For efficiency maybe you should be looking at

http://www.ebay.co.uk/itm/ES-INTEL-XEON-E5-26...s&hash=item1e833cb4f4

Like my first duallie (http://stats.free-dc.org/stats.php?page=host&proj=bwcg&hostid=2408756) or

http://www.ebay.co.uk/itm/Intel-Xeon-E5-2670-...s&hash=item51b63f2f80

if you want a few more points too. Then there are:

http://www.ebay.co.uk/itm/141115531789?ssPage..._trksid=p3984.m1423.l2649

If it is years you seek. then maybe one of these too:

http://www.servercase.co.uk/shop/components/m...ard-ssi-90-sxg190-a0uayz/

OK OC, I had a closer look after replacing the PSU in my main PC as it died!

I wish I hadn't as your 2P/8c/16t machines are returning better results/points than my 2P/6c/12t machines. OK so I would expect that as they have more threads available. However, I would expect them to do proportionally more based on the number of threads, 24 against 32 I would expect about 30% more points/results.

The reason I say that is the benchmarks are reported as about the same. so the main difference is the additional threads.

This is one of the better performng nodes - Here

What am I doing wrong?
I've not done any optimisation
Would the 48Gb of RAM hinder progress?
What can I do to optimise performance?

Any tips would be appreciated

Thanks again OC, I'll have a look at that thread...

I'm aware that running my MegaCruncher with the Ubuntu desktop client running costs in resources and performance. But it's getting time to learn enough to manage and run the system via the command line...

I will run it with the desktop stopped for a while to see if it makes a significiant difference/improvement.

Stay tuned...

Whilst it wopuld be 'green' it is likely to be of limited use for crunching. The low speed rating and the low core count will mean running several to get the same output as one machine with more cores or higher speed.

It just seems like going backwards! slowing the processor to reduce the power usage, is ok but for crunching you have to balance the output with the power usage. From many discussions, the more threads the better the machine will perform, however, if the speed of the processor is too low, all those cores are still not going to give you a good return for the power they use...

This is, of course, my humble opinion...

There are so many factors that make up efficiency.

the following figures compare 2 different rigs using 8c/16t cpu's

Based on cpu speed of (a) 2.4 and (b) 3.1 If the slow machine get points * 1 the faster machine should get points * 1.29

Based on figures for RAC it gets 1.40

based on figures for last 28 days it gets 1.35.

The inconsistency of work units and the fact that the faster cpu has a larger l3 cache lead me to think that the speed based estimate is off.... but not by much.

Consider if you will another cpu with a speed of 2.1 but with 12 cores I believe this would get points * 1.31.

I expect to move my rigs to their new home this weekend. when I do I will add power use figures to the first two. Maybe this will give an indication of what to expect from the third cpu mentioned above but I think it is worth mentioning that the Intel xeons with the Low Power "L" annotation are the slower clocked ones with slightly lesser tdp