Can I run multiple instances of BOINC on one machine?

I have <a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/hosts_user.php">two
computers</a> and my average credit for the fast one is 322.46. It does not seem to be bogged down one bit running BOINC in always mode. I see computers with slower processor and less memory getting 2511.98 average credits. How do they do this? I run it in sys tray only no screen saver and shut my monitor off after one minute. Nothing else runs except McAfee virus and firewall. I have shut these off before because I am behind a firewall in my router.

It does say it will take 35 days to run one model though, is this normal?

Any tips and recommendations would be appreciated,
Dale

xernon@msn.com Xernon is the planet I am from.

Hi Dale,

yes, there is a little you can do to run BOINC a bit faster. Try to defragment HD from time-to-time and also defragmenting memory (e.g. using TaskInfo) may help. You can disable screensaver, set BOINC you crunch when you are using computer and if running more that one project, set the leave application in memory to on.


&gt; Can I run multiple instances of BOINC on one machine?

There is not reason for this on single processor machine. It is not possible and even if it would, it would actually slow-down processing. On HyperThreading machine or real multi-CPU boxes, you can tell BOINC to use more thatn one procesor.

&gt; I have <a> href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/hosts_user.php"&gt;two
&gt; computers</a> and my average credit for the fast one is 322.46. It does not
&gt; seem to be bogged down one bit running BOINC in always mode. I see computers
&gt; with slower processor and less memory getting 2511.98 average credits. How do
&gt; they do this? I run it in sys tray only no screen saver and shut my monitor
&gt; off after one minute. Nothing else runs except McAfee virus and firewall. I
&gt; have shut these off before because I am behind a firewall in my router.

Link you your computer list should be http://climateapps2.oucs.ox.ac.uk/cpdnboinc/hosts_user.php?userid=46432
It is not good to measure performance using RAC. Acutally, there is a small bug in computing credit that gets corrected upon every stats database update (several times a day) but RAC remains affected.

&gt; It does say it will take 35 days to run one model though, is this normal?
Yes, it is normal. And it is a good measurement (together with sec/TS) for CPDN performance.
It is a know fact that Intel P4s with HT on are doing considerable better than AMD. This may make you feel that your AMD64 is a bit slower comparing to other machines (e.g. P4 at 3 GHZ) but one needs to compage AMD64 vs AMD64 etc.

&gt; Any tips and recommendations would be appreciated,
&gt; Dale

Your are welcome :-)
Hope to help..

Yes this helped a lot. I was looking at buying a P4 multi processor box to build for kicks and grins. Now if I do build it I will see how high I can drive my average credit. I now know I will need more then 4 processors for this, :-)

hehehehehe 6 sound slike a good round number.

Dale

&gt; Hi Dale,
&gt;
&gt; yes, there is a little you can do to run BOINC a bit faster. Try to defragment
&gt; HD from time-to-time and also defragmenting memory (e.g. using TaskInfo) may
&gt; help. You can disable screensaver, set BOINC you crunch when you are using
&gt; computer and if running more that one project, set the leave application in
&gt; memory to on.
&gt;
&gt;
&gt; &gt; Can I run multiple instances of BOINC on one machine?
&gt;
&gt; There is not reason for this on single processor machine. It is not possible
&gt; and even if it would, it would actually slow-down processing. On
&gt; HyperThreading machine or real multi-CPU boxes, you can tell BOINC to use more
&gt; thatn one procesor.
&gt;
&gt; &gt; I have <a>
&gt; href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/hosts_user.php"&gt;two
&gt; &gt; computers</a> and my average credit for the fast one is 322.46. It does
&gt; not
&gt; &gt; seem to be bogged down one bit running BOINC in always mode. I see
&gt; computers
&gt; &gt; with slower processor and less memory getting 2511.98 average credits.
&gt; How do
&gt; &gt; they do this? I run it in sys tray only no screen saver and shut my
&gt; monitor
&gt; &gt; off after one minute. Nothing else runs except McAfee virus and firewall.
&gt; I
&gt; &gt; have shut these off before because I am behind a firewall in my router.
&gt;
&gt; Link you your computer list should be
&gt; http://climateapps2.oucs.ox.ac.uk/cpdnboinc/hosts_user.php?userid=46432
&gt; It is not good to measure performance using RAC. Acutally, there is a small
&gt; bug in computing credit that gets corrected upon every stats database update
&gt; (several times a day) but RAC remains affected.
&gt;
&gt; &gt; It does say it will take 35 days to run one model though, is this
&gt; normal?
&gt; Yes, it is normal. And it is a good measurement (together with sec/TS) for
&gt; CPDN performance.
&gt; It is a know fact that Intel P4s with HT on are doing considerable better than
&gt; AMD. This may make you feel that your AMD64 is a bit slower comparing to other
&gt; machines (e.g. P4 at 3 GHZ) but one needs to compage AMD64 vs AMD64 etc.
&gt;
&gt; &gt; Any tips and recommendations would be appreciated,
&gt; &gt; Dale
&gt;
&gt; Your are welcome :-)
&gt; Hope to help..
&gt;

---

If you are serious about speed, Linux seems to be faster than Win XP :)

---

&gt; If you are serious about speed, Linux seems to be faster than Win XP :)


Yes, I agree (for AMD at least). I've seen speed increases between 5 and 15% when using Linux with an AMD optimised kernel. Although this was not tested by rerunning the same model, the change is outside the range of variance of individual models and is significant, I think. My thoughts are that the Fortran compiler used for the CPDN Linux release generates better code for AMD than the Intel one used for Windows which, not unexpectedly, generates code to favour Intel devices.

Andrew

---

Andrew

<a href="http://cpdnforum.info">CPDNforum<a>

&gt; I've seen speed increases between 5 and 15% when using Linux with an AMD optimised kernel.

Much the same increases have been seen with P4 machines.

---

&gt; Much the same increases have been seen with P4 machines.

It is by far more likely that the Kernal in the Linux distributions is faster than the kernal in WIndows XP.

As soon as they get the cross platform GUI running half way decently and I have time I will be installing Linux on most of my Farm, well, except the Macintosh ...

---

Hi, Paul,

Good to see you on the CPDN boards.

A thought on Linux. In my experience SuSE 9.1 Pro wouldn't work with a KVM switch and 9.0 Personal requires use of generic mouse driver -- it won't work with the proper driver for the M$ mouse. According to techs at the shop where my P4's were built, this is a common problem. (I like Linux kernel's speed and reliability but can't say the same for some of the add-ons.)

Yoper claims to be the fastest Linux implementation. Free download burns to a single CD for installation. (I made a CD but haven't installed it yet.)
<a href="http://www.yoper.com/">yoper.com </a>

Thank you for the comprehensive boinc documentation. Well done.

Jim

---

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

&gt; A thought on Linux. In my experience SuSE 9.1 Pro wouldn't work with a KVM
&gt; switch and 9.0 Personal requires use of generic mouse driver -- it won't work
&gt; with the proper driver for the M$ mouse. According to techs at the shop where
&gt; my P4's were built, this is a common problem. (I like Linux kernel's speed
&gt; and reliability but can't say the same for some of the add-ons.)
&gt;
&gt; Yoper claims to be the fastest Linux implementation. Free download burns to a
&gt; single CD for installation. (I made a CD but haven't installed it yet.)
&gt; <a href="http://www.yoper.com/">yoper.com </a>
&gt;
&gt; Thank you for the comprehensive boinc documentation. Well done.
&gt;
&gt; Jim
&gt;

Hi Jim -

I Installed YOPER 2.1.x last week. It worked absolutely fine using a KVM (Trendnet), and quickly &amp; easily detected my network, I/O devices, peripherals, &amp; got me out to the Internet, straight out, w/o any major probs.

I did have 2 'pilot error' issues, though:

1] I messed-up my LILO config, &amp; couldn't dual-boot to WinXP. Whenever I would boot, my comp default booted YOPER w/no WinXP option (I never got the LILO screen).

2] I couldn't get the BOINC CPDN Linux client installed.

Both of these probs were definitely due to my COMPLETE lack of Unix/Linux experience, since the install otherwise was flawless &amp; the OS ran perfectly, including network/internet access.

I didn't have much time, and finally just re-installed WinXP, so I really only played around w/it for one night. I'm going to give it another try when I've got some time to spare, but may install YOPER on it's own dedicated drive, so I can change the boot sequence in the BIOS, for either OS.

I'm really excited about YOPER, since Linux is faster than WinXP to begin with, and YOPER is not only supposedly optimized for speed, but also the only i686 Linux distro I've seen to date (&amp; free too!). Yoper was more-or-less a one man show by a guy who was active in the Overclockers.com forums (one of my hangouts), but recently has recruited some other peeps from Overclockers.com (especially the Folding team) in assisting, so it's definitely being developed w/speed in mind + also DC performance.

When the CPDN phpBB comes back up, I'm going to suggest a Linux sticky thread, 'cuz I think the interest is definitely there, and a little advice for us *nix noobs sure can't hurt.

I'm pretty confident if I researched a little more, &amp; hacked at it long enough, I could've gotten the CPDN Linux client to run. Probably was just a hint or two away.

Strat

BTW - Ver 2.2.x is very close to release, and I've seen a post at OCC from a guy now running it as a release on his servers, even though it's technically still beta. It's supposedly stable, according to him, &amp; several other peeps on OCC's forums also report the same.

---

Hi, Chicago, from Calumet City (though that was nearly 50 years ago),

Thanks for the info. I'm having a few regrets at having donated a PIII 733 recently. It would have been a good Yoper test box. (Oh well, it is surely being put to better use now. [It had the now antique SuSE 7.1 -- but I re-installed Win98SE because I didn't want to stick an unsuspecting person with SuSE 7.1.])

Having read the new installation/configuration pages at yoper, I recognize that it is not as user-friendly as SuSE YaST (Yet another Setup Tool), but YaST is not without issues.

You bit off a tough chunk to chew for a newbie (I dislike that term) and it is to your credit that you pulled it off. Best of luck on your next trial.

I share your excitement about Yoper. It seems sure to be the vanguard for the next wave of Linux.

Well, as the Thread topic is maximizing performance, I guess we haven't strayed too far!

Again, thanks for the input.

Jim

---

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

&gt; A thought on Linux. In my experience SuSE 9.1 Pro wouldn't work with a KVM
&gt; switch and 9.0 Personal requires use of generic mouse driver -- it won't
&gt; work with the proper driver for the M$ mouse.

Hi AstroWX,
I've been hitting the same problem which seems to be confined to Kernel 2.6+ (as used in SUSE 9.1 onwards) and Belkin KVM's especially the SOHO varieties. It is intermittent and is not a cabling or mouse problem - bypass the KVM and the problem disappears.

An Xserver reset clears the glitch (switch to Text terminal CTRL-ALT-F5 and back to current screen CTRL-ALT-F7 is a quick way in SuSE distros). I'm still digging round the see if Belkin will own up (they have removed the original acknowledgement and fix from their Knowledge base - was in Google archive - so I don't hold out much hope). Looks like time to add a replacement non-Belkin KVM to my pressie list...

Andrew

---

Andrew

<a href="http://cpdnforum.info">CPDNforum<a>

&gt; It is by far more likely that the Kernal in the Linux distributions is
&gt; faster than the kernal in WIndows XP.


Paul,
Although the kernel is undoubtedly faster in Linux than Windoze, I don't think that the CPDN speed increase can all be put down to that (unless you count implementation subroutine libraries as part of the kernel). CPDN is a major processor and memory cycle hog compared to its use of kernel services for such things as disc access, so efficiency of compiled code must play a part.

Andrew

---

Andrew

<a href="http://cpdnforum.info">CPDNforum<a>

&gt;I'm having a few regrets at having donated a PIII 733
&gt; recently. It would have been a good Yoper test box.

Actually, this is a very good idea, IMO. Part of the reason I cut my first Linux experience short was due to the fact I was trying to use it on my newest build, which is also now the fastest machine in my fleet. I hated to lose all its CPDN output while I putz around w/learning my way around Linux &amp; the Linux CPDN client. A spare parts, kludged together, Linux test/learning target machine sounds like an excellent idea!

&gt; Having read the new installation/configuration pages at yoper, I recognize
&gt; that it is not as user-friendly as SuSE YaST (Yet another Setup Tool), but
&gt; YaST is not without issues.

Even though it was the installation of the CPDN LINUX client that was the final 'show stopper', the embedded YOPER Linux tool for the partitioning of my HDD, the creating &amp; mounting of Linux 'logical' drives, and my trying to get LILO &amp; WinXP's MBR to play nicely together, where definitely time consuming &amp; required a much steeper learning curve than I expected. It was an unplanned spur-of-the-moment decision. I didn't even read the documentation prior (I know - RTFM!). I'll definitely be better researched &amp; prepared next time.

&gt; You bit off a tough chunk to chew for a newbie (I dislike that term)

I'll just consider myself "Unix challenged", at this point in time! ;-)

&gt; I share your excitement about Yoper.
&gt; It seems sure to be the vanguard for the next wave of Linux.

Somewhat of an evolutionary step, I suppose. From what I've managed to glean over time, from Overclocker.com's forums &amp; a few other sources, YOPER is supposedly not based on a specific core, like Fedora et al. Apparently its main proponent/architect picked modules from various cores &amp; distros which he felt ran the fastest for thier given function, then integrated them into YOPER, while attempting to further optimize them for speed whenever possible, and also adding i686 capability. Quite a feat, IMO!

BTW - Another guy at OCC has modified Knoppix with optimizations for overclocking, Distributed Computing (particuliarly the Folding at Home project), &amp; HDD intallation. If you ever run into something called "Overclockix", and wondered what it is, this is it. It's supposedly a somewhat faster version of Knoppix (streamlined with less 'eye candy' &amp; extraineous features), and was left cpu optimized at the same level as Knoppix (i286/386???). Pobably not a true YOPER contender, but interesting, &amp; thought I'd mention it if you ever where to stumble accross it.

&gt; Well, as the Thread topic is maximizing performance, I guess we haven't
&gt; strayed too far!

I think (hope?) not. This thread mutated to Unix/Linux after three posts, &amp; Linux definitely appears to be one of the components in attaining maximum CPDN performance. There seemed to be a growing CPDN Linux interest in the CPDN phpBB just as it went down. When it comes back up, hopefully there will be enough interest to sustain a Linux thread relating to CPDN.

&gt; Again, thanks for the input.

None needed, nor expected, but always appreciated! ;-)

/Chicago - out

---

&gt; Hi AstroWX,
&gt; I've been hitting the same problem which seems to be confined to Kernel
&gt; 2.6+ (as used in SUSE 9.1 onwards) and Belkin KVM's especially the SOHO
&gt; varieties. It is intermittent and is not a cabling or mouse problem - bypass
&gt; the KVM and the problem disappears.
&gt;
&gt; An Xserver reset clears the glitch (switch to Text terminal CTRL-ALT-F5 and
&gt; back to current screen CTRL-ALT-F7 is a quick way in SuSE distros). I'm still
&gt; digging round the see if Belkin will own up (they have removed the original
&gt; acknowledgement and fix from their Knowledge base - was in Google archive - so
&gt; I don't hold out much hope). Looks like time to add a replacement non-Belkin
&gt; KVM to my pressie list...
&gt;
&gt; Andrew

Thanks for that, Andrew. I was advised that it was unique to SuSE distros so my "solution" for next upgrade was to be "any except SuSE". (Been with SuSE since 7.1; Caldera OpenLinux before that.) It's good to know that the real solution is a different brand of KVM, not a different distro. Appreciated.


@ Stratcat,

Thanks for the Knoppix info. I doubt I'd go with it, unless there is a HD version. (I thought it was designed to run from CD, which is slow, with the HD mounted 'read only'. [I'm 68 but still learning! {Occasionally.}])

Cheers.

---

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