Can I use curve optimizer and offset undervolt at a same time?

[CodeError]

Last year I bought AMD CPU R5 5600X. I done some test and everything was fine. I don't game too often, this summer I played Helldivers 2 and temperature was going to max 85C. Room temperature was also high, around 34C. Now temperature in Helldivers 2 are in low 70C - room temperature 24C. Yesterday I done some test in 3DMark and I noticed that clock frequency now have lot of oscilation but benchmark score are the same.

Spoiler: Old tests - One Year ago

Spoiler: New tests - Yesterday

1. Why does clock frequency tend to oscilate now but before one year was stable?

2. Curve optimizer increases perfomance but also temperature. Can I use at a same time curve optimizer and offset undervolting? I would try these some combinations and put BIOS settings and benchmark progress here.

I can test stabillity in 3DMark, OCCT or any other program you want.

 

[ShrimpBrime]

If this is a stable setup for 1 year, you probably better off NOT making any changes and just enjoy the rig. Minor differences in PBO say 50 to 100mhz increase or decrease won't have a very large impact on benchmarking and gaming 3D.

The fluctuation is the CPU running where it needs to be, probably based on background tasks that may have been implemented through time via windows updates and user installed applications.
Generally to overclock it requires more power and not less. Of course this depends on temps and overhead, PBO only gives 200mhz increase either way, otherwise you run a static overclock.

It looks good, I wouldn't change anything.

 

[Ruru]

PBO only gives 200mhz increase either way, otherwise you run a static overclock.

Depends of the board though. My Crosshair VII for example can give moar than 200MHz.

 

[ShrimpBrime]

Depends of the board though. My Crosshair VII for example can give moar than 200MHz.

CROSSHAIR VII !!! So can give more, but what is the limit? And is that with a 5600, or it doesn't matter what cpu is installed?? My kids 5600G and ROG B450-I only allows 200mhz, but of course that's all I have to compare with.

 

[Ruru]

CROSSHAIR VII !!! So can give more, but what is the limit? And is that with a 5600, or it doesn't matter what cpu is installed?? My kids 5600G and ROG B450-I only allows 200mhz, but of course that's all I have to compare with.

I'll check on next reboot. Though I use the 200MHz option as it gives up to 5.05GHz on my 5800X.

 

[ShrimpBrime]

I'll check on next reboot. Though I use the 200MHz option as it gives up to 5.05GHz on my 5800X.

5600X isn't likely to hit some 5ghz even on that board lol.

 

[CodeError]

If this is a stable setup for 1 year, you probably better off NOT making any changes and just enjoy the rig. Minor differences in PBO say 50 to 100mhz increase or decrease won't have a very large impact on benchmarking and gaming 3D.

The fluctuation is the CPU running where it needs to be, probably based on background tasks that may have been implemented through time via windows updates and user installed applications.
Generally to overclock it requires more power and not less. Of course this depends on temps and overhead, PBO only gives 200mhz increase either way, otherwise you run a static overclock.

It looks good, I wouldn't change anything.

My primary goal is to lower temperature and try to maintain performance (+/-5%). I learned if you use only offset undervolt that will except temperature also lower CPU clock and with it performance. I will try to find sweet spot in the next week. If there is minimal difference I will return to default settings.


MBO is Asus x470-F, CPU cooler is Thermalright Phantom Spirit 120.

Nobody didn't mention that CO and offset undervolt aren't compatible so I will go with that.

I can't start with testing today, don't know for tomorrow but until the end of tuesday I will post pictures of BIOS setting and first CO and offset undervolt benchmark result.

What to use for stress test stability testing? OCCT is good but has a lot of options, don't know which one to set. Maybe go with 3DMark CPU profile test?

 

[freeagent]

You shouldn't need to touch the voltage at all. Use the curve with your own power limits in conjunction with boost override and you are off to the races.

I like Linpack Xtreme for all core loads, and SuperPi 32M for single core loads. I also use OCCT from time to time.

 

[ShrimpBrime]

My primary goal is to lower temperature and try to maintain performance (+/-5%). I learned if you use only offset undervolt that will except temperature also lower CPU clock and with it performance. I will try to find sweet spot in the next week. If there is minimal difference I will return to default settings.


MBO is Asus x470-F, CPU cooler is Thermalright Phantom Spirit 120.

Nobody didn't mention that CO and offset undervolt aren't compatible so I will go with that.

I can't start with testing today, don't know for tomorrow but until the end of tuesday I will post pictures of BIOS setting and first CO and offset undervolt benchmark result.

What to use for stress test stability testing? OCCT is good but has a lot of options, don't know which one to set. Maybe go with 3DMark CPU profile test?

It's not that CO and under-volt aren't compatible, they do the same thing essentially. One or the other. Most people seem to stick with CO and even at a per core level.

May I ask? Why the under-volting?? You have temps some people dream. Just keep the cooler clean and it will always perform well.

 

[Ruru]

5600X isn't likely to hit some 5ghz even on that board lol.

The PBO range goes up to 2000 as I checked

 

[ShrimpBrime]

The PBO range goes up to 2000 as I checked

2000Mhz!!
Does it ramp v-core accordingly? Have you tried like 500mhz? How does it go?? Fast?

 

[freeagent]

I always aim for +200, but usually +150 to be stable as a table

 

[Chomiq]

Curve Optimizer should help drop temperatures, not increase them as it's adjusting voltage on per core or on all core basis. I never used negative offset in combination with CO.

 

[Ruru]

2000Mhz!!
Does it ramp v-core accordingly? Have you tried like 500mhz? How does it go?? Fast?

Nah my CPU goes unstable even at +250

and I don't want to kill it with autovoltage spikes with too high PBO

 

[ShrimpBrime]

Nah my CPU goes unstable even at +250

and I don't want to kill it with autovoltage spikes with too high PBO

For sure, more speed requires more power = more thermal release. BUT if it's manageable, 90c or less, there's no reason not to go for it!!! (IMO)

2nd gen really liked cold temps from what I remember. It would pin all cores to 4.3ghz all the time with chilled water loop (tap to drain) which I did also try a TEC with too. And that was with the system on auto. Didn't have to mess with PBO or anything of the such.

 

[CodeError]

You shouldn't need to touch the voltage at all. Use the curve with your own power limits in conjunction with boost override and you are off to the races.

I like Linpack Xtreme for all core loads, and SuperPi 32M for single core loads. I also use OCCT from time to time.

It's not that CO and under-volt aren't compatible, they do the same thing essentially. One or the other. Most people seem to stick with CO and even at a per core level.

May I ask? Why the under-volting?? You have temps some people dream. Just keep the cooler clean and it will always perform well.

I forgot to mention above that when I had in hot summer 85C in Helldivers 2 that one of the core was at 90C. That is the main reason I want to lower temperature just a little with undervolting.

You say that curve optimizer is also undervolt, but I done test with default and CO -10 negative and with CO-10 temp is higher.

Spoiler: BIOS settings

Asus BIOS has two PBO option. I read that I should use only that in AMD Overclocking. This one on picture is not recommended to use. I left it on default. Should I turn it off?

Precision Boost Overdrive
Advanced
PBO limits -> Motherboard - Should I use Manual option?

Curve Optimizer
All cores
Negative -> -10

I didn't touch CPU Boost Clock Override +100-200MHz option.

Once again I measure temperature on default in 3DMark. Room tempereature was 22C.

Spoiler: Default

CPU-Z

CPU Profile benchmark

FireStrike Extreme benchmark

TimeSpy benchmark

Spoiler: Curve Optimizer -10

CPU-Z - Small increase in multi thread score

CPU Profile benchmark

FireStrike Extreme benchmark

TimeSpy benchmark

Why do I see temperature increase with Curve Optimizer -10? This is without offset undervolt.

What BIOS settings to change next?

 

[ShrimpBrime]

I forgot to mention above that when I had in hot summer 85C in Helldivers 2 that one of the core was at 90C. That is the main reason I want to lower temperature just a little with undervolting.

You say that curve optimizer is also undervolt, but I done test with default and CO -10 negative and with CO-10 temp is higher.

Spoiler: BIOS settings

Asus BIOS has two PBO option. I read that I should use only that in AMD Overclocking. This one on picture is not recommended to use. I left it on default. Should I turn it off?

View attachment 366714

Precision Boost Overdrive
Advanced
PBO limits -> Motherboard - Should I use Manual option?

Curve Optimizer
All cores
Negative -> -10

I didn't touch CPU Boost Clock Override +100-200MHz option.

View attachment 366715

Once again I measure temperature on default in 3DMark. Room tempereature was 22C.

Spoiler: Default

CPU-Z

View attachment 366719

CPU Profile benchmark

View attachment 366716

FireStrike Extreme benchmark

View attachment 366717

TimeSpy benchmark

View attachment 366718

Spoiler: Curve Optimizer -10

CPU-Z - Small increase in multi thread score

View attachment 366721

CPU Profile benchmark

View attachment 366720

FireStrike Extreme benchmark

View attachment 366722

TimeSpy benchmark

View attachment 366723

Why do I see temperature increase with Curve Optimizer -10? This is without offset undervolt.

What BIOS settings to change next?

For -CO to work (under-volting), needs to be used individual core usage.

 

[CodeError]

And PBO limits seems very important. When I switched from Motherboard to Auto in PBO limits, I saw drop in temperature.

Spoiler: PBO Limits - Motherboard; Invividual core -10 Negative

3DMark - CPU Profile

3DMark - FireStrike Extreme

3DMark -TimeSpy

Spoiler: PBO Limits - Auto; Invividual core -10 Negative

3DMark - CPU Profile

3DMark -TimeSpy

Auto in PBO Limits is for default CPU factory power limits?

How much average count can I go down before I encounter an error? -15? I will try later with -15 or -20 count.

 

[Taraquin]

Last year I bought AMD CPU R5 5600X. I done some test and everything was fine. I don't game too often, this summer I played Helldivers 2 and temperature was going to max 85C. Room temperature was also high, around 34C. Now temperature in Helldivers 2 are in low 70C - room temperature 24C. Yesterday I done some test in 3DMark and I noticed that clock frequency now have lot of oscilation but benchmark score are the same.

Spoiler: Old tests - One Year ago

View attachment 366351


View attachment 366352

View attachment 366353

Spoiler: New tests - Yesterday

View attachment 366354
View attachment 366360

View attachment 366355

View attachment 366359
View attachment 366356
View attachment 366358

1. Why does clock frequency tend to oscilate now but before one year was stable?

2. Curve optimizer increases perfomance but also temperature. Can I use at a same time curve optimizer and offset undervolting? I would try these some combinations and put BIOS settings and benchmark progress here.

I can test stabillity in 3DMark, OCCT or any other program you want.

I did that on my 5800X3D. I used -30 allcore + -56mv offset. It worked great

I forgot to mention above that when I had in hot summer 85C in Helldivers 2 that one of the core was at 90C. That is the main reason I want to lower temperature just a little with undervolting.

You say that curve optimizer is also undervolt, but I done test with default and CO -10 negative and with CO-10 temp is higher.

Spoiler: BIOS settings

Asus BIOS has two PBO option. I read that I should use only that in AMD Overclocking. This one on picture is not recommended to use. I left it on default. Should I turn it off?

View attachment 366714

Precision Boost Overdrive
Advanced
PBO limits -> Motherboard - Should I use Manual option?

Curve Optimizer
All cores
Negative -> -10

I didn't touch CPU Boost Clock Override +100-200MHz option.

View attachment 366715

Once again I measure temperature on default in 3DMark. Room tempereature was 22C.

Spoiler: Default

CPU-Z

View attachment 366719

CPU Profile benchmark

View attachment 366716

FireStrike Extreme benchmark

View attachment 366717

TimeSpy benchmark

View attachment 366718

Spoiler: Curve Optimizer -10

CPU-Z - Small increase in multi thread score

View attachment 366721

CPU Profile benchmark

View attachment 366720

FireStrike Extreme benchmark

View attachment 366722

TimeSpy benchmark

View attachment 366723

Why do I see temperature increase with Curve Optimizer -10? This is without offset undervolt.

What BIOS settings to change next?

Try reducing soc, vddg, vddp, vdd18, edc and tdc. Do them one by one. They all contribute to heat

 

[Zach_01]

And PBO limits seems very important. When I switched from Motherboard to Auto in PBO limits, I saw drop in temperature.

Spoiler: PBO Limits - Motherboard; Invividual core -10 Negative

3DMark - CPU Profile



3DMark - FireStrike Extreme



3DMark -TimeSpy

Spoiler: PBO Limits - Auto; Invividual core -10 Negative

3DMark - CPU Profile



3DMark -TimeSpy

Auto in PBO Limits is for default CPU factory power limits?

How much average count can I go down before I encounter an error? -15? I will try later with -15 or -20 count.

PBO on Auto means disabled, so power (W) and current (A) limits are the defaults.
When you set it to MB limits those limits are set freakishly high so then the main constraint is temperature.

If you want to control it you have to set PBO to advanced and manually set the desired limits.

You can monitor these limits on HWiNFO64 with CPU PPT, EDC, TDC sensors and PowerReportingDeviation(only for 100% CPU load).

 

[Chomiq]

I don't know what OP is trying to optimize here, Helldivers 2 is a heavy multithreaded workload, similar to running CP 2077. Seeing 70C is to be expected with CPU at 50-60% load probably at max frequency. Any temp drops OP is seeing right now are due to:
- ambient temps being reduced,
- power limits being gimped by turning off PBO.

SoC is already at 0.995 V, I doubt any changes to VDDG/VDDP will bring any temperature changes.

OP - fill in your system specs, that should be the first thing you do when you're asking for help.

Run Cinebench R23, check temps and scores with PBO2 off and enabled.

Also, check this thread:

5800x (and other Zen 3 chips) PBO settings/Temperature fix

The 5800x is well known for having heat issues on stock settings, because it's single 8 core CCX design shares the same power limits as the 2x6 (dual 6 core CCX) 5900x - so it's got a lot less space to cool itself with. A big part of the problem is that motherboards will raise the maximum TDC...

www.techpowerup.com

And apply the suggested PBO limits for higher performance on 5600X. Run R23 again and compare temps and scores.

Any additional improvements to temp should be achieved by adjusting Curve Optimizer, as it is pretty much down to individual samples no one can tell you when you can encounter errors. Easiest is to start at -30 and reduce it if you see instability.

 

[CodeError]

I did that on my 5800X3D. I used -30 allcore + -56mv offset. It worked great


Try reducing soc, vddg, vddp, vdd18, edc and tdc. Do them one by one. They all contribute to heat

Great! I will look into that!

I don't know what OP is trying to optimize here, Helldivers 2 is a heavy multithreaded workload, similar to running CP 2077. Seeing 70C is to be expected with CPU at 50-60% load probably at max frequency. Any temp drops OP is seeing right now are due to:
- ambient temps being reduced,
- power limits being gimped by turning off PBO.

SoC is already at 0.995 V, I doubt any changes to VDDG/VDDP will bring any temperature changes.

OP - fill in your system specs, that should be the first thing you do when you're asking for help.

Run Cinebench R23, check temps and scores with PBO2 off and enabled.

Also, check this thread:

5800x (and other Zen 3 chips) PBO settings/Temperature fix

The 5800x is well known for having heat issues on stock settings, because it's single 8 core CCX design shares the same power limits as the 2x6 (dual 6 core CCX) 5900x - so it's got a lot less space to cool itself with. A big part of the problem is that motherboards will raise the maximum TDC...

www.techpowerup.com

And apply the suggested PBO limits for higher performance on 5600X. Run R23 again and compare temps and scores.

Any additional improvements to temp should be achieved by adjusting Curve Optimizer, as it is pretty much down to individual samples no one can tell you when you can encounter errors. Easiest is to start at -30 and reduce it if you see instability.

I filled my system spec.

Yes, right now I am having great temps but because of the summer (one core was at 90C) I want to invest little time and study/apply Curve Optimizer to my CPU.

Thank you for the link, I will read that carefully with understanding. Tomorrow after work I am going on a trip until the end of weekend. After I return I will continue testing

​

 

[Zach_01]

Great! I will look into that!



I filled my system spec.

Yes, right now I am having great temps but because of the summer (one core was at 90C) I want to invest little time and study/apply Curve Optimizer to my CPU.

Thank you for the link, I will read that carefully with understanding. Tomorrow after work I am going on a trip until the end of weekend. After I return I will continue testing

​

You have options to tune the CPU with PBO (advanced+manual) settings to you liking.

Curve Optimizer
Boost Overdrive
PPT/EDC/TDC limits

And of course you can set a custom CPU max temp (90C is default), for example 80~85C or even lower if you like.

 

[CodeError]

I set every core in Curve optimizer to negative -20, I liked the temperature difference so I decided to stay on that value. Otherwise it would seem to me that it would be a very long testing because you need to test every core individually for a long time and I didn't want to squezze every last drop of voltage count, I was happy where I am.

So far I tested in:
- every core individually 8 hours in CoreCycler - Prime95 - Huge FFT
- 2X set of 3 cores in CoreCycler - Prime 95 - Small FFT - Core cycle - 5mn - 8 hours
- 1 hour - OCCT - CPU - Extreme Variable AVX2 - Core cycling every 6 seconds
- 1 hour - OCCT - CPU + RAM - Large Extreme Variable AVX2 - Core cycling every 3mn
- 1 hour - OCCT - CPU + RAM - Large Extreme Variable AVX2 - All cores
- Cinebench 30mn MultiScore
- Asus real bench - 8hours
- 3X - 3DMark; TimeSpy stress test (20 runs); FireStrike Extreme stress test (20 runs); Steel Nomad stress test (20 runs); CPU Profile - 5 runs
- Helldivers 2 - 2 hours

Not a single crash or WHEA error in HWiNFO64. Should I test more or this looks very stable?

 

[Zach_01]

I set every core in Curve optimizer to negative -20, I liked the temperature difference so I decided to stay on that value. Otherwise it would seem to me that it would be a very long testing because you need to test every core individually for a long time and I didn't want to squezze every last drop of voltage count, I was happy where I am.

So far I tested in:
- every core individually 8 hours in CoreCycler - Prime95 - Huge FFT
- 2X set of 3 cores in CoreCycler - Prime 95 - Small FFT - Core cycle - 5mn - 8 hours
- 1 hour - OCCT - CPU - Extreme Variable AVX2 - Core cycling every 6 seconds
- 1 hour - OCCT - CPU + RAM - Large Extreme Variable AVX2 - Core cycling every 3mn
- 1 hour - OCCT - CPU + RAM - Large Extreme Variable AVX2 - All cores
- Cinebench 30mn MultiScore
- Asus real bench - 8hours
- 3X - 3DMark; TimeSpy stress test (20 runs); FireStrike Extreme stress test (20 runs); Steel Nomad stress test (20 runs); CPU Profile - 5 runs
- Helldivers 2 - 2 hours

Not a single crash or WHEA error in HWiNFO64. Should I test more or this looks very stable?

Personally I think you’re ok with all these tests.
Now the ultimate test is to use it as you wish for a couple of weeks and see if it’s stable.

I usually run very few, very little synthetic tests and go quickly to use my PC as I want.