Nvidia Tegra

Is Moorestown only a threat from a manufacturing standpoint or is there a technology advantage there too? I don't see one today and by the time it hits Nvidia should be on Tegra II.
Look at it this way, right now Menlow is:
- Silverthorne chip with 80mW+ idle power on 45nm.
- Poulsbo chip with high idle power: Southbridge, Memory Controller, PowerVR SGX 535(?), PowerVR VXD on 130nm.
On the other hand, Moorestown looks like this:
- Main chip with Silverthorne, Memory Controller, PowerVR SGX 545(?), PowerVR VXD on 45nm with low idle power.
- Southbridge on 130nm with low idle power.

The technological threat is pretty simple: without wanting to sound biased, which I am not in this instance, PowerVR VXD is one hell of a video decode core that already manages 1080p H.264 High Profile in ~120mW on 130nm according to Intel, presumably excluding other idle power, the memory controller and external memories. Once you move that and the memory controller to 45nm, even 1080p video decode is basically going to be free except for the external Mobile DDR memory accesses.

From a CPU & 3D POV, I know enough about their architecture to be confident NV will be very competitive with Intel and everyone else. Video Encode & ISP, who knows, but those don't matter quite as much. So the only part I'm not sure about is video decode - certainly in isolation, it doesn't seem to be competitive with VXD, but with Menlow the system architecture is bad enough that it doesn't matter. On Moorestown, it's definitely going to show up, so they'll need to improve their video decode numbers in Tegra 2 quite a bit.

It's worth pointing out their current HD video decode numbers are still really good compared to other ones in the industry. I don't know exactly how good, but I'd suspect they're better than TI's & ST's, no idea about Qualcomm which uses an ATI-made Tensilica-based core in Snapdragon for 720p. In terms of video decode in general, they seem to be nicely ahead of most IP cores, including the ARC ones used by Broadcom. So in a way, everything points at VXD being the odd man out here, not the other way around... The good news for NV, of course, is that the only public licensees for it are Intel and Apple.
 
Without wanting to sound narrow minded, who'd want a 12" laptop which couldn't run XP/Vista/OS X?

If you got a REALLY good software solution together I could maybe see it making some headway in the Eee sized ultra-portable market up to say 7-9", but packaging and thermals are less critical than in say a 4" device.

I can imagine someone wanting a 12" laptop that runs Debian. pretty nifty list of architectures they support : [alpha] [amd64] [arm] [hppa] [i386] [ia64] [mips] [mipsel] [powerpc] [sparc].

12" widescreen is pretty small already I guess. I never owned a laptop anyway, a 486DX2 would make me happy already :p.
biggest problem I see for ARM on linux, no way to run warcraft III through wine.
To run some of the x86 stuff there's at least dosbox.

maybe there's way to somewhat use the spared power and space. have a bigger battery, along with lower power that gives you more lifetime, also you can more easily afford a 2.5" hard drive. I'd be happy enough with a cheap 12" tablet (detachable keyboard. so it becomes a laptop, or I might carry a full size one if I want :p ) with huge battery life and lots of gigs.
 
HTC going Tegra ?
Not a bad partner to be with. Immediate access to both Google Android and Windows Mobile development teams could spur further adoption of the chip lineup.
 
Nvidia has said they will support, atleast initially, only Windows Mobile. I don't think they will want to support Android in the foreseeable future.
 
Nvidia has said they will support, atleast initially, only Windows Mobile. I don't think they will want to support Android in the foreseeable future.

Nvidia only has to provide the chips and a basic driver.
Android's open-source nature and its community takes care or the rest (as do the handset companies), since Tegra is essentially just another ARM11 SoC with OpenGL ES 2.0 support.
 
Nvidia only has to provide the chips and a basic driver.
Android's open-source nature and its community takes care or the rest (as do the handset companies), since Tegra is essentially just another ARM11 SoC with OpenGL ES 2.0 support.

Emphasis added.

With linux, that's the problem with NVidia. If they ship a binary only driver, they get into trouble with GPL. Open source drivers from nvidia are obfuscated and have no 3D and no video decode offloading. Eg look here.
 
Emphasis added.

With linux, that's the problem with NVidia. If they ship a binary only driver, they get into trouble with GPL. Open source drivers from nvidia are obfuscated and have no 3D and no video decode offloading. Eg look here.

That desktop software and graphics chips example does not apply here.
The basic driver doesn't need to be shared with anyone, it could be provided solely to the handset manufacturer/OS developer.
Android is not a "true" open-source effort per se, like Open Moko. It has some strict guidelines too.
 
The basic driver doesn't need to be shared with anyone, it could be provided solely to the handset manufacturer/OS developer.

It does. When the handset manufacturer ships his phones with a binary only driver, he (handset maker) is obligated to provide corresponding sources to his consumers. And AFAIK, if GPL code can't be distributed under NDA.

Android is not a "true" open-source effort per se, like Open Moko. It has some strict guidelines too.
Right, but it can't override GPL on linux kernel.
 
It does. When the handset manufacturer ships his phones with a binary only driver, he (handset maker) is obligated to provide corresponding sources to his consumers. And AFAIK, if GPL code can't be distributed under NDA.


Right, but it can't override GPL on linux kernel.

What does the Linux kernel GPL have to do with Nvidia providing a binary driver under NDA'd licenses to a specific manufacturer ?
Like on desktops, they can choose not to follow the GPL, you know ? They could use an available Linux kernel, and provide the driver that works with it. :rolleyes:
 
Last edited by a moderator:
What does the Linux kernel GPL have to do with Nvidia providing a binary driver under NDA'd licenses to a specific manufacturer ?

The kernel's GPL means that if a binary module is loaded at runtime, it is a potentially a derivative work. This potential is the operative word here. Quoting from here,

A key dispute related to the GPL is whether or not non-GPL software can dynamically link to GPL libraries. The GPL is clear in requiring that all derivative works of code under the GPL must themselves be under the GPL. However, it is not clear whether an executable that dynamically links to a GPL code should be considered a derivative work (see Weak Copyleft). The free/open-source software community is split on this issue. The FSF asserts that such an executable is indeed a derivative work if the executable and GPL code "make function calls to each other and share data structures,"[27] with others agreeing,[28] while some (e.g. Linus Torvalds) agree that dynamic linking can create derived works but disagree over the circumstances.[29] On the other hand, some experts have argued that the question is still open: one Novell lawyer has written that dynamic linking not being derivative "makes sense" but is not "clear-cut,"[30] and Lawrence Rosen has claimed that a court of law would "probably" exclude dynamic linking from derivative works although "there are also good arguments" on the other side and "the outcome is not clear"[31] (on a later occasion, he argued that "market-based" factors are more important than the linking technique[32]). This is ultimately a question not of the GPL per se, but of how copyright law defines derivative works. In Galoob v. Nintendo the Ninth Circuit Court of Appeals defined a derivative work as having "'form' or permanence" and noted that "the infringing work must incorporate a portion of the copyrighted work in some form," but there have been no clear court decisions to resolve this particular conflict.



Like on desktops, they can choose not to follow the GPL, you know ?

Even on desktops, it's not for them to choose to not follow GPL. Kernel is already GPLed. And the driver runs in kernel space(parts of it atleast). NV is very careful about the manner of distributing it's linux drivers so as to not trigger that "derivative work" requirement which may lead to a lawsuit.
 
rpg.314 is spot on, and that is nearly certainly part of NV's justification for not bothering with Android (although hardly all, or even most, of it). Read this article on the subject if you still don't believe it: http://www.eetimes.com/showArticle.jhtml?articleID=206504233

Companies like TI have it easy. Their in-house video drivers are sufficiently unappealing they've got a third-party ecosystem developed around them! (for example, the OMAP3430 doesn't support 720p with official software, but does with third party DSP algorithms). NV doesn't work like that at all, and it'll be a cold day in hell when they give out their hardware and let people develop drivers for it. It's also much more exotic and would not be so viable to this approach anyway.

IMO, and this is entirely speculative and not based on any insider info, it's more likely that they add Symbian to their OS support list than Linux. This would also allow them to be a viable provider for every single manufacturer out there; with just WM and Android, this is obviously not the case... And even if it was, it'd represent individual players' product line, I suspect, making them less likely to gain a design win for just that part with no possible reuse in other areas.
 
Well, let's hope for Nvidia that Android doesn't take off.

Isn't Symbian going open source too (for everyone in 2010 or something like that)?

Edit: I understand the driver problem, but what's Nvidia doing in the OHA if they don't want their chips to support Android? Just hanging out with the cool people?
 
Last edited by a moderator:
Yup, it's going opensource. But it's not GPLed.

Neither is Android in it's full (part of it is licensed under Apache v2).
Here's why:

Why did you pick the Apache v2 open source license?

Apache is a commercial-friendly open-source license. The Apache license allows manufacturers and mobile operators to innovate using the platform without the requirement to contribute those innovations back to the open-source community. Because these innovations and differentiated features can be kept proprietary, manufacturers and mobile operators are protected from the “viral infection” problem often associated with other licenses.

Sources:
http://blogs.zdnet.com/Burnette/?p=428
http://www.openhandsetalliance.com/android_faq.html

Look who's also part of Android's OHA: Nvidia.
 
Last edited by a moderator:
Inkster: That protects them wrt integrated software and user interfaces etc., but not drivers. AFAICT you can't do much on that front... Unless maybe you could put the drivers on a separate microSD card? I think you'd get laughed out of the room if you proposed that at a sales meeting with a major OEM though! :D If someone has any clear evidence I'm wrong here, I'd be very interested however.

Regarding being part of the OHA, the way NV looks at is they are actively watching the ecosystem and will become a full part of it if they feel it's going in the right direction. The last time I talked to them, they felt it definitely wasn't; but this was before the G1's release, and I wouldn't be surprised if their answer was very different now. One of the main problems was fragmentation, and what I heard from other places is carriers were indeed trying not to adhere to Google's vision. Now given the G1's success, they're realizing they have to.

The other point with Android and Symbian is user interface customization just isn't that good or that easy AFAIK. Windows Mobile has a lot of flaws, but at least you can customize it very effectively and it seems this will be even more true with 6.5/7.0... When your main differentiators are your multimedia capabilities, it's pretty obvious you'd prefer your chip not to be used with an user interface as plain and uninteresting (although definitely both smooth and user-friendly) as that of the G1.

They could still start supporting Android with their 40nm SoC; however, as I said, I'd be very surprised if they decided to support Android before Symbian. I wouldn't exclude the possibility they announce support for both of them at the same time though, heh - who knows! :)
 
Back
Top