Update: AMD has responded to claims that some motherboard manufacturers may be tweaking power telemetry data and is investigating any possible effect it may have on a CPU’s lifespan. The company has said that it expects there to be no long-term impact, however.
Here’s the full statement from AMD sent to our friends at Tom’s Hardware:
“We are aware of the reports claiming that select motherboards may be under-reporting certain power telemetry data that could alter the performance and/or behavior of AMD Ryzen processors under certain conditions,” AMD’s statement says. “We are looking into the accuracy of these reports.
“We want to be clear with our customers: AMD Ryzen processors contain a diverse array of internal safeguards that operate independently of external data sources. These safeguards enforce the safety and reliability of the processor during stock operation. Based on our initial assessment, we do not believe that altering external telemetry in the manner described by those public reports would have a material impact on the longevity or safety of a user’s processor.”
Original story: Some motherboard manufacturers are reportedly feeding AMD Ryzen CPUs with biased telemetry data in order to bolster system performance, and that has potential to push some chips way beyond their factory-recommended specifications. But not to worry, the latest version of HWiNFO will tell you what motherboard manufacturers won’t.
To understand what’s going on here, you need to understand how AMD Ryzen reports power, and what determines a chip’s power limits. Luckily, there’s a detailed post over on the HWiNFO forums by overclocker The Stilt (via Overclockers) that explains all.
As The Stilt explains, “Ryzen CPUs for AM4 platform rely on external, motherboard-sourced telemetry to determine their power consumption. The voltage, current, and power telemetry is provided to the processor by the motherboard VRM controller through the AMD SVI2 interface. This information is consumed by the processors power management co-processor, that is responsible for adjusting the operating parameters of the CPU and ensuring that neither the CPU SKU, platform, or infrastructure specific limits are being violated.
“…it is the motherboard manufacturers responsibility to find the correct value for their motherboard design through the means of calibration, and then to declare it properly in AGESA during the BIOS compile time. In case the motherboard design specific correct value differs greatly from the declared value, there will be a bias in the power consumption seen by the CPU.”
The issue here is that motherboard manufacturers can, if they so please, feed data that doesn’t exactly match what is expected from any given chip, therefore allowing a chip to run above or below its rated power limit—that’s almost always above the rated limit, The Stilt says.
Within the new version of HWiNFO, v6.27-4185 Beta, this proposed leniency is reported under the value “Power Reporting Deviation”, which will report a value around 100% (100% being as close to an unbiased value as you can get).
Any deviation from 100% means your CPU is running a little below, or more likely above, its rated power spec.
In order to report a stable value The Stint recommends running the Cinebench R20 NT benchmark, with the HWiNFO sample rate set to less or equal to 1000ms.
Right now, only a few MSI motherboards offer user-configurable bias. Any other motherboards affected will need to have the functionality patched in, or the bias patched out, with a forthcoming BIOS update.
How much of an impact will a generous power envelope have on performance? It’s essentially a mild overclock, albeit one that your ‘stock’ CPU doesn’t know has been applied, but any performance benefit gained by such a change all depends on how great the deviation is per motherboard.
Another question put fourth is whether this could shorten a chip’s lifespan, as overclocking can be known to do. My gut tells me that any effect on CPU lifespan will be almost entirely negligible, and necessitates using the chip for far longer than most are wont to do, but further testing is required on a per mobo basis to draw anything near-conclusive—and we’re probably not going to wait around for 10+ years to find out.