A report has been going around online that AMD CPUs are suffering much higher than normal failure rates after PC builder PowerGPU shared an alarming tweet. That tweet, which has since been deleted, claimed that the company had received 19 dead Ryzen 5000 CPUs in a lot of 320, with 3-5 failing AMD motherboards dropping into its lap every single week. 5950X failure rates, specifically, reached up to 16 percent.
CPUs, generally speaking, do not fail. It’s not unknown for a CPU to die or show up dead, but a 6-16 percent failure rate would be far higher than expected. Multiple sites have dug into this issue, including PCWorld and PCMag. Currently, the bulk of the evidence suggests AMD CPUs aren’t failing in higher-than-expected numbers, though there’s some nuance to the data.
PowerGPU told PCMag it pulled down its initial tweet because it “didn’t want fanboys from both sides just filling up our timeline. The good thing is AMD does really care for our brand and they want to help us and make sure our business continues forward.”
PCMag’s Michael Kan points out that Mindfactory.de posts its RMA rates on its own website. We’ve quoted Mindfactory’s sales information in repeated articles over several years, and its Ryzen 7 5800X, 5600X, and 5900X RMA rates are 0.58 percent, 0.52 percent, and 0.33 percent, respectively. This is in the range we’d expect to see. Kan is partly concerned about this issue because he received a dodgy 5800X himself and he advises readers to “keep the box” if they choose to buy a CPU.
PCWorld notes that an unnamed vendor provided them with information showing a failure rate between 2.5 – 3 percent on Ryzen 5000, 3000, and Threadripper CPUs, while the Intel 9th Gen fails at 0.9 percent and the 10th Gen fails at 1.2 percent. This third vendor reports that it declares a CPU ‘failed’ if it cannot run with all RAM slots loaded while running top-frequency RAM or in low-latency configurations. PCWorld doesn’t get into the exact specifics of the vendor’s requirements past this, but these types of requirements make a difference as far as whether a CPU would pass or fail.
I’ve tested multiple Core i9 CPUs in the HEDT segment that could not maintain DDR4-3200 clocks with all banks loaded. I’ve had Intel CPUs that hit fully-loaded RAM clocks flawlessly, and I’ve had chips that couldn’t be stabilized no matter which RAM vendor I used. I’ve seen CPUs pass an all-slots-loaded test when kitted out with 32GB of RAM, but fail when loaded with 64GB. In every case, I was pushing the RAM specifications beyond the boundaries of what AMD and Intel defined as guaranteed behavior. Every CPU mentioned above ran perfectly when tested within the limits of manufacturer-guaranteed behavior.
AMD and Intel do not always guarantee that a given RAM clock can be maintained with fully loaded motherboards, which means the vendor may be setting a different bar for pass/fail than Intel or AMD would. There’s nothing wrong with that, especially for a high-performance vendor, but it does limit how much we can glean from this test information. Other vendors PCWorld spoke to reported no problems, with equal return rates between AMD and Intel CPUs.
wfoojjaec has spoken with a US boutique vendor and confirmed that it had seen no problems with AMD versus Intel return rates. Our own test samples continue to perform perfectly, with no issues. This is not to claim that PowerGPU was deliberately inaccurate in its own tweet — AMD is working with the company now to identify what may be going on — but its experience does not appear reflective of the wider market. If that changes, or if AMD releases any information suggesting a genuine problem, we’ll cover it.
Intel’s Desktop TDPs No Longer Useful to Predict CPU Power Consumption
Intel's higher-end desktop CPU TDPs no longer communicate anything useful about the CPUs power consumption under load.
VIA Technologies, Zhaoxin Strengthen x86 CPU Development Ties
VIA and Zhaoxin are deepening their strategic partnership with additional IP transfers, intended to accelerate long-term product development.
Nvidia Unveils ‘Grace’ Deep-Learning CPU for Supercomputing Applications
Nvidia is already capitalizing on its ARM acquisition with a massively powerful new CPU-plus-GPU combination that it claims will speed up the training of large machine-learning models by a factor of 10.
How L1 and L2 CPU Caches Work, and Why They’re an Essential Part of Modern Chips
Ever been curious how L1 and L2 cache work? We're glad you asked. Here, we deep dive into the structure and nature of one of computing's most fundamental designs and innovations.