Test Setup

All measurements were performed using two Chroma 6314A mainframes equipped with the following electronic loads: six 63123A [350 W each], one 63102A [100 W x2], and one 63101A [200 W]. The aforementioned equipment is able to deliver 2500 W of load, and all loads are controlled by a custom-made software. We also used a Rigol DS2072A oscilloscope kindly sponsored by Batronix, a Picoscope 3424 oscilloscope, a Picotech TC-08 thermocouple data logger, two Fluke multimeters (models 289 and 175), and a Yokogawa WT210 power meter. We also included a wooden box, which, along with some heating elements, was used as a hot box. Finally, we had at our disposal three more oscilloscopes (Rigol VS5042, Stingray DS1M12, and a second Picoscope 3424), and a Class 1 Bruel & kjaer 2250-L G4 Sound Analyzer which is equipped with a type 4189 microphone that features a 16.6 - 140 dBA-weighted dynamic range. You will find more details about our equipment and the review methodology we follow in this article. We also conduct all of our tests at 40°C-45°C ambient to simulate the environment seen inside a typical system with a higher accuracy, with 40°C-45°C being derived from a standard ambient assumption of 23°C and 17°C-22°C being added for the typical temperature rise within a system.

Primary Rails Voltage Regulation

The following charts show the voltage values of the main rails, recorded over a range from 60 W to the maximum specified load, and the deviation (in percent) for the same load range.

5VSB Regulation

The following chart shows how the 5VSB rail deals with the load we throw at it.

Hold-up Time

Hold-up time is a very important PSU characteristic and represents the amount of time, usually measured in milliseconds, a PSU can maintain output regulations as defined by the ATX spec without input power. In other words, it is the amount of time the system can continue to run without shutting down or rebooting during a power interruption. The ATX specification sets the minimum hold-up time to 16 ms with the maximum continuous output load. In the following screenshot, the blue line is the mains signal and the yellow line is the "Power Good" signal. The latter is de-asserted to a low state when any of the +12V, 5V, or 3.3V output voltages fall below the undervoltage threshold, or after the mains power has been removed for a sufficiently long time to guarantee that the PSU cannot operate anymore.



Registered hold-up time was pretty long and easily meets ATX spec.

Inrush Current

Inrush current or switch-on surge refers to the maximum, instantaneous input-current drawn by an electrical device when it is first turned on. Because of the charging current of the APFC capacitor(s), PSUs produce large inrush-current right as they are turned on. Large inrush current can cause the tripping of circuit breakers and fuses and may also damage switches, relays, and bridge rectifiers; as a result, the lower the inrush current of a PSU right as it is turned on, the better.



Inrush current is low for a 750 W PSU with such a long hold-up time.

Voltage Regulation and Efficiency Measurements

The first set of tests revealed the stability of the voltage rails and the efficiency of the Fury 750G. The applied load was equal to (approximately) 10% - 110% of the maximum load the PSU can handle, with 10% steps.

We conducted two additional tests. In the first test, we stressed the two minor rails (5V and 3.3V) with a high load while the load at +12V was only 0.10 A. This test reveals whether the PSU is Haswell ready or not. In the second test, we dialed the maximum load the +12V rail can handle while the load on the minor rails was minimal.

Voltage regulation on every rail but the 5V rail was average. We would definitely like to see lower deviations on the +12V rail, but the platform obviously can't perform much better. That said, the 5VSB's very tight voltage regulation was a pleasant surprise, although we saw the EVGA NEX750G deliver the exact same performance on its 5VSB rail since both use the same platform. The results of our CL1 test, conducted in accordance to Intel's Haswell-testing methodology, were catastrophic. Two rails failed to keep their voltages in spec. We should note that the Fury managed to pass this specific test successfully with 100 W maximum combined power on its minor rails. FSP's engineers were definitely wrong to suggest 160 W as the combined power total for the minor rails since it makes passing Intel's Haswell compatibility test, our CL1 test, incredibly difficult. Long story short, the Fury-750G passed our CL1 by keeping all its rails in spec with 100 W on its minor rails, so it is Haswell ready with no more than 100 W combined power on those rails. But its official specs boast a total of 160 W, so we had no choice but to put 160 W on the minor rails to determine whether it performed accordingly.

Another downside is this unit's fan profile as it is very aggressive. The fan rotates at high speeds once the ambient temperature reaches 40°C, which has the fan produce a significant amount of noise at even typical loads, becoming incredibly loud at full load. We think Bitfenix could have used a slower fan by utilizing bigger heatsinks internally, which would significantly reduce noise output. Yet the strong fan helped the unit deliver more than its full power at over 45°C ambient, which was obviously also highly dependent on the unit's highly tolerable electronic components.