cancel
Showing results for 
Search instead for 
Did you mean: 

Processors

Highlighted
Adept II
Adept II

So, here's something weird with my 3900X...

Computer Type: Desktop custom built (by me)

GPU: GIGABYTE AORUS GTX 1080 Ti (GV-N108TAORUS-11GD)

CPU: AMD Ryzen 9 3900X 12-Core Processor (Default, except FCLK at 1800MHz for memory OC at 3600MHz)

CPU Cooler: Cooler Master Masterliquid 240

Motherboard: ROG STRIX X570-F GAMING

RAM: Corsair Dominator Platinum 2x16GB 2800MHz DDR4 Samsung B-die module (CMD32GX4M2A2800C16)

PSU: Seasonic Prime Snow Silent 750W 80+ Platinum

Operating System & Version: Windows 10 Pro 64-bit Build 18362

GPU Drivers: GeForce 436.48

Chipset Drivers: 1.9.27.1033 (taken from AMD's website, as I can't figure out how to check this in my system)

Background Applications: Cinebench R15, Cinebench R20, Ryzen Master, HWInfo

Description of Original Problem and Troubleshooting:

About 2 weeks ago I decided to finally switch from Intel to AMD. I picked up a 3900X paired with the ASUS RoG X570 Strix-F motherboard and 2x16GB Corsair Dominator Platinum 2800MHz DDR4 kit with Samsung B-die modules. All went well, and after finishing the build, I immediately updated the BIOS to the latest available on ASUS's website, which is 1201 and this includes the ABBA AGESA update. I then also grabbed all of the necessary drivers, software suites, etc, Ryzen Master, and the chipset driver from AMD's website. I benchmarked the system and started overclocking my memory to 3600MHz with just 1usmus' DRAM calculator. These all went perfectly fine for me.

It was at this moment that I found out about the Ryzen 3000 boosting issue from Der8auer's video and from what I can gather the ABBA AGESA was supposed to somewhat mitigate this boosting issue. This intrigues me and I get to benchmarking the system matching Der8auer's polling requirement, just to see where my 3900X lies in Der8auer's distribution curve. I did the R15 single-core benchmark, with HWInfo for monitoring, clock speed monitoring only, 500ms polling, and no PBO. So, basically, the only change I have in BIOS at this point is just those related to my memory overclocking at 3600MHz and 1800MHz FCLK to run it at a 1:1 ratio. Now, I was quite happy with the result, as I was able to see in HWInfo, with R15, that my maximum boost during the single-core benchmark reached 4.5917GHz. For the sake of curiosity, I then tried the same thing with R20, and I was even more surprised since the maximum boost clock during the benchmark is 4.6167GHz. I repeated this many many times at the time and the results were consistent. That is to say that I can keep hitting 4.5917GHz and 4.6167GHz boost clocks in R15 and R20 single-core benchmark respectively.

Now, this is where things get weird for me and my understanding of the CPU behavior began to break down. Pretty much a week later up to the writing of this post, I occasionally randomly benchmarked the system with both R15 and R20, and I was unable to reach the same maximum boost clock in HWInfo. These are also reflected by the marginally low Cinebench score. I again repeated this many many times, and up to the writing of this post, the behavior is also consistent, in that, the 4.5917GHz and 4.6167GHz boost clocks are nowhere to be found. Now, the point of this post is to seek help in understanding why this is happening, what's wrong or if there's anything wrong with either my CPU, motherboard, memory, driver, BIOS, Windows version, monitoring software, or God knows what. I am completely lost now, and that's why I decided to post this here.

Do note that I have not changed anything in BIOS or in any other settings application, Ryzen Master or ASUS's AISuite, ever since I started benchmarking with R15 and R20 after seeing Der8auer's video. If anyone would like to see the "proof" for the 4.5917GHz and 4.6167GHz boost clocks, I can try and post the screenshots I took of it here.

If anyone knows or understands anything regarding this, in my opinion, strange behavior of the 3900X, do let me know! Any help would be appreciated! Thanks in advance!

*POST UPDATED TO INCLUDE MORE DEVICE INFORMATION.

0 Kudos
19 Replies
Highlighted
Adept II
Adept II

Re: So, here's something weird with my 3900X...

No comment people? Really?

0 Kudos
Highlighted
Exemplar
Exemplar

Re: So, here's something weird with my 3900X...

Sorry, jeffb0918, way too many words and too few facts.  HWinfo is the least reliable reporting application.  Please post only Ryzen Master screenshots for measurements.  Clear CMOS, run NO OC software, especially from your MB Vendor and remove any installed on your system.  Get NO drivers from the MB vendor - ONLY from AMD and other manufacturers.  Please run Cinebench and/or AIDA64 and post the RM screenshots.  Also please tell us all your parts, including CPU cooler and PS.  Sorry I see nothing weird yet.  Thanks and enjoy, John.

0 Kudos
Highlighted
Adept II
Adept II

Re: So, here's something weird with my 3900X...

Okay. I did what you ask me to do. Updated the original post with a specification list of my PC. I have cleared CMOS, no OC, no RAM OC, manual or with DOCP, no OC software from anywhere, all drivers are from AMD (for the CPU) and nVidia (for the GPU), BIOS is still 1201 with the ABBA AGESA in it from ASUS's website. Here's the Ryzen master screenshot with Cinebench R15 single core test. Do note that this is with a CMOS reset, which means my memory runs only at 2133 MHz. You can see that the peak frequency there is 4.410GHz, which is significantly lower than the 4.6GHz boost advertised. Now, shouldn't this be considered weird? Especially when I'm using the ABBA AGESA  already?

Now, my comments in my original post is that some week ago I was able to see a little over 4.6GHz when running the single core Cinebench test, why the behavior suddenly change/degrade, without me touching any settings related to that?Screenshot (438).png

0 Kudos
Highlighted
Exemplar
Exemplar

Re: So, here's something weird with my 3900X...

Thanks, jeffb0918.  I am still looking, but seeing 4 CCXs set to 4600 MHz.  Please try 'Current' speed instead of 'Peak' during the run and let me know what you see.  Also try AIDA64.  You can use the trial version if you do not have a license.  I have Gen 1 and Gen 2 TRs and am waiting for Gen 3, so I do not have a Gen 3 Ryzen yet.  I do not understand all this yet.  RM went through major changes for 3.  I want to look into your BIOS 1201 and ask AMD some questions - no weekend support.  So far I do not think this is weird because I suspect you were using HWinfo before, so the results are suspect.  Thanks and enjoy, John.

0 Kudos
Highlighted
Exemplar
Exemplar

Re: So, here's something weird with my 3900X...

OK, jeffb0918, here is something weird!  I DLed BIOS 1201 and examined it with a hex editor.  The embedded name is AGESA!V9.Combo-AM4 1.0.0.3.  It is not AGESA 1.0.0.3ABBA as advertised!  This could be an ID error only, but I suspect that your MB vendor has goofed!  This needs to be resolved first.  Please let me know what you learn.  Thanks and enjoy, John.

EDIT: Please look in your BIOS and see what AGESA it thinks is there.

0 Kudos
Highlighted
Adept II
Adept II

Re: So, here's something weird with my 3900X...

The 4 CCXs at 4600 was there by default, either by the BIOS or the CPU itself, I know not. As for the "current" tab, it's really hard to say because the number jumps every second. All I can tell is that it is not better or might even be worse than the "peak" tab since the "peak" reflects the highest clock speed during the polling time period, and this "peak" speed will be reached occassionally in the "current" tab, not persistently. If anything, I'm giving the 3900X the best case scenario to hit the 4.6GHz mark by looking at the maximum speed only instead of the "current" or "average".

For AIDA64, I can try that and I will post back with what I see.

The reason I find this "weird", aside from my use of HWInfo, is that my maximum single-core boost clock, even when monitored with Ryzen Master, fall behind significantly compared to other people's. Most people, statistically, seem to be able to hit 4.5, 4.525, 4.550, or 4.575 GHz, while my chip is way way lower. Now, my concerns are the following:

1. I got a defective 3900X.

2. I got a defective motherboard.

3. I got very bad luck at the silicon lottery somehow that my chip is way below spec (?)

4. Is the chip degrading over the few weeks I'm using it? Because by default, if I look at vCore, numbers are relatively high,  jumping around 1.3 all the way to 1.5V. Online, there has been numerous threads on this "idle" voltage subject and everyone seems to agree that this is totally fine and normal behavior, nothing to worry about.

So, I'm trying to get to the bottom of this boosting issue of mine, especially since if I look at it from the HWInfo measurements, it points towards degradation, because I'm losing speed without even changing any BIOS settings and whatnot in the course of 3 weeks-ish. And say that HWInfo is an unreliable monitoring software, assuming it has a certain degree of inaccuracy or offset in its sensor measurement if I perform the two comparative measurements (so, 3 weeks ago and now) with HWInfo, and only consider the difference between them, the error due to monitoring software on average should ideally cancel out almost entirely in my opinion.

0 Kudos
Highlighted
Adept II
Adept II

Re: So, here's something weird with my 3900X...

For the AGESA version, I was unable to find any entry regarding that in the BIOS. But, with HWInfo, it reports the AGESA version to be Combo-AM4 1.0.0.3.

From HWInfo, in the CPU tab, the microcode revision is 8701013 and SMU firmware revision is 46.49.0.

Now, I'm not entirely sure how to interpret these as the AGESA version, but I have read online that I should get the AGESA version from these numbers instead of the one listed in the motherboard tab, since that can be inaccurate.

0 Kudos
Highlighted
Adept II
Adept II

Re: So, here's something weird with my 3900X...

From what I can tell by looking through all the previous BIOS versions, they all have the same ID for the AGESA It is possible that they simply didn't bother changing the ID? But then again, it was suggested to determine AGESA version based on the CPU microcode version and SMU firmware version, but I'm not sure how to do this.

0 Kudos
Highlighted
Exemplar
Exemplar

Re: So, here's something weird with my 3900X...

jeffb0918, I have a bad attitude when I get down on an application - mark it off and blame all anomalies on it.  The first thing required is to get ASUS to explain.  Your results are too suggestive of what the ABBA AGESA corrected to be just

coincidences!  AGESA version:

pastedImage_4.png

pastedImage_5.bmp

I do not know those other versions.  You will need to ask an AMD support person.  AIDA64 also gives the AGESA version.

I suspect you do not have any defective stuff, with the exception of a BIOS!  CPU voltages jump around a lot as does temperature.  Bad SW is bad and their attempts to correct it may well make it worse.  I don't think you should assume that HWinfo remain constant and your stuff is dying.  Please let me know what ASUS says.  Thanks and enjoy, John.

0 Kudos