NVIDIA shows signs ... [2008 - 2017]

Status
Not open for further replies.
Would it? JHH was saying that they had taken A1 silicon to production for the past 7 GPUs. If taping out a chip constitutes production, his statement would be a meaningless tautology. Why stop at 7 GPUs, why not say 11? After all, they did tape out GF100-A1, and if that's all that's needed to claim taking A1 silicon to production, why not count it as well?

Maybe I'm misunderstanding what you mean.


it means its on 40 nm :smile:
 
Those are some delicious bitter tears from Charlie!

He still hasn't posted an article on GTX580/GF110 seemingly stunned by the performance and availability. If Nvidia can get GF110 derivatives out in short order (I think they will) old Charlie might have a heart attack!
 
Those are some delicious bitter tears from Charlie!

He still hasn't posted an article on GTX580/GF110 seemingly stunned by the performance and availability. If Nvidia can get GF110 derivatives out in short order (I think they will) old Charlie might have a heart attack!

This is the thing that enjoys me most about the GTX580 and probably the GTX570 and 560 launch - that this idiot will face people asking him: "Aeh.. what was the message of your latest Nvidia bash article btw?"

If you ask me sites like his and the one from Fuad should be shut down - plain and simple. They have nothing to do with serious journalism but instead just putting out marketing crap from either site and inventing stories.
 
A few months ago, JHH seemed intent on going to court, so either all those zeros changed his mind or Charlie is wrong.

Except that Jen-Hsun might be after something else, Intel excels in - apart from having x86 licensing rights, that is. But I think, if my assumption turns out to be true, then it a) would have leaked somewhere already and b) we won't be seeing it's effects not before around mid 2011.
 
Except that Jen-Hsun might be after something else, Intel excels in - apart from having x86 licensing rights, that is. But I think, if my assumption turns out to be true, then it a) would have leaked somewhere already and b) we won't be seeing it's effects not before around mid 2011.

Are you thinking about anything in particular?
 
That's more like it! :LOL: Then again, JHH said Q1 would see Tegra taking off, so who knows, maybe he was actually telling the truth this time.
.

Didn't Nvidia say that they had 42 design wins in June of 2009?

http://venturebeat.com/2009/06/16/nvidia-says-42-tegra-based-web-gadgets-coming-your-way-soon/
42 different devices are being designed with the Tegra chip. Customers include 27 different manufacturers as well as 27 different mobile phone carriers around the world. The devices include media players, web pads, and 12 smart phones. Some of them are more like a cross between a phone and a laptop, in a category called mobile Internet devices.
 
low end isn't where nV makes most of their money ;)

Nvidia loosing the low end of the laptop market and desktop market will be a huge problem for them going foward.

I don't know about you but if i'm looking for an ultra portable laptop i'm not going to be buying nvidia this time around. i'd look at bobcat or intels verison. If I want better performance for games. I might still want a fusion chip with then a dedicated gpu that can be switched off.
 
Alexko postet charts at the last page which I maybe don't understand completely so maybe I'm wrong, but the numbers for Nvidia for Q2/201 look as follows:

1) complete amount of units: 24.225 Mio units
2) stand alone graphics cards: 17.025 Mio units
3) desktop cards: 8.550 Mio units
4) desktop cards without lowend (> US$100): 2.835 Mio units

=> my interpretation is;

a; that Nvidia still has a IGP business: 7.2 Mio units ( 1 - 2 )
b; that Nvidia sells a lot of notebook graphic cards: 8.475 Mio units ( 2 - 3)
c; that only a small fraction of Nvidias business is based on performance to highend desktop graphics cards: 2.835 Mio units only

Also, AFAIK most notebook graphic cards are lowend to mainstream only.

So with Fusion and Sandy Bridge Nvidia could loose up to 88% of the volume.

Could that be true? If yes that looks scary cause without this volume you loose a lot of money to pay your engineers etc.. so the rest could go done too over time. Yes this looks doom and gloom, so I really hope I'm wrong.
 
a; that Nvidia still has a IGP business: 7.2 Mio units ( 1 - 2 )
Correct, many people underestimate it. Much of that is the fairly high ASP MCP79 and MCP89 for Socket 775, but quite a bit of the volume (wouldn't be surprised if it was >40%) is the ultra-low-cost low-ASP MCP61 for AMD, I suspect perhaps the single most successful chip NVIDIA has ever made. Volumes for both markets should shrink to practically nothing by the end of 2011.
b; that Nvidia sells a lot of notebook graphic cards: 8.475 Mio units ( 2 - 3) [...]
Also, AFAIK most notebook graphic cards are lowend to mainstream only.
Yes, but what's 'lowend to mainstream'? Today's low-end is tommorow's ultra-low-cost, and today's ultra-low-cost is tommorow's not-good-enough.

My point simply is that Sandy Bridge won't the dead kiss to notebook discrete GPUs you're expecting. Yes, IGP performance is improving faster than the sweet-spot of the marketing, so it's naturally going to be a bigger and bigger part of the market. But there's nothing magical about Sandy Bridge that will make that happen instantly; it's clearly a better product than the Core i5 IGP, though.

NVIDIA's discrete consumer GPU market is certainly shrinking slowly but surely (down to a fairly reasonable plateau I'd expect) and they're losing the IGP market. But that's precisely why they've invested such a disproportionate amount of money into Tegra and Tesla. They've known for a long time that their core markets were at risk even if they had hoped for a much better outcome in chipsets.

Now if Tesla/Tegra do not deliver (and they've falled short of their growth targets for a long time, especially Tegra) then they're in deep trouble long-term. I don't have visibility on Tesla above and beyond anybody else's, but regarding Tegra I think the design wins for T2 that are definitely going to ship at Moto and LG are better than some are still assuming, and their percentage of design wins for Android 3.0 tablets is excellent.

Qualcomm has little chance before QSD8672, which is a great chip but tape-out was delayed a bit too much due to other projects requiring more resources, and TI/ST-E haven't marketed very aggressively in that field while Marvell/Freescale are usually considered to be second-tier. So that's a big opportunity for NV, and Tegra3 is even massively more exciting on the hardware side - with the big risk factor for them being the software, so we'll see if they deliver better there than for T1/T2. If they do (big if), there's a pretty big chance they blow everybody's expectations for Tegra by a lot, and that'd way more than compensate for IGPs.
 
Basically, Sandy-Bridge is a huge threat to dirt-cheap DX10.1 crap (GeForce 205, 210, 310, 315, 305M, 310M…) but newer GF108-based cards should be fine, because they'll remain significantly faster, and DX11.

It's hard to tell how much of NVIDIA's low-end volume is DX10.1 stuff, and how much is GF108. The thing is that with a >100mm² die, GF108 cards can't be sold for $20~30 as very low-end models tend to, so that market is probably going to disappear for NVIDIA.

AMD might still be able to sell HD 5450s (and HD 6450s or whatever they will be called) throughout 2011 thanks to DX11, but after that? I don't know if Ivy Bridge will feature DX11 graphics.
 
That's what I've been having in mind for quite some time. I am still undecided, if it would make sense, but hey! :)

I doubt that Intel would be much more willing to give that away than an x86 license. Plus, while NVIDIA could make an argument about anti-trust issues for the x86 license, they have absolutely no grounds to request a manufacturing deal, so I don't think that's going to happen.
 
Intel would not need to give their mfg expertise away, but simply produce chips for Nvidia, who in turn, could have a significant advantage over time compared to AMD, given that Intel has led the mfg race between the two for quite a while.

And Nvidia would not need to sue Intel into it, but rather have an agreement over their own patents being used in Intels processors. Yes, right, processors.

Since they are die-integrated, a court ruling in favor of Nvidia (if that ever will come to pass) puts Intel in far greater risk than before, where they'd "only" had their chipset business endangered. Now, if such a ruling would exist, they could not sell their CPUs any more.

IMHO that makes quite an argument, given that going to court not necessarily bestows justice upon you.
 
Status
Not open for further replies.
Back
Top