1)
Message boards :
Number crunching :
Transit of Mercury Across the Sun Challenge
(Message 134805)
Posted 1303 days ago by  Mektacular
I was lucky as weather was good at the start of the transit. But after one hour it got cloudy. It IS pretty hard to spot.
Nice! Thanks for sharing this.
|
2)
Message boards :
Problems and Help :
Is there a way to abort accidentally abandoned tasks?
(Message 134556)
Posted 1310 days ago by Mektacular
While updating the configuration on a few machines, I inadvertently abandoned several tasks. Is there a way to abort them through the primegrid.com UI? Since the machines are effectively gone, I have no way to connect to them via BOINC manager.
Ten (or so) of them are tasks associated with the Mercury Challenge with deadlines of 11/24...
No, there's no way to do that, but even if there was, as a policy decision, we just let them time out. This isn't not a real problem for us, whereas "Oops, I didn't mean to cancel that one!" would be an actual problem.
Makes sense. Thanks for the quick response.
|
3)
Message boards :
Problems and Help :
Is there a way to abort accidentally abandoned tasks?
(Message 134547)
Posted 1310 days ago by Mektacular
While updating the configuration on a few machines, I inadvertently abandoned several tasks. Is there a way to abort them through the primegrid.com UI? Since the machines are effectively gone, I have no way to connect to them via BOINC manager.
Ten (or so) of them are tasks associated with the Mercury Challenge with deadlines of 11/24...
|
4)
Message boards :
Number crunching :
Transit of Mercury Across the Sun Challenge
(Message 134472)
Posted 1313 days ago by Mektacular
I did try using the google UI but I am having problems with the two cfg files. I tried editing them directly using vi but they are read-only. I tried doing a chmod (from 644 to 666) but it will not take my request to make them read-write.
Ran into that. You're not logged in as super user.
First thing you type when you connect to the instance 'sudo su'
Also, found out about custom instance types. You can setup a custom instance type that is 4 cpu's and 2GB of ram. It's a little cheaper. Must be an N2 region (central or europe west).
|
5)
Message boards :
Number crunching :
Transit of Mercury Across the Sun Challenge
(Message 134470)
Posted 1313 days ago by Mektacular
Sorry, I didn't write down instructions.
I basically followed the digital ocean instructions (the script commands are identical). The firewall only needs the 31416 entry, because you can open a terminal season from the google UI (you do need to do a 'sudo su' before executing the script commands). I forget where you enter your ssh key, but it only needs to go in once. Otherwise, they are pretty easy to setup.
There is a cpu limit of 24 per region for the free accounts and check to make sure you get an ax512 capable instance. The few I tried in the west region weren't ax512.
My go to instance type is - n2-highcpu-4 (4 vCPUs, 4 GB memory).
|
6)
Message boards :
Number crunching :
Transit of Mercury Across the Sun Challenge
(Message 134443)
Posted 1314 days ago by Mektacular
FYI, Google Cloud gives a free $400 credit for first time users. Their n1-highcpu-4 (4 vCPUs, 3.6 GB memory) vm has avx512 - at least in the one I created. It's a 2.0 GHZ and takes about 26 hours to complete the PSP LLR I started yesterday (it's not done).
The 3 CPU digital ocean takes about 18 hours to finish one.
A 4CPU Azure instance (their cpu optimized instances have avx512 also) takes about 21 hours to finish a unit (again not finished yet).
I also tried a 16CPU Azure instance. A 16 thread LLR appears to run at about 80% efficiency vs a 4 - 4 thread jobs. Ran 4 - 4 thread tasks overnight and they appear to take about 25 hours each (again not finished). Seems like 3-4 cpu is the sweet spot.
You're right Michael, click to spin up instances is addicting...
|
7)
Message boards :
Number crunching :
Transit of Mercury Across the Sun Challenge
(Message 134403)
Posted 1315 days ago by Mektacular
I'm clearly not functioning completely this morning. :)
Three processors, each with one core. Scrolling for the win. Now I have to figure out why only get 1 task...
|
8)
Message boards :
Number crunching :
Transit of Mercury Across the Sun Challenge
(Message 134399)
Posted 1315 days ago by Mektacular
Not sure if this is the right place to ask. Spinning up a few Digital Ocean instances for fun.
cat /proc/cpuinfo reports a gold cpu with the avx512 flags, but they all only report a single core. I've tried both the 3 CPU and 2 CPU sizes from a few regions. Primegrid will only download a single task, and won't download multi-threaded tasks.
Any ideas?
Here's the cpuinfo for a 2 cpu instance:
root@mek4-debian-s-2vcpu-2gb-ams3-01:~# cat /proc/cpuinfo
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 85
model name : Intel(R) Xeon(R) Gold 6140 CPU @ 2.30GHz
stepping : 4
microcode : 0x1
cpu MHz : 2294.608
cache size : 25344 KB
physical id : 0
siblings : 1
core id : 0
cpu cores : 1
apicid : 0
initial apicid : 0
fpu : yes
fpu_exception : yes
cpuid level : 13
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon rep_good nopl cpuid tsc_known_freq pni pclmulqdq vmx ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch cpuid_fault invpcid_single pti ssbd ibrs ibpb tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves pku ospke md_clear
bugs : cpu_meltdown spectre_v1 spectre_v2 spec_store_bypass l1tf mds swapgs
bogomips : 4589.21
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management:
|
9)
Message boards :
Problems and Help :
Selecting which GPU to use
(Message 131410)
Posted 1415 days ago by Mektacular
Also, there is a great guide over on overclockers.net I believe for running multiple instances of BOINC on the same computer for various reasons. I do this and run 3 instances. One is CPU only, one is for GPU0 only and one for GPU1 only. Each GPU instance has it's own venue at each project I run and I can select not only which projects are ran by that card, but which sub-projects from that project if it offers sub-projects. I set it up on Linux after only using it for maybe a week. If anyone needs any help, PM me and I'll do what I can.
John T
Thanks John, this was super helpful. It's a little tedious but I did eventually get it to work.
|
10)
Message boards :
Problems and Help :
Selecting which GPU to use
(Message 131227)
Posted 1420 days ago by Mektacular
Nah, that can be done with some config file magic:
<exclude_gpu>
<url>http://www.primegrid.com/</url>
<device_num>0</device_num>
<app>pps_sr2sieve</app>
</exclude_gpu>
<exclude_gpu>
<url>http://www.primegrid.com/</url>
<device_num>1</device_num>
<app>genefer17mega</app>
</exclude_gpu>
So is the response earlier in the thread simply wrong? It appears so...
Looks like like the most straightforward approach is multiple boinc instances. I'll have to give that a try. Windows 10...
|
