After a challenge I normally set up my systems to get a few of the next challenge wu's to test. So after the last challenge I did that. I got 4 wu's (wooLLR) and let them go. It has been some days so I went to check them. At around 55% done they have been running 137 hours. Calculating it out, it is going to take more than 8 days, which is the length of the next challenge.

Wondering if there is something I can do to speed them up?

They are running on somewhat older systems. Both q6600 cpu's. Each has a different motherboard and ram, but are running same in wu speed.

My main concern, even if they are long wu's, is finishing them in time during a challenge.

Thanks
Rick

Edit. I did want to add on almost all other wu types, my system will be close the the avg time listed in the sub-projects. But not these.

You should see what happens when you run 3 tasks instead of 4. *ALL* prime searches are searching larger numbers than they were searching only a couple of years ago. Bigger numbers mean more digits; more digits mean more memory is used.

The more memory you use, the larger the percentage of memory accesses that are going to not be fulfilled by the CPU's cache. Since the cache is MUCH faster than regular memory, cache misses can significantly slow down the program.

Running multiple copies of LLR uses even more memory, and increases the percentage of cache misses, which causes LLR to run slower.

When I started crunching here in 2010, multi-core CPUs were first becoming common, especially quad core CPUs, and people were starting to notice an inexplicable slowdown when running multiple copies of LLR. It's now pretty obvious that the cause is cache misses.

While the amount of memory used has increased over the years, the amount of cache in your hardware is exactly the same, so running multiple LLR's on your Q6600s is going to have a greater penalty than before. My computer is a Q6600, and it certainly seems that running 4 LLR tasks slows down more now than it used to years ago.

I am using a i5-2450M at 3.1Ghz and even with 3 tasks I find I cannot get one task finished it the eight days of the challenge.

The pair of older desktops are OK with under a hundred hours running 2 tasks at a time.

CPUID says I have 2 CPU on the i5 so perhaps I need to find out how to stop them doing double duty.

A lot of people are going to have trouble getting any work done in the challenge period. I make it that since the last WOO/CUL challenge the challenge has increased from 5 to 8 days but the tasks take about 3 times as long, but I am not certain I am actually using data from the last WOO/CUL challenge or an earlier one. The effect is the same -- some computers take longer than the challenge period to do any tasks.

We may have to have month long challenges before long.
____________
Member team AUSTRALIA
My lucky number is 9291*2^1085585+1

I am using a i5-2450M at 3.1Ghz and even with 3 tasks I find I cannot get one task finished it the eight days of the challenge.

Finally started on a new batch of them and limited it to 2 wu's at once. Running much faster. Right around the 100 hour mark by the time it ends. About 1% and hour. Will test the other machine with 1 @ a time next.

Rick

I'm also noticing very long expected work times.

On a Core 2 Quad Q9450, running 1 task, it's estimating another 141 hours
On a Core i7 930, that was running 5 tasks, estimated 178 hours.

My other two crunching systems (a Core i5 660 and an AMD FX-8120) running 2 and 4 tasks, are showing just under 70 hours remaining.

Thinking about HT, I just cut the i7 down to 3 tasks. We'll see what happens as it continues to run today.
____________
"Apollo was astonished, Dionysus thought me mad. But they heard my story further and they wondered and were sad."

I'm also noticing very long expected work times.

On a Core 2 Quad Q9450, running 1 task, it's estimating another 141 hours
On a Core i7 930, that was running 5 tasks, estimated 178 hours.

Manually calculating how much time I estimate (Time remaining = (TimeDone/%complete) - TimeDone) they should finish during the challenge, but still around 100 total hours (I aborted two WUs on the i7 to free up some cache and what not). Still seems like a long time on processors that aren't super old (the Q9450 blows away anyone that's using a Q6600 and a first gen i3/5, and the i7 isn't too shabby either). I think I tracked down my estimate problem though - the two systems that are reporting extra long times are running the nightly boinc from git (have to compile from source on Fedora for GPU computing to work), where as the other two are still 7.0.x (one is a Win box, the other I built quite a while ago).
____________
"Apollo was astonished, Dionysus thought me mad. But they heard my story further and they wondered and were sad."

Still seems like a long time on processors that aren't super old.