You just need to get one of these.

Of course, mine has 12 GB of RAM, not 2 GB. But yeah... eight 900 MHz Slot-2 PIII-Xeons running Windows Server 2003 Enterprise with PAE mode to access all that 12 GB of slow PC-100 Registered memory.

So, until BOINC became available, I was using only 1 processor with the UD client. It was rather depressing.

The methods for calculating points is a bit... confusing. UD one way.... BOINC another. It's like adding apples and oranges and hoping to make the result a bananna. Instead you just get a fruit salad. In this case people are wondering where the melon came from. I think I understand it, and then it gets confusing again. It's bringing back some bad memories from math class. What is "X" anyway? And I don't care when the two trains met. Anyway...

Don't mind me.... I'm in a weird mood tonight.


Here's the last 10 days of of results I had with UD....

Date         Run Time       points    results      points per result

---------------------------------------------------------------------------------
2/3/2006	0:001:05:31:13	839	2	419.50
2/2/2006	0:000:17:11:16	488	1	488.00
2/1/2006	0:001:03:45:37	789	2	394.50
1/31/2006	0:001:02:29:32	753	2	376.50
1/30/2006	0:001:04:53:43	821	2	410.50
1/29/2006	0:000:11:57:06	339	1	339.00
1/28/2006	0:001:08:12:13	915	2	457.50
1/27/2006	0:000:23:37:12	671	2	335.50
1/26/2006	0:000:14:30:59	412	1	412.00
1/25/2006	0:001:07:47:39	903	2	451.50
			
	Average points per result		~408

And here are the last 10 days results running BOINC....

Date         Run Time       points    results      points per result
---------------------------------------------------------------------------------
2/20/2006	0:009:18:08:52	6,921	15	461.40
2/19/2006	0:010:04:36:40	7,506	16	469.13
2/18/2006	0:006:16:11:22	5,053	10	505.30
2/17/2006	0:004:06:57:38	3,195	7	456.43
2/16/2006	0:008:12:17:03	6,758	13	519.85
2/15/2006	0:005:23:07:49	4,187	10	418.70
2/14/2006	0:008:05:30:27	5,906	13	454.31
2/13/2006	0:008:12:15:10	6,607	13	508.23
2/12/2006	0:006:23:08:38	5,100	11	463.64
2/11/2006	0:007:22:41:27	5,712	13	439.38

	Average points per result		~469

BOINC seems to have an advantage. Unless Microsoft Excel is fibbing to me (it is Microsoft afterall), then the percent difference between the two is ~15%

I guess the percentage of the point increase can really vary depending on the system. This was only a 10 day sample, but it holds with longer duration samples too.

As long as I'm returning valid results that can help people, I'm happy.


The only reason I want to get back to work tomorrow is to setup one of these. Except mine only has eight 550 MHz CPUs with 512k cache. Six or seven years ago, that was a kickin' system. Perhaps it's still useful afterall.


I'm just happy this grid exisits.


-j

Stage 2 of the experiment

I have suspended all network activity on all agents

(don't worry I'm still crunching)


Once I stop receiving points I'll open up the floodgates and then I will know exactly how many WU's and of what type I have returned to the WCG.

Immediately afterwards shut the sluces and then we will see another lump of WU's apear in my stats and once they have petered out we can compare and contrast in no more that 1000 words on the outcome of the question does WU's in = results on the WCG

Does it sound like a plan?



Dave

A rose by any other name will still smell as sweet.

The primary difference between the executables on BOINC and on UD are the code changes that were required to resolve differences in how the API's are difference for checkpointing, file access and reporting progress.

The science application will be run at essentially the same level of effectiveness on both platforms.

Just a little update on my tests:

On my Windows XP box using the AMD AlthonXP 2000+:
Average points per result returned UD: 203
Average points per result returned Boinc: 390

So almost but not quite a factor of 2 increase in points allocated using Boinc, which runs inline with other peoples findings, and I have no perceptable change in the average number of units I am crunching each day

For my next trick I shall switch OSs (always useful having windows and linux installed on the same machine ) and see if I still get the same sorts of results for Boinc on Linux.

On my straight Linux box:
Average points per result returned Boinc: 118

I need to have a look at this box and see what is in it, [since it was one I was given]

The moral of the story so far being if you are after points run Boinc on windows, and if you are crunching for the good of the cause it don't matter a fig which you run.

still on 865,999 at the 6pm update it's looking clean

a couple more of those just to be sure