Hello crunchers!

Recently a fellow cruncher stated that BOINC was optimized for Windows and would be less efficient on Linux. Is this true? I just finished converting two Windows XP systems to run Linux Mint 11.

If this is true, please explain why this is so.

Thanks.

---

BOINC is just the managing program, simply said it tells the projects science applications when they can run and which project can fetch work next. It doesn't do any crunching. So there's no need to optimize BOINC for any platform.

If you want to optimize anything, you'd do so with the project science applications, that's where the gain lies.

---

Sorry, I have a tendancy to use BOINC as a general term for the projects... gotta break the habit of doing that.

So where would I look for more info on optimizing individual project apps?

...and is this an issue for a "standard" computer configuration. I don't have GPU's or specialized hardware.

---

Processor or GPU specific optimizing of applications is most profitable on those projects that have open source code for their applications. Seti, Milkyway, Einstein (to a point), you best ask at the project(s) though, there's bound to be more knowledgeable people on their (Number Crunching) forums. Those can then explain whether or not Windows apps run better than Linux apps and vice versa, and why.

---

Recently a fellow cruncher stated that BOINC was optimized for Windows and would be less efficient on Linux. Is this true?

A while a ago I benchmarked a PC across as many projects as possible with a clean Win XP install. I then made a clean Ubuntu install and repeated the benchmarks.

On average Ubuntu was 33% slower than XP.

There were exceptions, 2 projects in particular were twice as fast. Some were notably agonisingly slow (Spinhenge, take a bow!).

Nothing to do with BOINC though. Just the apps from the projects, how they were compiled etc etc.

This was a while ago, can't even find the spreadsheet that I made as I was doing the tests, things have moved on since then anyway.

I only run a couple of linux boxes these days just to try projects that are linux only. Rest of the time I just let them crunch a couple of maths projects that they seem to do quite well.

Why not make your machines dual-boot or run VMs and try it for yourself? Would be both interesting and useful for all of us :)

Al.

@Chertsey, Interesting that Ubuntu was the the slow Linux OS. Before I decided on what Linux distro to convert to, I tried Ubuntu and Mint on an older machine. Ubuntu seemed to be sluggish on the older machine... so my very unscientific personal conclusion was that Ubuntu took more resources to run, one reason I chose Mint.

If I ran an OS from within a virtual machine, wouldn't the VM introduce system overhead, that would skew the results? (I have no experience with VM's)

Unfortunately I lost my Win XP Pro OS so I can't go back. Wouldn't it be difficult to benchmark projects? From what I understand of the nature of distributed computing, you wouldn't get the same data file to crunch on two different computers, right? Though I guess if you did it for a large sample of projects then the differences would average out statistically.

@Dagorath, sorry you had such a bad time with Windows. I was very happy with Win XP Pro/Home... and no cash for Win 7 right now. My main reason for switching to Linux is that I just wanted to try it out.

---

@Dagorath I had to chuckle at your comment on being lucky to lose XP. I do feel that the hand of fate was involved somehow as I was thinking of doing a full install, and when I realized that "something bad" had happened to the partition with the Windows XP restore files... the decision was made. ;) (I wrote about this in one of my blogs...)

I agree with your statements on GUI settings having an effect on performance. I am new to Linux so I don't know all of the tricks to optimize system performance.

It's quite possible that the default Mint config taxed the CPU less than the default Ubuntu config. Fedora wouldn't even run the full GUI on the older system.

If I win the lotto, I'll set up a room of computers with various OS configs and benchmark BOINC projects and share the results. You'll even be able to buy a custom built BOINC computer on my website. ;) Hmmm... maybe I could work out deals with some of the more popular projects to use their logos on custom enclosures so that crunchers could show off their grid networking box to their envious friends. ;)

Oh wait, I'm poor and using old computers with free Linux... nevermind.

(...got carried away there...)

---

@Dagorath Since you run Linux, your in the 1%. ;)

When you buy Microsoft, why not make it open-sourced? Hmmm... what better revenge? Maybe even keep all those programmers, etc., and use them towards building a super Linux distro? MS-Linux!

Then you could go to computer manufacturers and use your clout to force them to load MS-Linux on the computers they build! Buwahhwahha!

How soon we are corrupted and turn to the dark side...

Oh, and I think Mint ate my recovery partition when I tried to set up a dual boot Win/Mint system. I didn't notice it until later but my "Press F-11" restore option (set up by HP to access the recovery partition) was no longer displaying during boot. I don't blame Mint though, if the cheap/lazy buggers at HP had supplied me with a set of setup CD's, it wouldn't matter. I'll rant about HP all day long. :)

---

Next time assemble your own computer with off the shelf components for half the price of an HP.

The upside of that is that you are then allowed to open the PC case as well, without a) throwing your warranty in the bin and b) requiring some fancy newfangled screwdriver that no one else on this world uses.

---

Next time assemble your own computer with off the shelf components for half the price of an HP.

The upside of that is that you are then allowed to open the PC case as well, without a) throwing your warranty in the bin and b) requiring some fancy newfangled screwdriver that no one else on this world uses.

My Dell laptop was under warranty when it's keyboard died (ages ago). Dell sent me the keyboard kit, and actually allowed me to put it in! I was honored. And yes, it included the tiny screwdriver.

I won't even give breath to the unspeakable horrors that HP customer support put me through. I still have PTCSD from that. (Post Traumatic Customer Service Disorder).

Oh, Dagorath... burning the Microsoft campus would create a lot of carbon emissions. Instead, let's go green and turn it into a commune for homeless folks!

---

BOINC is just the managing program, simply said it tells the projects science applications when they can run and which project can fetch work next. It doesn't do any crunching. So there's no need to optimize BOINC for any platform.

Here is a nearby thread where FDF experiences problems with boinc mis-detecting user's activity. He runs bubuntu. Probably, bubuntu doesn't count as GNU/Linux anymore anyway.

---

I'm counting for science,
points just make me sick.

On CPDN each climate model type is compiled separately for Windows, Linux and Mac. The programmers try to ensure that a model type won't run much more slowly on any particular OS. But this isn't their highest priority. The most important thing is that each compilation should run stably, producing reliable climatic results.

Sometimes it proves impossible (at least with the manpower available) to achieve all these objectives. For example, the FAMOUS model type compilation ran super-fast on Mac, but tended to produce unstable climate and whenever this happened the model used to crash.

There have been reliable reports of some model types running faster on Windows, while other types are faster on Linux. This is to be expected because the model speed is compilation-dependent.

I get the impression that Macs run very fast anyway, whatever the compilation they have to crunch.