May I suggest that you have a web page that shows how to maximise the processing power for those individuals, like myself, who are lucky enough to have and run multi computers so as to increase the work units return rate i.e. HPC (High Performance Computing) or alike.

---

"All man born has a right to life and no man born has the right to take that life"

The BOINC client is expressly designed to run on a single machine, so as to favor home users that make up the backbone of DC projects such as this one. If you'd like to enlist a cluster for this project, probably the easiest way is to create multiple virtual machines on it. A second way is to have a master machine distribute a ready-made Linux OS and BOINC client to several diskless machines over a network. See this link.

If you're only interested in maximizing performance with multiple standard machines, then I offer the standard advice of ensuring adequate resources (with special attention to network bandwidth if your internet connection is slow), cutting down background applications, and perhaps running without a GUI on Unix/Linux machines.

The actual object of my idea is that computer selfbuilers like myself and others if they wish can merge our computers to provide extra power than an individual computer can't and therefore providing more data for the project scientists to mull over. It is not a credit score matter that I'm prepossessing this it is just purely more power for the institutions or universities who are dependent on us volunteers and our computers.

---

"All man born has a right to life and no man born has the right to take that life"

So my first reply holds: BOINC will only function on an operating system with a single motherboard/memory map--not an OS controlling multiple nodes (i.e. a cluster.)

If all you'd really like to do is centrally manage multiple BOINC clients there's BOINCView. I haven't used it personally though.

Edit: I'm pretty sure no one's hacked their way past this restriction, or we'd see crazy machines with thousands of CPU's at the top of the credit statistics.

AQUA@home is multithreading and runs on both cores of my Opteron 1210, thus not allowing other BOINC projects to run, each taking only one core.
Tullio

---

I think the OP is talking about running several stand-alone machines under one instance of BOINC. So if I had four Core i7 920 machines I could cluster them together and have my BOINC client recognize and assign tasks for 32 CPU's.

edit: And it's probably a good thing it isn't possible. I can see pallets of salvaged Pentium 4 machines being delivered into basements and sheds all over the world, creating monster machines which dim the neighborhood lights--Frankenboinc!

Belfry is correct. Each computer is 'stand alone'.

So the best way to maximise the throughput of climate models is DON'T run work from other projects at the same time.
And set ALL options to allow the computers to run flat out, not throttled in any of the many ways that current versions of BOINC allow.

Also, make all new builds with the latest versions of the Intel Core processors. They really scream along.

---

Backups: Here

They really scream along.

Thanks to competition from AMD.

So what Les is really saying is if you want to ensure future processors continue the trajectory of innovation and speed, make sure your new builds are AMD.

... if you want to ensure future processors continue the trajectory of innovation and speed, make sure your new builds are AMD.

Surely you jest!

---

"We have met the enemy and he is us." -- Pogo
Greetings from coastal Washington state, the scenic US Pacific Northwest.

I am running 6 BOINC projects on my Opteron 1210, including a VirtualMachine from CERN, all equally shared. All take only one core, except AQUA, which takes two. So to punish it when it starts, I start another VirtualMachine, Solaris, which is not itself a BOINC project and takes one core, leaving the other to AQUA. Then on the Solaris guest OS (my host OS is Linux) i start a BOINC client and a BOINC SETI app by a developer called Dotsch. Unluckily, they see my Opteron, which is SSE3 capable, only as a pentium_pro+mmx and the app is slow. When asked why I did this, I answered that it caused me to be asked to join the CERN project when I mentioned it in a message board of LHC@home.
Tullio

---

Why would you punish the AQUA app? Their multithreaded applications are just that, meant to run multithreaded. Boinc will automatically balance the resource share anyways, soooooo.

Why would you punish the AQUA app? Their multithreaded applications are just that, meant to run multithreaded. Boinc will automatically balance the resource share anyways, soooooo.

AQUA won't let me run my BOINC_VM Virtual machine by CERN, while all other BOINC projects coexist with it.

---

... if you want to ensure future processors continue the trajectory of innovation and speed, make sure your new builds are AMD.

Surely you jest!

I think he means that in order for innovation in processors to continue at a fast pace, and with reasonable cost, there needs to be competition. Supporting AMD helps ensure there is a competitor to keep overall processor costs down and make sure there is good reason to have to be innovative.