[SOLVED] 2700x not boosting max clock

[O-steal]

Hello

Specs:

• CPU: 2700x alll stock option
• Cooler: stock wraith prism
• PSU: focus + gold seasonic 650w
• RAM: flare X 3200MHZ 2x8G CL14
• Storage: nvme samsung evo 970 512G
• OS: win 10 1903 pro
• MB: MSI Mortar B450M
• BIOS 7B89v1B (same result with 7B89v1A , 7B89v19 , 7B89v18 and 7B89v17 )

Issue:
On any thread, the max clock speed ever attained is 4.2Ghz for a fraction of a second and this only happens when waiting idle on windows and all other cores are in “cool and quiet”state around 2.7GHz

On single thread tasks, I reach only 4.025 for more than 3 consecutive seconds

If I tried the integrated “game boost” mode from amd, the system crashes quickly

I want to RMA my cpu but the support says that this is expected

EDIT: People report temperature of 65C under full load with stock cooler. I have already 55C on idle and 85C on load (it thortells down to 3.8GHZ then).

Question for you
Would you say that my CPU is faulty? (Either because of boost and/or temps)

DATA
These are 10 min of logs of me:

• Browsing for 7 min
• Cinebench r20 single thread for 3 min

Spoiler: Logs

Cinebench

Here are some logs from a whole pcmark10 run as requested by support (msi afterburner file format)
https://filebin.net/cv9x7zf464b2egee

 

[O-steal]

I definitely hit 85C when running cinebench (case closed is 1C colder than opened). I ll try to file a new ticket but not complain about boost clock but rather temps then.
With PBO

Without PBO

 

[Rogue Leader]

PBO shouldn't be required to hit 4.3 GHz in lightly threaded loads, assuming other conditions are being met (e.g. cooling). PBO can improve max boost (I think up to 4.35-4.4 GHz with a 2700X, can't remember), and improve multi core boost.

PBO won't do 4.4ghz, it may hit "4.35" ghz but the exact measurement is questionable. According to AMD its 4.3ghz.

There actually aren't any X470 boards with true 8 phase; 6 is the highest. And there are only 5 mobos I see that have 8+ phases using doublers (and none of them ar mATX).

Edit: To expand on this, the mortar uses two pairs of FETs per phase. So as far as current capability goes, it should be on par with a doubled 4 phase VRM using one set of FETs per phase (assuming same FETs/cooling).

I stand corrected I thought that did exist. However I stand by the fact that a 4 phase VRM even doubled is not going to allow for max PBO without pushing some settings (and obviously his thermals are bad).

 

[Rogue Leader]

I definitely hit 85C when running cinebench (case closed is 1C colder than opened). I ll try to file a new ticket but not complain about boost clock but rather temps then.

Before assuming there is a failure in the CPU, maybe try removing your cooler, cleaning everything with alcohol and using new thermal paste to install properly. I think its quite possible the cooler is not fully seated or otherwise somehow compromised in its installation.

 

[drea.drechsler]

There actually aren't any X470 boards with true 8 phase; 6 is the highest. And there are only 5 mobos I see that have 8+ phases using doublers (and none of them ar mATX).

Edit: To expand on this, the mortar uses two pairs of FETs per phase. So as far as current capability goes, it should be on par with a doubled 4 phase VRM using one set of FETs per phase (assuming same FETs/cooling).

That's been the thing that sets the mortar apart, the VRM not only has the fully doubled FET's in the VRM but also a very big and well finned (for modern boards) heatsink on them. It's really quite cool running even with a 3700X running at 4.3Ghz in Prime95.

I know it's not saying much, but it IS the best AM4 mATX board. So if you want the smaller form factor and capable VRM, it's what you get.

View: https://www.youtube.com/watch?v=6qZW3-xZEHg&t=150s

 

[O-steal]

Before assuming there is a failure in the CPU, maybe try removing your cooler, cleaning everything with alcohol and using new thermal paste to install properly. I think its quite possible the cooler is not fully seated or otherwise somehow compromised in its installation.

I have done this already (see above). Repasted it twice and when i checked the previous pasting it was evenly spread. Lever properly installed and pushes with force on the cpu, no wiggling even just after pasting

 

[TJ Hooker]

PBO won't do 4.4ghz, it may hit "4.35" ghz but the exact measurement is questionable. According to AMD its 4.3ghz.

It's 4.3 GHz without PBO.
https://cdn.mos.cms.futurecdn.net/dP5hpEZQKxe9nAP9Zd2WJ7-650-80.png

I believe PBO can let it go a bit higher in at least some cases. There was a good HardOCP article exploring PBO on a 2700X, but unfortunately it seems to be unavailable at the moment.

 

[drea.drechsler]

I have done this already (see above). Repasted it twice and when i checked the previous pasting it was evenly spread. Lever properly installed and pushes with force on the cpu, no wiggling even just after pasting

It's easy to see something's wrong with cooling and that's holding the processor back. STI2/TFN Vcore even shows you're at 1.287 V, not really high under load.

PBO is giving you 25-75 Mhz more boost under load, but that's not much an based on TJhooker's linked chart you should be at least 4.05 G even without PBO.

With those temp's, if you had access to another big cooler would be great to test out and see what happens.

 

[Rogue Leader]

It's 4.3 GHz without PBO.
https://cdn.mos.cms.futurecdn.net/dP5hpEZQKxe9nAP9Zd2WJ7-650-80.png

I believe PBO can let it go a bit higher in at least some cases. There was a good HardOCP article exploring PBO on a 2700X, but unfortunately it seems to be unavailable at the moment.

I've read the article and seen the reports. Nobody who is actually honest is getting anything more than "4.35" ghz out of PBO.

 

[drea.drechsler]

I've read the article and seen the reports. Nobody who is actually honest is getting anything more than "4.35" ghz out of PBO.

I always understood the PBO benefit to be holding 4.3 longer under load with more cores, and even then only under a good CCL setup with a big waterblock and high-flow pump. Some people have claimed more but I'm sure it's under similar conditions.

But the benefit of PBO is you can reach for more than you're getting NOW, and it's really fairly safe. Whatever you get, it's more. Even OP is getting 50-75Mhz more, with something going wrong with his cooling.

 

[Rogue Leader]

I always understood the PBO benefit to be holding 4.3 longer under load with more cores, and even then only under a good CCL setup with a big waterblock and high-flow pump.

Agreed, XFR worked similarly if not as long (and only 2 cores).

 

[O-steal]

With those temp's, if you had access to another big cooler would be great to test out and see what happens.

Sadly I don't have access to a bigger cooler and that's kind of an investment I might not need because Wraith is supposed to be good is part of the reasons why I opted for an expensive 2700x (also amd says it voids warranty to use third party cooler).

Side note: really happy to see a real conversation about 2700x, boost clock, PBO and such. I have been worrying on my cpu for 6 months now and couldn't find a proper thread

 

[TJ Hooker]

I've read the article and seen the reports. Nobody who is actually honest is getting anything more than "4.35" ghz out of PBO.

Sure, but going back to my original point, it doesn't appear that PBO should be required to hit 4.3 GHz single core boost.

I stand corrected I thought that did exist. However I stand by the fact that a 4 phase VRM even doubled is not going to allow for max PBO without pushing some settings (and obviously his thermals are bad).

Given that there are 4 phase (non phase doubled) boards able to OC a 2700X to 4.2 GHz all cores with plenty of thermal headroom on the VRM, I don't see why something like the Mortar couldn't handle ~4.3GHz in lightly threaded loads.
https://www.guru3d.com/articles_pages/msi_b450_tomahawk_review,17.html
https://www.tweaktown.com/reviews/8751/msi-b450-tomahawk-amd-motherboard-review/index7.html

Heck, someone even achieved 4.2 GHz on the 3 phase Pro4.
https://www.vortez.net/articles_pages/asrock_b450m_pro4_review,6.html

 

[Karadjgne]

Afaik, Ryzen master uses an average of Tdie taken over 3 or so seconds to get its reported temp. The latest HWInfo may be similar or not an average as far as Tdie goes. Last I heard, the author was workin with AMD hard to get HWInfo on point, so not sure how far that's gotten yet. Tctl being hottest single core reading is kinda inaccurate overall as generally it's nothing more than a spike and not representative of the cpu as a whole.

Instead of jacking PPT/TDC/EDC, you can also lower those values, which lowers current applied, so uses voltage scaling more efficiently, which lowers temps. Could try 125/95/130 instead of the default 142/95/140.

In the midrange B450 mobo's, Msi had the better VRM's setup, using the same mosfets as used in the high end X470 boards, albeit a few less. Dunno where they dropped the ball, or what got changed, but in the midrange X570's, they are the worst, getting stomped on by the Gigabyte Aorus Elite and the Asus Tuf Gaming Plus.

 

[Rogue Leader]

Sadly I don't have access to a bigger cooler and that's kind of an investment I might not need because Wraith is supposed to be good is part of the reasons why I opted for an expensive 2700x (also amd says it voids warranty to use third party cooler).

Side note: really happy to see a real conversation about 2700x, boost clock, PBO and such. I have been worrying on my cpu for 6 months now and couldn't find a proper thread

Warranty should be the least of your worries. The chances of your CPU being defective are less than 1%. Also there is no way for them to determine what cooler you used, totally impossible. This is their current warranty requirements, the cooler requirement was old info.

https://www.amd.com/en/support/kb/warranty-information/pib

Something is very clearly wrong with your temperatures, unless your room's ambient temp is VERY high (standard room temp that stuff is tested at is 72F, if your room is 82F that can and will affect results), then something is wrong with your cooler, either the cooler itself or the installation. How are you installing it, and what thermal paste did you use?

 

[O-steal]

Something is very clearly wrong with your temperatures, unless your room's ambient temp is VERY high (standard room temp that stuff is tested at is 72F, if your room is 82F that can and will affect results), then something is wrong with your cooler, either the cooler itself or the installation. How are you installing it, and what thermal paste did you use?

(context i did watch hundred videos of cooler mounting and unmounting during the last 5 years (cooler tests and comparison, thermal paste comparison etc, I did not do that wrong))

The first time I just used the stock cooler out of the box with its stock paste. I applied it normally, on the cpu with nothing on it, not even touched by my fingers, just doped from its package, there was no trace of any dust or fiber. Put the lever in the brackets and sealed it. The rad was not moving after

Then when i realised the temp where wrong, I unlocked the lever and I twisted the cooler gently until it started to move, then I removed it completely vertically to make sure no paste fell anywhere. Cleaned both cpu and cooler with coffee paper filter and redid the process with 1 pea size of thermal NT-H1

Then redid it again, the spread from the previous application was good too. This time i tested a small cross

All three times, I get the same temps. My room is cold. The case does not overheat, there is plenty of fans directed on it, open or not, no big variation.

I guess I can find you picture of the surface of both cpu and cooler before and after cleaning

 

[Rogue Leader]

I'll take your word for it that you did it right. if you pated that way and its tight, you're good. I'm assuming the fan is plugged to the proper connector.

If thats the case something is still wrong, either the cooler itself or maybe the CPU. 55C at idle is absurd, and 85 at full tilt as well.

I'm assuming the fan is spinning up and down properly?

 

[drea.drechsler]

...
The first time I just used the stock cooler out of the box with its stock paste. ...

A thought...just a thought. The Prism uses mounting clips (the other Wraith's all use the mounting posts directly). One thing that happened to me on an AM3 was I got the plastic clip and frame attached incorrectly to the backplate so it was just a bit uneven. The result was the HSF seemed solidly mounted but it was cocked just every so slightly. I only noticed that because it was a Hyper T4 tower so it leaned quite noticeably. But mounted the way it was one corner would have been making contact and the opposite corner would have had a bit of a gap. Bad contact = bad heat flow.

That's just a thought, don't know if the AM4 clips can be skewed like that or not. Something to check.

 

[O-steal]

A thought...just a thought. The Prism uses mounting clips (the other Wraith's all use the mounting posts directly). One thing that happened to me on an AM3 was I got the plastic clip and frame attached incorrectly to the backplate so it was just a bit uneven. The result was the HSF seemed solidly mounted but it was cocked just every so slightly. I only noticed that because it was a Hyper T4 tower so it leaned quite noticeably. But mounted the way it was one corner would have been making contact and the opposite corner would have had a bit of a gap. Bad contact = bad heat flow.

That's just a thought, don't know if the AM4 clips can be skewed like that or not. Something to check.

Thanks for the lead but that something I checked when changing the paste, I used the screwdriver on the bracket just to make sure, and also how the spread looked nothing was skewed

 

[O-steal]

For info
I just tried to put all fans on max from the start (not just when reaching 70 the default) and put the case open in front of the window with air at around 10C.

I did get cooler temps : all core stress: around 68C without PBO and around 80C with PBO

But the boost clocks where identical (Which I understand as the throttling only happens at 85C and when i was testing single core, it did not went over 75C before when testing)

The all thread clocks were also identical except that they did not go down as they did when reaching 85 before. They stayed around 3950GHz

 

[TJ Hooker]

Are you still using cinebench for testing? Is that most recent temp and frequency result when running single thread, or 16 thread?

What OS do you have? What build/version (can type "winver" into search and run that to get the build version).

Do you have the latest AMD chipset drivers installed (1.9.27.1033 )? What power plan are you using?

Have you tried clearing CMOS? There'll be instructions on how to do this in the mobo manual if you don't know how.

 

[O-steal]

Are you still using cinebench for testing?

Use it for finding max temp all core

Is that temp and frequency when running single thread, or 16 thread?

The temps are for

all core stress

so 16 thread

What OS do you have? What build/version (can type "winver" into search and run that to get the build version).

see above, win pro 64 1903 build 18362.449 (but the issue is since the building in may)

Do you have the latest AMD chipset drivers installed? What power plan are you using?

cf question and thread : 650w gold seasonic, power plan balanced default 100% max cpu state, min cpu state 5%

Have you tried clearing CMOS? There'll be instructions on how to do this in the mobo manual if you don't know how.

Did clear cmos before before launching these last "very cold setup" tests. And since 6 months tried all bios options even though my issue is, as you explained in your first messages, that I should get proper single thread speed on default settings for BIOS as long as i don't reach above 70C on the CPU and am not limited power wise. Which I don't

 

[mdd1963]

Is there perhaps any sort of power limit within the BIOS? (I saw a reference to 120-plus Watts, but, perhaps a limit is keeping it at 95Watts sustained, etc, when we know the 2700X draws 135W+ at max boost under a heavy load like Cinebench or Blender

 

[TJ Hooker]

3.95 GHz under all core load doesn't seem unusual. You're not going to get 4.3 GHz on heavily threaded loads without manual overclocking (and a better cooler). It may not even be possible depending on your mobo and the quality of your chip.

 

[Rogue Leader]

Either way this isn't a power issue if hes hitting 85C (and 95C tCTL!!) under load. The system is throttling

 

[O-steal]

Is there perhaps any sort of power limit within the BIOS?

We already discussed MB compatibility and out of the box expected performance. Anyway, I don't reach out of the box perf using PBO which sets EDCat 168A and TDC at 114A (Power limit is at 1000W) and from the hwinfo64 screeshot we can see that 140W was easily achieved (and for single thread performance, wattage would not be the limiting factor as the 16 thread load was sustained)