Posts by Richard Haselgrove

1) Message boards : Cruncher's Corner : a new (?) netgear router. (Message 132441)
Posted 1 day ago by Richard Haselgrove


Your picture is interesting, I have been only swapping the middle two pairs of wires, not every pair of wires. But as you said as long as both ends of the cable are the same the router will just send the signal down the wire to the next connection. I just did a speed test on my laptop and while it is running 'g' it is only getting 25/18. My desktop, which is hard wired, is getting 43/34, so my wireless IS slower. Time to run more wires I guess!!

What's more, it's a standard - specifically, TIA/EIA-568 - and I've always found it's better to follow standards if it costs no more than doing it any-which-way.

Specifically, Mike has illustrated T568A (as it's usually known): I've always found that to be commoner in telephone systems. Computer network wiring is more commonly made up to T568B - which is electrically the same, but with the white/orange - orange/white pair connected to the top pins 1 and 2, and the white/green - green/white pair connected to the symmetrical pins 3 and 6.

I too have done the sort of networking which starts with an electric drill, a 18" masonry bit, and a set of stepladders.....

... and I've still got the boxes of Cat 5 cable in my cellar to this day.

Note that Cat 5 cable can be bought in two forms: solid core and stranded. The stuff which comes in 1,000-foot boxes is almost always solid core: it's designed for permanent, fixed, infrastructure situations. It's designed for connection to patch panels or the faceplates on wall outlets, using an IDC 'punchdown' tool. That's the sort of setup I'd use for your project: run the stiff solid-core cable as far as the skirting board (baseboard? floor moulding? kick board?), and terminate it with a fixed socket.

Those little RJ45 plugs with the catch which breaks off are much easier to fit on the softer, more flexible, stranded cable: use that to make up short-length patch cables to connect your equipment to the wall socket. I used to make those up too, until I got bored and found that if you purchase wisely, you can buy them ready-made and certified for much the same price as the cost of the parts. Just don't buy them in high-street computer shops, where the mark-up can be horrendous.

Personally, I like the ones with an integral strain relief like this:



- doesn't get in the way of the latch for unplugging, but does stop the wires working loose in the connector. I usually break the latches off by tripping over the cable and ripping it forcefully out of the socket: no sleeve or boot is ever going to stop it breaking under those conditions.
2) Message boards : Wish List : Improved handling of low speed internet links (Message 132127)
Posted 14 days ago by Richard Haselgrove
Could it be changed to allow the user to set a limit of one file transfer at a time, until this limit is no longer desired?

That exists already. Using a cc_config.xml file (see Client configuration), you can set (separately)

<max_file_xfers>N</max_file_xfers>
Maximum number of simultaneous file transfers (default 8).
<max_file_xfers_per_project>N</max_file_xfers_per_project>
Maximum number of simultaneous file transfers per project (default 2).
3) Message boards : Problems and Bug Reports : Work Running 100% Of Time Despite Settings? (Message 132000)
Posted 20 days ago by Richard Haselgrove
Or, from the official ClientSchedOctTen document,

This will tend to get large (max-min) clumps of work for a single project, and variety will be lower than the current policy.
4) Message boards : Problems and Bug Reports : question about last checkpoint (Message 131829)
Posted 26 days ago by Richard Haselgrove
The last batch I saw like that checkpointed 11 times during the run, or after each 9.09% of progress.

With the 'recommended' 7.2.42 version of BOINC that you are using, the early stages of progress (before the first checkpoint) are not accurately recorded: they are simply an estimate based on time spent so far (earlier versions of BOINC displayed no progress at all during this phase, which was even worse).

If you can, leave the computer running until the progress% shows exactly 9.090% (it should then stop changing, until a second big jump to 18.181%). You should then be able to look up the CPU time at the first checkpoint: you would need to leave the computer running for at least that long in each session, to make progress with these tasks. If that's not possible, you might be better off choosing a different type of Einstein or BOINC work.
5) Message boards : Cruncher's Corner : BOINC weirdness again (Message 131743)
Posted 30 days ago by Richard Haselgrove
That may have been me. They are *started* in FIFO order.

There is an override, in which tasks are marked "high priority" in BOINC Manager: we tend to call it 'EDF' for Earliest Deadline First on project message boards. That should have kicked in here, but clearly didn't.
6) Message boards : Cruncher's Corner : PC with 2 NVIDIA and 1 INTEL CPU/GPU, how to assign WU's (Message 131468)
Posted 44 days ago by Richard Haselgrove
The full list from http://downloadmirror.intel.com/23885/eng/ReleaseNotes_GFX_64.pdf is

IntelĀ® HD Graphics 5200 / 5100 / 5000 / 4600 / 4400 / 4200 / 4000 / 2500

7) Message boards : Problems and Bug Reports : "Run only the selected applications" problem (Message 131415)
Posted 47 days ago by Richard Haselgrove
I've bitten the bullet, and installed

08/06/2014 14:06:35 | | OpenCL: Intel GPU 0: Intel(R) HD Graphics 4600 (driver version 10.18.10.3621, device version OpenCL 1.2, 1298MB, 1298MB available, 184 GFLOPS peak)

from the win64_153322.zip: this time the setup.exe worked properly, and I didn't have to resort to a manual install.

First task is running now, apparently without problems. Timings won't be comparable, because the CPU is doing some heavyweight Albert testing at the moment, and I find that affects things. Previously, timings were tightly clustered around 660 seconds (11 minutes) when running 'lightweight' BOINC CPU tasks on three of the four cores: they rose to over an hour when all four cores were loaded. I suppose I should repeat that test with the new driver sometime.

OK, first task finished successfully in 13:38, against a 'current configuration' estimate of 13:21. Not fair to judge on that basis, because I was typing on the keyboard and had multiple windows open for at least some of the time - but not a disaster. Task 439781685, if you want to watch for validation.
8) Message boards : Problems and Bug Reports : "Run only the selected applications" problem (Message 131413)
Posted 47 days ago by Richard Haselgrove
My download file was win64_153318.zip: I see the current download for 64-bit Windows is win64_153322.zip. I might try that later - this is a test machine I can afford to mess with.

Please do report on how the new driver behaves if you decide to try it!

You wouldn't happen to have the installer for 9.18.10.3257, would you?
I'd be willing to test a newer driver if only I had the installer for the current one so I could back down again if the newer driver won't work.

Yes, I've got

Release Version: Production Version

Package: 206289

Intel(R) Graphics Driver: 9.18.10.3257
Intel(R) Display Audio Driver: 6.16.0.3112


Operating System(s):

Microsoft Windows* 7-64
Microsoft Windows* 8-64


Release Date: August 2, 2013

in both .zip and .exe formats - file is Win64_153117

It's about 140 MB in either format, so would take time to upload to a dropbox. Unless you have an FTP server?

I've downloaded 153322, but not tried installing it yet. Maybe after the next coffee...
9) Message boards : Problems and Bug Reports : "Run only the selected applications" problem (Message 131408)
Posted 47 days ago by Richard Haselgrove
A question for both Andrew and Richard or any one else running a Intel HD4000:

What driver version are you using for the iGPU?

I'm on version 9.18.10.3257 and that's working fine.

I've been looking to update the driver for other reasons than crunching but have had problems with the 10-series before and now I can't find a working download for my current one so I'm not comfortable trying a update that I might have problems reverting.

My HD 4000 is still on 9.18.10.3257, same as you, on the "if it ain't broke, don't fix it" principle.

But I also have an HD 4600 (host 8864187), which is running 10.18.10.3496 dated 11 March 2014. It's working fine, but I had the devil's own job installing it.

Machine is a Dell Optiplex 9020. People have reported having difficulty getting the automatic setup utility to run on Dells, but I'm beginning to feel that this is not a deliberate policy on the part of Dell - it's just a crappy installer provided by Intel. Anyway, I downloaded the .zip format driver package and tried running setup.exe, but that failed. The download page has a link to a 'Read Me' text file with manual installation instructions, and that did work (if you follow them exactly.....)

My download file was win64_153318.zip: I see the current download for 64-bit Windows is win64_153322.zip. I might try that later - this is a test machine I can afford to mess with.
10) Message boards : Problems and Bug Reports : "Run only the selected applications" problem (Message 131400)
Posted 48 days ago by Richard Haselgrove
In this particular case, there's a problem with the terminology - it is confusing, and they haven't found a succinct way of explaining it.

Binary Radio Pulsar Search (Arecibo)
and
Binary Radio Pulsar Search (Arecibo, GPU)

are effectively the same application: it's just that the 'big' GPUs - NVidia and AMD - are so darn fast that the tasks finish almost before they begin. So, the (..., GPU) variant packages 16 of the individual CPU tasks into one big bundle.

The Intel GPUs, on the other hand, are slower, and closer in runtime to the CPU cores. So, Intel GPUs get allocated tasks from the non-GPU queue.

That, of course, is the same queue that CPU tasks are drawn from, so you can't use that option to choose CPU but not iGPU. Instead, use the options towards the top of the preferences screen: uncheck "Use INTEL GPU".

I wonder why you can't use your INTEL Intel(R) HD Graphics 4000 GPU? Mine works just fine. Might simply be a driver issue.


Next 10

Home · Your account · Message boards

This material is based upon work supported by the National Science Foundation (NSF) under Grants PHY-1104902, PHY-1104617 and PHY-1105572 and by the Max Planck Gesellschaft (MPG). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the investigators and do not necessarily reflect the views of the NSF or the MPG.

Copyright © 2014 Bruce Allen