3900x super hot , gigabyte z570 master , vcore auto voltage.

[peter224722]

Hi , I've always had problem with my 3900x on my gigabyte z570 master being to hot.
With the default cooler anything other than windows 10 power saver mode would put the fan on full 747 mode.
Just got a 360mm AIO cooler and found that in AMD Balanced etc , the temperature would hit 74 on idle and the water temp would go up vertical.
Reading that dropping the vcore voltage may help I tried knocking off auto on vcore and used 1v from the 1.2v it send it was set to by auto.
Super cool , but only about 3.3ghz , so tried 1.1v and 3.5ghz and 45c idle so better.
Tried what the broad said was auto ( 1.2v ) and it's now 3.8ghz dropping down a little on 100% cpu after a long bender test.

Now on manual 1.2v and a setting the pump profile higher I'm getting a max temp of about 68c after 2 hours blender 100% work , and the task manager say at 3.68ghz .

WHAT I DO , up the voltage a little more , 1.25 ? , what is a best temp for the 3900x I should aim for , cooler the better or a higher temp , but higher clock speed ?

 

[gupsterg]

I've used 3x R9 3900X on ASUS ROG Crosshair VII Hero WiFi, I can't say I've seen a hot one as such. Cooling is EK Supremacy Evo AM4 block, Magicool G2 Slim 360mm rad, EK DDC pump, so I'd say very similar to your 360mm AIO.

I used Ryzen Balanced profile, set min CPU state as 0%, do a registry tweak to show core parking in power plan, set that to 10% so 2 logical CPU always remain unparked at idle. All idle ~30C (+/- ~2C).

I use mine at stock, but OC'd RAM, cooler CPU will boost higher and sustain a higher boost. For example usual CB20 run on multicore is ~7200 points, first of the day can be ~7350 points. If I load CPU with say Kahru RAM Test, I can see ~4.24GHz all cores sustain boost for a little while, as the average temperature rises it can be flitting between ~4.18-4.2GHz all cores boost. Realbench stress mode can initially be ~4.2GHz all cores boost, then as temp rise it can be ~4.1GHz with bounce downs and a 60 minute run will average ~4.05GHz across all cores.

 

[LtMatt]

Hi , I've always had problem with my 3900x on my gigabyte z570 master being to hot.
With the default cooler anything other than windows 10 power saver mode would put the fan on full 747 mode.
Just got a 360mm AIO cooler and found that in AMD Balanced etc , the temperature would hit 74 on idle and the water temp would go up vertical.
Reading that dropping the vcore voltage may help I tried knocking off auto on vcore and used 1v from the 1.2v it send it was set to by auto.
Super cool , but only about 3.3ghz , so tried 1.1v and 3.5ghz and 45c idle so better.
Tried what the broad said was auto ( 1.2v ) and it's now 3.8ghz dropping down a little on 100% cpu after a long bender test.

Now on manual 1.2v and a setting the pump profile higher I'm getting a max temp of about 68c after 2 hours blender 100% work , and the task manager say at 3.68ghz .

WHAT I DO , up the voltage a little more , 1.25 ? , what is a best temp for the 3900x I should aim for , cooler the better or a higher temp , but higher clock speed ?

What tool are you using to measure temperature?

Can you show me a screenshot from Ryzen Master Tool?
https://www.amd.com/en/technologies/ryzen-master

Here's mine with YouTube playing and a few Chrome tabs open, using the stock cooler.

 

[gupsterg]

OP be aware Ryzen Master likes to show averages. In HWINFO CPU Die (average) will be same as Ryzen Master. Only reason you'll see a difference between those two, for that sensor, will be polling interval, by default HWINFO is 2000ms and RM 1 second.

In this ZIP you'll find a WMV where I'm testing 3800MHz 1:1:1 for ~12hrs on a R9 3900X, HWINFO was the main monitoring tool used for test run, you'll see RM opened just to capture some aspects and other apps.

 

[peter224722]

What tool are you using to measure temperature?

Can you show me a screenshot from Ryzen Master Tool?
https://www.amd.com/en/technologies/ryzen-master

Here's mine with YouTube playing and a few Chrome tabs open, using the stock cooler.

it looks like I can't send a image direct , but need to host the images somewhere else.
When I set the vcore to 1.2v , I'm getting 4185mhz all cores 63c ( 35c idle ), and auto says 4038mhz all cores at 81.09c
But task manager say differant !!! .

 

[BlackDragon]

it looks like I can't send a image direct , but need to host the images somewhere else.
When I set the vcore to 1.2v , I'm getting 4185mhz all cores 63c ( 35c idle ), and auto says 4038mhz all cores at 81.09c
But task manager say differant !!! .

Host it on imgur.

 

[peter224722]

OP be aware Ryzen Master likes to show averages. In HWINFO CPU Die (average) will be same as Ryzen Master. Only reason you'll see a difference between those two, for that sensor, will be polling interval, by default HWINFO is 2000ms and RM 1 second.

In this ZIP you'll find a WMV where I'm testing 3800MHz 1:1:1 for ~12hrs on a R9 3900X, HWINFO was the main monitoring tool used for test run, you'll see RM opened just to capture some aspects and other apps.

I think i'm just chasing the rabbit , windows 10 task manager is wrong , if Hwinfo and rayzen master are correct.
So it look like I'm getting 4.1ghz all cores and under 70c on amd balanced profile , which is different again to the auto setting .

 

[gupsterg]

I think i'm just chasing the rabbit , windows 10 task manager is wrong , if Hwinfo and rayzen master are correct.
So it look like I'm getting 4.1ghz all cores and under 70c on amd balanced profile , which is different again to the auto setting .

I don't know how Task Manager assesses clocks, I don't really use it for that purpose. HWINFO shows two type, see this for more info.

Room ambient ~18C, in this ZIP is Goosebery half way mark sceenie and WMV near end. The screen video capture SW does place load on CPU, it's Faststone Capture I'm using. The rig was under load prior to Gooseberry run, HWINFO was zero'd and the blender run, rig had been ~18hrs uptime under load prior.

AMD have really improved on Ryzen Master from what it was for 1000/2000 series CPUs, but it is totally devoiding user from awareness of peak/max info as it averages temps, frequency, etc.

No real idea why your temps are as high as they are. I can only share I have not experienced what you have across 3x R9 3900X, all used on same mobo/RAM, etc as currently I do.

 

[LtMatt]

but it is totally devoiding user from awareness of peak/max info as it averages temps, frequency, etc.

The temperature sensor used in Ryzen Master Tool matches the CCD1 temperature in HWINFO64. The peak temperature for the sensor used in Ryzen Master is actually higher than the peak temperature for the Tdie reading on my system, as seen below.

It is perfectly acceptable to use the Ryzen Master Temperature in my opinion and that's what i have displayed and monitored on my overlays (CCD1/2).

 

[gupsterg]

AMD have really improved on Ryzen Master from what it was for 1000/2000 series CPUs, but it is totally devoiding user from awareness of peak/max info as it averages temps, frequency, etc.

The temperature sensor used in Ryzen Master Tool matches the CCD1 temperature in HWINFO64. The peak temperature for the sensor used in Ryzen Master is actually higher than the peak temperature for the Tdie reading on my system, as seen below.

It is perfectly acceptable to use the Ryzen Master Temperature in my opinion and that's what i have displayed and monitored on my overlays (CCD1/2).

Polling interval can play a factor, but from testing my 5x Zen2 and keeping close contact with author of HWINFO CPU Die (average) is meant to match Ryzen Master.

 

[LtMatt]

Polling interval can play a factor, but from testing my 5x Zen2 and keeping close contact with author of HWINFO CPU Die (average) is meant to match Ryzen Master.

I have my polling intervals at 1 second on both applications.

Added the average values in, but only CCD1 shows same temperature as within Ryzen Master Tool to within an accuracy of 0.5-1c.

 

[gupsterg]

I have my polling intervals at 1 second on both applications.

Added the average values in, but only CCD1 shows same temperature as within Ryzen Master Tool to within an accuracy of 0.5-1c.

You either have CPU Die (average) hidden or using an older HWINFO, you should have 4 sensors.

CPU (Tctl/Tdie)
CPU Die (average)
CPU CCD1 (Tdie)
CPU CCD2 (Tdie)

I linked a ZIP in post 7, download it and you will see the 4 sensors in JPG and also WMV.

 

[LtMatt]

You either have CPU Die (average) hidden or using an older HWINFO, you should have 4 sensors.

CPU (Tctl/Tdie)
CPU Die (average)
CPU CCD1 (Tdie)
CPU CCD2 (Tdie)

I linked a ZIP in post 7, download it and you will see the 4 sensors in JPG and also WMV.

No temperature sensors are hidden.

I am using version v6.12-3930. It probably isn't the latest version, but it was one of the HWINFO64 updates that became available once we released the SDK to third party developers so they could integrate Ryzen Master Temperature into their apps.

 

[gupsterg]

No temperature sensors are hidden.

I am using version v6.12-3930. It probably isn't the latest version, but it was one of the HWINFO64 updates that became available once we released the SDK to third party developers so they could integrate Ryzen Master Temperature into their apps.

Shall I link few posts from Martin stating how pants AMD were on providing info? AMD announced SDK and took ages to even provide them the info. It took several attempts for them to finally get anything out of AMD worthwhile.

In post 3 I clearly stated "In HWINFO CPU Die (average) will be same as Ryzen Master.".

v6.12-3930 is from months ago, some where around end of Sep/early Oct.

Spoiler

Even Ryzen Master has changed between then and recent versions on how it reports aspects.

All I can say is kudos to Martin and others efforts. HWINFO now shows FCLK/UCLK and PPT/TDC/EDC info and has been going strength to strength on monitoring of Matisse.

Another aspect I have noted between the 3x R9 3900X I used with same HW, some can have a larger CCD to CCD difference, post on OCN.

 

[LtMatt]

Shall I link few posts from Martin stating how pants AMD were on providing info? it took several attempts for them to finally get anything out of AMD worthwhile.

In post 3 I clearly stated "In HWINFO CPU Die (average) will be same as Ryzen Master.".

v6.12-3930 is from months ago, some where around end of Sep/early Oct.

Spoiler

Even Ryzen Master has changed between then and recent versions on how it reports aspects.

I think you missed the point of my original message, which was that it is acceptable to use and monitor the Ryzen Master Tool temperature as the primary temperature sensor for the CPU, regardless of what it is or isn't called in a third party app.

Nonetheless i have now updated HWINFO64 to the latest but nothing has really changed, though the temperature sensor you mentioned now appears but shows a slightly higher reading than Ryzen Master Tool, but there is not much in it.

Not that much changes between Ryzen Master Tool versions with regards to temperature reporting, you'd be surprised.

 

[peter224722]

Shall I link few posts from Martin stating how pants AMD were on providing info? AMD announced SDK and took ages to even provide them the info. It took several attempts for them to finally get anything out of AMD worthwhile.

In post 3 I clearly stated "In HWINFO CPU Die (average) will be same as Ryzen Master.".

v6.12-3930 is from months ago, some where around end of Sep/early Oct.

Spoiler

Even Ryzen Master has changed between then and recent versions on how it reports aspects.

All I can say is kudos to Martin and others efforts. HWINFO now shows FCLK/UCLK and PPT/TDC/EDC info and has been going strength to strength on monitoring of Matisse.

Another aspect I have noted between the 3x R9 3900X I used with same HW, some can have a larger CCD to CCD difference, post on OCN.

What screen capture are you using ? , looking at your result , mine is similar and very different.

 

[gupsterg]

I think you missed the point of my original message, which was that it is acceptable to use and monitor the Ryzen Master Tool temperature as the primary temperature sensor for the CPU, regardless of what it is or isn't called in a third party app.

Nonetheless i have now updated HWINFO64 to the latest but nothing has really changed, though the temperature sensor you mentioned now appears but shows a slightly higher reading than Ryzen Master Tool, but there is not much in it.

Not that much changes between Ryzen Master Tool versions with regards to temperature reporting, you'd be surprised.

I never said Ryzen Master was not acceptable . I just wanted a user to be aware that it shows averages .

You picked my line of "but it is totally devoiding user from awareness of peak/max info as it averages temps, frequency, etc.", which it is and is shown by your latest screenie.

CPU Die (average) sensor "Current" column is matching Ryzen Master, HWINFO 43C RM 42.8C in your new screenie.

CPU (Tctl/Tdie) sensor has no averaging on it, this is as Tctl is supposed to be, as per AMD developers guide information. A rotational value of the multiple sensors on die, showing highest value at an instance.

As said before in this post and my intention in original post highlighting how Ryzen Master was, it was meant to inform a user, not to cause a barrage of posting on the matter .

What screen capture are you using ? , looking at your result , mine is similar and very different.

Faststone Screen Capture.

 

[LtMatt]

I never said Ryzen Master was not acceptable . I just wanted a user to be aware that it shows averages .

You picked my line of "but it is totally devoiding user from awareness of peak/max info as it averages temps, frequency, etc.", which it is and is shown by your latest screenie.

CPU Die (average) sensor "Current" column is matching Ryzen Master, HWINFO 43C RM 42.8C in your new screenie.

CPU (Tctl/Tdie) sensor has no averaging on it, this is as Tctl is supposed to be, as per AMD developers guide information. A rotational value of the multiple sensors on die, showing highest value at an instance.

As said before in this post and my intention in original post highlighting how Ryzen Master was, it was meant to inform a user, not to cause a barrage of posting on the matter .

Faststone Screen Capture.

No worries.

I could see you going off on a tangent with wanting to start linking to posts, so i just wanted to reiterate what my original post was saying - that it's okay to use the temperature listed in Ryzen Master Tool to keep track of your CPU temperatures. It's the temperature sensor listed in Ryzen Master Tool for a reason.

Funnily enough CCD1 still seems to report closer to what Ryzen Master tool does once you add in the decimal point via HWINFO64, so i will probably just carry on using that but there is not much difference between all of them.

 

[gupsterg]

No worries.

I could see you going off on a tangent with wanting to start linking to posts, so i just wanted to reiterate what my original post was saying - that it's okay to use the temperature listed in Ryzen Master Tool to keep track of your CPU temperatures. It's the temperature sensor listed in Ryzen Master Tool for a reason.

Funnily enough CCD1 still seems to report closer to what Ryzen Master tool does once you add in the decimal point via HWINFO64, so i will probably just carry on using that but there is not much difference between all of them.

No worries . I only stated I'd link post in reply to your post #12 of this thread. Use whatever you think is best, I think you and I have posted more than enough on this now in thread for anyone to know what is what .

 

[LtMatt]

No worries . I only stated I'd link post in reply to your post #12 of this thread. Use whatever you think is best, I think you and I have posted more than enough on this now in thread for anyone to know what is what .

It's the first thing we suggest when troubleshooting temperature queries is to check which sensor in a third party app is reporting the same as Ryzen Master Tool, so that the user knows which temperature he should monitor for CPU temperature.

 

[Chrisc]

MY 3900X Would often go well into the 80's while pushing well over 1.4v. ive put a voltage offset of -0.100v so it sits around 1.3/1.35v most of the time while being a lot cooler.