While looking at all the machines listed under GRCPOOL.com I ran across an ARM CPU (android os) system with a 18,356.78 RAC!!!

https://universeathome.pl/universe/show_host_detail.php?hostid=626949

I am running several different iterations of ARM-based CPU system (not here though) and never thought I would see such a relatively large RAC on an ARM-cpu system.

So I wanted to celebrate this success!!!!

Tom M

---

A proud member of the OFA (Old Farts Assoc.)

And apparently I celebrated too soon.

Shortly after that listing are 4 MORE high-performing ARM-cpu systems using a variety of operating systems.

Maybe I need to redeploy my "best" ARM-cpu systems over here and see what happens?

Tom M

---

A proud member of the OFA (Old Farts Assoc.)

Just because it says it is running on the ARM architecture doesn't mean it is running on a 4 core phone.

There are ARM based server systems with hundreds of cores.

My best ARM based host is my Pi4 with four cores. It can only manage around 3300 RAC.

https://universeathome.pl/universe/show_host_detail.php?hostid=607777

---

A proud member of the OFA (Old Farts Association)

It does claim 8 cores. Not 32.

One of the other systems appear to be running rasberian.

Like I said. I am surprised by the RAC.

I have switched my highest performing (based on e@h RAC) arm CPU system over to u@h.

Tom M

---

A proud member of the OFA (Old Farts Assoc.)

Hi, All. My 4-core Pi-400 (running 64-bit raspian bullseye: Thank you, Keith Myers) is currently averaging ~16K credits per day. I think it could average quite a bit higher if it was running U@H only. For a $70 SBC (bought from Adafruit), I think that's pretty good for something that is so low budget. Fun! -Mike

Glad to have helped Dr. Hicks. I too am running multiple projects on my Pi4 so the Universe RAC is not maxed out.

I would have liked to get my Jetson SoC's running on Universe but alas there are no 32 bit libraries available in the Nvidia L4T distros.

I could change them over to a normal Linux distro like Ubuntu but I would lose the standard Nvidia CUDA development installation that allows me to use the Nvidia gpus at Einstein where they are quite productive.

---

A proud member of the OFA (Old Farts Association)

It sounds like my celebrating "high" RAC scores for Arm based systems was based on my ignorance of broad results of arm CPUs at u@h.

I am still looking forward to my current arm test even though I don't see a likely "arm"(s) race.
;)

Tom M

---

A proud member of the OFA (Old Farts Assoc.)

Glad to have helped Dr. Hicks. I too am running multiple projects on my Pi4 so the Universe RAC is not maxed out.

I would have liked to get my Jetson SoC's running on Universe but alas there are no 32 bit libraries available in the Nvidia L4T distros.

I could change them over to a normal Linux distro like Ubuntu but I would lose the standard Nvidia CUDA development installation that allows me to use the Nvidia gpus at Einstein where they are quite productive.

I sure hope the times come down quite a bit on my Arm CPU system. Right now they appear to be taking in the near range of one a day. Those could very well be E core processing speeds.

I am running 7 out of 8 cores.

https://universeathome.pl/universe/show_host_detail.php?hostid=634142

Tom M

---

A proud member of the OFA (Old Farts Assoc.)

I don't think they will. Those times seem on the same order of magnitude as my Kindle Fire tablet running the same application.

---

A proud member of the OFA (Old Farts Association)

ServeTheHome: The Most Important Server of 2022 The Gigabyte Ampere Altra Max and NVIDIA A100.

https://www.servethehome.com/the-most-important-server-of-2022-the-gigabyte-ampere-altra-max-and-nvidia-a100/

---

A proud member of the OFA (Old Farts Assoc.)

Hi everyone!

I did a test to see how the new Pi 5 works with U@H, and wow, the results are impressive, the new Pi 5 is very fast, comparable to old desktop CPUs that use 4-5 times the power. With the Pi 5 I got around 60 tasks per day. I published the benchmark results on my blog:

https://stfn.pl/blog/19-rpi4-rpi5-universe/

Hi everyone!

I did a test to see how the new Pi 5 works with U@H, and wow, the results are impressive, the new Pi 5 is very fast, comparable to old desktop CPUs that use 4-5 times the power. With the Pi 5 I got around 60 tasks per day. I published the benchmark results on my blog:

https://stfn.pl/blog/19-rpi4-rpi5-universe/

Thank you.

---

A proud member of the OFA (Old Farts Assoc.)

I have an ASUS Tinkerboard, similar to a Pi 3/4 with a 4-core arm CPU.
I use it exclusively on Universe at home. WCG used to give me some WU's and other projects don't seem to give me enough time to to complete them
I only keep it on for about 6 hours a day and just use it for tinkering with random projects

It takes about 7h's to process one Universe task and I do 3 at a time

I have an ASUS Tinkerboard, similar to a Pi 3/4 with a 4-core arm CPU.
I use it exclusively on Universe at home. WCG used to give me some WU's and other projects don't seem to give me enough time to to complete them
I only keep it on for about 6 hours a day and just use it for tinkering with random projects

It takes about 7h's to process one Universe task and I do 3 at a time

What happens if you run a single task? Does the processing speed up/slow down?

Thank you for your previous post.

Tom M

---

A proud member of the OFA (Old Farts Assoc.)

What happens if you run a single task? Does the processing speed up/slow down?

Thank you for your previous post.

Tom M

Interesting idea. I work FIFO so once I'm back next week I'll give it a try

So my Tinkerboard users a four independent cores so just running one task is done at the same speed as doing four tasks

As long as you aren't thermal or TDP throttling running four tasks.

---

A proud member of the OFA (Old Farts Association)

I have a 32-bit ASUS Tinkerboard (RK3288), cooled with an 52Pi Ice Tower low profile cooler, that manages between 7,800 and 9,100 credits/day for Universe@Home, running 24/7 on all cores.
I also have an Odroid-XU4 that manages the same daily scores, this while doing SiDock@Home on the side. It is cooled with a Xigmatek N881 Porter and 80mm Noctua NF-A8 5V fan.
My best scoring ARM system presently is a 8GB Odroid-M1, recently converted to be able to run 32-bit tasks by adding the 32-bit libraries to its 64-bit Debian Bookworm
It is still climbing in its daily scores and is almost at the 9,000 daily credits and needs no additional cooling due to its huge heatsink.

I would have liked to get my Jetson SoC's running on Universe but alas there are no 32 bit libraries available in the Nvidia L4T distros. I could change them over to a normal Linux distro like Ubuntu but I would lose the standard Nvidia CUDA development installation that allows me to use the Nvidia gpus at Einstein where they are quite productive.

Both my Jetsons crunch Universe and run L4T (the 4GB model upgraded to 20.04.6 LTS with Glibc 2.31). I got some warnings installing 32-bit libraries, but it is workable.
The 2GB model even has a both OpenCL and CUDA capable GPU, but still runs on 18.04.6 LTS and has Glibc 2.27

The new Pi5 is my fastest ARM device, faster than my Odroid-N2+.

CPU: BCM2712
OS: Linux Debian Debian GNU/Linux 12 (bookworm)
Results fetched: 100
Average duration (s): 5652.4
Average credit: 666.67
Number of reported cores: 4
Per core per day: 10133
Per system per day: 40532

CPU: Hardkernel ODROID-N2Plus
OS: Linux Ubuntu Ubuntu 20.04.1 LTS
Results fetched: 100
Average duration (s): 13930.1
Average credit: 666.67
Number of reported cores: 6
Per core per day: 4133
Per system per day: 24798