NVIDIA Tegra Architecture

Oh, seems I didn't think this through enough.
I sort of assumed you'd type "lspci" and see the nvidia GPU show up. Works on AMD and Intel hardware, afterall but as for a Tegra K1, I don't know. Some hardware will be behind PCIe and USB so it's discoverable (such as user input, at least)

Discoverability of hardware is a topic I'm slightly curious about. Might vary between specific tablets/computers/embedded platforms too.

They can but they don't have to, and when it comes to SoC specific stuff they generally don't.

Sure. You have things like Ubuntu announcing support for POWER8 and ARM versions of Debian and Ubuntu though, but the former is not SoC related and most often the disk images or such are provided "unofficial". The unofficial stuff can be well good enough I think.
 
Sure. You have things like Ubuntu announcing support for POWER8 and ARM versions of Debian and Ubuntu though, but the former is not SoC related and most often the disk images or such are provided "unofficial". The unofficial stuff can be well good enough I think.

Right, the ISA support comes from the distro. They have to maintain their repositories with binary support for the ISA. That sort of thing will (should?) work regardless of the SoC.

But someone else has to then configure the distro for the SoC. A kind of distro distro I guess?
 
I am now having doubts that the Nexus 9 will even have a Nvidia SOC.

With the HTC Desire 510 containing the "Qualcomm’s first 64-bit chip – the 1.2GHz Snapdragon 410" why would HTC ditch Qualcomm for the Nexus 9. Why not just use the Snapdragon 810.

http://www.forbes.com/sites/gordonkelly/2014/08/27/iphone-6-ipad-nexus-android-rival

The whole premise that Nvidia was going to win the Nexus 9 was that it was going to be an ARM 64bit SOC and many thought that only Nvidia would have a 64bit ARM SOC available in the time frame. It now looks like Qualcomm is also going to be available.

Snapdragon 810 is still sampling and is expected to be in devices that are shipped during the first half of 2015. Even then, the expectation is that the single-threaded CPU performance and the GPU performance of TK1-Denver will be even better than S810.

Note that the hardware inside HTC's new midrange phone has absolutely nothing to do with the hardware inside HTC's new tablet (the design requirements are very different). HTC has worked with NVIDIA in the past with the international HTC One X and One X+ models.
 
Snapdragon 810 is still sampling and is expected to be in devices that are shipped during the first half of 2015.

Well if the 410 (Cortex A53) is going to be available soon what reason is there that the 810 (Cortex A57) would not be far behind?

After all these are not Krait cores just cookie cutter standard cores.

Even then, the expectation is that the single-threaded CPU performance and the GPU performance of TK1-Denver will be even better than S810.
That logic didn't stop Google from going with the Snapdragon S4 Pro instead of the Tegra 4 in the Nexus 7 2013.

Cost also matters and QualComm has been known to undercut competitors on price just to keep them from gaining market share.
 
Well if the 410 (Cortex A53) is going to be available soon what reason is there that the 810 (Cortex A57) would not be far behind?

S810 is significantly more advanced than S410 with respect to fab process node (as mentioned above by Exophase), CPU core, GPU core, memory interface, modem core, multimedia processing, etc. And S410 is also coming to market first with Android 4.4 KitKat, not Android "L".

That logic didn't stop Google from going with the Snapdragon S4 Pro instead of the Tegra 4 in the Nexus 7 2013.

Tegra 4 was late to market and came after S4 Pro. On the other hand, TK1 32-bit and 64-bit variants are not late to market and will come before S810.

Cost also matters and QualComm has been known to undercut competitors on price just to keep them from gaining market share.

S810 should be more costly than Tegra K1 due to the more advanced fab. process node and due to the cost of the LTE modem built into the price of the SoC. The most cost-effective tablets tend to be models without built-in LTE modem.
 
Snapdragon 410 and 810 are not the only horses in the race.

The Snapdragon 615, a midrange SoC will be available in Sept. also:
http://www.slashgear.com/snapdragon-615-qualcomms-64-bit-octa-core-race-horse-27342960/

the HTC 820 will use this SoC.

And then there is the Samsung Exynos 5433, which is said to be a 64bit SoC also and coming in Sept. too (Samsung Note4).

So Nvidia better hurry up...otherwise the Denver-K1 will the the last of the upcoming 64bit ARM-Cores and not the first. :)
 
Last edited by a moderator:
That video is from TK O'Connor.
Here's what Evan Blass from evleaks had to say when he resigned the "leaking" business:
TNW: Are there any lesser-known leakers out there with the potential to step into your shoes?

EB: Someone who is immediately springs to mind is TK O’Connor, of TK Tech News. The man is a beast when it comes to leaking — and body checks — but he’s also the nicest guy you’ll ever meet.

He's not just linking to a video he didn't make. He took the screenshot and claims he's had the tablet in his hand and ran all kinds of benchmarks in it, which he's going to share when he leaves China and doesn't need to pay enormous amounts of roaming charges.

The HTC "Flounder" has been rumoured to be carrying TK1 Denver. Everything but the revision number in that CPU-Z seems to be spot on. The revision may not be in the CPU's microcode and it could be something that the CPU-Z doesn't read but just "assumes" as being a 32bit TK1 variant.
 
Even that CPU Architecture says "192-core Kepler" ?

That just adds to it being a FAKE.

As Metalhead2550 states:

there are a lot of discrepancies on that screen shot I would try to check them out because this sounds a bit fishy! CPU Architecture is not 192 Core Kepler, Its ARMv8 and GPU Renderer should not mention Denver because that's the CPU that's where it should say 192 Core Kepler!

https://www.youtube.com/watch?v=7W_utZdw5RY&feature=youtu.be
 
He's not just linking to a video he didn't make. He took the screenshot and claims he's had the tablet in his hand and ran all kinds of benchmarks in it, which he's going to share when he leaves China and doesn't need to pay enormous amounts of roaming charges.

Okay, let's say it came from a legitimate tablet and all, and that the leaker is honest. That still doesn't mean that the tablet was running an unmodified CPU-Z.

The HTC "Flounder" has been rumoured to be carrying TK1 Denver. Everything but the revision number in that CPU-Z seems to be spot on. The revision may not be in the CPU's microcode and it could be something that the CPU-Z doesn't read but just "assumes" as being a 32bit TK1 variant.

The revision comes from the MIDR register which the kernel will report if you read /proc/cpuinfo. I doubt CPU-Z is doing it any other way. So it should only be writing r3p3 if that's what the processor reports. Denver could legitimately report this but I doubt it, that's an awful lot of revisions for something that hasn't been in a product yet... And it's not like they have any kind of incentive to use the same thing as K1-32.

None of this means that there's no Flounder with Denver coming out soon, just because the screenshot looks suspicious. Everything else being spot on vs rumors and information already circulated doesn't mean anything.
 
1. The second core shows as "stopped". Hotplugging doesn't exist anymore in newer SoCs, so that's bullshit.
2. The r3p3 revision matches the A15 revision number
3. The already mentioned architecture name is bogus
4. The GPU renderer is bogus
5. The scaling governor "HYPER" is bullshit. Not only don't OEMs rename their governors to such names, but Hyper was actually some community renamed governor based on some other governor.
 
could NV be using the r3p3 for optimum code paths for applications?

Not saying the "leak" is real, but just like setting an AMD chip to be an intel chip so ICC code takes AVX/SSE3/4 etc path instead of SSE2.
 
could NV be using the r3p3 for optimum code paths for applications?

Not saying the "leak" is real, but just like setting an AMD chip to be an intel chip so ICC code takes AVX/SSE3/4 etc path instead of SSE2.

To elaborate, the version/revision number isn't an architecture caps string, it's purely a microarchitecture denotation. It'd be like ICC looking at the stepping number. The only diverging code paths based on this number would be for handling errata, and that'll almost always be done by the OS.

The whole thing would be blown anyway if they're not also reporting that they're a Cortex-A15, in which case that's what CPU-Z would report.
 
Sort of related to Tegra

NVIDIA NB106-N baseband module

NVIDIA NB106-N



  • Device Type:
  • Chipset / Module - NGFF

  • Network Type:
  • LTE / GPRS / EDGE / HSDPA 3.6 / HSDPA 7.2 / HSPA+ 14.4 / HSPA+ 21.1 / LTE CAT1 / LTE CAT2 / LTE CAT3

  • Release Date:
  • Q2-2014

  • Description:
  • The NVIDIA NB106-N is certified for use on the AT&T Network.

http://developer.att.com/developer/device_detail.jsp?id=f001f100-6dce-439f-953f-bb5b17a865b5
Google's Project Tango Tablet: From the Outside In

Notable ICs on the baseband module:
  • Nvidia ICE9245B-C2 low-power, multi-mode RF transceiver (red)
  • Micron 3PA98 JW806 NAND-based MCP (orange)
  • Nvidia Icera ICE9045T-A2 LTE Modem (yellow)
  • Texas Instruments TPS659121A PMU for Processor Power (green)
  • Skyworks 77621-11 Multimode Multiband Power Amplifier Module (blue)

http://www.designnews.com/author.as...umer,aid_274501&doc_id=274501&image_number=21
Project-Tango-Tablet-21.jpg
 
Back
Top