Intel ties AMD for most reliable CPUs in 2025 system builder report — Nvidia's Founders Edition GPUs dominate with the lowest failure rates
These parts have the least number of failures in Puget System's builds.
Get 3DTested's best news and in-depth reviews, straight to your inbox.
You are now subscribed
Your newsletter sign-up was successful
Puget Systems releases an annual reliability report detailing the reliability of the components that it uses for its systems. According to the data, Intel and AMD produce the best CPUs and are tied in reliability. The report also presents data on which motherboards, graphics cards, and memory modules have the lowest failure rates.
The tables have turned for 2025, as the company reports that the Intel Xeon W family of processors hasn’t seen a single failure this year and only one the previous year. For consumer CPUs, AMD Ryzen 9000 and Intel Core Ultra 200S have neck-and-neck failure rates: 2.52% for AMD Ryzen and 2.49% for Intel Core Ultra. Beyond these chip families, one SKU stood out: the Intel Core Ultra 7 265K, with a 0.77% failure rate. AMD’s X3D processors also delivered excellent performance, with a 1.51% failure rate, making them more reliable than the rest of AMD’s 9000-series CPUs.
Asus ProArt and TUF GPUs took the reliability crown last year, but because the company did not sell enough cards from a specific product line this year, Puget Systems chose the most dependable GPU manufacturer instead. Nvidia Founders Edition GPUs are at the top of the list, with a 0.25% failure rate, while Asus and PNY followed closely behind with 0.40% and 0.45%, respectively. The system integrator also said that Nvidia’s professional RTX Ada Generation and RTX Pro Blackwell GPUs were equally reliable, with low failure rates — except for the high-wattage RTX Pro 6000 Blackwell Workstation Edition.
Motherboards typically have higher failure rates than CPUs and GPUs, around 5% to 6%, as Puget Systems notes, because they have complex systems with multiple failure points. Nevertheless, two motherboards stood out: Gigabyte B860M Aorus Elite WiFi 6E Ice, which had zero failures (though the sample size was just 100 units), and Asus Tuf B850M-Plus WiFi, which had only a single failure over the year.
As for memory and storage, Kingston received the overall reliability award, with its ValueRAM DDR5-5600 32GB reporting a 0.9% failure rate, while its KC3000 SSD failed just 0.22% of the time. Nevertheless, Samsung also received a mention for one SKU—the Samsung 870 QVO 8TB SATA SSD—which reported no failures in 2025 and a 0.19% overall failure rate—significantly below the 0.74% average SSD failure rate for Puget Systems. For power supplies, the company has primarily used Super Flower Leadex units, which have a 0.47% failure rate, except for SFF builds. For these systems, Corsair PSUs had zero failures in both in-house testing and field deployment.
The company said these numbers do not reflect the overall industry, given its limited sales relative to the market. Furthermore, every unit it builds undergoes extensive testing, enabling it to detect component failures that would otherwise go unseen and remain unreported. Still, these numbers indicate the quality and reliability of PC components, so if you want to build a desktop computer that is more likely to last several years, consider selecting parts that Puget Systems uses.
Follow 3DTested on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.
Get 3DTested's best news and in-depth reviews, straight to your inbox.

-
Gururu So Intel and nVidia have won every year except 2024 (AMD Threadripper). No statistics this year, kind of lame.Reply -
ezst036 I wonder if this includes or excludes 12VHPWR.Reply
If they can fix the power connector, and send the very same card right back to the customer, is that considered an actual GPU failure? I'm not entirely sure it could be. 🤔
"The GPU" didn't fail, just the power delivery for a temporary moment. -
spongiemaster Reply
Puget Systems is an OEM builder. I have yet to see an OEM system with a burnt connector.ezst036 said:I wonder if this includes or excludes 12VHPWR.
If they can fix the power connector, and send the very same card right back to the customer, is that considered an actual GPU failure? I'm not entirely sure it could be. 🤔
"The GPU" didn't fail, just the power delivery for a temporary moment. -
ezst036 Reply
That is a good point. That's very interesting.spongiemaster said:I have yet to see an OEM system with a burnt connector.