I too had to try this out for myself, to find out what difference can be made by looking at BOINC vs UD.
I have a Linux machine, P4 2.4 512, obviously running BOINC, and it gets around 1000 pts for 24 hours of runtime.
At work I have a P4 3.2 1024 running UD for around 800 pts for 24 hours.
The other computer at home is a P4 2.6 512 running UD. This machine has hyperthreading, and on this machine I have had a look at what is going on.... according to the Task Manager, the UD application is only utilising 50% of the processor. When i run BOINC, it runs two applications simultaneously, and each use 45-50% of the processor. So, with BOINC, I should be getting twice as many points as I was with UD, as my computer is actually doing twice as much work!



dreamboat
now looking to get the work computer onto BOINC also, as it is also only using 50% of processor function.

Hi Dreamboat

The Intel Hyperthreading discussion has been had many times within the forum.

Although with UD and HT enabled it appears (to windows) that only 50% of the processor is in use. In actual fact 100% of the P4 is being utilised by UD.

As a consequence BOINC although it will run 2 threads on your HT Machine it shouldn't be any more efficient than the 50% apparant utilisation by UD's single thread.

I know this sounds counterintuitive but HT was a great marketing wheeze by Intel to have a pop at AMD who were gaining the upper hand.

This all said BOINC is producing some very strange results at the moment.

It currently appears to be twice as efficient on my single core non HT AMD cheap as chips processor !!!!

Dave

Didactylos

Sounds like a plan

It would make the BOINC - UD discussion obsolete as well

I noticed the mac client as well while downloading my Windows

It just needs a slicker installation gui (without the requirement for a host of hex) and we would be home and dry

Who and why would they call me that Spikey?

BOINC appears on paper to be slightly more efficient and it offers a number of other advantages such as preloading of multiple work units for those few and far in between days when the Grid is out of service - like the unofficial UDMon. It also supports Multi core and multiprocessor setups reaping the performance gains of those systems that UD cannot reach. It works with Linux and as Didactylos points out holds out the posibilties of Mac and other forms of Unix over and above Linux. It can also be installed as a service (officially) and so would run without anyone being logged onto the system creating further crunching time and truely invisible crunching and you can choose to continue to crunch when your Laptop is runing on it's batteries if that is OK by you.

All of these things add up to more crunching getting done which if it's the reason why we are here is a good thing

My only beef was with the points being allocated, which as we have discovered bear no relationship to those allocated by UD and so the points comparison of any team and any individual has been lost as a result. Points are now unfortunately a flawed currency - the price of progress

As results generated are completely random now the next closest benchmark that bears any reasonable scrutiny should be run time.

At least this measure shows your relative commitment to this great project

Dave

p.p.s

I'd just like to announce, and here is as good a place as any, that I am no longer a member of any team (with Didactylos's blessing) I have no bias, I am now unaffiliated. I will no longer be using any pseudonym or alternate account.

If you need help and I can spare the time (and I have the ability ) I'll lend a hand. If there are any general issues that affect the Grid with BOINC, with UD, the Website operation, Stats and Stuff I will flag them up to IBM. I am lucky indeed to still have the IBM hotline at my disposal.

Just needed to make that clear to all for reasons that will soon become apparent.

Thanks too for your show of support, for me, in recent weeks

They are appreciated (in a british understated way )

Keep 'em crunching folks

Dave