KWSN Orbiting Fortress Forum Index KWSN Orbiting Fortress
KWSN Distributed Computing Teams forum
 
 FAQFAQ   SearchSearch   MemberlistMemberlist   UsergroupsUsergroups   RegisterRegister 
 ProfileProfile   Log in to check your private messagesLog in to check your private messages   Log inLog in 

Preparation for a New Shrubber
Goto page Previous  1, 2
 
Post new topic   Reply to topic    KWSN Orbiting Fortress Forum Index -> Ye Olde Help Scrolls
View previous topic :: View next topic  
Author Message
The Knighty NI
Prince
Prince


Joined: 06 Sep 2007
Posts: 780
Location: Lost in space on a rather small Blue ping pong ball. :)

PostPosted: Wed Dec 01, 2010 3:29 pm    Post subject: Reply with quote

#ni-1 #ni-1 #ni-1 #ni-1 #ni-1 #ni-1
You guys are all freaking NI!

Cow_tipping Wrote:
Quote:
PostPosted: Tue Nov 30, 2010 10:47 pm Post subject:
Lots of memory can be very useful if you are planning to use VM's for projects as Dimes or Majestic-12.
Cool


Think you may have just come out with the Gem of an idea. Lol.

This is how I understand VM's work. Please please correct me if I am wrong in my thought's here because I don't know what effect this idea will have on the physical memory.

You take the attributes of the machine you want to have as your VM. These are passed into the VM as its characteristics.

EX.
Server with 32gb ram
4 processors 12 cores each @ 2.4ghz
Motherboard
Video card
any other attributes you want to include.

What I don't know is:

1. Do you now have the equivalent of two machines each running 48 cpu's running at 1.2ghz ergo exactly the same output as before with each WU taking double the length of time to complete

2. Do you now have the equivalent of two machines each running 48 cpu's running at 2.4ghz ergo double the output as before

If 2 is true would it then be possible to max out ram and run 8 32gb VM's on a MB maxed out at 256gb ram or would that equate to each VM running at 300mhz?
again resulting in exactly the same output only taking 8 times as long to do each WU?

Getting this right may make the difference between shrubbing about 57k to about 460k credits per day WOWWWWZERRS Now that is a shrubbing figure I am very hungry for Smile


By the way since I returned:
Team Position on 11th November 2161
Today 1st December 1210


I must be making a few dents in some of our team members and very slowly catching up Smile
Reckon that by 1st Jan 2011 I will be in the top 900 Smile
_________________
What is that in the Shrubbery?
Back to top
View user's profile Send private message
Hal9000x86
Baron
Baron


Joined: 06 Apr 2010
Posts: 217
Location: The 10 milllion year project that will find the question to the answer 42.

PostPosted: Wed Dec 01, 2010 4:00 pm    Post subject: Reply with quote

The first one would be true,you would be splitting the cpus evenly with the real os and one vm, or just 2 vms and a one or two cores for the real os. If you gave half the cores to one vm each machine would be running 24 cores at 2.4 ghz What Cow_tipping was saying is to make many small vms since dimes does not use a lot of cpu. Its more internet dependent. The memory is shared the same way it gets split between the real machine and all the vms. More memory means that each vm will run faster.
_________________

Back to top
View user's profile Send private message
The Knighty NI
Prince
Prince


Joined: 06 Sep 2007
Posts: 780
Location: Lost in space on a rather small Blue ping pong ball. :)

PostPosted: Sat Dec 11, 2010 9:59 am    Post subject: Reply with quote

Hal9000x86 said in Primegrid Winter solstice challenge 18-21 December
Quote:
GeForce 8400 GS is the lowest desktop add on card that offers cuda, the 7300GT is from the generation before that. You can get a lower end 400 or mid to high end 200 series and have a decent cuda card.


Its this sort of information that I need in order to chose the most credits for my bucks (play on bang for bucks) when I build my new shrubber next year.

Hell if a single GPU is churning out 415k credits per day I am beginning to think why spend building a server dedicated to shrubbing with lots of memory and CPU's unless there is a very good reason for doing so.

My workings out are based on Sophie Germain as a base line because the WU's all return the same credit and all take the same time within a few seconds. Been watching this very closely since starting the project so that I can get good average figures per day on WU time's to shrub.

The timing figures are based on 2300 WU's shrubbed at 2000Mhz and 2100 WU's shrubbed at 2100Mhz.

I made sure that I did a very stable overclock so that I could measure time per WU and extrapolate those figures forwards onto faster CPU's as accurately as possible. Its interesting to note that 100Mhz faster gives an extra 2.0864 credits per CPU per hour.

On my current machine the figures are like this

Mean Ave Time to shrub...........Proc speed..... WU's p/h per CPU...... WU's per 24 hrs .... X's 4 CPU's ....... Daily tot Credit

.............. 00:20:00 .................... 2000Mhz .................. 3 ............................ 72 .................... 288 ................... 4006.08
.............. 00:19:24 .................... 2100Mhz .................. 3.15 ....................... 75.6 ................. 302.4 ................ 4206.38


Based on the above figures it works out that building the server I was planning falls very short on shubbing power compared to a single decent GPU.

Est Mean Ave Time to shrub..... Proc Speed.....WU's p/h per CPU......WU's per 24 hrs ......X's 48 CPU's ...... Daily tot Credit
.............. 00:16:33 .................... 2400Mhz .................. 3.6 ......................... 86.4 ................... 4147.2 ......... 57 687.55

Is there any empirical evidence for shubbing on GPU's that I can use for comparison purposes. More importantly are there any comparisons between different GPU's and how they behave when shrubbing on specified WU's?
_________________
What is that in the Shrubbery?
Back to top
View user's profile Send private message
Hal9000x86
Baron
Baron


Joined: 06 Apr 2010
Posts: 217
Location: The 10 milllion year project that will find the question to the answer 42.

PostPosted: Sat Dec 11, 2010 3:16 pm    Post subject: Reply with quote

Depends on the project some are cuda only, and some are steam only. If they run both, the steam wus are normally faster. At least with the comparison of my two cards which are each others market competition, with the exception of these pps sieves which use open cl more then they use steam. Ati is going towards open cl now so it can be more universal between windows and mac without changing the hardware as much.

This might help, its a comparison of all the video cards in spec.
ATI/AMD
http://en.wikipedia.org/wiki/Comparison_of_AMD_graphics_processing_units
NVIDIA
http://en.wikipedia.org/wiki/Comparison_of_Nvidia_graphics_processing_units
_________________

Back to top
View user's profile Send private message
Display posts from previous:   
Post new topic   Reply to topic    KWSN Orbiting Fortress Forum Index -> Ye Olde Help Scrolls All times are GMT - 5 Hours
Goto page Previous  1, 2
Page 2 of 2

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum


Powered by phpBB © 2001, 2005 phpBB Group
Optimized Seti@Home App | BOINC Stats