25 December 2009
Ahh the memories
As mentioned in the previous post I was trying to finalise the memory upgrades. It turns out that 2 of the 5 sets are faulty. One set gave errors when used and I swapped it out in favor of another set. Another set reported 4Gb at boot-up instead of the 6Gb that I paid for.
Both sets were returned to the computer shop early in the week for replacement under warranty. While I was there I ordered another 2 sets, so a couple of the machines (if the memory works) will have 12Gb.
I was reading up on the Intel i9 processor which is slated for release in the 1st quarter of 2010. Its actually an i7-980x. Its a hex-core version of the i7 with hyper threading. Supposedly should be available as a new chip and a BIOS upgrade required to recognise it for my Asus P6T's.
Seti power outage
The Seti project is having a couple of power outages. One on the 27th of December and another on the 3rd of January. Apparently there is some fairly major electrical work going on "up the hill" where the Space Sciences Lab is located at UC Berkeley.
Due to this power outage I have been running down the Seti work and will concentrate on a few other projects for a week or two. For the machines that have 6Gb memory I have started running some climate prediction work (it uses a fair bit of memory).
20 December 2009
I updated the remaining P6T's with Windows XP to Windows 7 (64 bit). They have now all been done.
There are only two 64 bit science apps, the rest are 32 bit. They are the Seti multi-beam and the Milkyway GPU app (version 20b). The good news is that they are faster than their 32 bit versions, so there is a bit of a speed improvement.
After the issues getting the RAM working in Maul I downloaded the latest BIOS for the P6T's and then set about updating all of them. I found of the 5 P6T's they had 4 different versions. As expected Maul had the oldest.
I had some initial problems trying to get the program that flashes the BIOS to work, but it turns out its not needed anyway. There is a built-in utility in the BIOS to flash itself. I burned the BIOS image to a CD and updated all 5 of them using it.
The router continued playing up so I have thrown it in the bin and got the new one going. They are both the same make/model so it wasn't too hard an exercise, apart from having to find a machine where I could specify the IP address in order to talk to the router in the 1st instance. Once its been setup the router assigns its own IP addresses.
Windows 7 drivers
While I was updating things I downloaded a few of the drivers specific to the P6T and installed them, so instead of using the generic Windows 7 versions they have the manufacturer specific drivers. It doesn't make much difference but it might make them a little more stable.
The only remaining task for this year is the memory upgrades, now they all have a 64 bit operating system. This requires a bit of manual labour, pulling machines off the rack (and they aren't light weight). Two out of 5 have already been done.
19 December 2009
Another new version out this week, but I haven't installed it yet. There are a couple of complaints about the amount of CPU time its taking and of cuda tasks being stalled.
Christmas donation drives
Both Seti and GPUgrid have donation drives running. If people would like to donate please use the links below. As mentioned in last weeks post, I made a donation to Seti. In fact to keep things fair I even did one to GPUgrid.
Maul upgraded to Windows 7
Well I decided to be brave and upgraded Maul from Windows XP to Windows 7. The software side of things installed fine without any issue. I was expecting the fact that it had dual GTX260's might cause issues, but they were both found by the nvidia drivers and seem to work fine.
I then went to upgrade the memory and the 1st packet (3 x 2Gb modules) wouldn't work, nor the 2nd. The machine refuses to even POST with the 2Gb modules installed. I put the 1Gb memory modules back in, had to reset the BIOS and its back off and running. Its a standard ASUS P6T so they should work unless both sets are faulty. Possibly its got an older BIOS as it was the 1st of the P6T's that I got.
Yes its been warm most of the week, with one day being hot. So the all the machines were powered off for that day. Usually if its warm i'll only run one or two machines because of the heat they produce. This doesn't do too much for my crunching scores, with my RAC (Recent Average Credit) falling from about 60k per day to around 40k at the moment.
Even the router has been locking up on the warm days so its probably on its last legs. I bought another one today, so if it fails I can just swap it out.
12 December 2009
ATI drivers and things
The machine with the HD4850 (Chekov) had its beta drivers removed and I installed the Catalyst 9.11 drivers. I found a Milkyway 64 bit GPU app version 20b which I also put on it. This got the milkyway app back to working.
I reseated the RAM modules in Chekov as its been giving the "blue screen of death" as its known (a crash dump at start-up). This had the effect of removing the crash dumps but I have noticed the Einstein work units are getting computation errors. I guess the RAM is faulty so i'll swap out the set thats in there and replace with another set of 3 x 2Gb modules.
Nvidia 195.62 drivers
I've been installing these on the machines that don't have them. They support CUDA 3.0 and OpenCL 1.0 code. The DLL files for CUDA 3.0 however are still in beta testing, so they aren't available to the general public yet. Unfortunately the drivers didn't fix issues with the GPUgrid app getting errors on the GTX260 cards.
Another new one released. Well actually a few versions in quick succession. Currently its up to 6.10.24 with some fairly major changes to how it accumulates debt and therefore shares resources with multiple projects. It also has been built using a later version of the libcurl library that doesn't cache IP addresses, so it handles project servers going off-line better.
Its donation time for the Seti guys. I got an email from them asking for a donation to help support their plans for 2010. Interesting this one was "signed" by Dan Werthimer. He normally keeps a low profile on the web site. One of the volunteers (Blurf) who used to do the donation drives has decided to take a break from Seti for a while.
While many people are happy to do crunching for Seti only a small percentage (I think its about half a percent) actually donate to the project. Anyway its out with the plastic to help them out.
06 December 2009
NEZ forced resignation
This guy was the top ranking user at Seti@home. Apparently he was the IT director for Higley Unified School District. Its alleged that he was running BOINC on their computers without their permission. Seti@home project scientist Dr Eric Korpela has posted an FAQ and comments:
They have released the 195.62 drivers which support CUDA 3.0. Unfortunately the CUDA run-time and the FFT libraries are still in beta. I installed the drivers on Kirk and ran some Seti work and then some GPUgrid work using CUDA 2.3 libraries. This worked fine on Windows XP so I have now installed it on Spock. Its running a GPUgrid work unit at the moment on Windows 7 64 bit.
Windows 7 upgrades
I haven't updated the other machines yet as I am concerned about the multi-GPU issues with the Windows 7 drivers. From what most people have been saying it requires the use of dummy plugs on the back of the graphics card, otherwise windows won't recognise the 2nd GPU. I don't have any dummy plugs.
As mentioned last week Chekov has crashes at startup. I think this is being caused by the beta ATI drivers which worked fine in 32 bit, but after upgrading to Win 7 64 bit didn't. I will be putting the latest release catalyst drivers on it this week.
30 November 2009
Its been hot most of this week so most of the computers have been off. The old work horse Maul has been running every night when things cool down along with one of the other machines.
Einstein does cuda
The guys at Einstein@home have released a hybrid app that uses both a cpu and a gpu. It seems to saved all of one hour (from the usual 5 and a 1/2). So all in all not a particularly good improvement. I have disabled the gpu from use on their project as i'd rather run work that will fully utilise the graphics cards.
I have been concentrating on Einstein and GPUgrid in the hope of getting the Einstein scores up a bit. Unfortunately it doesn't seem to have made much difference, except to my pending results.
Chekov doesn't like beta
I have installed the ATI beta OpenCL drivers v2 on Chekov. Since upgrading to Windows 7 x64 it has been crashing on the 1st boot-up. After it reboots it seems to work. I didn't have this problem with the same drivers when running 32 bit. I presume its badly written drivers. As its been hot I haven't had a chance to install the latest release drivers (instead of beta) due to Chekov being off.
I did try the Milkyway gpu app but they all failed. Unfortunately their app was written using an older SDK and the CAL/Brook+ DLL supplied is rather old.
22 November 2009
I was going to upgrade Chekov to the 64 bit version of Windows 7 next, however they guys at Lunatics (they optimise the Seti apps) brought out a hybrid Astropulse application that runs on both the CPU and an ATI graphics card. I was beta testing it. I ran into some issues with earlier drivers needed for the Milkyway science app and the Seti app being built with a later SDK that required newer drivers. After resolving that issue it seems they take around 10.5 hours to run instead of the normal 16 hours on the cpu only.
Its summer here in Sydney, Australia so we are now into the hot weather. Today is forecast to get to 41 degrees. All the machines have been powered off during the day as the loft they are in gets even hotter. If its forecast to get over 30 I even turn off the file server and router.
Nvidia CUDA 3.0
Nvidia released a 2nd beta driver with cuda 3.0 support. The cuda run-time DLL's are only available for developers at this point and unable to be distributed to the public. It seems that the programs will need to be recompiled with the new cuda compiler, but there appears to be no additional speedup or improvements. The developers who have been looking at it suspect its additional functionality for the upcoming G300 (Fermi) chip.
I did try the 1st beta driver using existing CUDA 2.3 DLL's, however the GPUgrid app failed. It was run on Maul, which has dual GTX260 cards. They last behaved with the 182.50 drivers. I didn't bother with the 2nd beta and will probably wait until there is a "release" version.
07 November 2009
Spock updates to Win 7 x64
Spent a few hours upgrading Spock to Windows 7 Professional (64 bit) today.
The first part was to install windows, this went smoothly. I then installed the nVidia 190.62 (64 bit) drivers and a bunch of windows updates. Next I installed the 64 bit version of BOINC. This went fine too. I then copied the BOINC directories that I had backed up. When I started up BOINC it then complained that it couldn't connect to the core client. After a bit of fiddling I did a reinstall with the repair option. That fixed it. Still running the 32 bit apps at this point. I tentatively let it download some work for Seti, as I use an app_info file for it. It downloaded a bunch of cuda work so I stopped it asking for more. I used the ReSchedule tool to shift some of them over to the cpu to make sure they run fine.
I shut it down again and removed the existing 3 x 1Gb RAM sticks and installed 3 x 2Gb sticks. I restarted it and we're off! Everything seems to be running fine now. I am waiting for it to finish off the Seti work before I try updating the Seti multibeam science app to 64 bit. Its the only Windows/Seti optimised app to have a 64 bit version.
Replacement GTX295 arrived
I got a replacement GTX295 on Wednesday. It arrived unannounced. Its the older dual PCB version. So I complained to the company I purchased it from pointing out that I purchased the single PCB version. The guy on the other end tried to tell me they were newer. After that I had to send an email showing i'd ordered the single PCB version, which I did. The guy rang back and then tried to tell me there weren't any in stock and I was mistaken where they were shipped from (Brisbane). I suggested they might like to replace it with a different brand as long as its the single PCB version or refund my money.
31 October 2009
BOINC 6.10.17 release
This version became the "official" release version yesterday. I have it running on one machine and all the others are one minor version behind. There wasn't much difference between them.
I forgot to mention last week that the faulty GTX295 got returned to the distributor. I'll have to chase them up and see if they are going to give me a replacement soon.
Its 31 degrees in the loft at the moment, so all the number crunchers, except Maul are off. When things cool down overnight I usually fire up one or two of them.
Win 7 upgrades
No news from the shop, so they obviously don't want my money. Not that i'm in a hurry to upgrade the machines. There have been a few reports of work units taking longer to run, particularly on the GPU's after people have upgraded.
The guys at Seti have finally plugged in the SSD's I bought. No news yet on how well they are doing and if it has helped alleviate some of their MySQL bottlenecks.
Seti out of work
Seti ran out of work due to the radio telescope (Arecibo) being switched off for maintenance. No work coming from them for a week or two, except some old data the guys have been running through a software radar-blanker. They are prone to interference from radars located near the radio telescope. The data they are software blanking is from 2006, before they had a hardware based solution.
In the mean time I have been concentrating on Einstein as its scores are not very high compared to the other projects I run.
24 October 2009
We had a few hot days this week so had the entire farm powered off during the day. As a result not too much progress.
Seeing as it was finally released I bought an upgrade to put on my laptop. It had Vista which never worked properly. Its now running Windows 7. I have have a couple of issues since then:
- Wireless LAN keeps dropping out
- The fingerprint s/w (Verisoft Access Manager) isn't compatible and had to be un-installed
See previous posts about these. I spoke to the shop regarding the Windows 7 and memory upgrades. They are getting pricing for me. I was thinking of going to 6Gb per machine but with the i9 coming in Q2/2010 it might be worthwhile going up to 9Gb or even 12Gb.
We are up to 6.10.16 this week. It addresses issues for users that use remote access to their machines and BOINC (used to) kill their GPU tasks. As I don't use remote access software I never had this problem anyway.
ATI version of Astropulse
I've been trying to test this new hybrid application. As CAL/Brook+ doesn't have any FFT functions they have to be done on the cpu and the other bits can be done on the gpu. Unfortunately it won't work with the Catalyst 8.12 drivers. These older drivers are required for the Milkyway optimised ATI gpu app. The 9.x series Catalyst drivers crash.
17 October 2009
GPUgrid hits 6 million
I managed to hit 6 million credits on GPUgrid. Below is my certificate.
The donation drive at GPUgrid seems to be stuck on 2% of their target (50,000 euros by the end of this year). While I am the top donor, I am more than happy to lose the title if some company or individual would like to support them.
Another new version of BOINC came out today with a fair few changes so I will give it a try on one of the machines to see how it goes. The previous version (6.10.13) looks pretty good so far and I don't need the newer features, but some of them would be nice to have. The change history can be found on the forums at GPUgrid, Seti, Milkyway and of course the BOINC message boards.
Return of dud GTX295
I got a Return Authorisation number for the faulty GTX295, so I will have to drop it off with the distributor (Altech) next week.
ATI card settles down
The ATI seems to have settled down with the 8.12 drivers. While not the latest they seem to work with the Milkyway optimised app for CAL/Brook+. It completes a work unit in 65 seconds on my HD4850 card.
Unfortunately due to the policies over on their project it gives a maximum of 48 tasks at a time, making it almost impossible to build a decent cache of work. This means everybody hits the server all the time uploading, reporting and requesting new work.
Overall not a great deal of progress this week as there were a couple of warm days and I had the machines powered off during the day. Today had a cool breeze that kept the temperature down so I could have most of the machines running.
09 October 2009
Seti hits 7 million
I just hit the 7 million credits for Seti tonight. Below is my Setificate.
The software for the ATI has been causing some grief, with driver restarts whenever you click on anything on the screen. I have tried a number of the Catalyst drivers including 9.1, 9.2, 9.8 and 9.9 which all seem to have this problem while running CAL/Book+ applications. At the moment I am now trying the 8.12 driver.
Despite these driver restarts I have managed to get Milkyway@home (one of the BOINC projects that has a CAL/Brook+ application) up to 64,000 credits. Its taking approx 65 seconds per work unit on the HD4850 card.
A couple more BOINC versions out this week. Currently its up to 6.10.13 (which I have running on two machines). It all seems to be behaving well on both the ATI and the nVidia equipped machines. The developers think it might be stable enough for a release soon.
05 October 2009
This Saturday I picked up the other two i7 machines and got them home. One I had previously (Spock) but it had the wrong motherboard so it went back. It was already setup so all I had to do was take it out of the box and plug it in.
The other one has been named Chekov. It however differs from my standard setup in that it has an ATI HD4850 graphics card. Below is the GPU-Z screenshot for it. It was fate really. If all three of the GTX295's worked then I wouldn't have used it on its own, but seeing as one doesn't...
A couple more BOINC versions were released for alpha testing this week. I don't think this is the last of it as there have been a couple of bug fixes done since the last build. Its up to 6.10.11 which seems fairly stable so far. Having run it for a while on Kirk without any issues I have updated all of the machines to it.
ATI science apps
There are currently only two science apps available for the ATI cards. They are Milkyway@home (which I have attached to) and Collatz. There is a Seti@home Astropulse app being developed using CAL/Brook+ code but it has yet to be released.
CUDA has better software support and is much more advanced than CAL/Brook+ is. OpenCL, while supposedly being supported by both vendors, is even further behind. Unfortunately most of the projects see OpenCL as the longer term solution and don't want to develop new apps.
Future plans for "the farm"
There are a couple of things on the horizon, but not yet released so will have to wait a while for them to eventuate, if they happen at all. In expected release date order they are:
- Windows 7 x64 upgrades along with memory upgrades to 6Gb
- GT300 based graphics cards
- Intel i9 CPU upgrades
27 September 2009
This week saw Spock go back to the computer shop so it can be turned into a standard P6T instead of the deluxe v2. While it worked a treat they only have a single PS/2 port on the back which means the deluxe doesn't work properly with a KVM. They also have one of the PCIe slots spaced 3 slots away from the other so would be ideal for a dual graphics card setup, if you don't use a KVM.
I got one standard P6T back from the shop. Its been named Sulu. As you can see I am following a Star Trek theme here. Its off and running with my supposed standard setup of a P6T motherboard, Intel i7-920, 3Gb of RAM, XP Pro and a GTX295.
Dust storm hits Sydney
We had a dust storm hit Sydney on Wednesday. I woke up in the morning and the sky was a red color. As there was a strong wind blowing I had to close all the windows and doors to keep the dust out. I also had the "farm" powered off for the whole day as it would get too hot without opening the window or using the air conditioner.
The guys at TechPowerUp released an updated version of GPU-Z. I let it loose on Sulu now that it seems to understand the single circuit board versions of the GTX295. Below is the screen shot. Unfortunately my assumption that it had a 45nm GT200 chip was incorrect.
New ATI cards released
ATI released a couple of new cards to the market this week. The HD5850 has 1440 shaders and the HD5870 has 1600 shaders. This means the HD4850 I got about a month ago is no longer available.
Nvidia has also stopped making the GTX260's and GTX295 as they get ready to release a new card. This may mean I can't get the dud GTX295 replaced, but we will see. I might also want to see what they release to compete with the new ATI cards. A potential upgrade coming?
There were further BOINC versions out this week. We are up to 6.10.9 which I have installed on Kirk. The other machines are all on 6.10.7. The main difference from last weeks one is some bug fixes around the cuda tasks being immediately preempted and some additional work around supporting ATI cards.
Apparently ATI is not providing any help to the projects to support CAL/Brook+ so there are only a couple of projects that currently can use these cards. ATI seems to be concentrating on hardware. Its useless if you have a super fast card if there is no software (apart from a few games) that can use it.
19 September 2009
There were a couple of new versions of BOINC released, unfortunately both have an annoying bug where it preempts cuda tasks. I have only installed it on one machine which will be getting reset back to 6.10.3 after I finish updating this blog.
Ongoing issues with P6T deluxe
I am still having issues with the new P6T deluxe supplied by the computer shop. I have told them to supply the standard P6T motherboard seeing as this one is causing grief. Its also what was quoted and accepted. At the time their supplier substituted the P6T deluxe v2 motherboards as they didn't have stock. This morning I dropped one back to the shop (fortunately its been sitting in the garage all week still in its box). The third one hasn't been delivered yet, so it too will built as a standard P6T.
As mentioned last week, one of the GTX295's has problems. I have swapped it out for my GTX275 in the mean time to get the machine going. I downloaded a cuda memory tester app and when I get around to installing the faulty GTX295 back in a machine will run the test on it.
As I no longer have a spare nVidia graphics card I have given the ATI HD4850 to the shop to install in the third P6T. I was hoping to put a GTX295 in there (or the 275 temporarily).
A new record
And you thought I was a hard-core computing person... I spotted this link over in one of the GPUgrid forums. Its about a new world record computer sporting 4 x GTX295 cards. The company in question has an app which can utilize all those processor units.
The link: http://www.manifold.net/info/pr_gpu_record2.shtml
GPUgrid taking donations
This is a medical science project and they are looking to raise 50,000 euros to fund one of their staff for a full year. They are now taking donations. I got stuck while trying to do a credit card donation last week. I kept my promise this week as they now can take PayPal donations. The highest donor can choose which disease they will concentrate their analysis on next.
15 September 2009
Carrying on from the last post...
I went back to the computer shop and they supplied a couple of PS/2 to USB convertors. This got the machine up and running. I reinstalled Windows. Couldn't get it to recognise the LAN ports until I installed the drivers from the CD (the standard P6T doesn't cause this much grief). I then could see the network, apply service pack 3. I then downloaded another 51 updates from Microsoft and at least its up to date. Next was to setup windows the way I usually have it. Then installed BOINC. So far so good. Told BOINC to fetch some work so it gets a pile of cuda work from Seti at home and they all start erroring out. Suspend Seti work and let it get a GPUgrid task. It causes a blue screen of death. I reinstall the nvidia drivers and try again. This time it locks up the machine. At this point its gone midnight and I have to work the following day so I power it off and head to bed.
Today after work I come home and decided to swap out the GTX295 with a GTX275 that I know was working. Reinstall nvida drivers one more time. Start up BOINC again and it seems to be working so far. Its even screaming through Seti cuda work at about 2 minutes a pop. Pity the upload server isn't working so I can't upload results but they can wait a while.
My conclusion is the GTX295 is a dud and I will need to arrange for it to go back to where it came from (in Brisbane). I haven't taken the other machine out of its box yet. Hopefully its GTX295 will work. Oh and the name for this one is Spock.
Another strange thing about this one is the shop wired up the fans so they are controlled via the knob on the front of the case. That makes it quite messy inside as they daisy-chained power cables so they could reach. The fans can be plugged into the motherboard. If they are plugged into the motherboard then the computer can control their speed depending on how hot it gets. Thats they way I prefer them wired up. Looks like i'll have to rearrange the cabling next.
13 September 2009
I also got the extra power cable for the Corsair power supply and it doesn't seem to work either. I can only assume that with 3 graphics cards the power supply is overloaded (despite it being a 1000w).
It was also rather hot so all the machines were switched off. Overall not a very successful weekend.
New i7's arrive
Two more i7's arrived to join the "farm". I have yet to get them out of their boxes. I have rearranged the rack to accomodate them. Their specs:
Motherboard: ASUS P6T deluxe
RAM: 3Gb DDR3
Processor: Intel core i7 920 (2.67Ghz)
Graphics: Palit GTX295
You can see the two remaining quaddies on the middle shelf. The beige coloured machine is the file server. The quaddies are just finishing off the last of their work before I decommission them.
I have been testing the latest BOINC version as they now support the ATI graphics cards. We reached 6.10.4 but it unfortunately had some issues and I had to go back down to 6.10.3. There have been changes since then so I expect a couple of new versions next week.
ATI graphics card
The computer shop finally managed to get a Corsair PCIe modular power cable in. I haven't opened up Maul to see if it has the right connectors yet. The PCIe end is fine, its the power supply end that is questionable.
07 September 2009
And here is another shot, you can see it has a single fan in the middle, with a heatsink on either end, enclosed within the plastic casing.
03 September 2009
It seems to be running a fair bit cooler than the GTX275 that it replaced. This GTX295 is a single circuit board design with a single fan. The GTX275 has dual fans, so I suspect that being a newer card this one may have the 45nm G200 chips.
Two more i7's on the way
The shop has yet to get all the parts to make another two i7's, however they should arrive in the next few days. They still expect to be able to build them next week and have them ready for delivery next weekend.
ATI graphics card
Meanwhile the shop are still waiting for the modular power supply cable for the Corsair power supply in Maul. Once that arrives I might be able to get the ATI card working. I had to buy this as a spare part because I have used up all the PCIe power cables that came with it. Maul has two GTX260's and they use two power cables each.
28 August 2009
Palit GTX295 grahics cards
The 3 GTX295 cards I ordered arrived. Two have been dropped off at the computer shop to go into the new i7's. The 3rd is going into Kirk. I will be removing the GTX275 that is currently in there. It appears the computer shop forgot to give me the bag of cables for the modular power supply. They used a Y split one to run the existing card, but thats not going to cope with a GTX295.
ATI HD4850 graphics card
The shop have yet to order a power cable for the Corsair modular power supply that is in Maul, so I can't install it either. Apparently its getting ordered next week.
BOINC 6.10.0 testing
The CAL/Brook+ aware version of BOINC came out last week and has been installed on Maul. Its working fine although there are reports of numerous problems with the 64 bit versions. I run the 32 bit version. BOINC 6.10.1 was also released but I haven't tried it.
22 August 2009
Here you can see the 20cm fan in the top of the case, the 12cm Noctua fan on the heatsink and the 12cm fan on the back of the case. The card under the heatsink is the Palit GTX275. Just in front of the heatsink fan you can see the RAM modules.
In this shot you can see the end of the graphics card with the PCIe power connectors, the drive bays and on the front of the case another 20cm fan. Down the bottom an extra 14cm fan and at the back the Seasonic power supply.
This is another shot showing the fan at the top of the case and you can see the fans on the GTX275.
21 August 2009
Got myself an ATI HD4850 video card today. The shop must have seen me coming as it was sitting on the counter when I walked in. I'm not too sure which machine i'll be putting it in yet, probably try it in Maul and see if the ATI drivers will play nice with the Nvidia drivers.
Part of the reason for the ATI graphics card was because BOINC 6.10 was released earlier in the week with ATI support. Now all we need are some science apps that will work properly under BOINC using the ATI cards. There are science apps that use the ATI, they just doesn't run under BOINC.
I have ordered three GTX295 graphics cards to go into the new i7's. If they work out i'll replace the GTX275 thats in Kirk with one as well. These ones are supposed to be the newer single circuit board ones.
I have cleaned off two of the quaddies and shut them down. I'm not too sure if I should remove the GTS250 cards, power supply and put the old power supply back in and sell them or just sell them with the GTS250's. Its not urgent to dispose of them as the other i7's won't be available for a couple of weeks.
As mentioned before the current plan is to have 4 x i7's running. These will replace the existing i7 and all the quaddies, one of which has already gone to a new home. I had this vague idea that I might get an AMD 6 core machine and put 3 graphics cards in (the x2 variety so it has 6 gpus) and all it does is feed the graphics cards. I need to do a bit more research into this before it happens. The x2 graphics cards (ATI or Nvidia) use a lot of power so this may not be feasable.
Einstein have got their file server back online after almost a week. It doesn't appear to have lost any work. Seti are now also pumping out work, so I have finished the Climate prediction work, which was only added because I didn't have any other working projects at the time.
13 August 2009
GPUgrid out of space
GPUgrid have been having space problems with their upload server. This means I can't upload any results. Aparently it gets filled up by a log file. So no new work from them and can't upload completed work.
Einstein file server crash
The file server over at Einstein has crashed again (about the 3rd time in as many months). There is a simple message saying it will take a few days before it will be rebuilt. Strangely I can upload some work (ABP1) but not the other work units. I can't report any work though.
Seti out of work
And just to make things interesting Seti can't produce work units fast enough. Funny how you get a fast machine and then can't get work for it.
Given all of the above problems I have attached Kirk to Climate Prediction (their smaller work units take approx 100 hours). That should keep it busy for a while, when it can download the huge files.
12 August 2009
The case is a CoolerMaster Storm Sniper. Inside is an ASUS P6T motherboard. The CPU is an Intel Core i7 920 with 3Gb of DDR3 RAM. This one has a Palit GTX275 graphics card installed.
Another photo showing the controls and connectors on the top.
In this shot you can see how big it is compared to the quad-core machines it is replacing (Qui Gon Jinn).
Another shot showing the top panel. The twist-knob is a manual fan speed control. The push-button switch in the centre of the knob turns the fan LED's on and off.
And a front-on shot of it. As you can see lots of grills so plenty of airflow. It has a 14cm fan in the bottom blowing cool air up into the rest of the case. The power supply (a Seasonic 700w) is mounted at the bottom.
09 August 2009
Einstein cuda app
This weekend I have been testing the first beta of the Einstein cuda app. While its a 1st beta it seems to run fine on one of the quaddies. The drawback is that it uses a whole cpu to feed the gpu. I completed a 1st test run on Luke. At the time I write this I have a 2nd test running on Maul.
GPUgrid and 190.38 woes
It appears almost all the complaints about failed work on GPUgrid when using the nVidia 190.38 drivers are all GTX260 cards. It doesn't seem to matter if they have 192 or 216 shaders. I have two in Maul and can't use it for GPUgrid any more. Any driver version above 182.50 fails.
02 August 2009
I'm off for a weeks skiing so the "farm" will be switched off, except for Maul.
Maul overstretched itself last week by requesting too much work. This is caused by the ReSchedule tool swapping work to the cpu, so BOINC then thinks the gpu doesn't have enough and request more. The solution is to set all the projects to No new work (NNW) so that it can finish off what its got. Its been in this mode for the last 3 days.
The GPUgrid project (which does medical research) released a new cuda 2.2 application at the end of the week. I've already had one machine lock up and had to abort the work unit in question. Maul, who would be ideal for these work units has been in NNW mode for the last 3 days so it hasn't picked up any new work or the updated application.
After an on-going discussion over on the Seti message boards about the MySql log files being a point of disk contention we came up with the idea of using Solid State Disks (SSD's). These are of course fast and expensive when compared with a traditional hard disk. I made the offer of getting a couple of them for the project (Intel X25-M). I ordered 2 from a US web site with them shipping directly to the Seti@home project. They are due to be delivered on the 6th of August. Hopefully the project can put them to good use.
28 July 2009
Since the nvidia 190.38 driver came out last week, GPUgrid work has been failing on the GTX260's in Maul for all work units. The GTS250's in 4 of the quaddies work fine. This points to an incompatibility with the G200 based cards, which are the high-performance ones.
While its possible to go back to an older driver, Seti positively screams along now with GPU work units taking as little as 2 minutes when using the cuda 2.3 DLL's. If I downgrade I can't use cuda 2.3 and lose all that speed. I am hoping that the GPUgrid cuda 2.2 app, due later this week, will work properly with the current drivers. In the mean time Maul is concentrating on Einstein and Seti work.
As mentioned in my last post Yoda is up for sale. You can see it on eBay here: http://cgi.ebay.com.au/ws/eBayISAPI.dll?ViewItem&ssPageName=STRK:MESELX:IT&item=300333014651
Its my intention to also decommission all the other quaddies and sell them off too. This is mainly because of needing the space for the new i7's and also because of the amount of power needed to drive them. The energy bills have gone up 20% since the 1st of July due to price increases.
25 July 2009
Quaddies for sale
Yoda is for sale on eBay. I need the room (and money).
I have ordered 4 x i7 machines to replace all the quaddies. They are ASUS P6T's. In the mean time I have dropped off the 700w power supply and the GTX275 to be installed in the 1st one. If its reasonably quick i'll go with the same graphics card in all of them.
Cuda 2.3 released
Nvidia has released cuda 2.3 along with an updated driver during the week. Initial indications are that its about 30% faster than cuda 2.2.
I have upgraded 3 of the machines so far. The other two have to finish off their cuda work before I can upgrade them. Meanwhile GPUgrid are updating their app to use cuda 2.2 next week. Hopefully it won't take them too long to get to 2.3.
20 July 2009
Trashing cuda work
I upgraded the "stock" Seti cuda app to the optimised one and stuffed up the app_info. That resulted in all my Seti cuda work being deleted. To top things off I couldn't get any new cuda work units until much later due to the Seti servers being overloaded. It took me a short while to fix the app_info problem, but much longer to get any work unit to prove it worked. Once I had it working I upgraded the remaining machines with the new app and app_info files.
After the cuda app upgrade I decided to relocate the router and the 8 port switch. So I unplugged the switch, moved it to its new location and plugged it all in. No problem the machines only use the network during my ISP's off-peak times (1am to 7am).
When I checked the following day all the machines were complaining about no network connection. After checking a few things it seems the network cable had come out the back of the file server. I swapped the cable for another one and the network came back to life. After that I manually had to allow the machines to get work as they had all run out of work.
18 July 2009
Updated for Astropulse 5.05 and cuda 2.3
If you are not comfortable with editing an app_info then this is not for you. You would be better off using the Unified Installer for Windows available from the Lunatics web site see http://lunatics.kwsn.net/optimized-applications-release-news/lunatics-unified-installer-for-windows-v0-2.msg19610.html;topicseen#msg19610
BOINC is very unforgiving of an incorrect app_info and usually will delete all tasks if you get it wrong.
Do NOT use Internet Explorer to edit the xml files, it will stuff up your app_info. Use Notepad or another text editor.
Upgrade your BOINC client first and get it working before changing anything else. At the time of writing I am running the 6.6.37 client, although any client from 6.6.15 should be sufficient.
The app_info.xml below is based on a Windows XP platform (32 bit) and the cuda-capable card is a GTS250. If you are running on another platform you may need to add/amend the <platform_name> tags.
My computers support the SSSE3 instruction set. SSE2, SSE3 or SSE4.1 may be more appropiate for you. You will need to amend the program names as appropriate in the app_info.
I've assumed that you have your cuda-capable card up and running and have the necessary nvidia drivers (minimum version is 180.48). Use the 190.38 drivers (or later) if you want to use cuda 2.3.
a) Optimised multibeam and optimised Astropulse, available from the Lunatics web site
b) Cuda multibeam and support libraries
1. Download and install BOINC. Get this working before changing anything else.
2. Empty your cache of Seti@home work. This is best achieved by setting the project to No new work and letting it finish off its tasks. Make sure they are all uploaded and reported, there should be none on your tasks list. If you are feeling brave/confident then skip this step, but if all your tasks get deleted don’t say you weren’t warned.
3. Download the optimised multibeam and astropulse apps if you don't already have them.
4. Download the cuda multibeam app (from the Seti web site) if you don't already have them. If you run the stock cuda multibeam app then you should already have these in your projects\Setiathome.berkeley.edu folder.
You can also use the optimised cuda app available instead of the “stock” app. You will need to change the name in the app_info.
5. Disable network communications in BOINC.
6. Shutdown BOINC. Make sure it and the science apps are shutdown.
7. Browse your client_state.xml file (its in the BOINC data directory) and look for the entry <p_fpops>. We need to use this number. Do NOT change this file.
8. Browse the BOINC log file to get the estimated speed of your GPU or before you shut BOINC down click on the messages tab. This is usually given at the top of the log in Gflops. 9800GT’s are estimated at 60Gflops, GTS250’s at 84Gflops and GTX260 (216 shaders) at 96Gflops.
9. For each of the apps multiply the p_fpops value by the factor below and put this into the appropiate flops entry in the app_info given below. For multibeam 608 you need the estimated Gflops. The app_info given below has the values for a GTS250.
Astropulse 503 p_fpops x 2.6
Astropulse 505 P_fpops x 2.6
Multibeam 603 p_fpops x 1.75
Multibeam 608 Est.Gflops x 0.2
10. Make sure you have all the programs listed above in the projects\Setiathome.berkeley.edu folder. If not copy them there.
11. Save your app_info in the projects\Setiathome.berkeley.edu folder.
12. Start up BOINC. Check the messages tab to see if it lists any [file error] messages. If there are shut BOINC down, check you have the correct program names referenced. Go back to step 10.
13. If okay then enable new work for the Seti@home project.
14. Enable network communications again.
15. BOINC should now download work of all types. If not check your Seti@home preferences on the Seti web site, that Astropulse_v5, Astropulse_505 and Allow graphics processor are all ticked. If you have a slower computer you may not get Astropulse work units anyway.
Seti upload woes
Seti continues to have upload problems, but there have been a couple of suggestions made about resolving it without them having to run a 1Gbit fibre optic cable up the hill. They have a 1Gbit link which stops at the bottom of the hill, however they are at the top and only have a 100Mbit link from the bottom of the hill to the top where the Space Sciences Lab is located. There are also some software tweaks being applied on the BOINC client end to do with retries.
Einstein download throttling?
On the subject of file transfers I have noticed that Einstein@home seems to be downloading at 10Kbps (on average) when it used to be in the order of 50-100. Uploads however are quite quick, around 30Kbps. I did post a message in one of their forums but have received no answer. I suspect that they are throttling connection speed so their server doesn't fail again.
Astropulse 5.05 released
The optimised AP 5.05 app has been released for Seti@home. I installed that during the week on all the machines and updated my app_info files. Since then I have only seen a couple of Astropulse work units. Seti is mainly producing Multi-Beam work units at the moment, which is one of the reasons why they are having upload issues. Multi-beam work units don't take as long to process as Astropulse and are smaller. I'll post my app_info and instructions soon.
ReSchedule 1.9 released
The Reschedule program was updated to 1.9 during the week. It now leaves the VHAR (very high angle range) work units on the gpu instead of shifting them. The VLAR (very low angle range) work units are the ones that cause the Seti cuda app to take much longer to process. I ran around and installed it on all the machines with gpus.
11 July 2009
Seti bandwidth issues
Seti has been having continual bandwidth issues for the last 2 weeks. Its made worse by the fact they have a scheduled outage every Tuesday to perform database maintenance. After the scheduled outage the whole world tries to connect to it, so it hits the proverbial brick wall of 100Mbit. It has been taking most of the week to clear the backlog and get back to normal, just in time for the next outage.
I did try and install the GTX275 into Qui-Gon during the week but had a problem. The card just fits (there is about 1cm from the end of the card to the drive bays at the front of the computer). Unfortunately there are a whole bunch of connectors on the edge of the motherboard that it seems to sit on top of. It looks like it might have to wait until another machine comes along before I can use it.
KVM and case fans
The 25 foot KVM cable arrived during the week. I have now plugged in Maul to the new KVM even though its around the other side of the room. Suprisingly I can now use a PS/2 mouse, which the old KVM seemed unable to do.
I also picked up a bunch of Noctua 92mm case fans. I have 5 quaddies at the moment and only 3 of them have case fans. I'll be installing one this weekend for the machine that has a GTS250.
While I was in the computer shop I was told that there are some price cuts coming from Intel, so they will try and reprice my quote for 4 x i7 machines.
04 July 2009
This is really good because the VLAR and VHAR work units take forever to run on cuda, but run as normal on the CPU. I've installed it on all my cuda machines after 1st testing it on one machine.
I have Qui-Gon running out its cache of cuda work. I will put the 700w power supply and the GTX275 in there. That will free up a 500w power supply and GTS250 for Yoda. I haven't upgraded Yoda as its still having issues with cpu cooling. I need to get ACER out to look at it. If I don't upgrade Yoda then I haven't voided my 3 year warranty.
I am still waiting on the 25 foot KVM cable for the Belkin to arrive, but its not urgent as I have an old screen and keyboard/mouse on Maul at the moment.
28 June 2009
27 June 2009
This week has seen 2 of my 3 projects go down.
Seti with its various issues with database size and sheer number of users. Even now it can't hand out work fast enough to keep up with all the computers around the world. Its up now. Uploads take a long time to get through and there is no work available.
Just to cap things off Einstein's file server crashed a couple of days ago. No word yet on when it will be back, just a simple web page saying they are looking into it. The message boards and all file transfers are gone.
GPUgrid is still working, but I only have a few GPU's so there is only so much work I can download for it. All the machines, except one have run out of CPU work. The one remaining machine will run out of work in about 1 hour.
I still haven't put the new power supply into Yoda yet and therefore it hasn't got the GTX275 installed either.
Windows update has been giving out Internet Explorer 8, so i've installed it today on all the machines that have been offered it so far (about half of them). I don't install windows updates automatically as it can reboot the machines while they are working.
The new 8 port KVM is working fine, however i'm still waiting on the 25 foot cable to arrive. It will be used to control Maul on the other side of the room. At the moment Maul has an old 15" LCD screen along with a keyboard and mouse.
20 June 2009
Top shelf, Left at the back a thermometer and mouse in the foreground. Middle: A 22" LCD monitor, the (new) Belkin KVM and a keyboard on top. Right: Internet router at the back with the wireless antenna poking up and in the foreground an 8 port switch.
Middle shelf, Left is Yoda, Middle: Qui-Gon. Right: Luke
Bottom shelf, Left is a P4. Middle: Obi-Wan. Right: Anakin
The beige colored machine is a P4 used as a file server. The other 5 quaddies are identical except Yoda who doesn't have a graphics card (yet). Not shown is Maul as its moving to the other side of the room in order to distribute some power as this lot are all connected to the one power point.
19 June 2009
The long KVM cables ordered via eBay arrived today. The 4 port KVM I also bought at the same time (from a different seller) is yet to turn up. Not to worry - I ordered a Belkin 8 port KVM and cables anyway as I don't want to run more than one screen and keyboard. Its already arrived at the PC shop so i'll have to drive over there and pick it up on Saturday, as its not exactly small. I will be selling the old KVM and cables on eBay if anyone is interested.
While in the PC shop on Monday I ordered the above Belkin 8 port KVM and also a GTX275 card. I took the spare power supply that I hadn't opened back to the shop and swapped it over for a 700w one (they didn't have any 600w ones in stock).
It looks like back to rearranging the loft once I pickup the bits this weekend.
12 June 2009
I bought another 4 port KVM switch and a couple of cables on eBay. This should allow me to rearrange things a bit so I can put all the machines (except Maul) on shelves and run them all at once. I also need to buy another set of shelves.
I got a quote for 4 i7 machines to replace the 5 quaddies. Without any graphics cards the shop wanted $2,000 each. I think an i7 is about a 50% improvement over a quaddie. Using 150% of the price of a quaddie that means they should cost $1,800 each, so I suggested that if the shop wants the sale they need to get to this figure.
I have one CPDN work unit to upload and report, then I can detach from the project. Unfortunately their upload server ran out of disk space and has been off-line for over a week. I am unable to upload it.
One of the guys on GPUgrid had an interesting i7 machine. It was an asrock machine with 4 x GTX295 cards, giving a total of 8 CPU's and 8 GPU's. He said that it had 4 x 700 watt power supplies to drive it.
03 June 2009
In other news I have started testing the latest BOINC version. I know I said that I won't be testing it, but it has a couple of fixes in it that are useful. As of tonight its up to 6.6.33.
My Seti 10 year anniversary T shirt arrived in the post yesterday, so i'll see if I can organise a photo and put it online.
29 May 2009
GPUgrid still fails when running the current release drivers with their application. They believe its a problem with the cuda compiler and have passed details onto nvidia. As it might take a while until a fix is available I am considering going back to my original project mix (Einstein, GPUgrid and Seti) and downgrading the drivers on the quaddies. This should also relieve pressure on my ISP's monthly download limit as CPDN (climate prediction) work units are huge when they complete.
The BOINC development team are gearing up for some changes to facilitate BOINC and GridRepublic (an account manager add-on for BOINC) to use Facebook. As I don't use either Facebook or an account manager I am not planning on testing this time around.
I got another 500 watt Seasonic power supply for the last remaining quaddie to get upgraded but have yet to install it.
I have some issues at the moment preventing all the machines from running at once:
- The amount of power they all use, which would overload the circuit they are all currently hanging off.
- They are all in the loft which gets hot in summer so they overheat
- I don't have enough keyboards/mice and screens to run one file server, one i7 and 5 quaddies.
21 May 2009
At the beginning of the week I upgraded all the quaddies with the latest nvidia drivers (185.85). This made all the GPUgrid work promptly fail. Given the project were aware of these issues (other people were also reporting problems) I decided to run Seti cuda work only on these machines. As a consolation I downgraded the i7 to the 182.50 drivers and its running GPUgrid work on its two GTX260's.
BOINC 6.6.28 became the "recommended" version, which is good because I have been running it on all the machines for a couple of weeks now. It still has issues with work-fetch but overall its reasonably reliable.
In order to keep the CPU's busy I have attached 3 of the machines to CPDN (climate prediction). This is because BOINC doesn't get enough CPU work and gets too much GPU work. The smallest CPDN work units take around 5 days straight to complete.
15 May 2009
The quest for more credits continues...
GPUgrid are having 2nd thoughts about upgrading to the new cuda 2.2 drivers. Apparently in their testing on some 8800GT cards their work units kept failing. Of course this is announced after I have updated one machine to the latest drivers. I also had to download the latest cuda 2.2 DLL files from the nvidia web site.
It appears that you have to uninstall the old nvidia drivers before installing newer ones otherwise it doesn't quite work. If you don't it still thinks its running the old drivers.
14 May 2009
Highlights of this week
- Seti reaches 4 million credits
- GPUgrid looking to update to cuda 2.2
- Result file sizes likely to present a problem
As I mentioned above Seti hit the 4 million credits mark and kept on going. Also I am close to 7 million for all projects, but not there yet.
GPUgrid are going to update their science app to use the new cuda 2.2 standard. This is causing a bit of upset to some people who can't or won't upgrade their video drivers to the current version. I have updated one machine already and set all the machines to "no new work" for GPUgrid so they can finish off what they have.
Some of the result files produced by a couple of projects are large. Climate prediction has 43Mb result files when it finishes a work unit after 5 days processing. GPUgrid has recently been outputting 29Mb result files and yesterday had a 54Mb file. All this uses up my monthly transfer limit (25Gb peak, and 25Gb off-peak). I expect to exceed my monthly limit and then get throttled this month. I will have to investigate how to get a larger download limit and faster upload speeds.
07 May 2009
- Passed the 6 million credits un-noticed
- Further BOINC testing
- GPUgrid HIV protein analysis
It appears late last week or early this week I passed the 6 million credits for all projects. I didn't actually notice until yesterday. Its currently around 6.8 million at the moment so I will try and grab a shot of my BOINCstats image when I get to 7 million. Seti is also closing on 4 million credits.
BOINC came out with a few versions in the last week. We seem to have a version (6.6.28) that resolves the task "thrashing" (see my previous post). I have installed it on my i7 with dual GTX260 cards to see how it handles things. I have already installed it on a couple of the quad-core machines. Looking good so far.
GPUgrid are continuing their HIV protein analysis, complete with 29Mb result files to upload. They take about 20-25 minutes just to upload. I try and return these work units as soon as they have finished instead of waiting for my usual "off-peak" internet time (1am to 7am). I think my internet download limit will be exceeded this month.
01 May 2009
Highlights of this week
- Seti 10th anniversary
- GPUgrid starts analysis of HIV protein
- BOINC client testing continues
GPUgrid have started an analysis of a HIV protein which I considered a worthy thing to crunch. The work units are somewhat larger than normal. The output files are also rather large. The last one I saw had 29Mb of output to report back. Depending on which of the computers gets the work they can take days or 8 hours.
BOINC 6.6.25 was released with a couple of minor fixes in an attempt to track down the scheduling bugs that cause task "thrashing". That is it will start tasks then suspend them to run other tasks, only to resume the original tasks later. While a number of people have made suggestions (myself included) they all appear to be ignored by the developers.
26 April 2009
25 April 2009
As a result of Einstein being off the air the CPU's ran dry a couple of times, but the GPU's had plenty of work thanks to GPUgrid. I rejoined CPDN (climateprediction.net) and have 4 tasks running on one machine. They will be going all week.
Highlights of this week
- Additional pair of GTS250's installed
- nVidia drivers update to 182.50
- GPUgrid nearing 1 million credits
- Seti@home 10 year anniversary
- Further BOINC testing
While I was updating graphics cards I had to reinstall (in the 2 machines that had the 9800GT's) the drivers. I put the 182.50 driver on all the machines with the GTS250 cards. I am yet to update the drivers in Maul.
With all the new graphics cards my GPUgrid score is rising and not too far off the 1 million credits milestone.
Seti@home is coming up on its 10 year anniversary and they are having an open day. Not that I have been crunching Seti for that long. I did reach 10,000 "classic" work units on the 8th of July 2005. This was before they started using the BOINC platform. As I am nowhere near Berkeley, California it would be rather hard for me to attend. I will have to settle for an anniversary T shirt.
The BOINC software is having a few issues with swapping tasks out when it shouldn't be, so there is much discussion going on about the best approach to use when scheduling cuda tasks in particular. I will be testing it again.
20 April 2009
- Two GTS250 cards arrived and installed
- Ordered another pair of GTS250 cards
- BOINC 6.6.23 being beta-tested. Seems to be a bit better for GPU work
- Downloaded (but yet to try) nVidia drivers 182.50
18 April 2009
On Wednesday I swapped the 9800GT card for a GTS250 in Luke. I did this one first as it has already had its power supply upgraded. After a reinstall of the drivers its off and running.
On Friday I replaced the power supply in Obi-Wan. I then put the other GTS250 in there. As with Luke I had some fiddling with the black plastic clip that holds the metal plates on the back plane. I ended up filing it back for the GTS250. After installing the drivers its off and running.
You can see the heat pipes from the heatsink in this shot and the fan. Its hard to tell from the photos (the plastic shroud gets in the way) but there is a heatsink between the circuit board and the fan. It is slightly wider than the heat pipes.
Here is another shot looking down the length of the card.
And in keeping with the tradition of having the appropiate figurine, here is Obi-Wan standing on himself.
14 April 2009
Late last week I ordered a couple of GTS250 graphics cards for two of the quaddies that have yet to be upgraded. I already have 500 watt power supplies, but have been holding off the purchase of graphics cards pending nVidia releases. They should be available later this week.
In my absence it appears Einstein and GPUgrid have both gone off the air. Einstein's database ran out of memory. GPUgrid had a power failure, but appear to be back on-line as I write this.
BOINC 6.6.20 became the latest release version, so i'm up to date on all the machines. However there have already been a couple of development versions released since then. I will see if it warrants further beta-testing.
05 April 2009
04 April 2009
I have one machine sitting around idle as I haven't got power, screen or a mouse to have it working. Not to mention the amount of heat these things generate. It alternates between Yoda and Obi-Wan at the moment, taking turns to sit on the floor unplugged.
Still waiting for nVidia to release the GTX212. Hopefully that will be more suitable for all the quad-core machines. I still have two 500w power supplies sitting in their boxes waiting to be installed for GPU upgrades.
Highlights of this week
- GPUgrid passed the 500k milestone
- BOINC 6.6.20 is out, so will be upgrading the machines to that
- First lot of CPDN work finally finished (they take a week)
- The farm is getting close to 5 million for all projects
28 March 2009
This week has had its ups and downs.
On the up side Darth Maul joined the farm and is crunching away.
On the down side the Seti@home file server had a hard disk failure and hasn't been sending out any work for a while. It is back in a rather limited fashon and producing small amounts of work units. The Einstein@home file server also failed and so its not downloading or uploading anything and the website is gone too. That leaves GPUgrid and Orbit@home for work. GPUgrid is working but not all my machines have GPU's. Orbit hasn't had any work for months. So basically the farm is almost out of work. I have attached one machine to Climateprediction but it wants to download over 150Mb to get started, so that will have to wait until my ISP's off-peak hours (1am to 7am).
I have continued BOINC testing. The current version is looking stable enough now to be released. I have 6.6.18 on one machine and 6.6.17 on all the others.
27 March 2009
And here is the insides looking towards the back of the case to give you an idea how big the heatsink/fan is.
The CPU doesn't seem to run that hot. Not suprising considering the size of the heatsink/fan.
The graphics cards on the other hand run around 70+ degrees, and thats when I ramp up the fan speed manually. The nVidia drivers are not too good at increasing fan speeds while running cuda tasks. I'm currently running 182.06 drivers.
Here are some more pictures of inside. The motherboard is an ASUS P6T, Intel core i7-920 CPU, Thermalright ultra120 heatsink/fan and 3Gb of Corsair DDR3 RAM. In the pci-e slots there are a couple of Palit GTX26o graphics cards.
21 March 2009
I had a few teething problems installing XP which turned out to be a dud CD. Used a different CD and that fixed it. Applied SP3, download latest 74Mb of fixes from Microsoft and its ready to roll.
Installed BOINC and picked up a couple of GPUgrid wu and it seemed happy. It then picked up some Einstein work, cpu fans all ramped up and then it crashed. The shop had OC'ed it to 3.9Ghz. So hop into the BIOS, reset to factory defaults and off we go again at 2.67Ghz this time. Seems to be stable now.
Here is the GPU-Z screen shot of one of the 2 GTX260-216 cards.
Here is the CPU-Z screen shot of the i7-920 after resetting to factory defaults.
And here is the BOINC manager showing all the tasks.
Lastly I had to go and update the flops counts in the Seti app_info to reflect the different speed compared to my usual quad-core machines. It hasn't picked up any Seti work yet so we'll see how much of a speed difference it makes.
18 March 2009
If you are not comfortable with editing an app_info and using optimised apps then this is not for you.
BOINC is very unforgiving of an incorrect app_info and usually will delete all tasks if you get it wrong.
Do NOT use Internet Explorer to edit the xml files, it will stuff up your app_info. Use Notepad or another text editor.
Thanks go to Richard Haselgrove and Claggy for ideas in building this app_info.xml file.
Upgrade your BOINC client first and get it working before changing anything else.
The app_info.xml below is based on a 32 bit Windows platform. If you are running on another platform you may need to add/amend the <platform_name> tags.
My computers support the SSSE3 instruction set. SSE2, SSE3 or SSE4.1 may be more appropiate for you. You will need to amend the program names as appropiate.
I've assumed that you have your cuda-capable card up and running and have the necessary nVidia drivers (minimum version is 180.48).
a) Optimised multibeam and optimised Astropulse
b) Cuda multibeam and support libraries
1. Download and install BOINC 6.6.20. Get this working before changing anything else.
1a. If you have a cc_config file with a <ncpus>
2. Empty your cache of Seti@home work. This is best achieved by setting the project to No new work and letting it finish off its tasks. Make sure they are all uploaded and reported, there should be none on your tasks list.
3. Download the optimised multibeam and astropulse apps (usually from the lunatics web site) if you don't already have them.
4. Download the cuda multibeam app (from the Seti web site) if you don't already have them. If you run the stock cuda multibeam app then you should already have these in your projects\Setiathome.berkeley.edu folder.
5. Disable network communications in BOINC. Shutdown BOINC. Make sure its shutdown.
6. Browse your client_state.xml file (its in the BOINC data directory) and look for the entry <p_fpops>. We need to use this number. Do NOT change this file.
7. Browse the BOINC log file to get the estimated speed of your GPU. This is usually given at the top of the log in Gflops. My 9800GT was estimated at 60Gflops.
8. For each of the apps multiply the p_fpops value by the factor below and put this into the appropiate flops entry in the app_info given below. For multibeam 608 you need the estimated Gflops.
Astropulse 500 flops = p_fpops x 2.25
Astropulse 503 flops = p_fpops x 2.6
Multibeam 603 flops = p_fpops x 1.75
Multibeam 608 flops = Est.Gflops x 0.2 (eg 60,000,000,000 x 0.2 = 12,000,000,000)
9. Make sure you have all the programs listed above in the projects\Setiathome.berkeley.edu folder. If not copy them there.
10. Save your app_info in the projects\Setiathome.berkeley.edu folder.
11. Start up BOINC. Check the messages tab to see if it lists any [file error] messages. If there are shut BOINC down, check you have the correct program names referenced. Go back to step 9.
12. If okay then enable new work for the Seti@home project.
13. Enable network communications again.
14. It should now download work of all types. If not check your Seti@home preferences on the Seti web site, that Astropulse, Astropulse_v5 and allow graphics processor are all ticked. If you have a slower computer you may not get astropulse work units anyway.
Updated 14th April 2009