nec_v20

Let's talk about Ryzen Master and "Performance"

Discussion created by nec_v20 on Nov 14, 2019
Latest reply on Dec 3, 2019 by nec_v20

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:

 

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

 

 

___________________________________________________________________________________________________

 

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:

 

 

The system load during the single core run was:

 

 

The system load during the all-core run was:

 

 

___________________________________________________________________________________________________

 

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

 

 

 

The system load during the single core run was:

 

 

The system load during the all-core run was:

 

 

___________________________________________________________________________________________________

 

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

 

 

The system load during the single core run was:

 

 

The system load during the all-core run was:

 

 

___________________________________________________________________________________________________

 

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

 

 

The system load during the single core run was:

 

 

The system load during the all-core run was:

 

 

___________________________________________________________________________________________________

 

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:

 

 

The system load during the single core run was:

 

 

The system load during the all-core run was:

 

 

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.

Outcomes