Why you can trust Tom's Hardware
The 850 B3 achieves good overall performance, despite the 3.3V rail's bad transient response (hopefully EVGA fixes that as soon as possible, since the rail failed all of our tests). But regardless of a strong showing in most of our benchmarks, this PSU appears to have a serious flaw: our retail sample finished its OPP test and wouldn't fire back up. Apparently, something broke on the secondary side. Although we're glad the 850 B3 didn't blow up like the 450W model, this is our second (so far) B3-series power supply to die on the bench. And given that the 850 B3 is the family's flagship, we are worried about the platform's reliability.
EVGA still hasn't answered our concerns about the 450 B3 (or offered a second sample), and now we have another one dead B3 PSU. We can only guess that the over-power protection's triggering point is set too high in this unit as well. Although the PSU shut down under high stress, the fact it won't turn back on suggests an on-board component failed before the protection feature kicked in.
These aren't press samples we're working with here since EVGA didn't send any to us. That means we can be confident we don't have a hand-selected PSU. Beyond the 450 and 850 B3, we also bought the other models in this family and plan to explore whether the issue of premature failure plagues them as well. We are highly committed in our work, and when a company chooses not to send out a product that could be popular with our audience, we go out and get it ourselves.
The 850 B3 could be a very good PSU. But the fact that it stopped working once we pushed to find the OPP feature's triggering point makes this model much less appealing. Despite a low efficiency rating, EVGA's 850 B3 packs some nice features including a semi-passive mode and Japanese capacitors on the main PCB. Its power-good signal is accurate, and the hold-up time we measured is long enough to easily meet the ATX spec's demands. Fully modular cabling is another compelling advantage, along with lots of available PCIe connectors.
EVGA did make one egregious mistake on the cable selection, though: it only arms the 850 B3 with one EPS connector. We typically expect every PSU with more than 750W of capacity to have two, and we're fond of the 650W units (such as Aerocool's Project 7 650W) with more than one EPS connector.
Until we get to the bottom of failures across EVGA's B3 family (something that would almost require the company's cooperation), we do not recommend the 850 B3. While it's apparently less likely to treat you to a fireworks show (unlike the 450 B3), an 850W PSU selling for $100 that dies after OPP testing is not good. Super Flower's platform needs fine-tuning to improve its reliability (and safety). As of right now, it appears the 850 B3 wasn't tested extensively enough before it goes into mass production. After all, tuning takes time, increasing production costs, and that's not a good thing when it comes to budget-oriented PSUs.
MORE: Best Power Supplies
MORE: All Power Supply Content
Current page: Final AnalysisPrev Page Performance, Value, Noise & Efficiency
Stay on the Cutting Edge
Join the experts who read Tom's Hardware for the inside track on enthusiast PC tech news — and have for over 25 years. We'll send breaking news and in-depth reviews of CPUs, GPUs, AI, maker hardware and more straight to your inbox.
Aris Mpitziopoulos is a Contributing Editor at Tom's Hardware US, covering PSUs.
Most upcoming Arrow Lake CPUs will leverage TSMC nodes for compute, Intel 20A only for Core i5 and lower: Leak
Apple's MacBook Air gains M3 upgrade and support for two external displays — both 13-inch and 15-inch get improvements
Hardware hacker runs GTA Vice City on a router with a little help from an eGPU and Debian Linux
There's a typo in the Final Analysis: "Although we're glad the 850 B3 didn't blow up like the 450W model, this is our second (so fat) B3-series power supply to die on the bench."Reply
Shame on the PSU.
(so far) Thanks!Reply
Part of what I got from this review and the associated charts is that my next psu will likely be a Corsair unit.Reply
It seems like the general wisdom is still "Don't buy a PSU unless it has a 7- to 10-year warranty." The best products from the best OEMs always come with those warranties lately.Reply
If you ask me the warranty isn't an indication from the moments it is 5 years long. On the other hand when a company avoids sending some specific PSUs for review, then the whole thing looks suspicious.Reply
20217010 said:If you ask me the warranty isn't an indication from the moments it is 5 years long. On the other hand when a company avoids sending some specific PSUs for review, then the whole thing looks suspicious.
I don't necessary mean that longer warranty = better product. I just mean that there are some OEMs that are better than the rest, and their best PSUs are the best on the market, and all those best PSUs happen to have 7- to 10-year warranties on them. So if you see that warranty on a PSU right now, you can reasonably assume that it is one of the best models from one of the best OEMs.
I don't totally agree with the "Pros" and "Cons":Reply
* Six 8-pin PCIe-connectors are supposed to be capable of delivering up to 900W. This PSU can't handle that.
* The efficiency is as advertised. That's neither a "Con" nor a "Pro". (In fact I'm impressed with the efficiency at 20W load. Expected worse.)
* Sleeve bearing is what makes the fan relatively quiet, so it's not a "Con" per se. Had it been a ball bearing fan you'd written "High noise level" as a "Con" instead.
Efficiency is a con in general. It is lowReply
If the fan had DBB and an optimal fan profile then it wouldn't be a con. A sleeve bearing fan in a 100 buck PSU doesn't look good and those type of bearings are suited for horizontal installation so they should normally be avoided in PSUs.
Seasonic focus 850 currently for $10 more than this settles itReply
Hi Aris, congratulations for the wonderful review and all of your work.Reply
I've seen all of your reviews, but one thing I do not understand compared to other review sites and that's related to the advanced transient tests. For example, the be quiet! Pure Power 10 600W PSU, in Advanced Transient Test at +12V, 20% - 20ms has a voltage drop of 0.211V. At 50% load 20 ms is a voltage drop of 0.222 volts. Instead, at ardOCP, the same PSU records a 0.520 volt voltage drop at 20% load - 10 ms and 0.540 volts at 50% load – 10ms. And that's a huge difference. And I do not think this difference is because of the reduced time from 20ms to 10ms because at 1ms wich is tougher the drops it's not as high. What is wrong with ardOCP methodology?
Is it possible that a quality PSU like this to have a voltage drop in load so high (0.52 Volts)? And that's just an example. At all the PSU’s I've seen on both sites are differences like this. Same as when you were doing only 50ms or 200ms. Is there any explanation for these figures that do not coincide?