PrimeGrid
Please visit donation page to help the project cover running costs for this month

Toggle Menu

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other

drummers-lowrise
11) Message boards : Antarctic Crunchers message board (Message 138627)
Posted 570 days ago by Profile SoulflyProject donor
A few usefull tips on Google Cloud:

I use PuTTY to connect to google cloud or any other VM. Easy to use once you get familiar with it. You can use the same SSH key across multiple instances or beef up the security and create individual keys for each instance and even add another layer of security by applying password on the SSH key.
During creation of a VM click on Management, security, disks, networking, sole tenancy and chose security tab and scroll down to paste the private key you generate in PuTTYgen.

The most powerfull instance ive seen so far is the N2.
When creating a VM click the general purpose tab under machine family and chose N2 series (cascade lake 2.7 ghz). As machine type select custom witch enables you to chose vCPU(thread) count and memory size. LLR tasks require very little memory, so a 2 thread 1 gig VM will be quite productive.
Several 2 thread VM's will be less likely to suffer a shutdown in preemptive mode than say one 16 thread instance.

Creating preemptive VM click Management, security, disks, networking, sole tenancy and chose the management tab and scroll down to Preemptibility and select on
This will reduce price to about a third and your VM will shut down after max 24 hours depending on the load on the datacenter you have chosen. Keep in mind that preemptive instances can be shut down any time and sometimes even after 15 minutes (!) The key is to create a bunch of small VM's spread across different datacenters, and keep an eye on them a couple times a day to restart those that where shut down. Download Cloud Console on your phone if you away for a couple of days. Also be aware that price differs across datacenters.
12) Message boards : Antarctic Crunchers message board (Message 138553)
Posted 572 days ago by Profile SoulflyProject donor
I'm thinking about using cloud instances but I'm not sure about it as I haven't used cloud instances before. Would you still recommend DO? In the case of a "yes", would you guide me through the setup?


My DO account was blocked due to “atypical activity” with 10 droplets running and my response on the ticket has gone unanswered for 3 days. Have in mind that it is not possible to destroy the droplets on a blocked account, and they are still charging my account with 10 not reachable and not running droplets. I regard that as downright unprofessional. I am NOT looking forward to what kind of trouble this can develop into. However I can recommend to be very careful dealing with them upon scouting the net for other customers terrible experience with this provider.

However Google Cloud can almost match the price to performance when the VM is created in preemptive mode, though it requires more babysitting so to speak.
13) Message boards : Antarctic Crunchers message board (Message 138132)
Posted 581 days ago by Profile SoulflyProject donor
Welcome indeed welcome to the (frozen) force!


Thanks!

You reminded me of Teal'c from Stargate - indeed :)
14) Message boards : Antarctic Crunchers message board (Message 138127)
Posted 582 days ago by Profile SoulflyProject donor
I am in an geographical area where runtime type switches a lot between P4, P100 and T4 – haven’t seen K80 for a while – I suspect they are being phased out. Performance on these cards are currently as follows:

Currently a GFN23 50P chunk takes roughly:
33.000 sec on an P4
20.000 sec on an P100
10.000 sec on an T4

I am curious of the performance on the recently released 2060 KO that are based on the TU104 (at least EVGA promises this die on their KO versions) where I suspect the performance somewhat will match the T4/2070/2080 I certain compute scenarios.
Anyone got a 2060 KO who can give some kind of feedback?
15) Message boards : Antarctic Crunchers message board (Message 138126)
Posted 582 days ago by Profile SoulflyProject donor
Welcome two new teammates!


Soulfly and Clark Gaebel!


Happy crunching!


Thank you 😊

I am currently testing different cloud compute providers to compare compute capabilities, and so far, the difference in performance to cost ratio is huge. In addition, I found peculiar behaviour in VM environments. Ill address this in a thread dedicated to cloud experience very soon.

Br
Fred
16) Message boards : Problems and Help : CPU units assigned to subproject with only GPU units selected (Message 34795)
Posted 3828 days ago by Profile SoulflyProject donor
Take a look at this thread:
http://www.primegrid.com/forum_thread.php?id=3205
17) Message boards : Number crunching : Getting CPU WUs when not requested (Message 34794)
Posted 3828 days ago by Profile SoulflyProject donor
(trying) to participate in the Sophie Germain challenge i changed the preferences enabling only SGS, PPS sieve (CUDA), and GW sieve (CUDA).
However the preference change resulted in unententional participating of PPS LLR and 321 sieve...
Removing PPS LLR and 321 sieve from preferences at the same time didnt work, but removing them one at the time and saving preferences in between actually worked.
I guess a bug is around :)
[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2021 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 3.30, 3.71, 3.42
Generated 25 Sep 2021 | 20:53:39 UTC