Intel® Core™ i7-3630QM

[pinhodecarlos]

Which one to run?

collatz
mini_collatz
solo_collatz

[DrPop]

Which one to run?

collatz
mini_collatz
solo_collatz

I would run the solo_collatz.

[pinhodecarlos]

I ran 3 wu's of the mini_collatz application due to its size. Still waiting for their validation so I can see which boinc project gives more credit and uses less % of GPU usage without limiting the desktop interaction.

Carlos

[Slicker]

I ran 3 wu's of the mini_collatz application due to its size. Still waiting for their validation so I can see which boinc project gives more credit and uses less % of GPU usage without limiting the desktop interaction.

Carlos

While all three Collatz apps use the same credit calculation based on total steps, collatz is much larger and puts less stress on the server and network lines, so it gets 10% more credit than mini_collatz as an incentive for users to run it.

Since the solo app checks all results twice eliminating the need for a wingman, twice as much work gets done. That and since AMD's OpenCL performance sucks, it should pay about the same as the Collatz v2.09 ATI app. So, I thought it only fair to award double the credit. Those with NVidia GPUs should smile since they get the same credit bonus even though the CUDA app isn't much slower than the v2.xx versions.

[Fire$torm]

While all three Collatz apps use the same credit calculation based on total steps, collatz is much larger and puts less stress on the server and network lines, so it gets 10% more credit than mini_collatz as an incentive for users to run it.

Since the solo app checks all results twice eliminating the need for a wingman, twice as much work gets done. That and since AMD's OpenCL performance sucks, it should pay about the same as the Collatz v2.09 ATI app. So, I thought it only fair to award double the credit. Those with NVidia GPUs should smile since they get the same credit bonus even though the CUDA app isn't much slower than the v2.xx versions.

So I was correct when I told DrPop that Solo_Collatz pays better.... Good to know.

[Mumps]

While all three Collatz apps use the same credit calculation based on total steps, collatz is much larger and puts less stress on the server and network lines, so it gets 10% more credit than mini_collatz as an incentive for users to run it.

Since the solo app checks all results twice eliminating the need for a wingman, twice as much work gets done. That and since AMD's OpenCL performance sucks, it should pay about the same as the Collatz v2.09 ATI app. So, I thought it only fair to award double the credit. Those with NVidia GPUs should smile since they get the same credit bonus even though the CUDA app isn't much slower than the v2.xx versions.

But, correct me if I misunderstand it, didn't you say the "solo" is accomplished by doing the "Wingman" check using the CPU? Is that computationally intense? Meaning Carlos wouldn't want to dedicate the CPU to that rather than his NFS post-processing.

[pinhodecarlos]

Here's the deal, with Collatz running mini_collatz wu's I can do 53k points per day when compared to 63k per day with Moo! Wrapper. Both run smoothy on my laptop.
About the wu using a little bit of CPU, it doesn't matter because due to the fact that msieve, the program I use for post-processing jobs, isn't well optimized to take advantage of the cache and threads of an i7. Right now the job is running on 4 threads leaving 4 for the system.

Carlos

[Fire$torm]

But, correct me if I misunderstand it, didn't you say the "solo" is accomplished by doing the "Wingman" check using the CPU? Is that computationally intense? Meaning Carlos wouldn't want to dedicate the CPU to that rather than his NFS post-processing.

Good question since the app only allocates 0.02 cores per wu. At least on my boxes

[DrPop]

I would like to see the temps of your GPU cores when running Collatz vs Moo!. At least on my rigs, Moo! has always been quite a bit more demanding of the hardware and runs hotter - not something that goes too well with a laptop, although yours might be high end with good cooling so maybe it doesn't matter?
Try with the solo_Collatz as well, and see how much more you get than min_collatz. It will make a difference.

[pinhodecarlos]

I would like to see the temps of your GPU cores when running Collatz vs Moo!. At least on my rigs, Moo! has always been quite a bit more demanding of the hardware and runs hotter - not something that goes too well with a laptop, although yours might be high end with good cooling so maybe it doesn't matter?
Try with the solo_Collatz as well, and see how much more you get than min_collatz. It will make a difference.

I am not worried about the GPU temps because I use a cooling PAD and I noticed that temperature never raised above 63 ºC, in any kind of the tested GPU applications.
Another question is that having an Ivy Bridge processor I can use the internal GPU to help the calculations.
I think I will stay with Collatz because the RC5-72 is a waste of processing power, always had been since the days I crunched for them. They are attacking a problem by brute force and mathematically it is not the best way when there is another.

Here is the temps running mini_collatz:



Carlos