NVIDIA Tegra Architecture

[ams]

Nothing of what you say proves these boards are not using engineering samples of K1. I have already seen low priced dev kits running ES.

What would be the point of using potentially hundreds if not thousands (?) of TK1 engineering samples when TK1 production is supposed to commence at this time anyway? I doubt that would be the case here. Note that first silicon for Logan was available nearly 9 months ago.

 

[Ailuros]

I don't understand that logic. What would a developer gain from a 365 GFLOPS throughput variant as opposed to the 326 GFLOPS throughput variant in the Jetson TK1 kit? Since the Jetson TK1 kit is targeted for a variety of different development use cases, it makes sense to include active cooling for those who want to really push the hardware.

If there is nothing to gain from it then why did NVIDIA announce 365 GFLOPs in the first place?

NVIDIA never said anything about Erista being a replacement for Parker, nor did they say anything about Erista using FinFET 3D transistors. The SoC formerly known as "Parker" seems to have been deconstructed (the Denver CPU goes to "Logan", the Maxwell GPU goes to "Erista", and the FinFET 3D transistors will probably go to whatever is next).

No they didn't. In fact Ken just commented somewhare that Erista was pushed in between before Parker probably meaning that Erista will be on 20SoC after all.

http://www.computerworld.com/s/arti...essor_but_delays_Volta_GPU_Parker_mobile_chip

"Erista was moved ahead of Parker. We'll provide further updates later," Brown said.

What would be the point of using potentially hundreds if not thousands (?) of TK1 engineering samples when TK1 production is supposed to commence at this time anyway? I doubt that would be the case here. Note that first silicon for Logan was available nearly 9 months ago.

The SoC needed another spin if you haven't understood that so far and I'm not sure if they've entered mass production yet or not. Nothing of what you are stating here or in former posts proves or indicates that they've entered mass production yet. Nor has "first silicon" N months before anything any relevance to it either.

 

[Helmore]

There might not be anything to indicate that they have started mass production, but is there anything to indicate that they haven't? For many chips we don't hear much about when they start entering mass production.

 

[ams]

If there is nothing to gain from it then why did NVIDIA announce 365 GFLOPs in the first place?

Presumably the 365 GFLOPS throughput is the maximum throughput for a purpose built actively cooled gaming portable such as Shield.

In fact Ken just commented somewhare that Erista was pushed in between before Parker probably meaning that Erista will be on 20SoC after all.

Seems reasonable. NVIDIA cannot afford to have more gaps in their Tegra lineup such as the gap between Tegra 3 and Tegra 4.

The SoC needed another spin if you haven't understood that so far and I'm not sure if they've entered mass production yet or not. Nothing of what you are stating here or in former posts proves or indicates that they've entered mass production yet.

And so what if the SoC needed "another spin" beyond first silicon in order to enter mass production? First silicon was ready 9 months ago as I said earlier. Commercial availability tends to be 9-12 months after an SoC is first sampling. At the end of the day, NVIDIA made it very clear at 2014 CES and MWC 2014 that TK1 (Cortex A15 variant) will have commercial availability in 1H14. Obviously this means that mass production will commence in Q214.

 

[Ailuros]

Presumably the 365 GFLOPS throughput is the maximum throughput for a purpose built actively cooled gaming portable such as Shield.

And they halted the devkit at 326 just to piss developers off?

And so what if the SoC needed "another spin" beyond first silicon in order to enter mass production?

You didn't ask what the spin is for though.

First silicon was ready 9 months ago as I said earlier. Commercial availability tends to be 9-12 months after an SoC is first sampling. At the end of the day, NVIDIA made it very clear at 2014 CES and MWC 2014 that TK1 (Cortex A15 variant) will have commercial availability in 1H14. Obviously this means that mass production will commence in Q214.

All I care about at this stage is final silicon; when current variants do not fullfill N requirements or expectations IHVs have that dumb tendency to send a chip for another spin. The point here is NOT about availability which can be any time between N weeks from now until the 30.06.14. The point is that your expectations are too optimistic. Now you either can set your expectations a tad lower and get in the best case pleasantly surprised, or worst case should be rather self explanatory.

 

[ams]

I'm not sure what you are saying. Are you saying that my expectation of seeing Tegra K1 in a commercial device in 1H 2014 (and hence mass production by Q2 2014) is too optimistic? Are you saying that my expectation of seeing Tegra K1 with 326 GFLOPS throughput in a 10" tablet and 200 GFLOPS throughput in a 5" phone is too optimistic?

And sure, by all means please tell us more about this "spin" and why NVIDIA cannot reach the above targets.

 

[Ailuros]

Are you saying that my expectation of seeing Tegra K1 with 326 GFLOPS throughput in a 10" tablet and 200 GFLOPS throughput in a 5" phone is too optimistic?

Yes.

And sure, by all means please tell us more about this "spin" and why NVIDIA cannot reach the above targets.

Why don't you ask them yourself?

 

[ams]

Yes.

Right. So what do you think the GPU clock operating frequencies will be for TK1 in a 10" tablet and a 5" smartphone?

Why don't you ask them yourself?

Ask who? And why would I care to ask them about a spin (or any insider info) in the first place? I am a consumer.

 

[ams]

I already asked, but will ask again: was it confirmed that the Jetson TK1, as opposed to the Jetson Pro, is actively cooled?

I don't know if this is truly accurate or not, but according to bit-tech and Engadget, the Jetson TK1 is passively cooled:

http://www.bit-tech.net/news/hardware/2014/03/26/nvidia-jetson-k1/1
http://www.engadget.com/products/nvidia/jetson-tk1/specs/

 

[Ailuros]

Right. So what do you think the GPU clock operating frequencies will be for TK1 in a 10" tablet and a 5" smartphone?

Obviously less then you're expecting. Unless of course you mean with 10" tablet something that weighs a lot more than usual thin tablets and has active cooling for the SoC. In a few months we'll find out.

 

[Ailuros]

I don't know if this is truly accurate or not, but according to bit-tech and Engadget, the Jetson TK1 is passively cooled:

http://www.bit-tech.net/news/hardware/2014/03/26/nvidia-jetson-k1/1
http://www.engadget.com/products/nvidia/jetson-tk1/specs/

There are links in my former post to official NV documents that clearly indicate a fansink, it's size and consumption http://forum.beyond3d.com/showpost.php?p=1836985&postcount=2159

 

[ams]

Obviously less then you're expecting. Unless of course you mean with 10" tablet something that weighs a lot more than usual thin tablets and has active cooling for the SoC. In a few months we'll find out.

I might be off the mark, but I would be very surprised if I was off by more than 10%. Yes, we will see.

 

[GpuMonkey]

You didn't ask what the spin is for though.

Confident man who critizes with correct facts keeps others honest. Confident man who critizes with incorrect fact looks like fool.

 

[Rys]

There are two Jetsons. The automotive one which has a big heatsink and fan, and the newly announced one which is passively cooled.

 

[Deleted member 13524]

There are two Jetsons. The automotive one which has a big heatsink and fan, and the newly announced one which is passively cooled.

Thank you for putting an end to the "TK1 must be horrible so they're sure to use lots of heatsinks and fans everywhere" theories.

 

[Ailuros]

Confident man who critizes with correct facts keeps others honest. Confident man who critizes with incorrect fact looks like fool.

Since you crawled out under your stone you may as well deny that there has been a spin.

 

[Helmore]

AnandTech finally clarified the difference between Jetron Pro and Jetson TK1 for me. Here is their article: http://www.anandtech.com/show/7905/...on-tk1-dev-board-adds-erista-to-tegra-roadmap
In hindsight the naming convention seems like a bit of a giveaway. Jetson TK1 is a Tegra K1 dev board, Jetson Pro is a Tegra 3 customizable building block.

 

[silent_guy]

If Jetson Pro with a Tegra 3 has a bigger heatsink than Jetson TK1 with a Tegra K1, then the only obvious conclusion here is that Tegra 3 can't possibly fit inside a phone or tablet.

Right?

 

[Helmore]

I can't spot any heatsinks on the Jetson Pro image that AnandTech has put up, there is a small fan though. Their picture of a running Jetson TK1 board does show that it has a small active heatsink.

 

[Ailuros]

If Jetson Pro with a Tegra 3 has a bigger heatsink than Jetson TK1 with a Tegra K1, then the only obvious conclusion here is that Tegra 3 can't possibly fit inside a phone or tablet.

Right?

http://images.anandtech.com/doci/7905/Jetson_Pro.jpg

That's the Jetson Pro pic in Anand's article; the other two pictures in that writeup are both from Jetson K1 one without a fansink and one with one.