30 minutes into the challenge and just under 5000 tasks have been sent out. On normal days, we've been averaging 437 tasks per day.

I'm guessing that the very fastest computers, running the shortest of the TRP tasks may complete those tasks in about 5 hours. Perhaps a bit less if some cores are left idle to increase memory bandwidth.

Note that TRP tasks vary in length by about 50%; on my computer tasks will take somewhere between 8 and 12 hours. Most tasks are near the higher end of the range.

Let's try door number 3. I had only two of the channels populated. I moved 8g to each of the other two channels. Let's see if that helps.

Have you verified that system RAM is operating in quad channel mode by CPU-Z or similar utility?

Yep just did and it checks Quad. When I built the box I just populated 2 of the 4 channels with 4 sticks. This MoBo can handle 8 each 8gb modules. Today I moved 2 sticks of the ram 1 each to the other two channels...time will tell.

Let's try door number 3. I had only two of the channels populated. I moved 8g to each of the other two channels. Let's see if that helps.

Have you verified that system RAM is operating in quad channel mode by CPU-Z or similar utility?

It looks like the ones I just did have a larger FFT size. This from the stderr output...Using all-complex FMA3 FFT length 2560K. Does this affect the use of cache or ram? FYI my chip is a good one. I do not overclock but the chip I have is a good one from the substrate. For convenience the CPU is water cooled and is hard to get very hot.

The 2560k FFT is on the SoB unit you just completed. The recent Cullens are 1440k. If you were running the SoB at the same time as the Cullen work, then yes, it would have slowed things down some. Basic rule of thumb is that the size of the FFT increases so does the memory footprint of the work unit. That SoB (combined with multiple Cullens running at the same time) probably pushed you out of the cache and into the slower DDR4 RAM. When running my 33 hour Cullens, I wasn't running other types of primes and did not have any GPU work running either.

If you had only the Cullens running together and still got the 46 hour times, then I suspect that one of a few possibilities are slowing your machine down (46 hours is definitely slow for an i7-5820k):

1) other stuff was running as well (e.g., processor intensive stuff like gaming, CAD, etc. not just e-mail or a word processor).

2) heat...if your are getting overheating, the CPU may be throttling back.

3) your DDR4 RAM isn't running is quad channel

It looks like the ones I just did have a larger FFT size. This from the stderr output...Using all-complex FMA3 FFT length 2560K. Does this affect the use of cache or ram? FYI my chip is a good one. I do not overclock but the chip I have is a good one from the substrate. For convenience the CPU is water cooled and is hard to get very hot.

The 2560k FFT is on the SoB unit you just completed. The recent Cullens are 1440k. If you were running the SoB at the same time as the Cullen work, then yes, it would have slowed things down some. Basic rule of thumb is that the size of the FFT increases so does the memory footprint of the work unit. That SoB (combined with multiple Cullens running at the same time) probably pushed you out of the cache and into the slower DDR4 RAM. When running my 33 hour Cullens, I wasn't running other types of primes and did not have any GPU work running either.

If you had only the Cullens running together and still got the 46 hour times, then I suspect that one of a few possibilities are slowing your machine down (46 hours is definitely slow for an i7-5820k):

1) other stuff was running as well (e.g., processor intensive stuff like gaming, CAD, etc. not just e-mail or a word processor).

2) heat...if your are getting overheating, the CPU may be throttling back.

3) your DDR4 RAM isn't running is quad channel

1 Cullen WU with a Intel(R) Core(TM) i7-5820K CPU took 349,090.39 seconds...96.969 or some hours. I thought that horse might run faster....

Looks like you have HT running. Running 6 instead of 12 at a time would be more than a 50% improvement in times. It isn't just that the HT is engaged, but with 12 units you are filling up even that larger Haswell-E cache. Your times would be closer to my E-1650 v3 Xeon that is running them in about 32 hours or so doing 6 at a time.

I am trying 6 Cullen wu's on the I7-5820K with 6 cores no HT to see if it compares with your Xeon. Their architecture seems to be similar

Looks like I am headed to around 46 hours for 6 WU's so the Xeon is a bit faster for unknown reasons (more hardware registers etc.) as the cache is the same and ram is probably close (mine running 2133).

I just checked the last Cullen work units to run on the Xeon and they are about 33 hours (1440k FFT size). That FFT size could make a very big difference if the units you completed are using a larger FFT.

My RAM is also 2133, but I think your clock might be a little higher. That said, the Xeon and i7 may turbo mode differently, which might produce some difference also.

I am not clear on the FFT size issue. Can you fill me in please...

You can see the FFT size in the task column for any completed LLR workunit for your machine. This one, for example, is one of your completes with an FFT of 1440k. Your times will be similar once they complete...I think you are looking at an early estimate based on the percent complete, which is likely off a bit.

Also, I was thinking about a 5930k rather than a 5820k like you have. Your clocks are 3.3Ghz stock; my Xeon is at 3.5 stock so it will be just a bit faster unless you use that "k" designation to overclock your CPU.

1 Cullen WU with a Intel(R) Core(TM) i7-5820K CPU took 349,090.39 seconds...96.969 or some hours. I thought that horse might run faster....

Looks like you have HT running. Running 6 instead of 12 at a time would be more than a 50% improvement in times. It isn't just that the HT is engaged, but with 12 units you are filling up even that larger Haswell-E cache. Your times would be closer to my E-1650 v3 Xeon that is running them in about 32 hours or so doing 6 at a time.

I am trying 6 Cullen wu's on the I7-5820K with 6 cores no HT to see if it compares with your Xeon. Their architecture seems to be similar

Looks like I am headed to around 46 hours for 6 WU's so the Xeon is a bit faster for unknown reasons (more hardware registers etc.) as the cache is the same and ram is probably close (mine running 2133).

I just checked the last Cullen work units to run on the Xeon and they are about 33 hours (1440k FFT size). That FFT size could make a very big difference if the units you completed are using a larger FFT.

My RAM is also 2133, but I think your clock might be a little higher. That said, the Xeon and i7 may turbo mode differently, which might produce some difference also.

1 Cullen WU with a Intel(R) Core(TM) i7-5820K CPU took 349,090.39 seconds...96.969 or some hours. I thought that horse might run faster....

Looks like you have HT running. Running 6 instead of 12 at a time would be more than a 50% improvement in times. It isn't just that the HT is engaged, but with 12 units you are filling up even that larger Haswell-E cache. Your times would be closer to my E-1650 v3 Xeon that is running them in about 32 hours or so doing 6 at a time.

I am trying 6 Cullen wu's on the I7-5820K with 6 cores no HT to see if it compares with your Xeon. Their architecture seems to be similar