Project Information

Forum rules
User avatar
Alez
[ TSBT's Pirate ]
[ TSBT's Pirate ]
Posts: 10363
Joined: Thu Oct 04, 2012 1:22 pm
Location: roaming the planet

#1 Project Information

Post by Alez »

Project pages are here

Amicable Numbers is an independent research project that uses Internet-connected computers to find new amicable pairs. You can contribute to our research by running a free program on your computer.

Current goal of the project is to find all amicable pairs with smallest member < 264.
All new findings are published regularly on the Amicable pairs list page.
Image
The best form of help from above is a sniper on the rooftop....
Miklos M
Butterfly Whisperer / Milkweeder
Butterfly Whisperer / Milkweeder
Posts: 1452
Joined: Fri Aug 14, 2020 3:39 pm
Location: Florida, USA

#2 Re: Project Information

Post by Miklos M »

Just looking at the stats in Ami Numbers and looks to me like we can pick up a few points here also. I will give it go.
Image
davidbam
General Bitchin'
General Bitchin'
Posts: 6371
Joined: Wed Aug 15, 2018 1:15 pm
Location: Huntly, Scotland
Contact:

#3 Re: Project Information

Post by davidbam »

remember 8 or 9Gb system RAM per GPU WU - say 25Gb in your triple GPU system
I think this is fool-proof but could you just try it for me please? • There are 10 types of people in the world; those who understand binary, and those who don’t
Image
User avatar
Dirk Broer
Corsair
Corsair
Posts: 1962
Joined: Thu Feb 20, 2014 11:24 pm
Location: Leiden, South Holland, Netherlands
Contact:

#4 Re: Project Information

Post by Dirk Broer »

So, 8 or 9Gb system RAM per GPU WU -but how about the GPU's own RAM? I ask this because I found out that no matter how much RAM you have onboard, no matter what OS you are running, no matter what brand of GPU you have or whether it runs OpenCL or CUDA, you do not get more than 4GB 'available RAM'.
Isn't that strange? A reminiscence of old days 32-bit coding, a 4GB limit of RAM?
Image
User avatar
scole of TSBT
Boinc Major General
Boinc Major General
Posts: 5980
Joined: Mon Feb 03, 2014 2:38 pm
Location: Goldsboro, (Eastern) North Carolina, USA

#5 Re: Project Information

Post by scole of TSBT »

A lot of coders use int instead of long until they realize they need a long.
Image
User avatar
Dirk Broer
Corsair
Corsair
Posts: 1962
Joined: Thu Feb 20, 2014 11:24 pm
Location: Leiden, South Holland, Netherlands
Contact:

#6 Re: Project Information

Post by Dirk Broer »

But which coder to blame for the use of INT where LONG should/can be used? At Berkeley they think it's the (AMD and Nvidia) drivers, but as it affects all OS-es, under both CUDA (for Nvidia) and OPenCL (for both), I suspect that it is the Berkeley code itself that is responsible for the hated '4 GB available'.

An example: the top-7 computers for PrimeGrid have, besides their dual 24-core Xeons Gold 6248R, each 10 [!] NVIDIA Quadro RTX 6000, each with 4095 MB of RAM available.
Each NVIDIA Quadro RTX 6000 should have 24 GB of RAM...
Image
davidbam
General Bitchin'
General Bitchin'
Posts: 6371
Joined: Wed Aug 15, 2018 1:15 pm
Location: Huntly, Scotland
Contact:

#7 Re: Project Information

Post by davidbam »

Collatz seems to get around it
I think this is fool-proof but could you just try it for me please? • There are 10 types of people in the world; those who understand binary, and those who don’t
Image
User avatar
Dirk Broer
Corsair
Corsair
Posts: 1962
Joined: Thu Feb 20, 2014 11:24 pm
Location: Leiden, South Holland, Netherlands
Contact:

#8 Re: Project Information

Post by Dirk Broer »

davidBAM wrote: Tue Feb 02, 2021 12:32 pm Collatz seems to get around it
I see two examples here, one for a Vega 64 and one for a Vega 56, but take a look at #2: [3] NVIDIA Tesla V100-PCIE-32 GB (available: 4095 MB.....)

BTW: the two exceptions are both running Darwin OS....
Image
User avatar
Dirk Broer
Corsair
Corsair
Posts: 1962
Joined: Thu Feb 20, 2014 11:24 pm
Location: Leiden, South Holland, Netherlands
Contact:

#9 Re: Project Information

Post by Dirk Broer »

Update as to the project information, as the header still says 'CPU project for Windows, Linux'.

This is no longer the case:
Amicable Numbers up to 10^21
PlatformVersionVersion DateAverage Computation
Microsoft Windows running on an AMD x86_64 or Intel EM64T CPU3.02 (opencl_amd)30 Oct 2019, 16:39:57 UTC88,301 GigaFLOPS
Microsoft Windows running on an AMD x86_64 or Intel EM64T CPU3.02 (opencl_nvidia)30 Oct 2019, 16:39:59 UTC761,394 GigaFLOPS
Microsoft Windows running on an AMD x86_64 or Intel EM64T CPU3.03 (mt)9 Aug 2020, 12:30:27 UTC50,295 GigaFLOPS
Linux running on an AMD x86_64 or Intel EM64T CPU3.00 (mt)18 Oct 2019, 18:20:09 UTC10,604 GigaFLOPS
Linux running on an AMD x86_64 or Intel EM64T CPU3.02 (opencl_amd)30 Oct 2019, 16:39:58 UTC2,549 GigaFLOPS
Linux running on an AMD x86_64 or Intel EM64T CPU3.02 (opencl_nvidia)30 Oct 2019, 16:39:56 UTC120,124 GigaFLOPS
Intel 64-bit Mac OS 10.7 or later3.00 (mt)18 Oct 2019, 18:20:05 UTC6,396 GigaFLOPS
Intel 64-bit Mac OS 10.7 or later3.02 (opencl_amd)30 Oct 2019, 16:39:56 UTC27,297 GigaFLOPS
Intel 64-bit Mac OS 10.7 or later3.02 (opencl_nvidia)30 Oct 2019, 16:40:00 UTC724 GigaFLOPS
Linux running on 64-bit ARM (ARMv8-A architecture)3.00 (mt)18 Oct 2019, 18:20:11 UTC175 GigaFLOPS
I would like to warn against the Linux running on 64-bit ARM (ARMv8-A architecture) app though: It needs more than 2GB of system RAM for starters, and a 4GB Jetson Nano (ARM Cortex-A57) chokes on it in just a few seconds. My recommendation: A Jetson Xavier AGX might do, with 32GB of RAM.
Image
Post Reply Previous topicNext topic

Return to “Amicable Numbers”