Over the past few weeks, AMD has been firefighting a controversy with the PC enthusiast community involving its 3rd generation Ryzen processors. The issue people are reporting is that these processors are rarely hitting the advertised maximum boost frequency, not even for a single core, in a single-threaded workload like SuperPi that stresses out a core and doesn't scale across cores. It has been generally acceptable for a modern multi-core processor to apply its highest advertised boost clock speed to at least one of its cores and scale to lower boost frequency on other cores as a workload spreads across more cores.
The situation came to a head when prominent professional overclocker Der8auer published the results of a survey spanning thousands of respondents, from which only a single digit percentage reported that their 3rd gen Ryzen processors reach the advertised boost frequencies at stock settings. This prompted AMD to announce that it is working on a fix in the form of an updated AGESA microcode that will be distributed by motherboard manufacturers as BIOS updates specific to their models. The AGESA version at the heart of this mini review is ComboAM4 188.8.131.52 ABBA. Leading manufacturers have started pushing out beta BIOS updates for some of their models, and ASRock provided us with a beta version of such a BIOS for our ASRock X570 Taichi motherboard.
AMD detailed AGESA 184.108.40.206 ABBA in three aspects; an improved boost algorithm, changes to the processor's idle-state behavior, and an updated monitoring SDK. The boost algorithm is now more aggressive across the board and adds 20 to 25 MHz to the maximum boost frequency of a 3rd generation Ryzen processor. AMD believes this should cover the vast majority of users provided they're testing with workloads that actually max out a single core (while leaving others untouched), such as SuperPi, PCMark 10, or even real-world application such as a single instance of LAME, or a single web-browser process. AMD also refined the processor's idle-state behavior to disregard "intermittent OS and application background noise" seeking the processor to put up elevated boost states in response to workloads that don't really need them (such as video playback, hardware monitoring apps, chat clients, etc.). AMD also announced a late-September release of the new AMD Monitoring SDK, so software developers can make their apps correctly monitor AMD Ryzen processors.
In this mini review, we are testing the flagship Ryzen 9 3900X 12-core processor on our ASRock X570 Taichi motherboard with the latest 2.10 beta BIOS firmware that encapsulates AGESA 220.127.116.11ABBA to test whether our processor reaches its advertised maximum boost clock speed of 4.60 GHz, whether there are changes to the spread of boost frequencies under a multi-threaded workload, and how this change affects performance across our selection of 31 CPU tests and 10 games, with the game tests spanning three resolutions.
|Test System "Zen 2"|
|Processor:||AMD Ryzen 9 3900X|
|Motherboard:||ASRock X570 Taichi|
AMD X570, BIOS v1.30 and BIOS v2.10 Beta
|Memory:||2x 8 GB G.SKILL Flare X DDR4|
|Graphics:||EVGA GeForce RTX 2080 Ti FTW3 Ultra|
|Storage:||1 TB SSD|
|Power Supply:||Seasonic SS-860XP|
|Software:||Windows 10 Professional 64-bit|
Version 1903 (May 2019 Update)
|Drivers:||NVIDIA GeForce 430.63 WHQL|
AMD Chipset 1.07.07.0725
Clock Frequency and Boost AnalysisLet's cut straight to the chase with boost. The controversy surrounding boost with 3rd generation Ryzen processors focuses on a large number of users observing that their processors seldom or never reach their maximum advertised boost frequency, which is generally reserved for single-threaded workloads on both AMD and Intel. The advertised maximum boost frequency for the Ryzen 9 3900X is 4.60 GHz. With the older BIOS, the chip indeed fell short of the 4.60 GHz-mark in single-threaded workloads.
With the new BIOS that has AGESA 18.104.22.168ABBA, we see that AMD's claims do check out. Our 3900X processor not only reaches 4.60 GHz, but exceeds it by up to 110 MHz, reaching 4.71 GHz with the average ending up at 4.65 GHz. With up to 2 threads (1 core), the chip keeps up with this frequency. As we step up to 3T and 4T (2 cores), clock speed drops to 4.575 GHz, which is still 50-75 MHz higher than on the older BIOS.
As we move up to 5T, there is a sharp drop in boost frequency down to 4.40 GHz, which holds until 12T. Throughout this plateau, the new BIOS runs the chip roughly 20 MHz higher than the older BIOS. The next sharp drop happens at 13T, where the chip drops to 4.30 GHz, while still staying 50 MHz faster than on the older BIOS. From here onward, there is a gradual reduction in clock speed with increase in core counts until it settles down at 4.20 GHz all-core. Throughout, the new BIOS runs the chip at marginally higher frequencies. In theory, these slightly elevated boost frequencies should result in minor performance increases, particularly in less parallelized workloads. We put this to the test in the following pages.
Our Patreon Silver Supporters can read articles in single-page format.