NVIDIA's Jetson AGX Xavier Carmel Performance vs. Low-Power x86 Processors

[coder]

I am sorry, but nothing that you say justifies the price demanded by NVIDIA.

I think it's not a single aspect, but the aggregate of features that makes so expensive.

The fact that the board is intended for automotive applications does not justify the high price,

The automotive spec certainly adds cost.

https://blogs.nvidia.com/blog/2016/0...omotive-grade/

What is of course more difficult to match is the 512-core Volta GPU. I am not aware of any alternative with NVIDIA GPUs, but there are several choices with sizes from NUC up to Nano-ITX, which have comparable AMD Vega or Polaris GPUs at prices 2 to 3 times lower than Xavier.

You're missing the point, which is its tensor cores. Find an automotive-grade embedded computer with 22 int8 TOPS or 11 fp16 TFLOPS in 30 W. Oh and add 16 GiB of RAM with 137 GB/sec bandwidth.


Finally, independent of the GPU block, it has an ISP and Programmable Vision Accelerator.

https://www.anandtech.com/show/13584/nvidia-xavier-agx-hands-on-carmel-and-more

(my second time posting this link in the thread - I guess you didn't see it the first time)

[coder]

You've got quite a lot of opinions for the little information you know about this product.

I agree that for now there is no computer at this size that offers a PCIe 4.0 interface.
This might become important in the next years, but for the moment there exists nothing useful that could be connected to it.

PCIe 4.0 connectivity is ultimately for this:

https://www.anandtech.com/show/11913/nvidia-announces-drive-px-pegasus-at-gtc-europe-2017-feat-nextgen-gpus

If we take a much faster computer at half price, I (being a designer of embedded computer boards) am certain that I could design an add-in board (probably on PCIe), which would add the 16 camera inputs at a small fraction of the price difference between Xavier and that computer.

With an aggregate bandwidth of 109 Gbps? Okay.

Power consumption: not public, but equal or larger than that of an Intel NUC.

It is public.

https://wccftech.com/nvidia-drive-xavier-soc-detailed/

Who needs a faster GPU, can use either a Crimson Canyon NUC with a 512-core Polaris GPU.

No, that's not even close to the same ballpark.

[vegabook]

Yes. The system is much too expensive for your apparent type of use. So don't buy it? Nobody is forcing you.
You still seem to think that the target audience is Generic Joe for his desktop replacement needs.
"Devboards" like this can run several thousand dollars.

Show me ONE system with 16x MIPI CS2, PCIe 4.0, a cutting edge embedded CPU and GPU with this form factor and TDP for less than $1000.
Heck even a Matrox with a couple of SDI interfaces is several thousand dollars.
Or show me a x86 solution with the same TDP and form factor that will beat this solution costing your claimed "third of the price".

All that stuff should be on daughterboards. Nvidia's major problem is it doesn't have a CPU offering, just as AMD is resurgent (offering both) and just as Intel is about to put zillions behind its own GPU offering. Nvidia seriously needs to credibilize itself on CPUs but then prices its lame offering out of the market.

It's brilliant if you're Tesla, It's terrible if you're Joe-The-Hacker who would be instrumental in increasing your visibility/share.

[coder]

All that stuff should be on daughterboards.

Why? So you can have a cheap, ARM-based computer with Nvidia graphics?

Just for that, they can't integrate those functions into their SoC, and instead have to add a bunch more PCIe lanes and make a much bigger board?

You need to understand this: they did not build Xavier for people like you. Their Tegra product line was all about cell phones, tablets, chromebooks, and similar. That didn't work out so well for them, and they've found a more lucrative niche with a higher barrier to entry. They have ended the Tegra branding (actually, the X2 was their first post-Tegra SoC), so we're now in a post-Tegra world.

AdrianBc has outlined a number of good options for you. They happen to be x86-based, but maybe a couple years will change that. And remember that, for generic graphics tasks, Xavier's GPU is probably a little slower than a GTX 1050. A 7 nm AMD APU, or maybe even Intel's Gen11 graphics should mostly close that gap (any APU would still have < half the memory bandwidth).

Nvidia's major problem is it doesn't have a CPU offering,

This is their 3rd generation of custom ARM cores. That said, performance-wise, they'd have been better off if they just used a Cortex A76. Although, that didn't exist when they started Xavier, and probably building their own 64-bit core got them an automotive-compliant chip to market sooner.

It's brilliant if you're Tesla, It's terrible if you're Joe-The-Hacker who would be instrumental in increasing your visibility/share.

What about Joe the robot builder, who needs these features and processing power?