23 December 2011

23rd of December

Its Christmas eve eve and here is the news...

Farm news
Nothing much has been happening on the farm this week. I had only 1 machine running trying to get through those Full-res Ocean CPDN work units. They only take around 550 hours :-)

I am still trying to organise a case for the Adventure and its graphics cards. It always seems so much harder once you have most of the bits to get the last few parts. I have emailed some pictures to my usual shop and they have been in contact with one of their suppliers who specialises in rackmount cases.

Maul's heatsink and CPU fan seem rather clogged up with dust so I will take him into the shop next week so it can get cleaned. I also want them to rearrange the fan wiring so the fans don't just run at full speed.

GPUUG fund raiser
The GPU User Group ( http://gpuug.org/ ) are trying to finish off their 45 x 2Tb hard disk drive fund raiser. As I reported before they have already shipped 24 drives to Seti. I made another donation towards it. They need to finish this off so they can get onto the next item, which is an Astropulse server. At the moment they are $2,189 away from completing this one so if anyone would like to donate please help out.

Project news - GPUgrid
GPUgrid ( http://www.gpugrid.net/index.php ) updated their website and wow its quite fancy now and has the latest server software. It will take a little getting used to. Its mostly the same navigation as before but with the new server software some things are different, on top of the new look web pages.

19 December 2011

19th of December

BOINC testing
We got 7.0.3 this week. So I put it on Maul. Its crunching away at the moment. There have been some reports of weird work-fetch issues. I have found the manager has an issue with the projects tab and resizing if there is a scroll bar. Apart from that it seems to be behaving. The transfers tab now shows "Download:suspended". Before you didn't know if a transfer was going up or down when they are suspended. The file transfer backoff's have also been reduced after complaints they were too agressive in the current 6.12 clients.

The main new feature (apart from the task scheduling/work fetch) is support for OpenCL devices, primarily ATI and Nvidia graphics cards.

Not too bad, but not quite ready for release yet (in my opinion).

Farm news
This week has been warm with rain on and off, so I have to keep the window mostly closed, which also means it gets hot. I have had Sulu running all the time but mostly without GPU to keep the heat under control.

GPUUG funding drives
The GPU Users Group still have a couple of hardware funding drives going. One for a new server and one for 45 x 2Tb hard disk drives for the RAID array. Seeing as it was half-way there on the funding drive Seti asked if they could get half of the drives ordered. Various orders were placed (due to vendor limits) to get 24 drives. The remainder will get ordered once the funding drive completes. At the moment they need $2,999 to finish this one off.

The GPU Users Group have also started taking donations to Seti via PayPal. The University is unable under California law to accept PayPal donations. The GPUUG can as they are incorporated in a different state and not a government entity.

Project news - GPUgrid
They have just updated their server software. Some things (like web pages) are yet to be updated. This should allow them to support ATI cards in the future once they have a compatible science application.

The next update planned is to a newer science application which will only work on cards that have a compute capability of 1.3 or better.

12 December 2011

12th of December

Farm News
Well I have been buying presents again. Well its more that a bunch of them turned up today.

1st up a replacement monitor arrived. I have already swapped the old monitor out. It looks a bit clearer than the old one. Its also noticably lighter. The old one will go out in the next council clean-up.

It was followed a few hours later by the extra PCIe bridge cards and cable for the Adventure 2800. Along with the bridge cards came the 10 GTS450's. Even though they are from 2 different companies in Japan they happened to use the same courier company and arrived together.

For the last few days its been quite warm and then a storm rolled in yesterday and the temperature dropped again. I have only been running one machine without any GPU work. Now the temp has dropped I have fired up the GPU again.

Well I haven't got a case organised yet. I might just go for a rack-mount style case as they tend to be cheaper and more easy to find with the necessary cut-outs in the back-plane. I need to do some searching on the internet to see what is available.

BOINC testing
No new versions so far this week. There have been a lot of complaints about the backoff being too agressive in later BOINC clients so I have posted a request to the mailing list in the hope we can reduce them in the new version. There has been no response from the developers yet.

07 December 2011

7th of December

Back in Sydney.

Well its home with me. I haven't taken it out of the box. I have ordered a 2nd set of bridge cards/cable for it from the company in Japan.

I have asked my computer shop to see if they can find a suitable case for it (they need 12 cut-outs on the back plane for all the cards). I have a spare Corsair 1000 watt power supply which probably isn't enough for all the cards, but might be enough to get it going with say half the cards. The shop hasn't got back to me yet.

I need to decide which machine is going to host it. I have a number of Q6600's but they only have a single PCIe x16 slot. The P6T's have 3 slots (two are x16) but I would rather not use them as a host. I may need to get another machine with dual x16 slots. It wouldn't need to do much processing, it just needs to feed the PCIe bus.

The 10 GTS450 cards have been ordered. The latest email from the company in Japan is they shipped them today.

BOINC testing
Well the 6.13 line has ended and 7.0.2 is out for testing. There have been a few different issues reported for it already, so I will wait for another build before I install it on one machine (probably Maul).

Farm news
Sulu has just completed its Hi-res Ocean CPDN work unit that it had before I went away. Its also finished off the 3 regional models that it had. Since then its picked up another 3 Hi-res Ocean work units. Its also running Einstein CPU work.

I have a couple of machines running GPUgrid work so they should be getting a fair bit processed now as they try and catch up.

My 22 inch monitor has been playing up for the past 6 months. It switches the screen off after about 20 minutes of use. I think its got a dry joint and once it warms up it turns off. Some times it will do this while you are in the middle of using the machine, so its not power management. To fix this I normally just turn it off and on. The power switch on it broke yesterday. It would appear to not push out, so probably broke a spring. Now its on until it warms up, then all I can do is pull the power cable out and put it back. Time to order a new one I think.

GTS450 cards
I have installed the 3 of them in Maul, swapping out the GT240 cards that were there before. They run rather hot, not suprisingly being single slot cards. What is suprising is that they only appear to be a little faster. The peak flops reported by BOINC is approximately 1/3 more (401). For Einstein work the GT240's were taking around 80 minutes and the GTS450 around 60-70 minutes. It could also be a driver issue as I had to install later drivers.

If they overheat they make the machine reboot. I have added an extra fan near the cards to blow air down the length of the cards but it doesn't seem to make much difference. The top most card is 10 degrees hotter than the bottom card. The BIOS isn't agressive enough at increasing fan speed so I have had to do this manually to stop them overheating.

26 November 2011

26th of November

Happy birthday. Happy birthday. Happy birthday to me :-)

I couldn't help myself. I went back to Akihabara and got another GTS450 single slot card. It was in the display case. They didn't want to sell it to me on Wednesday, but today its marked down from 14,000 yen to 11,600 yen as it was display stock and their last one. When the total gets to 30,000 yen or more you have to pay a 5% consumption tax. This kept me under that by buying in two separate transaction.

That brings me up to 3 of them, which I will probably put in Maul to replace the GT240 cards. It should increase its throughput quite nicely without too much power increase.

I have been in contact with Elsa in Japan and they have managed to find some more. Due to some restrictions they have to sell through their distributor if I want to get them shipped to Australia. I will probably buy 10 to cut down on shipping costs. That will give me 8 to put in the Adventure 2800 and 2 spares. Didn't I mention I got the Adventure 2800? Now all I have to do is get them all home.

GPUUG update
They finished the fund raiser for the RAID box and it has now been delivered to Seti@home. They have two additional targets:
a) 45 x 2Tb server grade hard disks
b) Another server

I have kicked in a bit of money towards both. I am concentrating on the hard disks as they cost less and they are needed for the RAID box thats recently been delivered.

23 November 2011

23rd of November

While walking around an electronics department store called Akiba I found the rare single-slot GTS450. Its an ELSA branded one, the same company that I got the Adventure 2800 from. They only had 2 in stock, otherwise I would have bought more.

19 November 2011

19th of November

Well not too much to report this time as I am still in Tokyo.

I did order an Adventure 2800 from the company in Japan. I have done the necessary funds transfer. I haven't had a confirmation that they have received the funds yet. Hopefully it will show up early next week as I should be heading back to Sydney at the end of the next week.

I am not too sure what GPU's I will put in it at the moment. I am leaning towards stuffing it with 8 x GT440's. The single slot MSI GTS450's would have been ideal but I can't find anywhere that has any. I still have to buy a suitable case, but I have do have a spare 1000w Corsair power supply.

12 November 2011

Going on an Adventure

Geetings again from Japan.

BOINC testing - 6.13
The 6.13 client got up to 6.13.12 this week. It contains quite a lot of changes. Given my current location I am unable to test it at the moment.

BOINC testing - 6.12
The 6.12 client got up to 6.12.43 which has a couple of Mac installer fixes and an additional field for Virtual Box handling.

GPUUG website
The GPU Users Group have registered as a non-profit organization. People can donate using PayPal and they then use the money to purchase specifc hardware or software for Seti at Home. Their web site is at http://www.gpuug.org/

One of the items they were after was a RAID enclosure. I gave $1,000 towards it. Other people also donated to it, so that it has now completed that item. There are other items to buy such as 45 x 2Tb server grade hard disk drives.

Going on an Adventure
Remember some time back I posted about the LucidLogix Adventure 2800? Well it turns out they sell them over here through a company called ELSA Japan. They are located here in Tokyo. I have been discussing pricing and availability with them. I may be able to get one of these while I am over here.

They have them bundled in a rack mount case, but I can purchase the motherboard, PCIe interface cards and a 1 meter cable without the case and power supply.

29 October 2011

29th of October

Farm news
Greetings from Tokyo. The farm is off because I am overseas.

BOINC testing
We got an almost final version of the 6.13 client (6.13.10) which is expected to become the new version 7 client once testing completes. It sounds like its now stable enough to try this version.

iPad development
Well the developer has given up. He doesn't understand how to install BOINC on a Mac and use the iPad app that he managed to get compiled. I will try and post the Xcode project that he came up with in case its of use to another developer. Its not complete (he didn't do the GUI or convert the code to Objective C) and its fairly big at 144Mb. Its based on the Mac 6.12.36 client.

Project news - CPDN
The project was hacked and some user account details were compromised. The administrators shut the project down and have been investigating the extent of the problem. The project is now mostly running again but trickles aren't at the moment.

They have emailed the effected users to let them know. My understanding is that they were mostly moderators that had their account details compromised.

23 October 2011

23rd of October

Farm news
Well the farm has been running just a single machine for the last week. I have now shut it down until I return. As expected there hasn't been too much progress from a single machine. This single machine has 7 CPDN tasks on hand, 3 of which have completed but can't report. One of incomplete work units is a full-res ocean work unit that takes 550 hours and is about half done.

Hopefully I will have internet access from Japan and will be able to keep in contact.

Project news - CPDN
They were attacked and the project has been down for the last few days. No further news yet as to what happened.

Project new - Seti
They have upgraded the memory in a router which should have resolved the randon connection issues. Apparently it didn't have enough memory to cope with all the connections going at once. Hopefully it might also improve network throughput.

16 October 2011

16th of October

None again.

BOINC testing
We got 6.12.41 during the week. It fixes a couple of memory leaks.

I suggested a change to the way file transfer back-offs are handled. Dr Anderson has amended the process stightly and added these changes to the 6.13 client (which I am staying away from). I will ask if we can get it put into the 6.12 client as well.

Farm news
Well as you can tell the farm is still going, although in a reduced capacity at the moment as its starting to warm up in Sydney. The trip to Japan has been postponed for another week.

Project news - CPDN
Work has been rather sporadic as they wait for work units to be completed before generating more. They have a bunch of hadam3p (regional) work units available to download at the moment.

They are having a problem with the permissions on some of the files so they get stuck in downloading. Until somebody from the project can fix that we can't get the last 2 files that make up a work unit.

Project news - Seti
They are having their fall funding drive. I have made a donation to help them keep running.

We have been having some discussions around how to better use the limited network bandwidth over there. Hopefully some of the suggestions can be implemented. None of them involve hardware, just changes to data formats, compression of requests, reduced information in requests, etc.

08 October 2011

8th of October

Nothing new this week.

BOINC testing
We got 6.12.39 this week to give a run. Its pretty much the same as .34 but has a few changes around the virtual box handling and some Mac stuff.

I have made a request on the mailing list to amend the way BOINC handles back-off on file transfers. My idea is to back-off on a per URL basis rather than the whole project. This should alleviate the problem we had last week where Einstein GPU work had download issues and we couldn't get any CPU work which was available from other servers. Dr Anderson has said he'll look into changing the way it works.

Farm news
Things have been chugging away during the week. I have purchased a couple more shelving units. The file server has moved into a rack since taking last weeks photos.

The place where I am working want to send me to Japan for 4-6 weeks as early as next week. That means the farm will be off while I am away. It also means I won't be updating this blog for a while.

The 3 Climate Prediction full-res ocean work units finished yesterday and have been uploaded. Just as well seeing as they take 550 hours and farm will be off.

iPad development
The developer isn't sure how to test it, so I have repeated what I told him before, which was to install BOINC on another machine and use the iPad BOINC manager that he has managed to compile to talk to it. It gets very frustrating when you have to keep repeating instructions, all of this by email. Anyway we will see if he can get it to work.

02 October 2011

1st of October

A pinch and a punch for the first of the month.

Nothing new yet, but will probably get to 45 million for GPUgrid tomorrow.

BOINC testing
Well I didn't do any this week. 6.13.6 came out today, but I still don't think its stable yet.

Farm news
They say a change is as good as a holiday. So I have reorganised things. Instead of having the i7's stacked on top of each other I have given them their own shelving unit. This should hopefully reduce the heat a bit. I have also raised them off the floor some more to allow for power boards, which I also replaced.

These CM Storm Sniper cases have a fan in the base blowing air up from the bottom and onto the graphics card which is another reason why I moved them up a bit (less dust/hair from the carpet getting sucked up).

As you can see there are 2 machines on the left and 3 on the right. The last one doesn't have a shelving unit at the moment and hasn't been plugged in.

The cabling isn't tied off. I need to get a couple more lan cables and then I can use cable ties to make the cabling a bit more secure and neat. While I do have enough cables I use color coded ones for some parts of the network.

You can't see it but to the right (and in the middle of the room) is the file server with its UPS and one of the quaddies. I need to sort out some shelving units for them too. The UPS was full of dust so gave it a bit of a clean. The dust filters on the file server were also clogged up so I gave them a clean as well. I was going to wash them but couldn't work out how to get them off.

You can see the graphics from a Climate Prediction work unit on the screen. Its one of those full-res ocean work units that take 550 hours. The figurine on the right is Darth Maul, so you can guess which machine he is standing on.

Project news - Einstein
This weekend they are having download problems with the GPU work. Its been fine most of the week, but this weekend downloads have been backing-off and when they do get through they are very slow. Its probably because of the number of users all wanting GPU work.

25 September 2011

25th of September

Nothing new to report this week.

BOINC testing
We got a 6.12.37 for testing as well as the major re-write version 6.13.4. I am not currently testing the 6.13.x versions until they become a bit more stable.

I will be trialling 6.12.37 in the coming week although there is no useful functionality that I am interested in just yet. Its mostly around Mac installer fixes for OSX and some additions to the Virtual Box handling.

Farm news
Maul managed to trash all its Einstein cuda work on Saturday, complaining that one of the exe files had disappeared. When I noticed it I suspended the project and detached and then re-attached. This has the effect of deleting all the project files (when you detach) and then downloading them all again (when you re-attach). This got it going again. I have since run a disk diagnostics on it but it hasn't found any issue. The drive is one that I bought earlier in the year, so isn't too old.

The weather turned cold over the weekend so I have fired up a couple of the GTX570's for GPUgrid work. I normally don't run them during the day due to the amount of heat they produce, but being cold its not an issue.

BOINC Manager on iPad
I am paying for an iPad developer to port the BOINC manager to the iPad. Part of the problem is the supporting libaries aren't available on the iPad. While he has managed to get a working version of wxWidgets (used extensively in the manager) its more recent than the version BOINC has been using and various wx functions have changed. He is trying to work around these isses.

I am not trying to get the core client or other parts of BOINC working on the iPad at the moment, the exercise is just to get a working manager so one can access your computers via the iPad.

LucidLogix Adventure
I have sent another email to Edom in Taiwan asking for pricing on the Adventure 2800 and 2500. These are PCIe bus expansion chassis. The Adventure 2800 offers 11 PCIe connectors single spaced and the Adventure 2500 has 4 PCIe connectors double spaced. No response yet from them. I last heard from them in December 2010.

I have been upable to find a source for these single slot cards. It seems Newegg  no longer has them available. The computer shop have spoken to their local MSI representative but they don't know when they will be available either.

The idea was to replace the GT240's in Maul with these. If I ever manage to get an Adventure 2800 they could also be used in there. There is a GT545 which has 144 cuda cores and looks almost the same as as the old GT240's.

17 September 2011

17th of September

Passed 10 million for Einstein earlier in the week.

BOINC testing
As mentioned last week 6.13.3 came out. It has been withdrawn due to not being able to upload. It also has various issues around fetching work.

Farm news
One of my old Acer Veriton v7900's has a failed motherboard, so I took the memory out of it and put it into Obi-Wan. As it had a 32 bit version of Windows XP it wasn't able to use all the memory. This weekend I installed Windows 7 x64 on it. Its off and running again.

The Drobo also got some new 3Tb drives installed. Rather than just re-copying the files across to it I let it recover the raid onto the new drives and then repeated the process a couple more times as I removed the older 2Tb drives one at a time. The 2Tb drives are now up for sale.

11 September 2011

11th of September

Nothing new again this week, although getting close to one for Einstein

BOINC testing
6.13.3 was released this week. It is still not stable, so I don't suggest people try it. From reading the change log it looks like they have just about rewritten it from the ground up. I haven't bothered with it yet. Oh and you can upgrade from an earlier version, but you can't go back unless you have a backup of the data directory.

Farm news
As mentioned last week Obi-Wan had a clogged up heatsink. I got the shop to clean it out and reapply thermal paste. They did it on the spot for me and it is now back in action.

While I was at the shop I dropped off both of the faulty GTX460's to arrange warranty repair or replacement.

Project news - CPDN
CPDN has produced another batch of full-res ocean work units and I have picked up 3 on one machine. So I need to leave it going for about 550 hours to finish them off.

Project news - Einstein
They are now up to 10 new Pulsars found in the Parkes data. This goes to show that the southern hemisphere probably has a lot more undiscovered pulsars and other bodies.

Einstein is having problems with the number of users after cuda work. Not only have they had to setup another machine to generate work units because they couldn't keep up, but now their network connection is getting hammered (probably as a result of having more work available). They had a CPU and a GPU version of the app for the Binary Radio Pulsar Search, but are now only giving out work to GPU hosts.

Project news - Seti
They were trying to process some older data, however it has turned out to be un-usable. Various processes that split the data into work units kept crashing, so they have given up on it and are processing newer data. This meant that they didn't have any work available for a while.

There is a newer v7 app in the works for multi-beam searches. The optimised GPU version is already available but the CPU version is not.

04 September 2011

4th of September

Nope nothing new this week

BOINC testing
6.12.34 (Windows/Linux) and 6.12.35 (Mac) were promoted to release status. No real change for me as I am already running 6.12.34 on all the boxes except one.

Farm News
Things have been fairly quiet. I finished off the two remaining CPDN full-res ocean work units and then couldn't get any other CPDN work until a couple of days ago. They are only running 2 experiments, the regional (HADAM3P-EU) and the Rapid-Rapit (HADCM3N also known as the full-res ocean), so there isn't much work getting generated and when it is they all get taken quickly.

I got fed-up with Chrome not displaying web pages when IE works fine, so I removed Chrome from most of the machines. I will remove it from the 2 remaining machines today.

I got a new router, which has improved download speeds a bit. The IP addresses on my network changed so that created a bit of work.

27 August 2011

27th of August

Nothing new this week.

Farm News
The GTX460's have been removed from the machines. As mentioned previously they both seem to have failed memory. It turns out they have a 2 year warranty so they should be able to be repaired/replaced under the warranty.

In the mean time the GTX560Ti has been installed in Obi Wan as you can see from the pictures below. It looks almost the same as the GTX460 that it replaced but is almost 1 inch longer.

While I was poking around in there I noticed the CPU heatsink was covered in dust, so I gave it a bit of a blow-out using a can of air. It really needs to be removed and the fan taken off to be done properly. Something I will get the shop to do.

Project news - GPUgrid
They are preparing a new science app for release in Fall 2011. The main differences:
- 10% faster on Fermi.
- It should support parallel gpu runs later on
- It drops support to compute capability 1.1 and 1.2 (so it supports from 1.3 inclusive)

Project news - Seti
Their raw data storage raid got itself tangled up last week so no work was getting created while they fixed up the raid and restored the data from an off-site backup. They have now got it going so work is flowing again, although the usual network congestion issues remain.

Speaking of network congestion they have identified a problem with one of their routers. They think it needs a memory upgrade to cope with the number of connections it needs to handle.

21 August 2011

21st of August

I passed 42 million for GPUgrid.

Farm News
The GTX460 in Obi-Wan seems to have died. It started by failing work units after 4 seconds. It then was drawing purple squares all over the screen. I ran the cuda memory tester on it and it covered the whole window with purple squares. I swapped it out with another one and it too failed the cuda memory test, so I assume both are dead.

At the moment Obi-Wan is running without a GPU (it has on board video). I have ordered a GTX560Ti as a replacement.

As the GT240's in Maul are getting a little dated I am looking for suitable replacements. They need to be single-slot cards. I came up with the GT545 which looks pretty much like the GT240. MSI also make a GTS450 available as a single-slot card, although a number of complaints about fan noise on them.

Project news - Einstein
They have discovered another 3 binary pulsars from the Parkes radio telescope data. That brings them up to 9 so far.

We need another decent size radio telescope in Australia to cover the southern hemisphere given these previously undiscovered binary pulsars. There is a proposal to build a wide-area array (similar to the Allen telescope) in Western Australia but that may not happen.

14 August 2011

14th of August

No new milestones to report this week.

BOINC testing
Updated Chekov to 6.12.34. There is a 6.12.35 out, but its for the Mac only.

I have put a request out on Freelancer to get the BOINC manager ported to the iPad. This (if it happens) will allow me to control the machines remotely. I can do this from my laptop but I prefer to use the iPad as its ligher, more portable and the battery lasts longer.

Farm news
Things have been running smoothly for most of the week. In fact I haven't even looked at the machines for most of the week.

Two of the four CPDN full-res ocean work units finished and the other two are running with 200+ hours left to go.

I updated the nVidia drivers to 280.26 on all the running machines. This supports CUDA 4.0 and is slightly faster than previous versions.

New project: Radioactive@home
This is a new project being run out of a university in Poland. They are mapping radiation. To use it you'll need an external sensor that plugs into a USB port, pretty much like EQN does. It runs as a non-CPU intensive task in BOINC the same as EQN does.

The project is in start-up mode at the moment and doesn't have sufficent sensors. After prototyping a sensor, they are now getting some manufactured, however their initial oder was only for 50 and they have all be taken. Their URL is: http://radioactiveathome.org/boinc/

06 August 2011

6th of August

Passed 41 million for GPUgrid.

BOINC tasks
This is a monitoring program that is great for keeping an eye on all the machines from a single computer. Its similar to BOINCview. Its been updated to version 1.09. It can be found at http://www.efmer.eu/boinc/boinc_tasks/

BOINC testing
There is a 6.13.1 being tested. It has some fairly major changes from 6.12.x around the work fetch algorithm and its early days. That is its too risky to try it out just yet.

There is also a 6.12.34 being tested. This has some minor fixes as well as a couple of installer fixes. The windows installer had an incorrect DLL file in the package. The Mac installer has been updated to work with OSX 10.7 (Lion). Maul is running 6.12.34 and I will update the others later as it seems pretty stable.

Farm news
Weather has been warm despite being winter here in Sydney. As a result I've suspended GPUgrid for a few days as the GTX570's generate a fair bit of heat. Instead I have left a few machines (Maul, Chekov with Obi-Wan intermittently) running CPU only work.

Seti is still plagued with download issues (not enough bandwidth and too many users). I can't see any easy fix for this and its likely to take some time before anything can be done.

Chekov has 4 of the Climate Prediction full-res ocean work units running. These take 550 hours or so. Two of them are around 80% done and the other two are a long way off completing.

Project news - Einstein
They have a new gamma ray search running, so work for that has now started flowing. Its a CPU only task at the moment although they have said they are looking at a GPU version.

We're also crunching the gravity wave data as well as the binary pulsar search using Arecibo data. We've pretty much finished off the binary pulsar search using the Parkes data (which has so far found 6 new pulsars).

24 July 2011

Well this week I managed to knock over another million for GPUgrid. Below is my 40M certificate.

BOINC testing
While I was away 6.12.33 became the recommended version (I was already running it).

We're on another round of testing now for the 6.13.x version. I have yet to download and install it as there are quite a few changes that need to become a bit more stable. The main change in this one is to make the skins, used in simple view, more translatable for languages other than English. There are also some scheduler changes in this.

Farm News
This week was a short week for the farm as I only got back on Tuesday. Nothing much to report other than the weather is better in Vancouver than here in Sydney, where its been raining all week. Now we're into the weekend its finally stopped raining.

The farm powered up first go, except for when I plugged Maul back in and the earth leakage circuit breaker (safety switch) tripped and that was before I had even turned it on! After resetting the breaker and retracing things step by step it worked fine.

Project news - Einstein
We have been processing data from the Parkes radio telescope. They have discovered 5 new pulsars in the data so far. Here is a link to message thread on the Einstein@home web site regarding this: http://einstein.phys.uwm.edu/forum_thread.php?id=8979

As we have now processed all the Parkes data we are transitioning to a new run of BRP (Binary Radio Pulsar) search using new data from Arecibo. There is little or no GPU work available at the moment, but CPU work units are available.

24 June 2011

Run down

The farm finished off all its work now. All the machines have been powered down. I am off on a vacation soon. I will be away for a few weeks.

This week GPUgrid hit the 39 million mark. The GTX570's can really chew through their work units.

Seti has still been plagued by network issues, so the farm has been concentrating on Einstein  and GPUgrid work. Unfortunately they are so popular and can't handle the number of users they have via a 100Mbit network connection. They really need a faster connection to the outside world, but unless somebody with deep pockets is willing to come up with the cash it isn't likely to happen.

19 June 2011

19th of June

Well its been a fortnight since the last post. I managed to pass 38 million for GPUgrid and 8 million for Einstein.

BOINC testing
We have 6.12.33 to test at the moment. Not much different to the previous couple of versions other than bug fixes. I have put it on 3 machines at the moment, the others haven't been updated from the release version (6.12.26) yet.

Farm news
There is a new optimized GPU app available for Seti (known as x38g), which is somewhat faster than the previous optimized GPU app (x32f). It caters for the newer V7 format work units which will be introduced later this year, as well as handling the current V6 format. It requires cuda 3.2 capable drivers to be installed, which I already had. I have it running on the 2 machines that do Seti work.

I am running the cache down on all of the machines as I will be taking a trip soon. The farm will be off while I am away.

Project news - GPUgrid
They brought out a revised app known as 6.14 for cuda 3.1 capable machines. Unfortunately they set the task priority to Above Normal so the whole machine is sluggish. Their main developer has gone off on a 10 day (work) trip to the US, so we have to put up with it for the next week or two.

Project news - Seti
They have been having network congestion issues for the last week or so. Now they are starting to run low on the data to send out due to having a whole bunch of short work units. They can't get their raw data off their backup and split it into work units fast enough.

Last week they did their Green Bank observing and have lots of data to process. It will probably require a new application as their data format is slightly different to what they get from Arecibo. We might get to process this data towards the end of this year depending on how fast they can come up with a new science application.

05 June 2011

5th of June

As you can see from the certificate below I managed to pass 35 million for GPUgrid.

BOINC testing
We got 6.12.28 during the week. There us a bug fix where 6.12.27 was corrupting the output files for all work units.

There is also a feature when setting a project to No New Tasks it will report on every upload (rather than last upload). I queried this as it will add overhead to the scheduling servers, but Dr Anderson thinks its minimal impact.

Farm news
Nothing much to report for the farm this week. I have been keeping a couple of the GTX570 machines busy all week, hence the GPUgrid milestone.

Still no work from CPDN at the moment.

Project news - Seti
Their network connection has been maxxed out for the whole week. Downloads keep going into back-off mode and uploads too. They have a 100Mbit link but with 100,000+ computers all trying to connect at the same time. Their problem is being compounded by the fact they are giving out short work units that get processed quickly, which means they are being returned faster and then more work is requested.

For the time being I have set my couple of machines that do Seti work to concentrate on other projects.

29 May 2011

29th of May

I passed 34 million for GPUgrid. See I am making those GTX570's do some work.

Farm news
I went away last weekend and as the tradies were doing the bathroom again on Monday I left everything off for 3 days.

I finished off my other Full Resolution Ocean work unit for Climate Prediction. It took 528 hours to complete.

I am now getting the S6Bucket work units from Einstein. They are the S6 run of the gravity wave data.

Project news - Climate Prediction
They still don't have any work available. They are reconfiguring servers and moving files around. This is taking longer than they expected. Until the reconfigure has finished no new work will be generated.

Project news - Seti
Their storage server died and so they stopped work unit creation as well as uploads and downloads being  effected. Apparently its now fixed but due to the number of people trying to upload and download the network is maxed out and nothing is moving.

16 May 2011

Power nap

This week wasn't that productive. I've had various builders running around demolishing a leaking shower and the rotten floor under it. The machines were all off and unplugged at the beginning of the week. Its the same again this week as they have decided the whole of the bathroom floor needs to be replaced.

Despite the destruction of the bathroom I managed to pass 33 million for GPUgrid.

CPDN full ocean work units
I managed to finish off my first one. It took 511 hours to complete. Its been uploaded and reported. I still have another one running on a different machine and its 65% through at the moment, so another 9 days of crunching before its done.

Project news - CPDN
They are out of work. They have some projects in the works but due to storage issues (the work units product some huge files) they might be holding off creating more work units until they can get some of the ones being crunched back.

Project news - Einstein
They are having an open day on Saturday the 2nd of June at AEI Hannover.

Apart from that they also have a new type of work unit designated the S6Bucket. I haven't managed to pick any of these up yet so can't provide any details about them.

08 May 2011

8th of May

Passed 32 million for GPUgrid during the week.

GTX570 installed
I put the other Palit GTX570 Sonic Platinum card into Sulu during the weekend. I hit a minor snag as the power connector had an 8 pin-only PCIe plug, but it turned out to be a 6 to 8 pin adaptor. The GTX50 needs two 6 pin PCIe power cables plugged in, whereas the old GTX295 needed 6 and an 8 pin power cable.

QCN sensors
The company I bought these through shipped me another pair of them. I only placed one order in March. They seem to have processed the order twice. Rather than send them back I will keep them. I have let them know and they'll give them to me at a discount seeing as I didn't make two orders.

Project update - Seti@home
They have started recording data from the Green bank radio telescope this weekend. While we don't expect to be able to process it immediately, as there is some mapping of the data to a common format and some other software development required, there is some hope it will happen soon.

We've also been prompting them to see if they can use the data from the Parkes radio telescope. At the moment they don't have any agreement for data sharing or a data recorder in place.

BOINC tasks
This is an add-on program that was written by a fellow cruncher for monitoring a group of machines (up to 60). Its quite similar to the old favorite BOINCview, which is getting rather out of date. I have installed it on my old laptop to see how it goes, and its quite good now, supporting some of the new functions such as Notices which are only in the development versions of BOINC.

You can grab a copy from http://www.efmer.eu/

01 May 2011

1st of May

Passed 32 million for GPUgrid. Einstein is also coming up on one.

Allen Telescope Array hibernation
This was initially funded by Microsoft founder Paul Allen, however the SETI institute has run out of money to keep it running. This has no impact on Seti@home. Despite press reports saying S@H is effected, its not. The SETI institute and Seti@home are two separate organisations. S@H doesn't even get data from the ATA. For more details see the Q&A link below from Dr Eric Korpela of the Seti@home project:


BOINC testing
After last weeks issues caused by the Microsoft security patch we now have 6.12.26. So far its got a few minor things that I have picked up, but nothing that would prevent it being released to the public.

CPDN full resolution ocean wu
As I mentioned last week I have picked up one last week. It has been crunching away for the last 138 hours and is about one third complete. So another couple of weeks at least before its done. I have also managed to get another one on a 2nd machine, so it will be running for the next 21 days as well. The current estimate is 509 hours to complete them on a standard i7-920.

I still haven't put the 2nd Palit GTX570 into a machine yet. I am planning on swapping out the GTX295 that is in Sulu. I will be selling off the older card.

Processing 24/7
Now the cooler weather is with us in the southern hemisphere, I have been running more of the computers. In additional the CPDN work units need so long to crunch that I have left those (now 2) machines running 24/7 so they can meet the project deadlines.

25 April 2011

25th of April

Passed 11 million for Seti. GPUgrid is closing in on one too.

BOINC 6.12.25 released
6.12.23 came out on Thursday and failed to run. Its seems Microsoft's recent C run-time security fix earlier in the week has broken a number of programs, BOINC included. On Saturday we had 6.12.24 for Windows but it too failed and today we got 6.12.25 which resolves the issues.

EVGA precision is also getting errors since the Microsoft security patch.

CPDN full resolution ocean wu
They have these new work units with a full resolution ocean. I managed to pick one up which is currently estimating 490 hours to completion. Its been running for over 2 days already. That means the machine will need to be on 24/7 for at least 20 days to finish it. They have a 3 month deadline, so that doesn't leave much leeway.

GTX570 Sonic
Well its off and running. As expected its faster than the reference design. It also runs cooler than the reference design. Without having to resort to manually controlling the fan speed it seems happy to sit on 66 or 67 degrees C while crunching. Some of the "long" GPUgrid work units take around 11.15 hours on the Palit and 12.75 hours on the EVGA, so a bit of an improvement.

Kirk and the disappearing video card
As mentioned before I was going to remove the SSD from Kirk. I have now removed it. I had a few issues with software referencing the D: drive (which is now the C: drive) and had to reinstall the video drivers and BOINC. I still need to run it for a while before I can say its behaving but it would appear to be fixed.

Nvidia 270.61 driver
These were released during the week. There have been some reports of them down-clocking. The main advantage of these is they support Cuda 4.0. The Cuda 4.0 toolkit is at release candidate 2 status at the moment (ie hasn't been publicly released).

21 April 2011

Palit GTX570 Sonic

Here are some pictures of the Palit GTX570 Sonic card. The Sonic range from Palit are factory over-clocked graphics cards.

And here are some screen shots from GPU-Z with it running a GPUgrid task

And lastly a screen shot from Vtune (available from the Palit website). As you can see despite the dual fans it still runs hot.

19 April 2011

18th of April

Nothing new this week, but closing in on one for Seti.

Seeing as Microsoft is pushing IE9 down the Windows Update pipe I have removed IE from all the machines. As mentioned in my previous post there are a few reasons not to use it. I have now installed Chrome on all machines instead.

File Server
After being back for a couple of weeks its now been fully commissioned and the old one is tucked in a corner of the loft. Backups and things are somewhat quicker than the old machine.

I still need to get WSUS installed on it to save on Windows updates. The virus scanner also has an update server module that can be installed to save on updates.

GPUgrid long work units
As if they weren't long enough they have a new batch of the long work units. The short ones take about 6 hours on my reference GTX570 and 10 hours on the GTX295 and GTX275. The long ones seem to be about 8.5 hours. I haven't tried running them on the old graphics cards.

GTX570 installed
I got a pair of Palit GTX570 Platinum Sonic cards. These are factory over-clocked versions of the GTX570. I have installed one in Chekov and got it going. It completed its 1st GPUgrid work unit yesterday. As expected it finished it off somewhat quicker than the reference GTX570. I will post more information about the new card later.

Kirk woes
I got the kingston memory late last week and installed it in Kirk. Its staying its having trouble with the graphics card now. Maybe the PCIe slot has gone faulty or maybe something else is causing the issue. I have swapped the graphics card with another one, but that didn't solve the problem. I'll look at removing the SSD next, once I have copied the data off it.

09 April 2011

9th of April

Nothing this week.

Slow internet
I tracked this down to IPv6. Seeing as the old file server was Windows XP based it knew nothing about IPv6, however Windows Server 2008 R2 does. My network hardware and ISP don't support IPv6.

There is a registry change one can make to disable it, or you can right-click on your network adapter and choose properties -> IPv6 and untick it. This works fine until you reboot and then its back again.

Microsoft also have a "Fix it" that applies the registry fix, called MicrosoftFixit50444. The registry fix doesn't appear to work whatever way you apply it. Its probably been broken since Service Pack 1 as the fix pre-dates SP1. Despite that i've run it on all the Windows 7 machines and the file server.

It seems you need to power-off the machine for the fix to work. A reboot isn't sufficent. Having powered them all off and restarted them seems to have resolved the issue.

BOINC testing
We got 6.12.22  this week. Mostly screen saver fixes, but looks like they might get this one released soon. I've upgraded all the machines with it this weekend.

The memory for the i7-970 machine is on back order with Kingston. They think it will ship next week. Until it arrives i've been leaving it switched off.

Project news - Einstein
They are transitioning to a new run of the gravity wave search, which will no doubt add pressure to their servers as we need to download new data packs. They are currently looking at their scheduler with a view to not removing old data packs from client machines until all old work units have been completed.

Project news - Climate Prediction
They have a new type of work unit modelling the role of the Atlantic meridional overturning circulation (AMOC). They use more detailed ocean rules, as such they take more time to complete. Previously they would only run these on a super computer, however they are now doing this via distributed computing.

07 April 2011

6th of April

File Server
This weekend was spent setting up the file server. Everything has been copied over and its running quite happily at the moment.

I do however have a problem with the proxy server, Squid, taking a while for DNS lookups. It seems to take about 10 seconds or so before it finds a web site. After that the pages load nice and fast. The only thing thats changed is its gone from a Windows XP box to a Windows Server 2008 box. The rest of the network is unchanged and the configuration was unchanged.

From searching on the internet most people say its an issue with the DNS addresses not being defined, but its being picked up from the registry and pointing to the router's DNS address. I tried using the ISP's DNS server address but it doesn't make any difference. I'm now thinking its probably got more to do with the firewall on the server, so will do some experimenting with it.

Kirk issues
Kirk crashed last week and now seems to get errors on every Einstein CPU work unit, so i've stopped it running Einstein for the time being. I expect its the memory again. I've ordered some Kingston memory modules for it and they should arrive next week (currently they are out of stock).

Project news - Einstein
They are transitioning to a new search run with different work units, so its likely there will be lots of downloads of new data packs.

IE9 issues
It seems that it's not ready for prime time, even through its now being offered via Windows Update. A few problems I have had with it so far:
  1. It clobbers the existing Java, so you need to reinstall java after installing IE9
  2. There is no flash player for the 64 bit IE9. The adobe web site suggest using a beta version, which no longer exists.
  3. Its not available for Windows XP, so you lose compatibility with XP machines
Chrome doesn't have these issues, and its still faster than IE9.

27 March 2011

All home

Nothing new this week, creeping up on something for GPUgrid.

BOINC testing
6.12.19 came out at the end of the week, so I haven't had a chance to put in on anything yet. Mostly minor fixes. There is a fix for the screen-saver crashing, but I don't use the BOINC screen saver anyway.

Kirk graphics card
I had this machine off during the week. When the weather cooled down on Friday I went to start it up and it refused to boot and just gave error beeps. After digging out a manual to find out what the beep codes were it was complaining about the graphics card. I removed the GTX295 from it and put a GTX460 in there to get it going.

It seems I might need to look at getting the Palit GTX570 Sonic Platinum cards. I am hoping the release on Thursday of the new GTX590 will force the price of the GTX570 to drop, so I can get them a bit cheaper.

New server back
Its finally been sorted out by the shop. The faulty hard drive got replaced a couple of weeks ago and now the memory has been replaced as well. I picked it up on Saturday, but have yet to get him out of the box and swap the functions over from the old P4.

Windows 7 SP1
The DVD arrived just before the weekend, so I ran around updating the machines on Friday and Saturday. They take around an hour to do each machine, except Kirk which has an SSD. It only took about 15 minutes to complete.

20 March 2011

20th of March

I passed 50 million for all projects this weekend. Below is my BOINCstats image.

BOINC testing
We are still on 6.12.18 this week. No major issues. There is some discussion going on about the notices and what people would like to see.

Project news - ClimatePrediction
They finally sorted out their space issues and I have managed (after 2 weeks) to upload the last results for the EU work units that I was running. I have since downloaded another 8 to crunch. I run them 2 at a time and suspend the others.

The Famous experiment they were running has almost finished. They have processed 200,000 work units and will generate a some more to finalise the experiment.

Project news - Seti
They seem to have run out of work this weekend. No major issue if you have a couple of projects running like me.

IE9 is out
I'm running it on one machine. Unfortunately its not available for Windows XP so I probably won't be updating the other machines. Its definately faster than IE8 and probably a bit more secure too.

I was going to try Google Chrome as well and see how it works out. At least it should work under Windows XP.

13 March 2011

13th of March

Nothing this week, although I am close to 50 million credits for all projects.

It stands for Quake Catcher Network. Its a BOINC project that use a motion sensor (either in a laptop or attached via USB). Last week I ordered a couple of USB sensors. They come in kit form as you can see below. Some assembly required. Actually you need to solder the 4 wires onto the circuit board and then put it into the plastic box. The circuit board is approx 1 inch square.

So mine is now setup, although not in a very good location at the moment. The software runs within BOINC as a non-CPU intensive app (doesn't seem to use any CPU according to task manager) just monitoring the sensor.

BOINC testing
We got up to 6.12.18 this week. Its mainly fine tuning the notices and work-fetch at the moment. I've put it on a couple of machines and its working okay for me.

There was one user who has a 160 CPU machine that doesn't seem to be able to get more than 60 CPUs recognised. If only the rest of us had this problem :) It seems Windows groups the CPU's into processor groups of not more than 64, so scheduling tasks across all of them is somewhat different than a smaller system.. They have made some coding changes to allow for this, but we will have to wait for another build.

07 March 2011

7th of March

Kirk back home
It seems they might have actually fixed him up. It has spent the entire weekend doing work, mainly Einstein which is particularly memory intensive.

I had a couple of BSOD's but they were caused by GPUgrid and the nvidia driver commiting suicide. I deleted a file out of the GPUgrid slot directory to stop it crashing everytime I started BOINC. When I next started BOINC up it deleted everything to do with GPUgrid. I then had to reattach to the project.

I had this problem with the other machine that has a GTX295. I believe its an issue with cuda 3.1 not knowing which GPU is doing what, as it doesn't happen on my single-gpu machines.

05 March 2011

5th of March

None this week.

BOINC testing
We got 6.12.16 today. It has a number of changes around the notices area. I will be putting it on a machine later tonight.

Kirk home
Yes he finally came home. It also has a SSD in it. It doesn't really need one. I'm not too sure why I got it, but anyway it there now. So far its managed to finish off its work successfully, so i've given it a full load along with a couple of GPUgrid tasks. We'll see how it handles them.

The 120mm fan on the back of the case was rattling, so I got them to replace it with a Noctua while I was there. It would seem the CoolerMaster fans aren't too reliable as its the 2nd one I have had to replace so far.

Project news - Climate Prediction
Its seems we can't upload or get any work from them at the moment. I have a couple of completed EU tasks on one machine that have been trying to upload for a couple of days and the trickles can't get through either.

From checking on their web site it appears they are doing some database maintenance that is running longer than expected.

Project news - GPUgrid
It seems nvidia have jumped to Cuda 4.0 (well its in release candidate mode for a March release). The GPUgrid guys are just going to skip 3.2 and go to 4.0 for the next app. They will still support Cuda 3.1 for those people running older drivers.

Project selection
Given Orbit has gone into hibernation I have detached all the machines from it except Maul. I left him on it even though its not likely to get any work, so that way I will know if they come back online.

At the moment Maul and Obi-Wan are doing Seti work almost exclusively, the others are running a mixture of Climate Prediction, Einstein and GPUgrid.

It turned cooler today, so I have 4 machines running flat-out at the moment, including Kirk. They have managed to get through a fair bit of work so we might get near a milestone or two in about a week.

27 February 2011

27th of February

Fellow cruncher msattler passed 100,000,000 credits for Seti@home this weekend. My congratualtions to him on such a remarkable achievement.

Windows 7 SP1 out
Microsoft released it officially on the 22nd of Feb. I don't want to download it multiple times (ie for each machine). About a month ago I ran the service pack blocker, which stops it automatically downloading via windows update. Yesterday I ordered the service pack on DVD. The DVD is free, but I have to pay for postage.

26 February 2011

26th of February

I passed 6 million for Einstein during the week. Below is my certificate.

Machines in the shop
The new file server had one of its brand-new hard drives diagnosed as faulty. I will be dropping it back to where it came from for a replacement during the week. Unfortunately I got it online so I have to organise the swap over.

The 1Gb memory in the file server is getting replaced under warranty, hopefully with Kingston this time. Apparently Corsair don't make 1Gb sticks any more.

The 2Gb memory in Maul is getting replaced with more of the same. The distributor the shop purchased from doesn't want to replace it with Kingston.

BOINC testing
They have tweaked the testing criteria. Over all 6.12.15 seems fairly okay but there have been a few fixes applied that we are waiting for a new build. The next major version will address installer issues.

Project news - Einstein
We are currently analysing the S5 gravity wave run at the top end of their spectrum. They have decided they are going to use some of the data from S6 gravity wave run. Apparently its rather "noisy" but they believe they can use some of it but there will be varying run time due to tests they will use on it.

Project news - Orbit
Their initial grant has been used up. They applied for another one from NASA but it has been knocked back due to funding cuts to the NASA budget. That leaves the project without funding to continue.

It is probably one of the more useful projects to improve search strategies for near-earth objects that potentially will impact with the earth. Even political leaders might want to know they are going to be flattened by a rock before its too late to do anything about it.

Project news - Seti beta
They are looking at adding an extra search algorithm to their existing multi-beam work units. This will be tested on Seti beta before getting deployed to the main site.

19 February 2011

19th of February

Coming up on one for Einstein, but just a bit shy of 6 million at the moment.

BOINC testing
We got 6.12.15 this week. It addresses some installer issues for the Mac. It has a couple of tweaks around work fetch that apply to all versions. It won't ask for work if it uploaded in the last 5 minutes. I suspect this will lead to issues where people won't be able to build up a decent cache of work, but time will tell. Its meant to reduce the load on project servers that are overloaded.

Its a copy utility that comes with Windows 7 and Server 2008. There is also a version for Windows XP and Server 2003. It was designed as a reliable copy utility, however its much quicker than Xcopy or Windows Explorer.

I setup batch jobs (it runs from a command prompt) to backup the BOINC data directory to the file server. Depending on the amount of data this was taking up to 20 minutes to copy a couple of Gb. When using Robocopy it can do the same job up to 20 times faster. It can compare file timestamps and only copy newer or changed files a whole lot faster and skip the ones that haven't changed.

Machines in the shop
The technician who was working on both of them has been off sick for a lot of last week so they haven't done anything with the 2 machines. I'm expecting to bring them home next week.

Project news - Einstein
They have released a newer cuda cuda app that addresses all known issues with being able to cross-validate work done on a different platform. That is a Binary Pulsar work unit would fail the comparison if the version were run on a different operating system. For example the result files for a Windows work unit wouldn't match a Linux work unit.

Project news - Seti
Its up, but the network bandwidth has been totally used up since the middle of the week. Looking at the cricket graph its been pegged at 95Mbit for downloads and a smaller amount getting through for uploads. Oh to be a popular project :-)

Well the week started off cooler so I had a few machines running at the beginning and by today we're back to Maul only and running overnight. The forecast for next week is looking similar. As we head into autumn things should start to cool off and we can get more machines up and running.

12 February 2011

12th of February

Nothing new again this week.

BOINC testing
6.12.14 came out yesterday. I've upgraded a few machines with it. Nothing new really for windows users in this one.

File server
It went back to the shop this morning. They worked on it for a few hours, with me looking on. They flashed the BIOS (that took a couple of hours as it has to be done from a floppy and the machine doesn't have a floppy drive). That didn't make any difference and it still insisted one drive was degraded. We used the Adaptec controller and it doesn't do raid 5, so back to the Intel one. It still thinks one drive is degraded, after doing a high-level format as a non-raid drive. At that point I left them to it. They we going to low-level format the offending drive when I left.

The memory is still an issue, and they will address that later. I provided invoice numbers and dates so they can return them to their supplier for a credit and give me memory that works.

It was sitting on the same bench at the shop as the file server when I went in this morning. The shop has yet to replace the memory in it.

Project news - Seti
They are having trouble with one of their storage servers with it refusing to resync raid drives. Sound a bit like my problems. There is no new work being distributed and those that have completed work can't report it at the moment.

We have asked what additional hardware they need as it seems hard drives are failing quite regularly on the older machines, but they haven't been able to give a definitive list.

Software updates
Windows update has been giving out a fair few fixes in the last couple of weeks, so i'll run around and update each machine. Maul is fairly up to date as its running most days. There is also the nvidia driver and BOINC to bring up to the latest version on each one. Not a big deal but it all takes time and usually a reboot or two.

Update 13 Feb: It seems the nvidia control panel won't work on Win 7 x64 with a GTX295 when using 266.58 drivers. So i've had to go back to the 260.99 drivers on this machine.

08 February 2011

8th of February

Nope, nothing new this week. Actually I haven't looked at my stats for a week or so.

Once again it was hot. Saturday got up to 40 degrees, so of course everything was off. Sunday fortunately cooled off, so I managed to get a couple of machines going.

BOINC testing
We got 6.12.13 during the week. The main difference is a change to the user accounts that it creates for the sandbox installation (ie running as a service). I don't run it that way as it doesn't allow cuda apps to run under Vista or Windows 7 due to Microsoft's security features.

Well it went back last weekend as it kept locking up. It seems the Corsair memory is not what it claims to be. Allegedly they are rated for 1600Mhz (its printed on each stick), but running a memory test on them at that speed gets errors. The memory tester identifies them as PC10600, which should mean they are rated for 1325Mhz. Somewhat shy of 1600Mhz as printed on them.

The i7-920 never had a problem with them, but the i7-970 can presumably push them to their limit and beyond. The good thing is they have a limited lifetime warranty, so the shop will be replacing them with Kingstone Memory - when they get stock.

File server
My what a pretty case it has. The case is a Fractal Designs R3 in white. Its beautiffuly painted in white enamel inside and out.

Pity it too had problems. Apart from only thinking it has 2Gb of RAM when it actually had 3Gb it also decided the raid drives were degraded. Only problem is it wouldn't repair them which is the whole idea of having a raid 5 array.

It was painfully slow copying files across the network to it. On top of that the proxy server seems to take forever to do its job. Does anyone else run Squid 2.7 on Windows Server 2008?

All in all rather a let down. It will be off to the shop this weekend. It might be time to find another computer shop given all the recent problems with upgrades.

29 January 2011

29th of January

Nothing again. Its been pretty warm all week, so they have been off.

BOINC versions
We had 6.12.12 this week. This one is a release candidate. So far only a few minor problems reported to do with translations into German and French. I have it on 3 machines and have yet to experience any issues with it, but I need to give it some more testing around the work-fetch area.

GTX560 released
Nvidia released the GTX560 during the week. This uses a new gpu designated GF114. It has 384 cuda cores (aka shaders). Palit have a nice looking dual-fan overclocked model called the GTX560 Ti Sonic. Its specs looks quite impressive for the $210 (AUD) price tag, so it may become the "standard" card in most of my number crunchers.

At the moment its a toss-up between an overclocked GTX560 or an overclocked GTX570, both from Palit. I'll wait and see if we can get some real-life run times from fellow crunchers before buying anything. Besides nobody has any stock.

Kirk in the shop
Well it worked last week when it came home after its upgrade. Last night when I fired it up, after having a week off, it had problems. It started off alright and when I downloaded some new work errored out the whole lot. It also locked up a couple of times and I had to use the reset switch.

This morning its back to the shop for them to look at. Of course it decided to work straight out of the box. Anyway I left it with them doing a memory test and will pick it up next week.

Server bits
The shop are supplying a case, power supply, motherboard and I am supplying the other bits. This morning I dropped off the memory kit and Windows Server 2008 to put in it. I managed to find a shop selling Windows Server 2008 fairly cheap while I was in Melbourne during the week.

I ordered some hard disks for the new server on Friday and they should arrive next week. Its getting 3 x 1Tb WD caviar black drives which will be configured as a raid 5 array.

22 January 2011

22nd of January

None this week. Its been too hot.

BOINC testing
We got 6.12.11 during the week which is meant to address work fetch issues.

Unfortunately I left Maul fetching work for a couple of hours during the week. It went beserk and downloaded too much work. As soon as I noticed I disabled work fetch and promptly put 6.12.11 on it, but now it has to try and finish off the work its been allocated. This has been compounded by the fact its usually off during the day due to the hot weather.

Kirk upgraded
I got an i7-970 cpu during the week. This morning it was off to the computer shop for a quick upgrade. This coupled with a memory upgrade as I had a couple of memory kits lying around from 9 months ago. So now he has 12 cores and 12Gb of memory. This is what it looks like with task manager.

File Server
The old one is starting to have some issues, which isn't suprising considering its 8 years old. Its a hand me down that I was given because the power supply had failed. So i'm getting the shop to build a new box for me using some spare parts that I have and the i7-920 that came out of Kirk this morning.

Project news - Seti
Their upload server died during the week. No, its not one of the 3 new ones. They had to reconfigure the latest new server as the upload server temporarily. To quote one of the Seti staff "its much too good a machine to use as an upload server". Once they get the old server sorted out (appraently the 3ware raid controller seems to have stopped working) they will swap them back.

16 January 2011

16th of January

None this last fortnight. Not much work done at all. The computers have mainly been off due to summer. Maul sometimes gets to run overnight but thats about it.

Project news - Einstein
They have sorted out performance issues with their cuda app (which turned out to be the worker thread wasn't raised in priority). Maul which is equipped with 3 x GT240 cards can process a cuda work unit around 2 hours 10 minutes.

Project news - Seti
The seti guys got another server, which was built and paid for by the GPU users group. This arrived at UC Berkeley last week and is now in operation.

Graphics cards
I have been looking at getting a Palit GTX570 Sonic. Unlike the EVGA GTX570 that I have its a factory overclocked model. More importantly it doesn't have the reference design cooler, but rather a dual fan one. Currently nobody seems to have stock of these in Australia. As usual i'd get one, see how it goes and if fine replace the GTX295's with one of these in each i7 except Maul. That is it would become the standard card for the i7's.

Speaking of graphics cards i'll be selling off a bunch of the old ones on eBay. I have a number of GTS250's, an old GTX295 (1st generation) and an ATI HD4850.

Intel core i7's
It seems Sandy Bridge has arrived, so the older generation i7's are now slowly running out. I'll be looking at updating the 920's with 950's and selling off the old cpus. As mentioned before i'd love to get an i7-970 but they are still around $950 each.

I might even update the old file server (it is an old P4 @ 1.8Ghz given to me for free as the power supply had blown) to something more modern and use one of the old cpus in it.

01 January 2011

End of 2010

Well this is the position at the end of 2010. Unfortunately didn't get over 50 million, but will work on that for 2011. Let see if I can get to 75 million!

There has been no crunching on the farm for the last couple of days as it been so hot. Even 30+ degrees overnight, too hot to consider starting up one or two machines even. Monday is expected to have a temperature drop so maybe then I can get a few going to make up for lost time.

All the best wishes for the new year.