No gain from Radeon 4870 + Core i5 2500k folding

[frontier204]

Hi all,

I'm folding with my primary PC (Core i5 2500k 3.8 GHz / 8GB / Radeon 4870 512MB), and I noticed that no matter what I do with the settings, having my GPU + CPU folding together can't improve over just the CPU folding SMP units.

My setup is as follows:

• SMP3 Client folding with no advmethods or bigadv, with -smp 4
• GPU client folding with its default parameters, using Catalyst 10.12 DLLs
• Environment variables...
• BROOK_YIELD 2
• CAL_NO_FLUSH 1
• CAL_PRE_FLUSH 1
• FLUSH_INTERVAL 384

If my FLUSH_INTERVAL variable is set to 256 or higher, the GPU client eats ~5% of the CPU time (or 20% of 1 core). I get ~2600 PPD from the GPU but I lose 3000 PPD from the SMP.
If I set my FLUSH_INTERVAL variable to 32-48, the GPU client reportedly does not take any CPU time in Task Manager. GPU gives me ~1050 PPD as it's being starved, but SMP still loses around 1100 PPD.

In both cases, the GPU client takes up an extra 63W (according to my UPS), for a LOSS of total PPD. Does anyone else have an SMP + ATI combo that actually gets positive gain from the GPU?
...or should I stop GPU folding and suffer from the Radeon 4870's bad idle power consumption?

Thanks

[boggtrotter59]

Every ATI card I've owned sucked too many cpu cycles to be worth using... Combine that will the low PPD the cards puts out... it' a no win situation. I'd fold only on the cpu or get a non fermi nvidia card. Fermi cards suck quite a few cycles as well but not as bad as ATI cards.

[frontier204]

Thanks; as I thought... :(
So the extra wattage is better spent on OCing the i5 further I assume? (given that I'm at 46 Celsius with CPU fan at 1400RPM and my top exhaust fans unplugged, there's headroom left )

[boggtrotter59]

Ya that's what I would do... O.C. it as far as she'll go and see how much more power it consumes... shouldn't be that much more. You should also see a hefty increase in PPD with the increased clock speed...