Overclock.net banner
1 - 20 of 30 Posts

Prophet4NO1

· Registered
Joined
·
3,727 Posts
Discussion starter · #1 ·
I have been playing with my curve and power settings off and on since getting my 5800x almost two years ago. Aside from the PBO and power limits, are there other settings to help keep temps down? Just cranking the fans on my loop to max plus a degree or two from water temps, getting more stable higher clocks and better scores on benches. Are there other power sucking heat intensive settings that are good to turn off or tweak deeper in the BIOS? I have a Crosshair VIII Hero and most of the settings have little to no documentation to really explain what they do. I already have most of the boards features I do not use turned off like network stack, since I have a 10Gig NIC in the machine. I think I have pretty much squeezed all this chip has to give with out chilling my water to see if it will boost higher.

Basically looking for the things people have found to help get a tiny bit more. This is manly a tinker and gaming rig. I have been poking around but have not really found anything in my own tinkering that seems to help and no real threads anywhere about streamlining the BIOS settings. I know anything I am missing will only give tiny gains. I am just poking around for fun.
 
Discussion starter · #3 ·
Yes. Like I said above, I have been playing with curves and power limits since day one. Test, retest, tweak some more. It is dialed in about as good as it will get. Right now it will do 4.7-4.8Ghz all cores loaded. Seems to be temp limited for the most part, since getting my loop a tad cooler keep the all core in the 4.8ghz area more. I am not keen on doing a water chiller, so looking for ways to drop heat that are more fringe areas. I can not really go any lower on power limits or negative offsets and get the same clocks.
 
Yes. Like I said above, I have been playing with curves and power limits since day one. Test, retest, tweak some more. It is dialed in about as good as it will get. Right now it will do 4.7-4.8Ghz all cores loaded. Seems to be temp limited for the most part, since getting my loop a tad cooler keep the all core in the 4.8ghz area more. I am not keen on doing a water chiller, so looking for ways to drop heat that are more fringe areas. I can not really go any lower on power limits or negative offsets and get the same clocks.
Sounds like you basically have it tuned to the max already. I'm not aware of anything else at least that would give you lower power and temps while keeping those clocks.
 
Discussion starter · #5 ·
Yeah, it is kind of grasping at straws. It would be nice if more of the stuff in BIOS had some form of documentation. Lots of switches to flip. No explanations on what they do. Been that way forever. Way back in the day it was a big deal to turn off as much stuff as possible that you did not need to take load off the north bridge as best you could. Help with OC a fair bit on some chips. With the modern SoC setups, I wonder how much is buried deep in BIOS for things the SoC is doing that you do not need sucking power and pumping heat into the whole package. But it has some cryptic label on it only the engineers and AMD understand.
 
Run a static clock, get rid of the boosting. Curve pretty much just ramps up the cpu, PBO helps sustain it..
 
Discussion starter · #7 ·
I have not played with static clocks much. I would assume since PBO and curve tuning can get me around that 4.7-4.8 range on all core loads, a manual clock would be as high if not a little higher. I have a profile in my BIOS that I started playing with. I just ended up focused on curve tuning when I had spare time here and there.
 
Static at 4.6 is extremely easy to cool. You will lose out on the high single turbo but at 4.6 is a happy medium. But both 5800Xs I have owned will do 4.6 with 1.25 or less on the vcore.
 
  • Rep+
Reactions: Owterspace
Aside from the PBO and power limits, are there other settings to help keep temps down?
If you worry much about the temps it might be worth to consider the low leakage parts like 5700x.
The 5800x designed to reach higher clocks at the expence of the power consumed.
Hence it has the 60% higher TDP than that of 5700x.
Try to limit its EDC supply a bit as the quick fix, this has less of impact on performance.
 
Aside from making sure you're using the least aggressive (most droop) load-line calibration setting, there isn't much you can do for CCD temperatures outside of PBO.

That said, using PBO to cap maximum boost with a negative boost clock offset can allow much deeper negative COs than stock or a positive boost offset. Depending on sample 4.7-4.8GHz all-core boost, that will hold in all but the heaviest tests, should be possible.
 
Discussion starter · #14 ·
What CPU block do you have? Saw a good drop on my x3D going from EK to Optimus.

EK. I do not recall the name of the block off hand. But, they are all pretty similar besides tops and finishes. I may look at the Optimus when ever I get around to getting a new GPU. Will need to tear the loop apart anyway. Good time to look at changes.
 
Discussion starter · #15 ·
If you worry much about the temps it might be worth to consider the low leakage parts like 5700x.
The 5800x designed to reach higher clocks at the expence of the power consumed.
Hence it has the 60% higher TDP than that of 5700x.
Try to limit its EDC supply a bit as the quick fix, this has less of impact on performance.

I pretty much have all my PBO limits as good as they are going to get. Any lower and clocks drop or there is more instability in boost. I keep revisiting them since getting the curve dialed in. I may try playing with them some more, but it seems to be about as good as they will get.
 
Discussion starter · #16 ·
Aside from making sure you're using the least aggressive (most droop) load-line calibration setting, there isn't much you can do for CCD temperatures outside of PBO.

That said, using PBO to cap maximum boost with a negative boost clock offset can allow much deeper negative COs than stock or a positive boost offset. Depending on sample 4.7-4.8GHz all-core boost, that will hold in all but the heaviest tests, should be possible.

Vdroop may be on auto. I will have to look. At one point I was playing with it and anything not auto was causing issues. A lot of weird crashing. But that my have been on an older BIOS. Not sure if I have messed with it on the one I have now. I will have to revisit that.
 
I pretty much have all my PBO limits as good as they are going to get. Any lower and clocks drop or there is more instability in boost. I keep revisiting them since getting the curve dialed in. I may try playing with them some more, but it seems to be about as good as they will get.
Not sure I get what "instability in boost" means, but anyways,
can you clarify what exactly is your concern about temperature?

Honestly, I'm vainly trying to understand, why this triggers so much user attention.
 
Not sure I get what "instability in boost" means, but anyways,
can you clarify what exactly is your concern about temperature?

Honestly, I'm vainly trying to understand, why this triggers so much user attention.
He probably means negative CO high single core boost.

I can't answer the question as to why this triggers so much user attention. But I think that Ryzen 5000 chips are temperature limited. For example my 5700x with either static OC or unlimited power draw can get quite hot. More so than 3800x that I had, in the case of 3800x it was also temperature limited but it shown very clear stopping point with frequency.

I can't max out my 5700x before maxing out it's temperature ( single core still 4850 limited, thanks AMD ). Regardless, I think these chips are capable of pushing higher all core clocks but temperature is limiting factor.
 
He probably means negative CO high single core boost.
Unlikely, as the latter isn't affected in any way by PBO limits (current/power).

But I think that Ryzen 5000 chips are temperature limited
This is not quite true.
What you observe is the reliability limit in action which is not controllable from user space.
Tech-wise it's just a function of temperature and voltage: Fmax = f (Vdd, t°).
The greater any of the two, the lower allowed max allcore frequency.
 
Unlikely, as the latter isn't affected in any way by PBO limits (current/power).


This is not quite true. What you observe is the reliability limit in action which is not controllable from user space.
Small Ryzen core dies cannot transfer all that heat into the heatsink or cold plate of AIO / liquid cooler. At 1.25v, 1.21v adjusted for droop I already hit thermal limit with my 5700x running Prime95 AVX2 small FFTs. Board is reporting like 250 watts of power draw but I don't believe that at all. CPU is sitting under Arctic Liquid Freezer II 360mm AIO.

But that aside if I had more freedom to go up to 1.3v or 1.35v I would be able to push higher clock. I personally would prefer having static overclock that is AVX2 Prime95 stable but that limits core clock quite a bit. So using negative CO is more practical, to drop 300 - 400 ~ MHz in AVX2 high power draw loads and still get good all core boost for rendering and such loads.

More voltage = Higher stable core clock ( Probably caps out at like 5 GHz - 5.1 GHz all core for Ryzen 5000, depending on temperature )
Less heat = more voltage ( also less heat will result in somewhat lower power draw, possibly more stable it depends )
 
1 - 20 of 30 Posts