After a strong start boasting exceptional reviews and sales performance, AMD's 3rd Generation Ryzen platform has in recent weeks come in for some major criticism. Beginning with reports of strangely high idle voltages in July, and lately followed up by more focussed discussion of Boost Clocks and whether they meet advertised levels, AMD PR has been running damage control while their AGESA development team hurried to develop and validate BIOS code for distribution to motherboard manufacturers. In their latest Community Update (#6 for those counting at home) they address criticism head on, detail the status of the latest AGESA, and outline plans to release an SDK for Ryzen CPU telemetry monitoring in the near future.
Boost Changes
AGESA 1003ABBA has now been released to motherboard manufacturers for integration into vendor and model-specific BIOSes. The purpose of this update is two-fold: to improve Ryzen CPU Boost capabilities beyond that of the previous algorithm, and modify the CPU boost algorithm to introduce an 'activity filter' to allow more consistent behaviour during idle periods.
In the latest Community Update it's claimed that the previous AGESA introduced an error into the CPU Boost algorithm which prevented the CPU from pushing the frequencies to the maximum level. In of itself that's quite understandable. AMD's Zen 2 architecture incorporates a huge amount of telemetry into the Boost algorithm from a wide array of temperature and voltage sensors to adjust frequencies on a millisecond-by-millisecond basis; it's easy for subtle flaws to make their way into such complex systems. Luckily the real performance impact isn't likely to be too high.
Clearly however, debauer's survey and investigations by members of the tech. press into a maximum Boost frequency shortfall for 3rd Gen. Ryzen have struck a nerve. AMD are claiming that the fixed algorithm rolled out with the new AGESA will increase Boost Frequencies by 25-50 MHz, an amount that's not particularly meaningful in terms of performance, but significant in that a far greater number of users would reach the advertised Boost frequencies.
Assuming a 50MHz improvement in Boost frequencies, 28% more respondents to debauer's survey who owned a Ryzen 9 3900X would be able to reach the expected 4.6GHz boost frequencies under the new AGESA, from 5.6% to 33.6% of the total. Tweaks to the testing methodology could also serve to increase that number a little. From a PR perspective, that's a big deal.
AMD's projections are based on a spec system with beta BIOS updated to AGESA 1.0.0.3 ABBA, running PCMark 10 and Kraken JavaScript Benchmark workloads. These apps. have been chosen as they offer 'bursty' single-threaded workloads rather than the more consistent load generated by Cinebench, the application more typically used when investigating this behaviour.
Revisiting Idle Voltage Concerns
Although sufficient for some users, July's software update to power management profiles which helped the processor ignore requests for voltage/frequency boost from lightweight applications hasn't been enough to ameliorate the concerns of all users. The new AGESA brings forward modifications to the CPU Boost algorithm by incorporating an 'activity filter', allowing it to disregard certain load patterns recognised as background activity.
The result should be lower peak and average voltages being reported under low load conditions, and hopefully it will be a more holistic solution than continuing to tinker with the Windows Ryzen Balanced power plan. It also shouldn't impact more general boost behaviour and will not cap voltages, so you will still expect them to vary as usual between 0.2 and 1.5V depending on workload.
New Monitoring SDK
Continued uncertainty among consumers and some elements of the tech. press (particularly the problems cited above) has highlighted the need for more robust tools to monitor Ryzen CPU sensor telemetry. With that in mind, AMD are planning to publicly release a new SDK that exposes sensor data more transparently than before via new API calls, making monitoring tools developed by different parties more consistent in how they gather and report this information. Among these API calls are:
Current operating temperature: Reports the average temperature of the CPU cores over a short sample period. Filters transient spikes that skew temperature reporting.
Peak Core(s) Voltage (PCV): Reports the Voltage Identification (VID) requested by the CPU package of the motherboard voltage regulators. This voltage is set to service the needs of the cores under active load, not necessarily the final voltage experienced by all of the CPU cores.
Average Core Voltage (ACV): Reports the average voltages experienced by all processor cores over a short sample period, factoring in active power management, sleep states, VDROOP, and idle time.
EDC (A), TDC (A), PPT (W): The current and power limits for your motherboard VRMs and processor socket.
Peak Speed: The maximum frequency of the fastest core during the sample period.
Effective Frequency: The frequency of the processor cores after factoring in time spent in sleep states (e.g. cc6 core sleep or pc6 package sleep). Example: One processor core is running at 4GHz while awake, but in cc6 core sleep for 50% of the sample period. The effective frequency of this core would be 2GHz. Intended to give a feel for how often the cores are using aggressive power management capabilities that aren’t immediately obvious (e.g. clock or voltage changes).
Various voltages and clocks, including: SoC voltage, DRAM voltage, fabric clock, memory clock, etc.
Peak Core(s) Voltage (PCV): Reports the Voltage Identification (VID) requested by the CPU package of the motherboard voltage regulators. This voltage is set to service the needs of the cores under active load, not necessarily the final voltage experienced by all of the CPU cores.
Average Core Voltage (ACV): Reports the average voltages experienced by all processor cores over a short sample period, factoring in active power management, sleep states, VDROOP, and idle time.
EDC (A), TDC (A), PPT (W): The current and power limits for your motherboard VRMs and processor socket.
Peak Speed: The maximum frequency of the fastest core during the sample period.
Effective Frequency: The frequency of the processor cores after factoring in time spent in sleep states (e.g. cc6 core sleep or pc6 package sleep). Example: One processor core is running at 4GHz while awake, but in cc6 core sleep for 50% of the sample period. The effective frequency of this core would be 2GHz. Intended to give a feel for how often the cores are using aggressive power management capabilities that aren’t immediately obvious (e.g. clock or voltage changes).
Various voltages and clocks, including: SoC voltage, DRAM voltage, fabric clock, memory clock, etc.
Ryzen Master v2.0.2.1271
Users will get a small preview of the SDK's capabilities by updating AMD Ryzen Master to version 2.0.2.1271, which now includes the new Average Core Voltage API for 3rd Generation Ryzen processors.
Addressing Reliability Worries
Following the uncovering of lower than expected boost frequencies in the most recent AGESA, concern was raised in an unofficial capacity by 3rd parties working with motherboard manufacturers in the development of new performance-grade BIOSes for Ryzen. It was alleged that AMD deliberately reduced operating voltages under maximum boost frequency states to thereby increase the expected life-span of the CPU, echoing very early criticism levied at the platform's operating voltages. TSMC's 7nm process, it was argued, wasn't suited to such high voltages over extended periods.
AMD's response is forthright:
We perform extensive engineering analysis to develop reliability models and to model the lifetime of our processors before entering mass production. While AGESA 1003AB contained changes to improve system stability and performance for users, changes were not made for product longevity reasons. We do not expect that the improvements that have been made in boost frequency for AGESA 1003ABBA will have any impact on the lifetime of your Ryzen processor.
Expected Delivery
Regular updates to the AGESA have been welcomed by enthusiasts keen to maximise the CPU performance of the still relatively unexplored Zen 2 architecture, but there has to come a point where the frequency and significance of updates starts to concern mainstream and novice users. For all the tools available today, updating the BIOS is a process not without risk. With every update there is a chance the end-user bricks their system, particularly is the motherboard is without dual-BIOS functionality. Such is often the frustrating price for early adoption of a new platform but AMD need to be mindful that this entire process is eroding rather than building up trust.
AGESA 1003ABBA is now in the hands of motherboards manufacturers who are adapting BIOSes as we speak. The typical turn-around time for development and validation is 2-3 weeks, so we should expect finished variants before the end of September. Hopefully this will not impact the Ryzen 9 3950X's anticipated September launch.
The AMD Monitoring SDK should see a public release before the end of September. Click here to download the latest version of AMD's Ryzen Master utility.
SOURCES: debauer RYZEN 3000 Boost Survey Results, AMD Community Update #6
