rebirtherVolunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 2 Jan 13
Posts: 7733
Credit: 44,382,238
RAC: 11,440
Thx to Ian who made this possible!
Both apps are now available for cuda10.2 and 12.2
/ 
Reply
Quote
Thank You!
to anyone wondering these apps mainly target things like Nvidia Jetson devices, but also should work on Arm64 desktop platforms with dedicated Nvidia GPU.
the CUDA 10.2 app supports CC 5.0 - 7.5 (including the 5.3/6.2/7.2 for jetsons), and will work on Ubuntu 18+/glibc 2.27+
the CUDA 12.2 app supports CC 5.0 - 9.0 (including the 7.2/8.7 for jetsons), and will work on Ubuntu 20+/glibc 2.31+
/ Reply
Quote
What are the prerequisites, besides having a Jetson Nano and/or Xavier running L4T (Linux-Ubuntu for Tegra)? And what is CC supposed to mean?
Last L4T version for the Jetson Nano is 32.7.4, last JetPack 4.6.4
(see https://jetsonhacks.com/jetpack-and-jetson-linux-l4t-versions/)
/ Reply
Quote
What are the prerequisites, besides having a Jetson Nano and/or Xavier running L4T (Linux-Ubuntu for Tegra)? And what is CC supposed to mean?
Last L4T version for the Jetson Nano is 32.7.4, last JetPack 4.6.4
(see https://jetsonhacks.com/jetpack-and-jetson-linux-l4t-versions/)
the latest L4T version for Nano is actually 32.7.6 with JetPack 4.6,6, that's what I have. look at the nvidia site, not a third party.
https://developer.nvidia.com/embedded/jetson-linux-archive
CC means compute capability, it roughly defines the Nvidia GPU generations and capabilities of the GPU hardware. it's a metric used in compilation so that the code can be optimized and enabled for that specific GPU hardware. my comment about CC versions included was just to say that it's compatible with the CC version for the jetsons, which does differ from the CC version used on PCIe cards
for the Nano TX1/TX2/Xavier, the prereqs are that you need the CUDA 10.2 (or newer) drivers and Ubuntu 18.04 (or newer). jetpack 4.6.6 has both of those. I dont know what drivers are in jetpack 4.6.4, a quick google search says you have cuda 10.2 drivers so you're probably good.
for Xavier/Orin, which can run newer drivers. you need either of the following configurations:
Jetpack 5 + upgrade the drivers to CUDA 12.2 (nvidia enabled the ability to upgrade driver package on Jetpack 5+ starting with CUDA 11.8)
Jetpack 6, which comes with CUDA 12.x drivers already.
do you have many of these Jetson devices? I can't see your computers list (hidden). if you're interested in other projects, I have also ported optimized CUDA applications for Einstein BRP7-Meerkat (not just BRP4, all jetson generations, Nano TX1 and up) and Asteroids (only JP5/6 Xavier/Orin devices). Send me a PM if interested in these and I can set you up.
/ Reply
Quote
My 4GB Jetson Nano is crunching away on a TF atm. Only he has downloaded 80 of them, with an estimated runtime of 4 seconds each while he has been busy with WU 1 for more than 7 hours now.
/ Reply
Quote
rebirtherVolunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 2 Jan 13
Posts: 7733
Credit: 44,382,238
RAC: 11,440
My 4GB Jetson Nano is crunching away on a TF atm. Only he has downloaded 80 of them, with an estimated runtime of 4 seconds each while he has been busy with WU 1 for more than 7 hours now.
This is a boinc server issue with new apps and updates, the server has no datas to work with, 0 estimated time = wrong progress bar, should be better if you send some results back.
/ Reply
Quote
My 4GB Jetson Nano is crunching away on a TF atm. Only he has downloaded 80 of them, with an estimated runtime of 4 seconds each while he has been busy with WU 1 for more than 7 hours now.
they will take about 19hrs to complete. on the Jetson Nano TX1
/ Reply
Quote
My 4GB Jetson Nano is crunching away on a TF atm. Only he has downloaded 80 of them, with an estimated runtime of 4 seconds each while he has been busy with WU 1 for more than 7 hours now.
they will take about 19hrs to complete. on the Jetson Nano TX1
69,390.75 sec. = 1,156.5125 min. = 19 hr 27 min
CPU type: ARMv8 Processor rev 1 (v8l) [Impl 0x41 Arch 8 Variant 0x1 Part 0xd07 Rev 1]
Aantal processoren: 4
Extra processoren: NVIDIA NVIDIA Tegra X1 (3964MB)
OS: Linux Ubuntu 20.04.6 LTS [4.9.253-tegra|libc 2.31 (Ubuntu GLIBC 2.31-0ubuntu9.16)]
BOINC version: 7.16.6 (due to old kernel)
/ Reply
Quote