My computer ID: 5940401 has been running for a long time with an Athlon 64 x 2 5800+ slightly overclocked via FSB only, to just about 3GHz.

I was in my office looking around at stuff I was going to give away or put-away and I saw this old board with a Phenom 9550 4-core in it and thought to myself, "that might be sort of interesting..." and so I swapped the Athlon 64 for the Phenom.

So I've taken the number of crunching processors up from 1 to 3, but I've dropped the speed from 3 to 2.4Ghz. (in each case I left a core idle to feed the video card).

Also, let's take into account that it is running XP Pro 32-bit and is bound at 3.25GB of memory.

Anybody want to speculate as to whether the very steady 18-19k RAC (nVidia GTX 460) goes up or down on that computer?

Should go up. As long as the GPU's RAC stays somewhat steady, the CPU crunching has gone from 1x3GHz to 3x2.4GHz (effectively 7.2GHz). The task durations will increase, but you are doing ~233% the amount work in the same period of time on CPU.

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

Should go up. As long as the GPU's RAC stays somewhat steady, the CPU crunching has gone from 1x3GHz to 3x2.4GHz (effectively 7.2GHz). The task durations will increase, but you are doing ~233% the amount work in the same period of time on CPU.

But I've slowed the CPU-portion of the processing for the GTX460.

I noticed that the "free core" was staying about 50% busy, which is a little high.

If I slow the GPU, even a little, there's no way that the extra 2 cores at 2.4GHz can possibly overcome the loss.

I'm not even sure why you stopped using the 1 core in the first place. The amount of CPU use is minimal on Multibeam tasks. You may notice a few minutes added to your crunching.

I have an FX-8150 that I keep 1 core idle for Poem @home to run on the GPU. I've not noticed any time differences for my CPU work

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

A GTX460 doesn't stress a dual core 3200+, never mind a quad core phenom, so don't bother reserving one core to service the GPU, just let everything crunch and gain a few hundred more a day.

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

A GTX460 doesn't stress a dual core 3200+, never mind a quad core phenom, so don't bother reserving one core to service the GPU, just let everything crunch and gain a few hundred more a day.

Good advice. I can personally vouch for that. No need to reserve a core.

Regards,

A

---

A GTX460 doesn't stress a dual core 3200+, never mind a quad core phenom, so don't bother reserving one core to service the GPU, just let everything crunch and gain a few hundred more a day.

Good advice. I can personally vouch for that. No need to reserve a core.

Regards,

A

Unless you're running the Nvidia OpenCL Astropulse app, and have 270.xx or later drivers, then a full core is required to feed the app, otherwise it just stutters along,

Claggy

I'm not even sure why you stopped using the 1 core in the first place. The amount of CPU use is minimal on Multibeam tasks. You may notice a few minutes added to your crunching.

I have an FX-8150 that I keep 1 core idle for Poem @home to run on the GPU. I've not noticed any time differences for my CPU work

Seems reasonable to me, so next time I'm at the computer I'll add the other core to crunch and we'll see how that goes.

Hey!

I found-out!

The RAC is falling like a stone!

<ok, ok... I just needed to amuse myself>

confused... So the RAC dropped after adding the Core or before?

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

confused... So the RAC dropped after adding the Core or before?

skildude - I'm being "funny." I didn't get a chance to add the last core before I ran out of CUDA work, which is funny because I was worried about the CPU not being able to feed the GPU fast enough.

The RAC is falling because we're out of work.

It was "gallows humor."

Should go up. As long as the GPU's RAC stays somewhat steady, the CPU crunching has gone from 1x3GHz to 3x2.4GHz (effectively 7.2GHz). The task durations will increase, but you are doing ~233% the amount work in the same period of time on CPU.

Not only the clock is important. The Phenoms always do more MIPS than Athlons at the same clocks on my systems.

EDIT: Example: This Athlon x2 runs @2200 and this Phenom x3 runs @2100.
Compare the benchmark results for yourself.

---

confused... So the RAC dropped after adding the Core or before?

skildude - I'm being "funny." I didn't get a chance to add the last core before I ran out of CUDA work, which is funny because I was worried about the CPU not being able to feed the GPU fast enough.

The RAC is falling because we're out of work.

It was "gallows humor."

I thought that was reserved for the RIP skildude thread

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope