More cores or higher clock speed?

[retina]

Originally Posted by nitf

Is there a way I can set a smaller FFT on an assignment I get?

No.

[nitf]

Originally Posted by retina

No.

Clear and succinct, just the way I like it. :)

The takeaways I've got from this are to not be so impatient and be content with a 3.2 ms/iter on the cpu, and to let the cpu take its foot off the gas and save me ~$56 in electricity costs year round.

(142-87 = 55 watt difference, /1000 = .0.055 kilowatts, * 24 * 365 = 481.8 kwh/yr, * $0.116116 rate= $55.94)

[axn]

Originally Posted by nitf

DDR4 PC4-25600 3200MHz.

Can you buy/get some 3600 (or 4000) sticks? Or alternately, can you try changing the FCLK to higher frequency?

[VBCurtis]

Originally Posted by nitf

Clear and succinct, just the way I like it. :)

FFT size is, roughly speaking, the level of precision for the main calculation loop in the program. The precision needed is determined by the size of the number being operated on; the program has been optimized and tested such that the FFT chosen is reliable, but the next-smaller one introduces unacceptable chances of error in the calculation.

So, you can go *bigger* FFT and the calculation works, but you can't go smaller.

As for your conclusion about your CPU- yes. If your BIOS allows you to take axn's advice, you can get a little more speed out of it (perhaps without adding much or any wattage). Prime95's performance is mostly limited by how fast it can get data (memory and cache bandwidth), rather than how fast the chip computes the data (clock speed of CPU).

[nitf]

Originally Posted by axn

Can you buy/get some 3600 (or 4000) sticks? Or alternately, can you try changing the FCLK to higher frequency?

I swapped out for 3600 (2x16 DDR4-3600 PC4-28800 CL16) and that brings me down to ~2.8 (anywhere from 2.775 to 2.95, with 25k iterations between screen output) ms/iter on the same exponent as before. Next step would be to fiddle with the fabric clock but I may coast a bit and enjoy this improvement for an assignment or two first.

Thanks everyone for your help.

[dcheuk]

Originally Posted by nitf

I swapped out for 3600 (2x16 DDR4-3600 PC4-28800 CL16) and that brings me down to ~2.8 (anywhere from 2.775 to 2.95, with 25k iterations between screen output) ms/iter on the same exponent as before. Next step would be to fiddle with the fabric clock but I may coast a bit and enjoy this improvement for an assignment or two first.

Thanks everyone for your help.

I found that tinkering with primary and secondary timing helps too.

[dcheuk]

Originally Posted by nitf

I swapped out for 3600 (2x16 DDR4-3600 PC4-28800 CL16) and that brings me down to ~2.8 (anywhere from 2.775 to 2.95, with 25k iterations between screen output) ms/iter on the same exponent as before. Next step would be to fiddle with the fabric clock but I may coast a bit and enjoy this improvement for an assignment or two first.

Thanks everyone for your help.

You might have better luck with the fabric clock - I wasn't able to get my system stable at >3600mhz memory clock + >1800mhz fclk 1:1 ratio with any memory kits lol.