up bad because of those portions of the chips, bad enough so they can't be sold off as lower GF varriants, and is that enough to offset the additional ~$10 million it would take for them to have an entirely seperate line. As of now one of the reasons why they can be so shockingly effective in the performance/$ column is because the GF brand is subsidizing Tesla development. Yes, GeForce users are paying a premium for that, but how much of one and is it enough to make a meaningful impact are more difficult questions to answer.

 

[Schmide]

Honestly, I think you are a bit over your head in this discussion, when I stated explicitly what numbers were reported I meant exactly what I stated

Why do you keep doing this? You did it in the Anti-aliasing discussion we had. It does you no justice. You seem to put your ego ahead of your argument.

I will stand by my statement. NO HOLIDAY SALES ARE ON THIS FISCAL REPORT.

A company sells a buncho product. They report on their sales approximately a quarter after the sales. So Oct-Jan will come in April-May. I'll put it another way. No 5000 series cards made it on the books at ATI either.

Edit: Here I'll translate the chart for you.

                            Quarterly Highlights     Fiscal Year Highlights
($ in millions except
 per share data)     Q4 FY2010  Q3 FY2010  Q4 FY2009    FY2010     FY2009
[COLOR="Red"](date of product)   (July-Sept) (April-Jun)  (July-Sept) (Oct-Sept) (Oct-Sept)
[/COLOR]Revenue                $982.5     $903.2     $481.1     $3,326     $3,425
                     =========  =========  =========  =========  =========
GAAP:
Gross margin            44.7%      43.4%      29.4%      35.4%      34.3%
Net income (loss)      $131.1     $107.6    ($147.7)   ($68.0)    ($30.0)
Income (loss) per
 share                  $0.23      $0.19     ($0.27)   ($0.12)    ($0.05)
Non-GAAP: (1)
Gross margin            44.7%      40.7%      28.1%      38.6%      39.9%
Net income (loss)      $131.1      $77.4    ($145.3)    $141.4     $160.3
Income (loss) per
 share                  $0.23      $0.13     ($0.27)     $0.26      $0.29
                     =========  =========  =========  =========  =========

Give it a rest Schmide, Ben's right

-ViRGE

 

[BenSkywalker]

They report on their sales approximately a quarter after the sales.

As a heads up, if you work for a publicly traded company and you know they are handling things this way you may want to talk to someone at the SEC, what you are describing is not allowed unless you have it spelled out why you are not taking charges for sales/losses in the quarter in which they occur. This isn't PR spin talk, it's criminal. Enron/Worldcom changed a lot about what is allowed to be reported, what you are talking about isn't.

 

[MagickMan]

You think a 480 GTX will beat hd5970?

How would it beat a multi-GPU card? I very much doubt that. OTOH, for single GPU solutions, it'll be on top.

 

[RussianSensation]

If the lead is only 5% ATI might be able to beat that with a memory swap (7 Gbps GDDR5) and a small GPU core speed increase.

However, I think Fermi's major advantage could be minimum FPS (with full DX11 turned on) rather than average FPS.

I find the 5% performance advantage very hard to believe.

- A lot of games simply run better on NV's architecture such as Far Cry 2, World in Conflict, Borderlands. Simply doubling GTX280 would net NV a lead in these games which is a lot more than 5% over 5870. GTX280 was more or less a beefed up G80. GTX280 performed really well. In this architecture though, NV increased the efficiency of the shader units/architecture in general. The card is also packing a significant advantage over 5870 in memory bandwidth which will help with high AA modes (8AA, etc.)

No one predicted 9700 Pro to be the card that it ended up. NV could have simply doubled everything on GTX280 as ATI did over 4870 and had been ahead of the 5870, but they didn't take this approach. This leads me to believe they have chosen a design that is superior to simply "doubling" of the previous generation. I am guessing 20-30% performance increase over 5870 on average for 480, with 470 about 5-10% faster.

 

[SolMiester]

That is true, but think about this:

Why does Nvidia need to beat ATI in games? How will being faster than HD5970 help anything? Maybe Nvidia is doing the right thing by focusing on HPC instead of gaming?

The following excerpt from Tom's hardware really brings the situation into perspective.



http://www.tomshardware.com/reviews/future-3d-graphics,2560-3.html

meh, Toms is just brown nosing again....eye infinity, nv has had that for years, if the ATi GPU was so damn flash, they could have done more with compute....it still doesnt have as many features as nV...

Like having a powerful motor with tiny wheels................

 

[zerocool84]

I find the 5% performance advantage very hard to believe.

- A lot of games simply run better on NV's architecture such as Far Cry 2, World in Conflict, Borderlands. Just doubling GTX280 would net NV a lead in these games over 5870 that's a lot more than 5%. But NV increased the efficiency of the shader units/architecture and is also packing a significant advantage over 5870 in memory bandwidth.

No one predicted 9700 Pro to be the card that it ended up. NV could have simply doubled everything on GTX280 as ATI did over 4870 and had been ahead of the 5870, but they didn't take this approach. This leads me to believe they have chosen a design that is superior to simply "doubling" of the previous generation. I am guessing 20-30% performance increase over 5870 on average for 480, with 470 about 5-10% faster.

And there's games that run better on ATI hardware so what???

 

[Schmide]

I'm sorry. I'm wrong. My bad.

 

[RussianSensation]

And there's games that run better on ATI hardware so what???

Yes, but DX11 games will focus on geometry and tessellation as well. In this area, NV will be superior to ATI based on the Fermi architecture for future games. They have also shifted their focus towards complex shaders as they increased Stream processors from 240 to 512. NV was generally was superior in texture heavy games, while ATI has a historical bias towards performing well in shader heavy games. This may not be true anymore.

I just think since NV's tech is more forward looking, but it may be 12 months or more before the advantages of this architecture are actually visible over 5870 (we would need DX11 designed games from the ground up).

 

[Red Storm]

... eye infinity, nv has had that for years...

Really? I was looking for such a solution (a good performing video card that could run 3 or more displays) a little less than a year ago and I couldn't find anything. Please show me.

 

[cbn]

meh, Toms is just brown nosing again....eye infinity, nv has had that for years, if the ATi GPU was so damn flash, they could have done more with compute....it still doesnt have as many features as nV...

Like having a powerful motor with tiny wheels................

Well they make a good point.

What good are today's video cards when they are overkill for single monitors?

I guess we could crank up AA to x8, but at this point I am not sure I find the extra anti-aliasing useful. Is not being able to see little teeny tiny jaggies while in the game action is completely stopped really worth $100s of extra dollars in video card costs? LCD monitor pixels are pretty small these days.

 

[zerocool84]

Yes, but DX11 games will focus on geometry and tessellation as well. In this area, NV will be superior to ATI based on the Fermi architecture for future games. They have also shifted their focused towards complex shaders as they increased Stream processors from 240 to 512. NV was generally was superior in texture heavy games, while ATI has a historical bias towards performing well in shader heavy games. This may not be true anymore.

I just think since NV's tech is more forward looking, but it may be 12 months or more before the advantages of this architecture are actually visible over 5870 (we would need DX11 designed games from the ground up).

and in those 12 months it won't matter cus it won't be able to max out games anyways. The Nvidia cards have to give up other things for good tessellation performance. They don't have dedicated units like ATI hardware.

 

[cbn]

I just think since NV's tech is more forward looking, but it may be 12 months or more before the advantages of this architecture are actually visible over 5870 (we would need DX11 designed games from the ground up).

Hopefully Nvidia sponsors some Tessellation heavy games, because I am guessing a lot of the console ports will be favoring ATI cards goes (re: ATI has the hardware contract for Xbox 3)

Speaking of consoles, I have noticed most of the new TVs being sold today are 1080p. This begs the question how much extra GPU power will we really need for our PCs? (it looks like there won't even be a difference in resolution pretty soon, if anything I would actually expect the console to need more GPU power than a PC since the pixels on large 1080p TVs would be larger)

 

[SmCaudata]

Hopefully Nvidia sponsors some Tessellation heavy games, because I am guessing a lot of the console ports will be favoring ATI cards goes (re: ATI has the hardware contract for Xbox 3)

Speaking of consoles, I have noticed most of the new TVs being sold today are 1080p. This begs the question how much extra GPU power will we really need for our PCs? (it looks like there won't even be a difference in resolution pretty soon, if anything I would actually expect the console to need more GPU power than a PC since the pixels on large 1080p TVs would be larger)

Look at a Blu-ray movie vs a video game. You can get A LOT more detail at 1080p resolution than is currently capable. There are issues of storage, but just because TVs are 1080p doesn't mean that we don't need graphics power.


As to the main article..what is interesting to me is that for the last few months everybody has been saying that Nvidia parts won't affect the ATI parts because they will be much more powerful and therefore more expensive. Sadly though, I have a feeling that the 470 cannot price compete with 5850. If this article is correct I have a feeling that ATI will have no price pressure.

 

[ronnn]

If these numbers are correct, should make for exciting times. Of course if Charlie can make performance numbers say 5%, Nvidia pr should have no problems having a focus on specific games/settings that give at least 20%. Expect the fud and bans to fly.

 

[cbn]

You can get A LOT more detail at 1080p resolution than is currently capable. There are issues of storage, but just because TVs are 1080p doesn't mean that we don't need graphics power.

Well sure there are advances in Graphics APIs that will require more GPU power at 1080p, but how fast will these advances come? At the moment it seems the console industry gets most of the programmers attention.

 

[edplayer]

..., if anything I would actually expect the console to need more GPU power than a PC since the pixels on large 1080p TVs would be larger)

What?

why would a tv display (1920x1080) that is "larger" need more gpu power than a computer monitor that is also 1920x1080?

 

[cbn]

What?

why would a tv display (1920x1080) that is "larger" need more gpu power than a computer monitor that is also 1920x1080?

The pixels on the larger TV display are bigger, therefore they would benefit from more anti-aliasing in order to look smooth to the human eye.

Anti-aliasing requires more GPU power.

 

[blanketyblank]

The pixels on the larger TV display are larger, therefore they would benefit from more anti-aliasing in order to look smooth to the human eye.

Yes, but you forget the fact that you sit farther away from a large TV. I don't think anyone sits 1' away from a large 1080p set like you would with a normal monitor.
Possibly someone might with a 32" set, but even that is probably a bit too close.

 

[edplayer]

that makes sense (that you would notice AA differences on a larger display easier) but it doesn't lead to a TV display "needing" more gpu power.