Crunching continues mostly overnight for Asteroids and Seti on the Intel-GPU machines
Raspberry Pi's and Parallella's are running Einstein work.
The B+ Raspberry Pi is playing up. Its probably worn out the micro-SD card, so I will get a new one and copy across the image from the current card.
The Nvidia GPU machines are off.
I upgraded one of the Intel GPU machines this morning. I'm running some work on it now to get an idea how much faster (or slower) it is from the i7-3770 that it replaced. The BOINC benchmarks indicate lower floating point but greater integer performance. The GPU is noticeably faster than the i7-3770.
I've replaced the power supply with a lower wattage (450w) gold-rated unit. I also replaced the motherboard (ASUS H170), CPU (i7-6700), memory (8Gb DDR4 at 2133Mhz) and CPU cooler. Its in the original case using the original DVD-ROM and hard disk.
Windows decided it was no longer activated. I called Microsoft to get an activation code and after going round in the phone queue 3 times providing my Win7 product code I get told that I have to reinstall Win7 activate it and then upgrade to Win10. That was my afternoon written off.
I will order the parts for the remaining Intel-GPU machines (5 of them) next week and probably be in the computer shop doing them a pair at a time for the next few weeks.
The BOINC 7.6.12 from the Debian Stretch repository seems to have an issue. It no longer writes out to the log files. The stderrdae and stdoutdae are no longer created and the std*gpu files are empty. We think this may be caused by other Debian updates. The Parallella version which runs under Ubuntu doesn't have this problem.
We got 7.6.14 today, so I have that installed on 1 of the Intel-GPU machines at the moment. It has some tweaks to reduce the I/O priority of itself and the tasks (not to be confused with the CPU priority) for windows only. There is some further refinement around number and time displays in the manager and tweaks for showing multi-GPU tasks.
ClimatePrediction went off the air a fortnight ago. They had a problem with their storage sub-system controller that required a firmware update and then they had to restore/rebuild their 200Tb storage array. They came up partially last week and then took most services off the air. They seem to be back online now although its very slow, probably due to large numbers of climate models being returned.