Posts by MarkJ

1) Message boards : Number crunching : Don't know where it should go? Stick it here! (Message 1940883)
Posted 2 days ago by Profile MarkJ Special Project $250 donor
Post:
AMD are offering the first 40 winners in the US to swap their Intel i7-8086 with an AMD epyc. See article here
2) Message boards : Number crunching : Power9 anyone? (Message 1940599)
Posted 4 days ago by Profile MarkJ Special Project $250 donor
Post:
Raptor Computing have a Talos II Lite which is somewhat cheaper. Link to the Phoronox artice here.

Its a single CPU EATX machine in a tower case with power supply for $1399 (USD) and then you buy a CPU and memory.

The Power9 chips have 4 or 8 threads per core, so the 22 core CPU gives 88 threads. Expected to ship in July 2018.
3) Message boards : Number crunching : Power9 anyone? (Message 1940049)
Posted 7 days ago by Profile MarkJ Special Project $250 donor
Post:
IBM have an AC922 compute server using their latest Power9 processors. It can take up to 2 CPUs and they have a couple of different models, up to 22 cores with 4 threads per core. It also supports PCIe gen4 and has a 120MB level 3 cache.

Price tag looks like $5100 (USD) for the system, then $2999 for the base model CPU or $3999 for more cores. Link here

There is also the Talos Raptor II system reviewed by Phoronix. Its only a 16 core/64 thread machine. Link here

They both look quite inpressive as compute servers. You can get an open source Linux for it and then you’d have to compile the Seti apps and threads galore.
4) Message boards : Number crunching : Boinc 7.10.2 problem on Linux (Message 1937355)
Posted 29 days ago by Profile MarkJ Special Project $250 donor
Post:
I have seen issues with boinccmd which go away if you cd to /etc/boinc-client. From memory they seem to have been fixed with a kernel update, even though it shouldn’t have anything to do with it. I think 4.15 or 4.16 work. What kernel and linux flavour are you using?

I assume your gui_rpc_auth.cfg has the same password for both local and remote computers and the machine making the request is listed in remote_hosts.cfg on the remote one.

Can you use BOINCtasks to query the remote host? If the local machine is a Linux one you might need to run it under wine.
5) Message boards : Number crunching : Back into the party! (Message 1934819)
Posted 10 May 2018 by Profile MarkJ Special Project $250 donor
Post:
Now Pi-1 - shows all it's tasks, both completed and in-progress, but Pi-2 shows no tasks at all, but it is currently happily working on 3 tasks. Also, Pi-1 shows 3 abandoned tasks in it's listing for May 9 (today). I never abandoned any tasks today for any computer.

Does this make any sense to anyone? I wonder if the Seti software that adds computers has a weird bug?

That sounds like its got the same host id as the other one. Did you copy the SD card from one to make the other? You need to have each one start off with a clean BOINC data directory if possible. When the tasks finish exit boinc-client and nuke the client_state.xml and client_state_prev.xml files then start it up again. It will recreate them and should assign a new host id.

I have seen problems with Seti giving me new hostid's when it should retain the old one (the opposite of what you have). Its supposed to use the mac address to assign a hostid. I have heard of it happening when people use the internet sharing feature in windows.
6) Message boards : Number crunching : Back into the party! (Message 1934401)
Posted 8 May 2018 by Profile MarkJ Special Project $250 donor
Post:
Powering the Pis
I use a number of 5 port USB chargers to run them and a couple of official power adapters as well. The USB chargers are technically rated at 3 x 2A plus 2 x 1A output but I have successfully run 4 Pi3's off a single one plus the fans on the top of the case using the 5th port. They're Astrotek AT-UPS-008 or 008B.

Cooling them
One solution that our team founder over at Einstein uses is to stack them and have a couple of 80mm fans on one side of the stack. He has made a few videos about his Bramble as they're called. You can see the video here
7) Message boards : Number crunching : Back into the party! (Message 1934176)
Posted 7 May 2018 by Profile MarkJ Special Project $250 donor
Post:
You can get the temp from the Pi via vcgencmd:
vcgencmd measure_temp

The Pi3's need active cooling as they get quite hot under load. I have a few running Einstein work. I designed a 3D printed case with room for 4 of them side by side and fans on top. Pictures here
8) Message boards : Number crunching : Building on Arm Linux (Message 1930678)
Posted 18 Apr 2018 by Profile MarkJ Special Project $250 donor
Post:
Have you read https://setiweb.ssl.berkeley.edu/beta/forum_thread.php?id=2275?

As you will see we had an ARM app done with the help of Claggy and Tom Rinehart so the existing one should simply work out of the box and use neon if the CPU supports it. They may be able to assist you with compiling your own version.
9) Message boards : Number crunching : Debian: Lightdm stopped working after kernel update (Message 1930052)
Posted 15 Apr 2018 by Profile MarkJ Special Project $250 donor
Post:
Last night I got an updated kernel from stretch-backports (4.15). I was running the 4.14 from there without any issues. On my boxes with Nvidia drivers they all refused to start up the desktop complaining about lighdm has stopped working. My intel-GPU machines were fine so it seems to be the combination of the Kernel + Nvidia drivers

After reinstalling one box I discovered an easier way assuming you can login to the machine by some other method (ssh).

After it reboots and the desktop gets stuck with the lightdm failure message, login via ssh to it. As root remove the nvidia-kernel-dkms and reinstall it:

apt remove nvidia-kernel-dkms -y
apt install -t stretch-backports nvidia-kernel-dkms linux-headers-amd64 -y
nvidia-xconfig

Reboot and lighdm should now work. You probably don't need to redo the nvidia-xconfig but that's what I normally do after a Nvidia driver update.
10) Message boards : Number crunching : What happened to Hard Drive Quality????? (Message 1925962)
Posted 23 Mar 2018 by Profile MarkJ Special Project $250 donor
Post:
Laptop HDDs tend to fail more often due to the laptop moving around. I replaced the mechanical drive in my current laptop with a SSD. Apart from being faster there is no moving parts so less chance of failure as a result. Then it all comes down to the quality of the electronics, pay a bit more for a decent one.

I only have WD hard drives in my rigs. Haven’t had any issues with them in the last 5 years or so. If a cruncher died I would swap out the drive and reinstall it. For a daily driver you need regular backups of it, and you only realise how good your backup plan is (or lack of) when it dies. Heck I even backup my file server which has enterprise grade drives on a raid controller.
11) Message boards : Number crunching : Anyone still using a raspberry pi 3 (Message 1924948)
Posted 16 Mar 2018 by Profile MarkJ Special Project $250 donor
Post:
Here are the BOINC benchmark results from a 3B and the 3B+

Rpi 3B
Number of CPUs: 4
975 floating point MIPS (Whetstone) per CPU
15364 integer MIPS (Dhrystone) per CPU

Rpi 3B+
Number of CPUs: 4
1123 floating point MIPS (Whetstone) per CPU
17834 integer MIPS (Dhrystone) per CPU

Over on Einstein I am seeing run times of approx 42,000 seconds for the 3B and 37,000 for the 3B+
12) Message boards : Number crunching : Anyone still using a raspberry pi 3 (Message 1924866)
Posted 16 Mar 2018 by Profile MarkJ Special Project $250 donor
Post:
For those that don't know the Raspberry Pi Foundation released the Pi 3 model B+ on the 14th of March. Its a bit faster (1.4GHz instead of 1.2GHz) and has better networking than the 3B (330 Mbit/sec plus dual-band WiFi). It also supports power over ethernet (PoE) by way of a PoE HAT module. Memory is still the same 1GB of DDR2.

13) Message boards : Number crunching : Linux CUDA 'Special' App finally available, featuring Low CPU use (Message 1922044)
Posted 2 Mar 2018 by Profile MarkJ Special Project $250 donor
Post:
Hi Mark, all I can say is WOW! So I guess I can chalk my low integer scores up to my kernel. Are those 4.14.0-0.bpo.3-amd64 kernels special just for Ryzen? Or just the normal kernels for Debian Stretch?

Its the standard kernel from Debian, currently in stretch-backports. I was using the 4.9 kernel before and getting similar numbers.

Machines are standard Ryzen 1700's, nothing OC'ed. Memory is DDR4 2400MHz. BIOS is one version behind their current offering (they're ASUS Prime X370-Pro motherboards). I just take the optimised defaults and adjust fan settings whenever I update the BIOS.
14) Message boards : Number crunching : Linux CUDA 'Special' App finally available, featuring Low CPU use (Message 1922024)
Posted 2 Mar 2018 by Profile MarkJ Special Project $250 donor
Post:
After your question, re: boinc manager benchmark score, I looked again at my "reported" integer speed of 82+ Gips. Something is odd there. It is reported as "speed per core" but it seems to me unrealistically high.

@Gene, I get this on my Ryzen 1700's. All four of them are pretty much the same.
Starting BOINC client version 7.8.4 for x86_64-pc-linux-gnu	
Processor: 16 AuthenticAMD AMD Ryzen 7 1700 Eight-Core Processor [Family 23 Model 1 Stepping 1]		
Benchmark results:	
 Number of CPUs: 16	
 4139 floating point MIPS (Whetstone) per CPU	
 50370 integer MIPS (Dhrystone) per CPU	

Although when the first version of 7.9.2 came out it was getting 100,000 integer. The second version had it back to 50,000 so they probably tweaked a compiler flag.
15) Message boards : Number crunching : Multiple Different Seti Version Running on Same Machine - Why? (Message 1912856)
Posted 13 Jan 2018 by Profile MarkJ Special Project $250 donor
Post:
I expect BOINC is trying the different apps to find out which is fastest for each machine.

You have your computers hidden so you’ll have to login to the Seti website go to your account, click on view computers and then for each one details and then applications. Under there it will list each app and a figure for APR (Average Processing Rate). It takes at least 11 valid tasks to work out. It will cycle though the available apps and then pick the one with th highest APR.

If you want help un-hide your computers. If you want to see what information is shown click on any users name and then view computers. Its only fairly basic info like OS, memory, number of CPUs, etc.
16) Message boards : Number crunching : Anyone still using a raspberry pi 3 (Message 1909792)
Posted 31 Dec 2017 by Profile MarkJ Special Project $250 donor
Post:
Pictures? :)

Well I don’t know about him but I designed a 3D printed case to hold four side by side with active cooling. I am calling it the Pi ^4 (Pi to the Power of Four) case. I have done another version but am waiting on some parts before I can splash it all over the internet :-)

Pics of the prototype can be found here.
17) Message boards : Number crunching : SETI@home Data Activity page (Message 1909744)
Posted 31 Dec 2017 by Profile MarkJ Special Project $250 donor
Post:
I wonder what happened with LOFAR?

Parkes isn’t even on the map even thought its been talked about over the years and we should start to see some data for it soon(ish).
18) Message boards : Number crunching : SIGSEGV: segmentation violation (Message 1909741)
Posted 31 Dec 2017 by Profile MarkJ Special Project $250 donor
Post:
If they updated the Seti apps to use GCC 6 then the OP wouldn’t need the the vsyscall=emulate, at least for Seti.

The Einstein developers recently rebuilt one of their apps using libc 2.15 (from around 2012) to work around the issue. I would have thought they would update to a later version rather than going to even older code, but maybe it’s too hard.
19) Message boards : Number crunching : I need a little assistance with GPU's (Message 1909312)
Posted 28 Dec 2017 by Profile MarkJ Special Project $250 donor
Post:
while I have found a few places that state this, none actually stated where the data dir resides.

var/lib, usr/bin, etc/boinc-client....

It varies from one Linux to the next. Your BOINC event log/messages tab will tell you at the beginning. For Debian based systems it’s /var/lib/boinc-client.
20) Message boards : News : Special Fundraiser for Parkes Data Store and GPU development system (Message 1906093)
Posted 10 Dec 2017 by Profile MarkJ Special Project $250 donor
Post:
It's a good thing we have a head start here in Sydney, if I get to the bank early I might still squeak in under the rope.

If you have one the donation page does accept Australian Visa cards. It says it will also take Amex and MasterCard but I have no first hand knowledge with either of them. Another alternative if you have a PayPal account is to donate through the GPUUG web site. Exchange rate isn’t in our favour unfortunately.

You know you want one of the spinning hard drive or GPU icons next to your name...


Next 20


 
©2018 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.