Processors

[expresso]

Hi -  its been a long time since i built a AMD system -  glad to say i am back and just built a 5700G setup -  my question is about the temps  -  

i am using an phanteck mini itx case evolve -  200mm front fan -  140mm top fan and 140mm rear fan -   be quiet dark rock slim cooler

my room is warmer than most -   but my temps are around 50C not doing much -  and hit 84C under full load all cores with handbrake and other video converters - 

its all stock -  i didnt touch nothing in the bios other than memory XMP -   is this normal - avg. temps for this chip -  ?

coming from an intel 3770K  OC 4.5  all 4 cores -  it never hit more than 70C under full load -    i am just wondering if this is normal for AMD -  or  could it be the termal paste -  maybe i added a bit too much -   its been a while since i built one - maybe 8 years ago for the 3770K -

i like some input from other AMD users -  what can you say about my temps ?

thanks

[ryzen_type_r · Accepted solution]

I haven't used a 5700G, but Ryzens in general run fast and hot.  They will boost up higher and higher until it bumps into the power, thermal, or frequency limits.  Your 5700G has a PPT of 88W, a max boost clock of 4.6 (I think?), and a thermal limit of probably 90 or 95C.  Plus it has onboard GPU, so yeah, it's probably not strange to see mid 80's under heavy load especially if in a warm room.

My previous build was an Intel Haswell which wasnt known for cool running so believe me, I know how you feel.  First time I ran Prime95 FMA3 on my 5800X and saw temps hit 90C and stay there I thought there was something wrong.  But, after I did some research and realized that this was normal (AMD even said it was by design).  These CPU's monitor themselves and will try to run as fast as possible within the given limits.

But, if you don't like seeing such high temps, theres a number of ways you can reduce them.  You can disable the boosting completely (in BIOS by turning off CPB, or in the Windows power plans) which lowers power consumption and temps by a crapload.  Or you can lower the Platform Thermal Limit setting in BIOS to whatever you want and the CPU won't exceed that temp, pretty neat.  Or you can enable PBO and lower the PPT limits to lower the power consumption limit, which would also lower temps too.  Just be aware, you'll lose performance if you do any of this.

P.S. This is the new normal for CPU's these days.  Latest Intels basically do the same thing.

[ryzen_type_r · Accepted solution]

I haven't used a 5700G, but Ryzens in general run fast and hot.  They will boost up higher and higher until it bumps into the power, thermal, or frequency limits.  Your 5700G has a PPT of 88W, a max boost clock of 4.6 (I think?), and a thermal limit of probably 90 or 95C.  Plus it has onboard GPU, so yeah, it's probably not strange to see mid 80's under heavy load especially if in a warm room.

My previous build was an Intel Haswell which wasnt known for cool running so believe me, I know how you feel.  First time I ran Prime95 FMA3 on my 5800X and saw temps hit 90C and stay there I thought there was something wrong.  But, after I did some research and realized that this was normal (AMD even said it was by design).  These CPU's monitor themselves and will try to run as fast as possible within the given limits.

But, if you don't like seeing such high temps, theres a number of ways you can reduce them.  You can disable the boosting completely (in BIOS by turning off CPB, or in the Windows power plans) which lowers power consumption and temps by a crapload.  Or you can lower the Platform Thermal Limit setting in BIOS to whatever you want and the CPU won't exceed that temp, pretty neat.  Or you can enable PBO and lower the PPT limits to lower the power consumption limit, which would also lower temps too.  Just be aware, you'll lose performance if you do any of this.

P.S. This is the new normal for CPU's these days.  Latest Intels basically do the same thing.

[expresso]

thanks for that info.   i guess i leave it alone and not worry about it -    its been a long time since i built a system and first time i am back with AMD - i love it and will stick with AMD -  just wasnt sure about the temps much -  i didnt even try any OC -

need alot to learn and follow some OC guide if i can find one on this chip with a B550I MB Gigabyte -

i may try replacing the termal paste at some point just to be sure i didnt add too much - 

[Gwillakers]

I pretty much agree with what Ryzen-Type-r just posted.

It's your decision, but I would not let it alone.  

An old engineering adage stated, that for every 10C you can drop the Temp on a Chip you double its life.

I believe you should take a straight forward approach and lower the Thermal-Throttle-Limit for your processor.

Then test its performance.     I don't think you will lose much performance.    It's those last few MHz that cost a lot in voltage and temps.

Set PBO to Advanced.

Set Limits to Manual

Set Thermal-Throttle-Limit to Manual

Set the new field "Thermal Throttle Limit"  to a temperature (in Celsius) that you feel comfortable with.   (Me I like 75)

[expresso]

thanks - my understanding is - its fine working within its limits of 95C -   i am just not used to seeing those temps

my previous system was an 3770K 4 core  4.5 OC -  i maxed it at about 70C  full load - now my idle temps are 48C more or less

i do have a warm room - thats for sure -  since i havnt done this in 10 years - i need to go slow and easy adjusting in the bios

i love to find a nice step by step to OC this or adjust in simple terms   -  to be safe - last thing i want to do is clear the cmos to boot

MB dosnt have a clear cmos in the rear - too bad for that -  so i want to follow and tested  OC or adjustment instructions

are they any around to follow the steps  ?

[ryzen_type_r]

Yeah I noticed these Ryzens do consume a bit more power at idle than my old Haswell, I believe it's actually the I/O (SOC) side that's responsible for most of it as the CPU cores are mostly asleep or running at very low speeds at idle.  Unless you've got a lot of background processes in Windows running.

How warm is your room?  My 5800X looks to be running about 12-15C higher than ambient temp at 'idle'.  But, it doesn't have built-in GPU like yours, so not sure how much that will affect idle temps.  Maybe not a whole lot.

My mobo is an Asus, so the settings locations might be different to yours, but modern UEFI BIOS's should have a search function.  Just search for 'Platform Thermal Limit' or something like that and you can just key in the temp limit you want directly.  Might even be in more than one location, it is in 2 on my Asus but I only have to enter it into just one location and the CPU will respect the new temp limit.

If you only have a clear CMOS jumper, you can temporarily connect your reset switch to it and then that becomes your clear CMOS button.

[expresso]

would i be able to leave a cable to the cmos and just use it if needed -  i did that before long ago -   if i plug in a reset cable and leave it there - would it be fine -  just use it if needed  ?

thanks

[elstaci]

I just upgraded to a Ryzen 7 3700X and has the same Maximum Operating Temperature as your 5700G at 95C.

I am using a Hyper212 EVO CPU Cooler with Push-Pull Fans on it. One is a very high RPM fan (5000 RPM) and the other is a 2000 RPM fan and my processor temperatures are between 45C and 60C for normal computer use like browsing and watching video and Multitasking. 

So a very high quality TDP Rated CPU Cooler will help keep your Ryzen running at a cooler average. The 5700G TDP ratings is 65 Watts.

The Temperatures are constantly changing between 45c & 60C unless I am stressing the processor using a program like OCCT which causes the temperature max out at 72C and stays there the entire time during the Stress test.

But I notice that when my Room temperature (Ambient Temperature) is above 84 - 85F the processor temperature increases about 3-7C higher but as soon as the room temperature is around 80F it goes back to the regular temps.

So the point I am making is that Ambient Temperature and good  air flow inside your computer case does have a direct effect of how hot your processor will run.

I have my bios in the same state as yours where everything is mainly default settings except RAM and fTPM settings. PBO and all the rest of the BIOS settings that affect the processor I have it set on default.

I am certain if I disabled PBO my temperatures averages would be much cooler.

[ryzen_type_r · Accepted solution]

---

@Gwillakers wrote:

Then test its performance.     I don't think you will lose much performance.    It's those last few MHz that cost a lot in voltage and temps.

 Yep, for sure diminishing returns.  I remember I did a quick Cinebench test with CPU running normally vs. disabling boost and the multithread score dropped 20%, but the CPU package power consumption dropped by 50%.  And the peak CPU temp was 30C lower!

Nice thing is, you can disable boost easily in the Windows Power control panel and you can create a separate power plan with the boost disabled and switch power plans using a batch file which you can assign to a hotkey combination.  It would be the modern version of a Turbo button on an 80's PC LOL.

Now, it would be great if Ryzen Master could control Platform Thermal Limits on the fly without rebooting, and imagine being able to switch RM profiles using the command line....you could have an energy sipping cool running 45W computer for normal everyday desktop tasks one minute, but instantly unleash the beast when you need it with a keypress.  Or have it switch automatically when running certain applications.

[Gwillakers]

I just believe in the direct approach.   Ryzen is great at managing it's temp.

Set thermal Throttle limit to what you want.

No need to figure out how much to lower Power, or Current,  or Multithreading, or frequencies...

Ryzen does it.   AND IT DOES IT WELL.   WITHOUT LOSING MUCH PERFORMANCE.

Thats what I think most people don't understand.  How little it cost to lower temp from 90 to 75.

What I believe happens, is that when just 1 core hits the Throttle point, the cpu might just look to dispatch the work on a cooler core.   Thus none of the cores have to endure the torture of 90C.    But I could be wrong here.

On a lighter topic,  why can't the "Im not a robot" ask you more interesting questions ... like check all the blondes....

[DRDOS]

Did you ever try to compare the values returned by both - CPU and integrated GPU thermal sensors, if yes, then what was the result?
I did and found it very odd. An integrated GPU thermal sensor returns quite realistic values that very well suit to what i measure on the CPU lid by the external thermal sensor. Unlike integrated GPU thermal sensor, CPU thermal sensor returns such a values that bring some questions...
1. If CPU and GPU is located on a single die, then how come the temperature difference in a different parts on the die is so much (up to 25-30*C)? Even if CPU and GPU are not in a single die, but in a 2 standalone ones, even then the temperature difference between them can't be that much.
2. Why the temperature measured on the CPU lid is so much lower than the value returned by the CPU thermal sensor (up to 30*C)? E.g., when there's only 55*C on the lid, CPU thermal sensor returns 85*C.
3. There're obvious mismatches between power consumption measured by an external measurement device, a power dissipation constant of the cooler and a temperature value returned by CPU thermal sensor.

Trying to answer these questions, i've found only 2 explanations:
1. The value returned by the CPU thermal sensor is inaccurate, contains significant static error. This explanation though does not explain some questions, e.g., why the values returned by both CPU and GPU thermal sensors are almost equal in case if CPU is idle and GPU is loaded by a stress test.
2. For some reason a thermal conduction interface between the CPU die and the CPU lid has too high thermal resistance. But, if it will appear CPU and GPU are in a single die (i don't have exact information at the moment), then this explanation also doesn't answer these questions.

At the moment i still don't have clear understanding why does this happen. Maybe will figure it out later.
Official answer from AMD says that it is normal. But inspecting this APU myself in my service lab, i don't see it is normal, there should be an exact explanation that explains these questions.

Delidding it could probably bring some light here, but i didn't want to do it, at least yet.


___

One else supposition i'm currently thinking of is that there probably could be a small areas with relatively high power dissipation in the die, so that thermal energy can not be conducted fast enough out of there, as a result a local temperature in those areas appears to be significantly increased, and probably intentionally a thermal sensor(s) is located exactly in those ares. If this is the case, then it is question only this CPU developers can answer, and then i'd agree it is normal, or a "new normal" as somebody said here above.

This last supposition would bring another question... An architecture design that implies quite a high temperature differentiation between different areas in the die may probably have negative impact upon a lifespan due to thermal deformation effects, so that these modern processors probably might not be so 'immortal' as those manufactured 20-25 years ago and work flawlessly so far. This is a supposition though, the time will give an answer.

[DRDOS]

Got a little bit of free time to collect some info today...
So, this processor contains 3 dies (CPU, GPU and I/O), and one part of an issue probably is that CPU die (that amongst the 3 dies has the highest power dissipation - about 65W) is displaced from the center towards the corner (as on the attached picture - in the left top corner), which means most of the coolers designed to take the heat effectively from the center of the processor lid only, in this case will work less effectively as the most hot area is located nearby the corner, not in the centeral part. A processor lid of course isn't thick enough to deliver such amount of the heat from the small spot in its corner to its central part (where it has thermal contact with the cooler) without significant thermal resistance.



Therefore, in order to take the heat out from the CPU die effectively a cooler should have quite a wide area of the thermal contact with the processor lid, in the best case it must have a thermal contact with entire surface of the processor lid.

However, there are a lot of cooler models that have 3 or even 2 thermal pipes. Even those which have 4 or more thermal pipes won't be working effectively enough in this case unless they have a good thermal contact between all of the pipes (a proper cooler design either should have a solid copper thermal conductive part that connects the pipes, or the pipes must be located very close to each other).

Examples of the cooler design that should be working the most effectively in this case:

      

Examples of the coolers that should be avoided in this case:

   

In my case i used an old IceHammer cooler, which of course doesn't take the heat effectively out of the corners of the processor lid:



Later when i'll have a chance to test it with proper cooler, will post the results here.
Also, will try to measure a temperature in different parts of the lid to detect a temperature differentiation/gradient when using a cooler with narrow thermal contact such as IceHammer one on the photo above (previously measured it only in a single spot on the lid).

[ryzen_type_r]

That chip photo is of the non-G Ryzen 5000 series processors which use separate SoC and CCX chiplets.  Ryzen 5000's with integrated graphics use a different, monolithic, die:

https://www.techpowerup.com/285534/amd-ryzen-5-5600g-apu-die-shots-published

Still, it's always a good idea to get the best cooling you can afford/stuff into your case.

[DRDOS]

Very interesting... Thanks for this info.

So, this desktop Ryzen 7 5700G processor (unlike desktop X-versions without integrated GPU) is a single monolithic crystal die same as its mobile version.
This means a cooler that has larger thermal contact area with the processor lid won't fix an issue. This also ruins my whole supposition stated above, as according to that supposition a mono-crystal processor must be free of such an issue, but both - G (mono-crystal die placed in the central part) and X (separate CPU die, in 5600X/5800X models, displaced towards the corner) versions have the same issue with the CPU temperature.

Sure, with more efficient cooler a temperature should be lower of course.
However it's still a question whether that CPU temperature value reported in BIOS and in software tools is correct (now i doubt it is).

[raidersofall1]

That's not normal at all. My 5700g idles at ~25 to ~33 degrees, and under a full sustained load, it never goes over 65C, and that's will a gpu overclock that brings the wattage to peak at 95 watts. I'm only using a single 140mm AIO in the nzxt h1 case. 

[mikenz]

Hi I see this was awhile ago but I have a slight tip as I have a couple of ryzen cpus including this I run project hydra and most cpus I have get about a 10 degrees drop in temp I have xmp and pbo on and running now at 56c was at 68c while mining 

[expresso]

Being bored lately with the colder temps outside -  i started to tinker a bit again in the bios -  just for fun -  updated to latest bios -   XMP is on -  and boosted to 4000 - with just a bit of a bump in volts -  cant complain there -   now with the CPU and temps   -   my room  is always warmer than most i know  - my case is small Mini Itx -  have one large 200mm fan in front - thats it -  with a Artic cooler 240 Rad -   i have managed to get high clocks and much lower temps -   adjusting just the PBO max override - using MB limits with x4 scaler   -  i actually boost  up to 4.8 just a sec. but still not too bad - and full load all cores sits around 4.5 - 4.575 -  sometimes i hit a 4.6 - all cores -  and temps dont go past 65C   -  normal use of PC  44C   -  the volts were set manual -  that did it -  

[expresso]

 i wanted to play with the FSB  - its set at 100 FSB - Auto -  but no matter what i change it to in the bios - it always sits at 100 on CPUZ -   when i go check in the bios - its 100 -  the changes i make there dont stick -   maybe have to turn off PBO to raise the FSB on this chip ?   was just curious -  to know why i cant adjust it -  was going to try to just change the FSB to see what it gives me in clocks and temps -      Core Temp - is giving iffy readings on the FSB  sometimes it  shows 104 - sometimes 88  -   sometimes 100 - while CPU Z and HW monitor show 100 FSB even when core temp shows higher -   Core temp cant be trusted 100% - but the temps are correct they match all the others -