I noticed a long time ago that CPU tasks take more than twice as long when I run the Intel GPU tasks. I captured some data, but not enough to model whether it's more advantageous to just turn the damned thing off.

Caveat: I'm sure more cores (like on i7 CPUs) help compensate for this, but it also slows down tasks on ALL of those cores...so...

Anyway, has anyone captured ~48 hours worth of credit using CPUs and the Intel GPU, and then done the same with just the CPU tasks? This would be easier with standardized work unit credit, like on Einstein@home. The variable credit means I can't just use the run time for the comparison.

Even if you don't have data, any thoughts on this, or the methodology I intend to use (if nobody has looked at this)?

---

Of course, the model of Intel GPU would matter and I forgot to mention that.

I have three machines running HD4000 GPUs
2 i5s (4 CPU threads) running Linux, and 1 i7(8 CPU threads) running Mac OS X.

I also have an i3 with an HD4400 running Windows 10, and it looks like it cranks out twice the credit per clock cycle that the HD4000 does.

So from the data I do have, it looks like the HD4000 GPU equates to the production of about two processor cores. Which seems to put things at break even when it makes CPU tasks take twice as long (and I run another CPU project). So it appears using the HD4000 is definitely a losing proposition with the i7 since it doubles the processing time for all 8 CPU threads and only gives me back about 2 CPU cores worth of performance (it looks like).

---

(if nobody has looked at this)?

Done > 2 years ago:
http://setiathome.berkeley.edu/forum_thread.php?id=75215
http://setiathome.berkeley.edu/forum_thread.php?id=75971
http://setiathome.berkeley.edu/forum_thread.php?id=76814

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

Thanks! It didn't just give a short answer, I can see how various configurations affected things and why this is a problem.

---

I have an I7 with Intel GPU
When NOT running the GPU task, my CPU's run at 2.7 Ghz
When running the GPU taks, my CPU's run at 1.5 Ghz
=> this explains the longer time needed !