I have a new build consisting of a 7950x, and x670 gigabyte aorus elite ax rev 1.0, using curve optimizer does not seem to be changing my power draw at all. I have tried in ryzen master, as well as the BIOS, during r23 multicore, it seems to pull between 220 and 230w according to Core Temp, and boost between 5.1 and 5.3, no matter if I put the negative offset 5, 10, or 15. Also it seems turning PBO on to motherboard, auto, manual etc makes no difference either.
Under load, the cpu quickly goes to around 90-92 deg, consistantly, I have never hit 95c, but is there throttling going on between 90-95c also? Any help would be appreciated.
The cooler I am running is a EVGA CLC 280 with noctua fan swap. I am going to be adding a push/pull setup to it this weekend to try to further lower the temps.
more cooling will result in more power draw because it’ll just keep boosting unless you have a limit set somewhere else. Could you try set the lowest ECO mode and see if your power draw goes down at all just to double check that the changes you’re making are actually being applied?
Mine also never reaches 95c but the benchmark results are exactly what they should be so I’m not worried. From reviews I was expecting it to reach 95c after 10 seconds and never cool down. In practice it’s several Cinebench runs to get to 92c and then it stays there
I’d also recommend setting ryzen master back to default and making all your changes in the BIOS. There might be some sort of automatic overclock turned on which is overriding your Ryzen master settings or the other way around.
That’s not what curve optimizer does - it lowers the voltage for a particular boost speed by your offset, which lowers your temps and allows you to boost higher, which in turns uses more voltage.
CO neg offsets will allow you to boost higher and longer, but they won’t reduce power draw. If you want to reduce power draw you need to set a boost or power limit. CO neg offsets will allow you to boost higher for a given power limit.
As far as ryzen master vs bios, I do prefer to use the bios settings to adjust, and have been clearing out any changes in ryzen master before I change anything in bios. I will try the eco mode in a little bit to see if it will limit the power.
I think i don’t have a complete understanding of the curve optimizer still, as I am not seeing any increase in boost speeds in multicore r23 at any settings. If the CO is going to allow me to boost higher, because of less heat, that is not the behavior I am observing.
I also think my performance numbers are not off at all in general, as I am getting right around 39k in r23 multi-thread, with XMP turned off on my memory as its not stable with my kit unfortunately.
I will do some more testing and report back shortly, thanks again.
It won’t really be noticeable. A small CO negative offset will result in a few hundred points in CB, because you were able to sustain clocks for say 20 seconds longer before it throttled due to heat.
I’m getting 38k with a -18 offset and a power limit of 190W. If I remove the offset, the extra heat lowers my score by maybe 1000 points since it hits 95C earlier and throttles down earlier.
That’s kinda weird i see you have the revision 1.0 of the board.
I know there also is a revision 1.1 of the said board.
But how does the board behave on auto overclocking?
I do know that static overclocking is not really a strong point,
of the revision 1.0 board.
I have found that I can maintain 5.2-5.3 all core around 91 deg now at -12 offset, pbo manual motherboard, 150 mhz.
System is pretty stable at this point. I upgraded the cooling to a push/pull setup on the 280 AIO, and I am mostly happy with how it is performing. Thanks for the help.
My next issue is Exfo memory stability with the bundled gskill memory kit from Micro Center. I can not get memtest86 to pass after about 30 min with 2 dimms installed, so far. Last night I was able to run a single dim with exfo on without any issues. Going to be testing the other stick in the furthest slot later, to see if its the memory itself, or a signaling issue, then expanding the test to other slots on the motherboard if it passes.
the trick for all core is to actually use a negative freq override capping the freq so it doesn’t boost into the area of unstability
this will let you use greater negative curves