5600G runs hot?

[cluelessgenius]

so had to replace my old 7700k and and went pretty mild with a 5600G

 

yesterday i watercooled the rig again and the cpu idles at 40-ish C° 

 

is that normal or do i have to check the cpu block? or maybe theres some special bios settings i gotta looke out for?

 

gpu is at 27° and cpu at 43°? 

 

i got 2 xspc tx 240 ultrathin hooked up , an ekwb x99 block (with amd mount plate obviously)  and everthing on a gigabyte b550i board

[RONOTHAN##]

Ryzen 5000 series has a high idle voltage, so temps do run hot. 40-50C idle is about what you'd expect on air, and with the way PBO works, it just increases the voltage and clock speed to keep it at about that range at idle, with load temps always ending up at ~80C (give or take ~5 degrees)

[cluelessgenius]

1 minute ago, RONOTHAN## said:

Ryzen 5000 series has a high idle voltage, so temps do run hot. 40-50C idle is about what you'd expect on air, and with the way PBO works, it just increases the voltage and clock speed to keep it at about that range at idle, with load temps always ending up at ~80C (give or take ~5 degrees)

really? so do i understand that right? no matter how much cooling i throw at it itll always increase voltage and clock until its 40C idle and 80C load temp? whats PBO?

[Alvin853]

Is it really idle, or is there some small background task going on? My 5600G idles between 30-33°C with an air cooler. Also what Windows power plan are you using? "High Performance" prevents the CPU from powering down properly during idle, I prefer "Balanced". Check Ryzen Master to see what the boost algorithm is doing.

[cluelessgenius]

2 minutes ago, Alvin853 said:

Is it really idle, or is there some small background task going on? My 5600G idles between 30-33°C with an air cooler. Also what Windows power plan are you using? "High Performance" prevents the CPU from powering down properly during idle, I prefer "Balanced". Check Ryzen Master to see what the boost algorithm is doing.

master shows 3 cores literally idling, one is full clock, 2 are fluctuating. thats the image i remember from yesterday. temps were 39C and 28 on the gpu doing nothing but having ryzen master and discord open while watching a youtube video

[cluelessgenius]

11 minutes ago, RONOTHAN## said:

Ryzen 5000 series has a high idle voltage, so temps do run hot. 40-50C idle is about what you'd expect on air, and with the way PBO works, it just increases the voltage and clock speed to keep it at about that range at idle, with load temps always ending up at ~80C (give or take ~5 degrees)

ok so it sounds like PBO is not what i want activated. will check bios later and see what i can do. 

[Alvin853]

5 minutes ago, cluelessgenius said:

one is full clock

That does not sound like it's idling properly. Here is what my 5600G looks like when idle (and I have a very aggressive PBO enabled, so it can still power down when using PBO)

[FliP0x]

PBO is a nice feature and I would not turn it off, just adjust it manually.

 

While in the BIOS, look for Precision Boost Overdrive and tune the settings starting with a negative offset of 10 and increase it by increments of 5 until you notice instability or a performance hit (cores not boosting as high anymore). Then revert to the last stable offset.

 

In general, offsets of 10-20 should be safe to use and won't impact your performance, but will noticably decrease your CPU temps.

[RONOTHAN##]

10 minutes ago, cluelessgenius said:

ok so it sounds like PBO is not what i want activated. will check bios later and see what i can do. 

You want PBO on, yeah temps go up but it also increases clock speed to go along with it. Instead of the chip running at 4.4GHz under full load it will boost up to 4.6GHz or higher. The more you cool it, the higher those clock speeds go. 

[cluelessgenius]

17 minutes ago, Alvin853 said:

That does not sound like it's idling properly. Here is what my 5600G looks like when idle (and I have a very aggressive PBO enabled, so it can still power down when using PBO)

hmmm ill have a closer look and post a pic when i get home today

 

13 minutes ago, FliP0x said:

PBO is a nice feature and I would not turn it off, just adjust it manually.

 

While in the BIOS, look for Precision Boost Overdrive and tune the settings starting with a negative offset of 10 and increase it by increments of 5 until you notice instability or a performance hit (cores not boosting as high anymore). Then revert to the last stable offset.

 

In general, offsets of 10-20 should be safe to use and won't impact your performance, but will noticably decrease your CPU temps.

what offset? dont remember all the settings right now but what value do you offset? is there a temperature target? or do you offset voltage like the oldschool days?

 

9 minutes ago, RONOTHAN## said:

You want PBO on, yeah temps go up but it also increases clock speed to go along with it. Instead of the chip running at 4.4GHz under full load it will boost up to 4.6GHz or higher. The more you cool it, the higher those clock speeds go. 

i dont need it to go higher. the gpu is doing the heavy lifting anyway. i just need it to chill out, feed the gpu with data and shut up about it and not warm my loop up with 80C or so. and in idle it should completly fuck off. i dont want to open browser tabs with 5ghz clock, id rather turn the fans off if i could

[RONOTHAN##]

3 minutes ago, cluelessgenius said:

i dont need it to go higher. the gpu is doing the heavy lifting anyway. i just need it to chill out, feed the gpu with data and shut up about it and not warm my loop up with 80C or so. and in idle it should completly fuck off. i dont want to open browser tabs with 5ghz clock, id rather turn the fans off if i could

Do remember that these chips are also very thermally resistive, so while they may be running at 80C, they're only outputting 50-60W of heat, not really heating up the loop at all. Still, if that's what you're worried about, look into undervolting those chips.

[cluelessgenius]

Just now, RONOTHAN## said:

Do remember that these chips are also very thermally resistive, so while they may be running at 80C, they're only outputting 50-60W of heat, not really heating up the loop at all. Still, if that's what you're worried about, look into undervolting those chips.

you mean it runs at 80C internally but not towards the waterblock? 

[FliP0x]

5 minutes ago, cluelessgenius said:

what offset? dont remember all the settings right now but what value do you offset? is there a temperature target? or do you offset voltage like the oldschool days?

I don't remember the exact name of the setting, as I didn't touch it again after setting it up.

 

Made a quick search and it should be the Curve Optimizer that you set negative offsets to. Everything else should be set to Auto.

 

Depending on the Motherboard Manufacturer, it should look something like this:

 

or

 

[RONOTHAN##]

5 minutes ago, cluelessgenius said:

you mean it runs at 80C internally but not towards the waterblock? 

Yeah, the actual core temps are at ~80, but it's probably closer to 40-50 at the block.  

[cluelessgenius]

18 minutes ago, FliP0x said:

I don't remember the exact name of the setting, as I didn't touch it again after setting it up.

 

Made a quick search and it should be the Curve Optimizer that you set negative offsets to. Everything else should be set to Auto.

 

Depending on the Motherboard Manufacturer, it should look something like this:

 

or

 

alright thanks ill have a look later

 

14 minutes ago, RONOTHAN## said:

Yeah, the actual core temps are at ~80, but it's probably closer to 40-50 at the block.  

well then show me that. why would they show me the scary number if they dont really seem to matter anyway? idk but this sound stupid to me would we not want them to be the same temperature. that means it keeps the heat inside like a brick oven instead of dissipating it? 

[RONOTHAN##]

Just now, cluelessgenius said:

why would they show me the scary number if they dont really seem to matter anyway?

Because it is the number that matters. If the internal temps are too high, that's what causes the chips to melt themselves to death. The temp at the heat spreader really doesn't make a difference to the health of the chip, and if you're monitoring temperatures that's mostly what you care about.

 

3 minutes ago, cluelessgenius said:

idk but this sound stupid to me would we not want them to be the same temperature

Yes that is what you want to happen in an ideal world, but this is not an ideal world. Every chip out there has some thermal resistance to it, some are high, some are low, and the temps on the heat spreader will always be lower than the temps at the center of the die. There are a number of reasons why these Ryzen chips in particular are so resistive to heat, from the fact that they're so dense to the fact that they are stacking quite a bit of silicon IIRC. It's a phenomena that every modern CPU has to deal with due in no small part to how small the nodes are now. Ryzen CPUs are just particularly vulnerable to it because of some choice made in the design stage (I'm not a hardware engineer who worked on it so don't know the exact reason, they're just a bit worse than Intel currently for thermal resistance), plus PBO basically has the chips run at redline out of the box. 

 

Gone are the days of X58/X79 where you could have a CPU pulling back 300-400W and it still run at 70C under water. It's part of the reason overclocking is just kinda dead at ambient, because in order to cool stuff down enough for them to start scaling with voltage, you need to be quite a bit below 0C. 

[cluelessgenius]

17 minutes ago, RONOTHAN## said:

Because it is the number that matters. If the internal temps are too high, that's what causes the chips to melt themselves to death. The temp at the heat spreader really doesn't make a difference to the health of the chip, and if you're monitoring temperatures that's mostly what you care about.

sure but kinda stupid to show me internal 80C when i cant seem to do much about it if the temp doesnt transfer to the block. like, cool its 80C. what do they want me to do about it if i cant really seem to affect it at all.

 

19 minutes ago, RONOTHAN## said:

PBO basically has the chips run at redline out of the box.  

thats the weirdest thing to me. im cool with them offering oc mechanics like this but enabling them out of the box? when i buy a chip and board i expect them to run at whats on the box and them if i want to take the clocks higher i happy to find features like this but if i dont then i dont want to see high temps if the cooling is proper. i installed this thing and immediatly started questioning wether i screwed the cpu block down enough or put too much paste on or whatever even as far as questioning the quality of my now rather old cpu block design (x99 block). especially because of the difference between cpu and gpu. if the gpu is chilling at sub 30s and the cpu is at 50C doing nothing really you start thinking of diagnosing what the issue is. 

[Alvin853]

18 minutes ago, cluelessgenius said:

thats the weirdest thing to me. im cool with them offering oc mechanics like this but enabling them out of the box? when i buy a chip and board i expect them to run at whats on the box and them if i want to take the clocks higher i happy to find features like this but if i dont then i dont want to see high temps if the cooling is proper. i installed this thing and immediatly started questioning wether i screwed the cpu block down enough or put too much paste on or whatever even as far as questioning the quality of my now rather old cpu block design (x99 block). especially because of the difference between cpu and gpu. if the gpu is chilling at sub 30s and the cpu is at 50C doing nothing really you start thinking of diagnosing what the issue is. 

It's not enabled out of the box, AMD has 2 different technologies, that sound almost the same, and many people get confused: There is Precision Boost and Precision Boost Overdrive. The first one is the stock Ryzen boost behavior and is enabled out of the box, it can change the clocks up to 1000 times per second depending on load, temp, power limits and many other factors. Precision Boost Overdrive is enabled manually, and it's a way of overclocking while still using the really good boost algorithm Ryzen has. Basically PBO allows you to reconfigure the boost behavior, alter limits or remove them entirely.

 

For example, for every 65W TDP Ryzen CPU the stock PB limits are:
PPT: 88W

TDC: 60A

EDC: 90A

And in case of the 5600G, Clock limit: 4450 MHz (at least on my unit, I know it's listed as 4.4GHz max boost, not sure if there is a small variation between units or if they just leave out the extra 50 MHz)

 

As you can see in my screenshot above, I raised those limits to:
PPT: 100W

TDC: 90A

EDC: 140A

Clock limit: 4650 MHz (Boost Override +200)

 

Also I'm using the curve optimizer to undervolt each core, the PB algorithm does not only determine what clock to run, but also what voltage, and that voltage can be modified per core, lower voltage -> lower power consumption, but may lead to instability, so you really need to be careful or test each core individually. That's what people have been talking about, when they say "offset". All of these values will depend on your CPU, you can use my values as a starting point if you want to fiddle with it, but I'd recommend starting with -10 on all cores on the curve optimizer. My worst core can only do -10, so while all other cores can do bigger offsets, if I set it to -15 all core, my Windows will probably boot, but when a heavy load hits the worst core, I might end up with a bluescreen.

PBO can also be used to decrease the limits. If you don't want your CPU to heat up your loop too much, you could limit the PPT to 50W for example, then the boost algorithm will pick lower frequencies to stick to the power limit. PBO makes it really easy to fine tune your CPU, only thing it can't do is extreme overclocking, because the Boost Override is limited to +200 at max. 

[cluelessgenius]

8 minutes ago, Alvin853 said:

...

how do you stress test individual cores?

[Alvin853]

Just now, cluelessgenius said:

how do you stress test individual cores?

I'm using OCCT (free for personal use, not my software), left menu "Test" -> select CPU in bottom left tab (top left and entire right side are for monitoring), I do "Large", "Extreme" and "Variable" for the settings, then Threads set to Advanced and configured as in this popup, this will start a single thread stresstest on core 0 and cycle the core every 5 seconds.

[cluelessgenius]

1 hour ago, Alvin853 said:

I'm using OCCT (free for personal use, not my software), left menu "Test" -> select CPU in bottom left tab (top left and entire right side are for monitoring), I do "Large", "Extreme" and "Variable" for the settings, then Threads set to Advanced and configured as in this popup, this will start a single thread stresstest on core 0 and cycle the core every 5 seconds.

interesting. why would you cycle them?

[Alvin853]

4 minutes ago, cluelessgenius said:

interesting. why would you cycle them?

In my experience, the errors/crashes are most likely to happen, when a core ramps up or down, not when it's under constant load. So I'm cycling them to make each core ramp up and down multiple times per minute.

 

Edit: also this way, I can run the test once and test all cores and I don't have to redo the test for each core individually. I'm not testing for thermal stability, single core loads should never overheat the CPU, I'm only testing for voltage stability, does the voltage respond fast enough, when the clock speed changes. For thermals, you'd want to run a constant load.