cancel
Showing results for 
Search instead for 
Did you mean: 

Red Team Discussions

Challenger
Challenger

Let's talk about Ryzen Master and "Performance"

GigaByte brought out a revised version of their AGESA 1.0.0.4 B (F10c) for my GigaByte X570 AURUS XTREME motherboard which still falls short of the previous 1.0.0.3 ABBA version in that the memory clock can only be set to a maximum of 3600 (MCLK 1800, FCLK 1800) whereas in the prior BIOS version there was no problem running the RAM at 3733 (MCLK 1866, FCLK 1866).

It is however an improvement over the prior AGESA 1.0.0.4 B (F10a) version which would not even run at 3600.

That being said, I decided that I would spend an entire day (from 10AM to 10PM) yesterday finding out the state of play with regard to my system and the configuration options offered by Ryzen Master Version 2.0.2.1271.

For this test I decided that I should do a completely clean installation of Windows 10 Enterprise, with only the latest patches for the OS, drivers and Ryzen Master installed and running and using CineBench R20 as the benchmark for all the tests.

Here is my Windows version:
Windows Version.PNG

The following Ryzen Master parameters were the same for all the tests:

the same parameters.PNG

___________________________________________________________________________________________________

I followed the same testing methodology for each of the options offered by Ryzen Master - Eco-Mode, Default, Precision Boost Overdrive, Auto Overclocking and Manual - which consisted of running CineBench twice before recording the CineBench Score (if there was a discrepancy between the first and second run I would do a third one) in both All-Core and Single core.

After this I would run the All-Core and Single-Core tests again but this time I would screenshot Ryzen Master whilst the test was running and for each screenshot I chose the same part of the Benchmark run for all the different options to show the system load.

The only thing running on the system for the score was CineBench R20 and Ryzen Master, whereby CineBench was always in the foreground (focussed) and Ryzen Master was behind that window.

___________________________________________________________________________________________________

The first test up was Eco-Mode and the Cinebench score was:

Eco_Cinebench_Result.PNG

The system load during the single core run was:

Ryzen_Master_Single_Core.PNG

The system load during the all-core run was:

Ryzen_Master_All_Core.PNG

___________________________________________________________________________________________________

The second test was the Default mode, the benchmark score was:

Default_Cinebench_Result.PNG

The system load during the single core run was:

Ryzen_Master_Single_Core.PNG

The system load during the all-core run was:

Ryzen_Master_All_Core.PNG

___________________________________________________________________________________________________

The third test was the Precision Boost Overdrive mode, the benchmark score was:

Cinebench_Result.PNG

The system load during the single core run was:

Ryzen_Master_Single_Core.PNG

The system load during the all-core run was:

Ryzen_Master_All_Core.PNG

___________________________________________________________________________________________________

The fourth test was the Auto Overclock mode, the benchmark score was:

Cinebench_Result.PNG

The system load during the single core run was:

Ryzen_Master_Single_Core.PNG

The system load during the all-core run was:

Ryzen_Master_All_Core.PNG

___________________________________________________________________________________________________

What is beyond my ken is that in all four of the above scenarios, there is next to no difference between the various CineBench results, but what flabbergasts me entirely is that in the First and Second scenarios I am running my system "In Spec" according to AMD; however in the Third and Fourth scenarios, I am, according to AMD, voiding my warranty on my CPU by implementing them.

There is however next to no difference in the exorbitant amount of voltage being applied - especially the single core CineBench runs result in voltages being applied to the CPU cores which I would consider to be reckless.

TSMC - the creators of the 7nm Node and the people who should know what they are talking about - have, according to reports I have read, specified a maximum voltage of 1.30 to other reports I have seen 1.325 Volts for that Node.

Under no circumstances in the scenarios above are any of those maximums even remotely adhered to.

What is worse in my view is that not only is a ridiculous amount of voltage being pumped into the CPU, but the results, in terms of CineBench results does not even remotely justify it, as can be seen when I manually configure Ryzen Master to remain within specs:

___________________________________________________________________________________________________

The fifth test was the Manual Mode, the benchmark score was:

Manual_Cinebench.PNG

The system load during the single core run was:

Ryzen_Master_Single_Core.PNG

The system load during the all-core run was:

Ryzen_Master_All_Core.PNG

By manually clocking the cores to 4.225 GHz and setting the Peak Core(s) Voltage to 1.325V I do take a performance hit on the single core score of around 3.5% I do however gain over 2.6% in the all-core score.

And this at significantly less voltage and temps.

Aside from synthetic benchmarks, how often are single core workloads applied to the CPU, let me think now, ah yes, that would be NEVER.

All things considered, any Ryzen Option aside from manual configuration results running the CPU in what AMD itself considers an unsafe and potentially damaging way.

The thing is that applying too much voltage over time will not normally suddenly kill the CPU, but rather degrade the performance, and then, if AMD is lucky, not kill off the CPU that people have spent quite a lot of money on within the three year warranty.

The Default amount of voltage will be applied whether Ryzen Master is loaded or not, and the so-called "Eco-Mode" can only be considered a sick joke on the part of AMD.

0 Kudos
10 Replies
Adept I
Adept I

Re: Let's talk about Ryzen Master and "Performance"

This is a timely post, since I'm having similar issues on my ASUS ROG X570 Crosshair VIII Hero with 3700X.  My testing isn't as complete as yours, but I'm seeing no difference in performance between modes of operation, but big differences in power draw:

https://www.reddit.com/r/Amd/comments/dwenk1/pbo_and_ecomode_worthless_anyone_else_seeing_this/

(I'd post that information here, but reformatting it for this forum doesn't sound like much fun).

One thing I'd like to point out, though, is that your screenshots of the Eco-Mode run show Default as being the Control Mode (not Eco-Mode).  It's possible you've just grabbed the wrong screenshots.  But, I ran into a case where Ryzen Master just reverted away from Eco-Mode when I shut it down and restarted it (as opposed to resetting it myself).

Also, did you have to do anything special to get Eco-Mode to show up in Ryzen Master?  On my Asus board, it won't show up unless PBO is turned on in the BIOS.  And, PBO seems to override anything Eco-Mode wants to do.

0 Kudos
Challenger
Challenger

Re: Let's talk about Ryzen Master and "Performance"

I had not noticed that, but I just activated Eco-Mode once more, and here are the two screenshots from the actual profile and then the home page where the various readouts are shown:

Here is how the Profile page is displayed:

Profille Page.PNG

And this is how it is displayed on the Home Page:

Home Page.PNG

So unless "Eco-Mode" is completely bogus, it could well be that it hasn't been linked to the Home Page readouts.

I would strenuously warn against setting values in BIOS for two reasons:

1) It interferes with the working of Ryzen Master (the one exception to this that I have found is setting the LLC - Load Line Callibration) in my experience. The reason for this is that you will have noticed that changes you can make in the BIOS section involving tweaking is mirrored in the "AMD Overclocking" and "AMD CBS" sections of the BIOS, the actual changes you make however does not seem to be reflected there.

From what I have seen, it appears that Ryzen Master is unaware of the setting entered into the "AI Tweaker" area of the BIOS and if the settings you enter into the Ryzen Master software conflicts with what you have entered into the Tweaker area of the BIOS then you are going to be entering a world of pain.

2) Some changes you can make in in the "AMD Overclocking" and "AMD CBS" section of the BIOS cannot be undone with a "Clear CMOS" and if you screw something up by putting in a wrong value, it can brick your system and then you better hope that you have a mobo that has a flashback function to reflash the BIOS.

If you are going to be using Ryzen Master, then, in my experience, it is imperative that you activate "Load Optimised BIOS Defaults" and then only change the the LLC (Load Line Calibration) values for the CPU and Vcore SOC.

The AGESA versions are shipped to the motherboard manufacturers as a binary, and not as source code, so the other parts of the BIOS have to treat the AGESA portion as a "Black Box" and see what changes in their BIOS values have on whatever the hell is going on in the AGESA portion. The ironic saying, "What could possibly go wrong", comes to mind

In my case I have set the Vcore LLC to "Turbo" which is the third highest value and for the Vcore SOC I have set it to "High" which is also the third highest value. The rule of thumb for setting these values has always been never to set the highest or next highest value, unless you have a really good reason for doing so.

This generation of Ryzen CPUs is my first experience with an AMD Ryzen BIOS and it has been a steep learning curve compared to the Intel BIOSes I have become used to over the decades (I have been a computer techie now for 37 years and am 60 years old) so it quite literally has been a case of "Teaching an old dog new tricks"

My experience has been with the GigaByte Ryzen BIOS; however I am confident that my experience is applicable to your ASUS BIOS. It has taken me days/weeks of trial and error to come to the conclusions I have shared with you.

The bottom line is, that you can either set the values in the BIOS by hand and avoid Ryzen Master, or load the BIOS defaults, set LLC and nothing else and then control the BIOS via Ryzen Master. Any other combination has resulted in my system reacting erratically and unpredictably.

I wrote something else which you may find interesting here:

Definitive guide to configuring the Ryzen 3900X 

P.S. If you want to format something from Reddit to post here, then just copy the text and then and paste it into Notepad, turning "Word Wrap" off under the option "Format", then copy that and paste it here. It takes a lot of the hassle out of the whole procedure.

0 Kudos
Challenger
Challenger

Re: Let's talk about Ryzen Master and "Performance"

So I decided to play the game, "How low can you go" with regard to the voltage and still be able to do multiple Cinebench runs without crashing.

The lowest it would run at stably after multiple reboots and Cinebench R20 runs was 1.275 Volts at an all core of 4.225 GHz, going down to 1.26875 Volts resulted in an almost immediate crash, as in black screen and reboot.

Here is the Cinebench score at 1.275 Volts

Cinebench_1_275_Lowest_Voltage.PNG

The system load during the single core run was:

Ryzen_Master_Single_Core_1_275.PNG

The system load during the all-core run was:

Ryzen_Master_All_Core_1_275.PNG

The ambient temperature was 25° C as measured by the thermistor that came with my new motherboard and that is the normal ambient temperature which is applicable to the results I got in the first post of this thread.

I realised that posting the results with no indication of the ambient temperature made the results somewhat ambiguous. The results were, and are, valid with regard to each other in my own environment, but not comparable with results others have when benchmarking in a different environment.

When I get the 3950X (hopefully at the end of this month if there are no shortages) I will be supplementing this post with those results conducted in the same way.

0 Kudos
Challenger
Challenger

Re: Let's talk about Ryzen Master and "Performance"

Further testing with my system showed that it would only be rock steady stable (running CineBench on a loop for 10,000 seconds 2 Hours 46 Minutes 40 Seconds and completing multiple runs) at an all-core clockspeed of 4.225 GHz at 1.29375 Volts.

My own Ryzen 3600X has a very sharp cutoff, with regard to required voltage, at 4.225 GHz. Just increasing the all-core clockspeed by 25 MHz to 4.25 GHz required 1.35 Volts to run stably.

Upgrading to the newest Ryzen Master software (the version 2.1.0.1424) decreased the maximum achievable all-core CineBench score by approximately 60 Points.

What the new software did not however do is change the stupid amount of voltage it pumps into the CPU at any setting aside from Manual.

To all of you who think, "This guy is a nut, AMD knows what it's doing", I will wait until next year when reports of CPUs failing, or no longer performing as they should.

The thing is that overvolting a CPU will degrade the CPU over time, meaning that it will either require more voltage to achieve previous performance (accelerating the decay of the CPU) or the clockspeed will be reduced.

And yes, I will say, "I TOLD YOU SO".

I know how I will be configuring my 3950X when I get it, and I am confident that I will have a lot of joy out of the CPU for the next five years or so (my system runs 24/7) with the clockspeed I set now still being achieved in five years time.

To those of you who are of the "AMD know what they are doing" mindset, I just hope that your CPU dies within the three year warranty period, and not, as Murphy's Law dictates, exactly one day after.

Just don't be surprised if the CPU you bought has to make do with worse and worse performance as time goes on. I estimate that the complaints will start to become prevalent next year.

0 Kudos
Adept I
Adept I

Re: Let's talk about Ryzen Master and "Performance"

This is interesting stuff.  On my system, I've only followed the 1usmus settings for overclocking my memory and setting up his power plan (which I'm not even running any more).  I've changed the settings in the BIOS, not Ryzen Master.  Besides the actual memory timings, here's what I've changed:

Global C-state Control = Enabled (from Auto)

CPPC = Enabled (from Auto)

CPPC Preferred Cores = Enabled (from Auto)

Voltage Block (voltage range)3200 FAST Rec
DRAM Voltage1.365
SOC Voltage1.025
cLDO VDDP Voltage0.9

Misc Items3200 FAST Value
Power Down modeDisabled
Gear Down modeDisabled
Command Rate1T
BGSDisabled
BGS altEnabled
FCLK1600

Termination Block Omega3200 SAFE/FAST Rec
procODT34.3
RTT_NOMOFF
RTT_WROFF
RTT_PARKRZQ/5(48)

CAD_BUS Block Omega3200 SAFE/FAST Rec
CAD_BUS ClkDrv24
CAD_BUS AddrCmdDrv20
CAD_BUS CsOdtDrv20
CAD_BUS CkeDrv24

I don't THINK I've changed anything which would change the peak core voltages, but I'm seeing the same kind of thing you are.  Here's Ryzen Master during a run of Cinebench single core:

20191203 -- Ryzen Master Cinebench Single Core.JPG

So, single core is showing a Peak Core Voltage of 1.43v.

And, here's one during all core:

20191203 -- Ryzen Master Cinebench All Core.JPG

Multi-core shows a Peak Core Voltage of 1.34v.

And, just for grins, here's Ryzen Master while my system is "idle:"
20191203 -- Ryzen Master Idle.JPG

Idle Peak Core Voltage is 1.17v

The resulting 3700X Cinebench Scores:

20191203 -- Cinebench Scores.JPG

EDIT:  Never mind the following.  Apparently, the 3600X is supposed to have higher numbers here than the 3700X.  Sorry.

OP:  But, I also wanted to bring up the differences between your 3600X and my 3700X in Ryzen Master's reported PPT, TDC, and EDC:

Your 3600X (low)Your 3600X (high)My 3700X
PPT128W1300W88W
TDC80A700A60A
EDC125A840A90A

Even with your lowest numbers, my 3700X's are lower.  Your high numbers are probably because PBO is on.  But, have you changed something to make the low version of those numbers higher?

0 Kudos
Challenger
Challenger

Re: Let's talk about Ryzen Master and "Performance"

Here's an experiment for you.

In your BIOS click on "Set Advanced BIOS Defaults", and now change ONLY the values for the CPU LLC (Load Line Calibration) and SOC LLC to two levels below the highest value. Optionally you can turn off the motherboard logo as well.

There, that's your BIOS configured.

Start your system and now kick off Ryzen Master.

Create a new profile (or use Profile1 or Profile2) and set the Control Mode to "Manual"

Under "Cores Section" set all the cores to 4.225 (just as a start off point, if that is stable you can go higher - without of course changing the Peak Core Voltage value below).

Set the "Voltage Control" to "Peak Core Voltage" of 1.3V

In "Memory Control" set the "Memory Clock" and the "Fabric Clock" to 1600.

Set the MEM VDDIO to 1.35 Volts, MEM VTT to 0.675 Volts and VDDCR SOC to 1.1 Volts

Leave everything else on AUTO in the memory section except under "DRAM Controller Configuration" set "Cmd2t" to "1T"

That's it.

Now hit "Apply&Test" and get back to me with your results

0 Kudos
Challenger
Challenger

Re: Let's talk about Ryzen Master and "Performance"

As another update.

I was speaking with someone from an overclockers Discord and invited to join.

In the discussions I was having someone told me that Ryzen 3000 could not be clocked to 4.1 GHz when running Prime95 with small FFTs (which produces more heat than smallest FFTs according to them), so I did the experiment and as you can see from the screenshot I have my Ryzen 3600X running Prime95 with small FFTs at 4.125 GHz

The screenshot was made over three hours into the Prime95 run as you can see from when it was started and if you look down at the bottom right of the screenshot you can see my current time.

The cooler I am using for this test is the Noctua NH-U12A.

The ambient temperature was 26°C, in case someone is wondering if I was conducting this test in a chilled room (aka "Doing an Intel" )

prime95_4125_3hrs.PNG

Also notice that MCLK and FCLK are clocked to 1867 for this run and at no point was there any throttling.

0 Kudos
Challenger
Challenger

Re: Let's talk about Ryzen Master and "Performance"

One more thing that I considered was the positioning of the fans on the cooler.

Out of the box the fans were configured that they were poking above the cooler.

This does not make sense to me, and the cooler itself has absolutely no problems with RAM clearance so I have placed the fans so that they are poking below the actual cooler as you can see in this photo:

20191207_223914.jpg

My thinking is that by having the fans positioned as they are I prevent a dead space between the cold plate and the actual cooler which could function as a heat-trap

0 Kudos
Adept I
Adept I

Re: Let's talk about Ryzen Master and "Performance"

Quick question, is there a setting that disables the sleep button in my power options. It only shows shutdown and restart. Thanks! 

0 Kudos