Ryzen 5 3600 95ºC

[Derek12]

You don't need to disable core performance boost because PBO will still make it boost (If Im not Mistaken). The thing you need to do is to go to Windows Power Plan and set 'Maximum CPU state' to 99% that will disable boost. For idle you set the 'Minimum CPU state' to 0% that will make it idle as long as there is no background process taking your CPU cycle.

You're right, I didn't thought about that, I am using the Ryzen balanced power plan but I've set the maximum to 99 and maximum temperature in CB is 69ºC though the CPU clock is a bit lower and Windows says 97% CPU usage... I got 7870 score (versus 8676 with 100%)

 

[ixi]

The 3600X & XT come with the Wraith Spire which is not actually bad at all. The problem is that the 3600 comes with the Wraith Stealth which is about half the size (about the same size as an Intel Celeron/Pentium stock cooler). Even the 3400G which is a quad-core comes with the Spire cooler.

Ohh, didnt know that 3600 has different cooler.

 

[Apocalypsee]

You're right, I didn't thought about that, I am using the Ryzen balanced power plan but I've set the maximum to 99 and maximum temperature in CB is 69ºC though the CPU clock is a bit lower and Windows says 97% CPU usage... I got 7870 score (versus 8676 with 100%)

View attachment 186190

Yep about right, without boost the CPU will run at 3600MHz instead of above 4000MHz. That is what I do before change the cooler. The temps looks nicer but I hope it won't affect the thing you do at least until you buy new cooler. I play games on my rig and even at 3600MHz its plenty fast for 1080p gaming

 

[Derek12]

Yep about right, without boost the CPU will run at 3600MHz instead of above 4000MHz. That is what I do before change the cooler. The temps looks nicer but I hope it won't affect the thing you do at least until you buy new cooler. I play games on my rig and even at 3600MHz its plenty fast for 1080p gaming

Thanks! I was referring the clock is a bit lower than 3600, it's 3500MHz and Windows task manager says 97% CPU usage but I think the difference is minimal

 

[Mussels]

Thanks for the replies

I tried disabling core performance boost, and the temps reached 80º at full.

Maybe I will get a new HS that fits in my case, and if it's still hot, then a new case (that fits in the desk space)...

Another thing, is it possible to reduce idle clock further? Hwinfo says the minimum is 500MHz but even in power saving mode and minimum clock at 0 or 1% it only gets to 2100... it consumes about 20W in idle with its temps at 40's, while the i3 6100 consumed 1-2W and temps were basically ambient's (it's true the R5 is in order of magnitudes more powerful than the i3 but I would like to reduce that power consumption...

CPB sounds like the issue, its over volting it past what the stock cooler can handle

Idle clock reporting doesnt work when the cores are *off* - you wont see low clock speeds because instead of clocking down, they power off completely (and dont report anything to monitoring software, which shows the last speed they reported)

Theres a whole lot of weirdness when it comes to ryzen 3000 and idle, but my experience boiled down to: update to the latest chipset driver, and watch background programs. Icue was a particular PITA, as it totally breaks ryzen idle states unless you delete certain monitoring files it seems to use extremely outdated versions of.

 

[Derek12]

Thanks! I was referring the clock is a bit lower than 3600, it's 3500MHz and Windows task manager says 97% CPU usage but I think the difference is minimal

I also forgotten to say the fans weren't at full speed when I made the graph...

CPB sounds like the issue, its over volting it past what the stock cooler can handle

Idle clock reporting doesnt work when the cores are *off* - you wont see low clock speeds because instead of clocking down, they power off completely (and dont report anything to monitoring software, which shows the last speed they reported)

Theres a whole lot of weirdness when it comes to ryzen 3000 and idle, but my experience boiled down to: update to the latest chipset driver, and watch background programs. Icue was a particular PITA, as it totally breaks ryzen idle states unless you delete certain monitoring files it seems to use extremely outdated versions of.

Many thanks, I didn't know that, it could like Android's AIDA64 which sometimes say when cores are turn off...


I played a bit with fan's curve, setting the intake ones as silent, but CPU and exhaust to use standard settings (they don't make as much noise as two 12mm at full speed...)
Now with CPB disabled and windows max CPU to 100 the temp is 74, I don't know why it's higher than using Windows max CPU power plan as 99%

Maybe I'll try to play undervolting a bit, but this computer will be used for teleworking so I don't want it to be unstable, but maybe I can live without CPB until I get a new HS and/or case that can support CPB.

I've researched Ryzen a bit (the last AMD CPU used was an Athlon 64 several years and finally I will finally settle with this settings for now with the current setup...

Left CPB enabled
Enabled PBO.
PPT 55W ...
Left voltage at stock
All fans on silent profile.
Windows power settings 100% CPU

And now max temperature is 76º and fans aren't even at full CB score is 7950. I believe this may be the best performance I can get out of the CPU without high temps and without killing single core perf with the stock HS and this case, but if anyone has ideas (beyond replacing HS and case) I would be grateful

 

[Zach_01]

@Derek12 keep in mind that Core clock on HWiNFO (or any common software) does not contain all the core c-states but the active one (C0). And this is called the discrete clock.
All C-states together (C0/C1/C6) are stated only by the effective clock. The higher the effective clock the higher the performance and vice versa, no matter what discrete clock says.

Simple explanation:
C0 = active state
C1 = sleep state
C6 = deep sleep state

You can also see this on the "Core X C0/1/6 Residencies" How much time (in %) each core spends on each state.

FEATURE - Effective clock vs instant (discrete) clock

It has become a common practice for several years to report instant (discrete) clock values for CPUs. This method is based on knowledge of the actual bus clock (BCLK) and sampling of core ratios at specific time points. The resulting clock is then a simple result of ratio * BCLK. Such approach...

www.hwinfo.com

I've done in the past tests with different power plans and the 99~100% max CPU state. With 1usmus's "Universal Power plan" (CPB/PBO enabled via BIOS) when I set max state at 99% the CPU boosted frequently just under 3600MHz, but if anything was heavy enough (intensive CPU gaming for example) the boost would hit 4100~4200MHz briefly.

 

[Derek12]

@Derek12 keep in mind that Core clock on HWiNFO (or any common software) does not contain all the core c-states but the active one (C0). And this is called the discrete clock.
All C-states together (C0/C1/C6) are stated only by the effective clock. The higher the effective clock the higher the performance and vice versa, no matter what discrete clock says.

Simple explanation:
C0 = active state
C1 = sleep state
C6 = deep sleep state

You can also see this on the "Core X C0/1/6 Residencies" How much time (in %) each core spends on each state.

FEATURE - Effective clock vs instant (discrete) clock

It has become a common practice for several years to report instant (discrete) clock values for CPUs. This method is based on knowledge of the actual bus clock (BCLK) and sampling of core ratios at specific time points. The resulting clock is then a simple result of ratio * BCLK. Such approach...

www.hwinfo.com

I've done in the past tests with different power plans and the 99~100% max CPU state. With 1usmus's "Universal Power plan" (CPB/PBO enabled via BIOS) when I set max state at 99% the CPU boosted frequently just under 3600MHz, but if anything was heavy enough (intensive CPU gaming for example) the boost would hit 4100~4200MHz briefly.

Thanks for the explanation! Using Ryzen Master it shows the real estate of the cores...
I tried disabling PPT and using the power plan with 99 CPU and this is the CPUz score

However with PPT enabled at 55W and max CPU 100 I get this

It seems that there is a performance penalty with the 99% but on the bright side the temps are lower.

 

[Zach_01]

Of course there is a penalty, since the boost is cut.

Ryzen Master is so complicated... that only AMD knows and understands what exactly is reporting for CPU temp, speed, voltage and states.
Im sticking to HWiNFO. This is how I manage to know that much about the system.

This is me...

Completely stock (CPB/PBO enabled)

 

[Derek12]

Of course there is a penalty, since the boost is cut.

Ryzen Master is so complicated... that only AMD knows and understands what exactly is reporting for CPU temp, speed, voltage and states.
Im sticking to HWiNFO. This is how I manage to know that much about the system.

This is me...

Completely stock (CPB/PBO enabled)

View attachment 186268


View attachment 186269

I agree. I am also sticking with Hwinfo, most reliable, easy and complete info software I've used

 

[HD64G]

55W limit is very low but you lost 2% on singlethreading and ~10% on multithreading performance to the Zach_01 stock, so you are good enough imho. I would test the 65W limit also and if it allows temps to stay below 80% then ti will be your optimum with your case's airflow which seems to be your main limit of your CPU performance.

 

[Zach_01]

I do not know how @Derek12 measures power but here are 2 examples of temp effect.
Cooler: H110i 280mm, Ambient 25°C (air to radiator). Min~Max Fans/Pump rpm (550~2050/2340~2820)
Notice the frequency, voltage and power difference. Its not the best test as the cooler's capacity is high for the CPU under almost any circumstances, but it will give you an idea.

---------------------------------------------------------

1st test CPU-Z stress test. Keep in mind that this is considered a light load.
Temp diff 5°C. Freq diff 35~40MHz (effective)

1a. Low cooling capacity


Max reported power (PPT) 70W, 87.5% PowerReportingDeviation(Accuracy)
True power 70 / 0.875 = 80W

1b High cooling capacity


Max reported power (PPT) 72.2W, 89% PowerReportingDeviation(Accuracy)
True power 72.2 / 0.89 = 81W

-------------------------------------------------------

2nd test Prime95 FFTs 128KB. This is one of the heaviest and worse load the CPU can face. I usually avoid running it under any conditions but I do briefly and rarely for sake of observation.
Temp diff ~7°C. Freq diff 60~65MHz (effective)

2a. Low cooling capacity (wrong voltage circled sorry, its the above)


Max reported power (PPT) 96W, 90.5% PowerReportingDeviation(Accuracy)
True power 96 / 0.905 = 106W

2b High cooling capacity


Max reported power (PPT) 103W, 92.2% PowerReportingDeviation(Accuracy)
True power 103 / 0.922 = 111.5W

-------------------------------------------------------

Cinebench R15/20 are right in the middle of the above 2 tests in terms of CPU stress and power consumption.

 

[Space Lynx]

Thank you all
I am using the preapplied thermal paste (I had a tube somewhere but I misplaced it, if I find it I will repaste and check if it gets lower.
I will use this computer for gaming and working (but the games I am currently playing are not CPU hungry and it runs about 70 but the load is very low).
I will try to pick up an aftermarket HS, maybe that Hyper 212 Evo or some Noctua that is not too big for my case (my case fans are Noctua and they are very fine). But as I am in the eve of upgrading to a M2 SSD maybe I will pick up a cheap one

21 euros https://www.amazon.es/NOX-NXHUMMERH...per+212&qid=1612058308&sr=8-4#customerReviews

or 27 euros

DEEP COOL GAMMAXX GTE V2 Blanco, Disipador de CPU, con 4 Tubos de Calor, un 120 mm CPU Ventilador PWM de LED Blanco, Compatible con Intel/AMD : Amazon.es: Electrónica

Compra online DEEP COOL GAMMAXX GTE V2 Blanco, Disipador de CPU, con 4 Tubos de Calor, un 120 mm CPU Ventilador PWM de LED Blanco, Compatible con Intel/AMD. Envío en 1 día GRATIS con Amazon Prime.

www.amazon.es

those should help a lot over stock cooler.

 

[Mussels]

My dad has one of the gammax coolers, they're quite good for the price

 

[Derek12]

55W limit is very low but you lost 2% on singlethreading and ~10% on multithreading performance to the Zach_01 stock, so you are good enough imho. I would test the 65W limit also and if it allows temps to stay below 80% then ti will be your optimum with your case's airflow which seems to be your main limit of your CPU performance.

I tried 65W and in CB temps gone up to 83º,. Fans are almost at full speed. I got 8309 pts.

CPUZ score, now closer to @Zach_01

I do not know how @Derek12 measures power but here are 2 examples of temp effect.
Cooler: H110i 280mm, Ambient 25°C (air to radiator). Min~Max Fans/Pump rpm (550~2050/2340~2820)
Notice the frequency, voltage and power difference. Its not the best test as the cooler's capacity is high for the CPU under almost any circumstances, but it will give you an idea.

---------------------------------------------------------

1st test CPU-Z stress test. Keep in mind that this is considered a light load.
Temp diff 5°C. Freq diff 35~40MHz (effective)

1a. Low cooling capacity
View attachment 186298

Max reported power (PPT) 70W, 87.5% PowerReportingDeviation(Accuracy)
True power 70 / 0.875 = 80W

1b High cooling capacity
View attachment 186299

Max reported power (PPT) 72.2W, 89% PowerReportingDeviation(Accuracy)
True power 72.2 / 0.89 = 81W

-------------------------------------------------------

2nd test Prime95 FFTs 128KB. This is one of the heaviest and worse load the CPU can face. I usually avoid running it under any conditions but I do briefly and rarely for sake of observation.
Temp diff ~7°C. Freq diff 60~65MHz (effective)

2a. Low cooling capacity (wrong voltage circled sorry, its the above)
View attachment 186300

Max reported power (PPT) 96W, 90.5% PowerReportingDeviation(Accuracy)
True power 96 / 0.905 = 106W

2b High cooling capacity
View attachment 186301

Max reported power (PPT) 103W, 92.2% PowerReportingDeviation(Accuracy)
True power 103 / 0.922 = 111.5W

-------------------------------------------------------

Cinebench R15/20 are right in the middle of the above 2 tests in terms of CPU stress and power consumption.

Thanks for the info! Yes, I also measure power with that item. Running P95 with 128k FFT, you can clearly see the PPT limited to 65W, with this test, the CPU clock has throttled down to 3.5 so Windows say 98% CPU usage (that means as you clearly say, this is the most demanding load).

21 euros https://www.amazon.es/NOX-NXHUMMERH212-Nox-Base-refrigeradora/dp/B01BE7AW28/ref=sr_1_4?__mk_es_ES=ÅMÅŽÕÑ&dchild=1&keywords=hyper+212&qid=1612058308&sr=8-4#customerReviews

or 27 euros

DEEP COOL GAMMAXX GTE V2 Blanco, Disipador de CPU, con 4 Tubos de Calor, un 120 mm CPU Ventilador PWM de LED Blanco, Compatible con Intel/AMD : Amazon.es: Electrónica

Compra online DEEP COOL GAMMAXX GTE V2 Blanco, Disipador de CPU, con 4 Tubos de Calor, un 120 mm CPU Ventilador PWM de LED Blanco, Compatible con Intel/AMD. Envío en 1 día GRATIS con Amazon Prime.

www.amazon.es

those should help a lot over stock cooler.

I see and it's and interesting product, the only problem is that it needs the original brackets for AM4 and I think I lost 'em or even thrown them, after removing them for the stock HS I suppose I need a cooler that has its own bracket or comes with those

 

[HD64G]

So, with the 65W limit you have almost the full performance and much lower temps. All good me thinks.

 

[Derek12]

So, with the 65W limit you have almost the full performance and much lower temps. All good me thinks.

Correct, maybe a new HS is not a priority thing anymore with these settings (but I am going to run a torture test for CPU and GPU at the same time), so I could save money for an M2 drive, or more RAM.

 

[damric]

The wraith stealth is weak, but 90% of the problems are people mounting it wrong. I see people not get every corner tightened down correctly all the time. Verify each corner spring screw clicks. The cooler will function, but very poorly, if not every corner is secure. The stealth is plenty fine for stock normal operation.

 

[Derek12]

The wraith stealth is weak, but 90% of the problems are people mounting it wrong. I see people not get every corner tightened down correctly all the time. Verify each corner spring screw clicks. The cooler will function, but very poorly, if not every corner is secure. The stealth is plenty fine for stock normal operation.

Yes, I remounted several times and made sure all 4 screws are tight until they don't turn and made sure the backplate didn't fell out. What i didn't notice is any click (beyond the spring noises and maybe the MB creaking...

 

[Zach_01]

I tried 65W and in CB temps gone up to 83º,. Fans are almost at full speed. I got 8309 pts.
View attachment 186314

CPUZ score, now closer to @Zach_01
View attachment 186315



Thanks for the info! Yes, I also measure power with that item. Running P95 with 128k FFT, you can clearly see the PPT limited to 65W, with this test, the CPU clock has throttled down to 3.5 so Windows say 98% CPU usage (that means as you clearly say, this is the most demanding load).
View attachment 186316

I see and it's and interesting product, the only problem is that it needs the original brackets for AM4 and I think I lost 'em or even thrown them, after removing them for the stock HS I suppose I need a cooler that has its own bracket or comes with those

The actual power consumption of that run is
65 / 1.062 = 61.2W PPT

Not an issue... just so you know.

 

[cst1992]

There's a site called userbenchmark.com that has a benchmark tool which you can use to compare your results with others' R5 3600s. Technically you should get above 50 percentile, but with your results so close to others' now I don't even doubt it.

About the fans, 80mm are indeed very small.
Airflow is a function of the fan's area, curvature of the blade and speed of the motor. Let's assume the same speed and curvature, so this is the comparison of various fans in airflow relative to your 80mm fan:
80mm: 1x
92mm: 1.26x
120mm: 2.14x
140mm: 2.91x
200mm: 5.94x(!)

Of course, since the 200mm fans are so large, they also move much slower and still move a LOT of air.
You should get a 114% improvement in airflow just by upgrading to 120mm fans.

 

[Mussels]

There's a site called userbenchmark.com that has a benchmark tool which you can use to compare your results with others' R5 3600s. Technically you should get above 50 percentile, but with your results so close to others' now I don't even doubt it.

About the fans, 80mm are indeed very small.
Airflow is a function of the fan's area, curvature of the blade and speed of the motor. Let's assume the same speed and curvature, so this is the comparison of various fans in airflow relative to your 80mm fan:
80mm: 1x
92mm: 1.26x
120mm: 2.14x
140mm: 2.91x
200mm: 5.94x(!)

Of course, since the 200mm fans are so large, they also move much slower and still move a LOT of air.
You should get a 114% improvement in airflow just by upgrading to 120mm fans.

terrible website, and double terrible for AMD ryzen users. google search 'userbenchmark scandal' - the lengths they'll go to, to protect their favoured hardware is nuts.
(ryzen faster in every benchmark, so intel wins! - but then the tweaks to the algorithm to do so, make i3s faster than i9s in their ratings. theres a good reddit thread that pops up in the start of the search)

 

[cst1992]

That's not good.
However, a numbers-to-numbers comparison of the same model of CPU should still be valid, right?
Unless their benchmark software is flawed in itself?

 

[Mussels]

That's not good.
However, a numbers-to-numbers comparison of the same model of CPU should still be valid, right?
Unless their benchmark software is flawed in itself?

i simply cant recommend or trust it, when they're fudging numbers behind the scenes - UB also only does very brief short tests, which dont give time for weak cooling to get over whelmed (or for CPU power limits to come into play - another choice they made that benefits intel)

 

[RJARRRPCGP]

Its as tall as the screw wtf lol..

Looks like the same fan though. Nice and quiet at 3500RPM, I thought it would have been louder

I still like those heatsinks better than the Intel push-pin heatsinks!