A major goal of this website is to provide quality and reliable power supply unit (PSU) reviews. As a computer system component, similar to the CPU, the PSU also has its technical specifications and performance measurements. We believe that most readers of this article understand the important role that PSU plays in computer system. In order to make it easier for our readers to “digest” the reviews, it is necessary to have a user manual as the guidance, so we have it here. In this article, we will introduce the equipment and methodology we used, to help the reader, you, to establish a basic concept for understanding each technical specification and how does these terms affect the operation of the power supply system behind their measurements.
Good equipment is the cornerstone for delivering accurate reviews. Every computer component has a suitable testing software and/or instrument, for example, the CPU and GPU can be tested with some benchmark software. But this becomes much more complicated when it comes to test PSU. In addition to having the proper knowledge of electronics, expensive and precise external electronic load and other instrumentations are required to comprehensively and accurately test various performance measurements.
The Chroma power supply test system is widely used by manufacturers in the PSU industry in their R&D, testing, and production processes, as well as Intel’s testing laboratory. The Chroma 6000 ATS was used for our early reviews. Starting with the EVGA SuperNOVA 1600 G2 reviewed at the end of 2016, we upgraded to the more advanced Chroma 8000 SMPS ATS system.
Since December 2019, another set of Chroma 8000 test system has been added and used together with the previous set of Chroma 8000, the equipment strength has reached a new level.
The programmable high-performance Chroma testing system can simulate various input and output states on the PSU under test, then we use appropriate instruments to measure different testing subjects. The electronic load module is the key component of this system, it simply determines the maximum power of PSU that can be tested. The electronic load module in our Chroma 8000 system consists of nine 63640-80-80 programmable DC electronic load (400W each), two 63630-80-60 programmable DC electronic load (400W each), and one 63610-80-20 (100W ×2), which allows us to have up to 4400W of load, two sets are up to 8800W load capacity.
Our Chroma 8000 system also includes one Chroma 66202 digital power meter, one Agilent 34401A digital multimeter, one Chroma 61605 AC source (4KVA), one Chroma 80611 timing/noise analyzer, one Chroma 80612 OVP/short circuit tester, one Chroma 80613 ON/OFF controller, and an industrial PC.
Our external instrumentations include the following: one Tektronix MDO3014 oscilloscopes, and one Tektronix TDS3014 oscilloscopes to analyze ripple, hold-up time and other subjects; one Suwei SW826 non-contact tachometer (resolution: 1RPM/min) to measure speed of fans; one FLIR One Pro thermal camera (accuracy: ±3°C or ±5%, resolution 0.1°C) to capture thermal images.
With the help of these expensive toys, we are able to perform the comprehensive test in accordance with the “Desktop Platform Form Factors Power Supply Design Guide” developed by Intel, which includes the often said “Intel ATX12V” specification.
Performance, Specification & Methodology
We will introduce each specification and methodology based on the structure of our PSU reviews. The test methodologies reference to “Intel Desktop Platform Form Factors Power Supply Design Guide”, which is a common guideline for R&D, production, and testing in PSU industry. The test program is developed by our tester “fc (fcpowerup)”.
Our PSU reviews are structured as follows:
1.Conclusion & Rating
2.Packaging, Contents & Exterior
3.Teardown & Component Analysis
4-1. Voltage Regulation
4-3. No Load & Light Load
4-4. Fan Speed & Temperature
4-5. 5V Standby
4-6. Cross-Load Tests
4-7. Ripple & Noise
4-8. Inrush Current & Hold-Up Time
4-9. Dynamic Tests
5. FCP Graphics Compatibility Certification
1. Conclusion & Rating
The content replaces the previous introduction chapter, introduces the market background of product appearance, performance, positioning, price, quality and after-sales, etc., and evaluates product performance, work noise, material quality, whether it passes FCP graphics card compatibility test certification, etc. to sum up.
Introduce the market background and market positioning of the brand and PSU itself.
2. Packaging,Contents & Exterior
Introduce and analyse the packaging, exterior design and build quality, cables, and other accessories.
3. Teardown & Component analysis
Disassemble the PSU to examine and analyse the interior design and components.
4. Performance Tests
Please see the detailed introduction in the next section.
5. FCP Graphics Compatibility Certification / FCP Gaming Ready Certification
Subsequent we will add a separate introduction article, mainly for the power compatibility test of the current high-power graphics card, build a PC game platform, use the graphics card for testing, including large 3D games and mainstream 3D test procedures, test whether the power supply can run safely and reliably. .
5. Conclusion & Rating Give conclusion and rating according to the PSU’s exterior design, performance, build quality, acoustic noise, market positioning, price, and warranty.
Performance Tests in Details
4-1. Voltage Regulation
Different hardwares in computer request different operating voltages, usually CPU and graphic cards use +12V, motherboard and hard drive use +5V and +3.3V. Intel has clear requirements in their ATX12V specification: at all loads, the regulation ranges of +12V, +5V, +3.3V and +5VSB shall always remain within ±5%, and the regulation range of -12V is ±10%. Since current computer system has little use of -12V, -12V output is optional.
Beside the load regulation in Intel ATX standard, we also test voltage accuracy in our reviews.
Voltage accuracy refers to the maximum deviation of the voltage from the standard value over the entire load range. Generally, the factory setting of voltage output will be slightly higher than the standard voltage to offset the transmission loss along cables.
Load regulation refers to the changes of output voltage depends on loads over the entire load range. The smaller the variation, the stronger the voltage stability.
For example, the following graph shows the 5VSB of a PSU. Its standard voltage is 5V, the maximum measured voltage is 5.118V, the minimum measured voltage is 5.074V. Thus, the calculated maximum voltage deviation is 2.36%, and the load regulation is 0.88%.
Common efficiency certifications include Blue Angel, ENERGY STAR, and 80 PLUS. Especially 80 PLUS, which has affected the entire industry in the past decade, is already the benchmark for PC PSU. The following chart is a simple demonstration of 80PLUS requirements,
and we will have an article to talk about 80PLUS in detail in the future In “What is 80 PLUS, can you replace the 80 PLUS Gold Power Supply with the 80 PLUS Titanium Power Supply to earn the electricity bill cost back? “In this article, I have introduced this certification in detail.
Our efficiency test is different from other common certifications such as 80PLUS, which based on percentage load of rated maximum power. We test efficiency in fixed steps of load as the following table to cover both light loads, typical loads, and heavy loads and match the actual usage scenarios. Readers can find the corresponding power usage according to their hardware configuration. The efficiency is tested in both 230VAC 50Hz and 115VAC 60Hz.
The testing results are included in the “Input & Output Test” table, and are also display as efficiency curve.
Intel ATX12V specification has efficiency requirements for 20%, 50%, and 100% loads are 65%, 72%, and 70%, respectively. These requirements are easy to achieve for current mainstream PSUs.
4-3. No-load & Light Load Test
No load test measures the power consumption of PSU itself with zero output. We test the idle power consumption by shorting PS-ON and GND to turn the PSU on without having any output load. The PSU should has no damage or hazardous condition in the situation.
Light load test examine the performance of PSU when the PC is idle with desktop or when it is lightly used. We evaluates the voltages, efficiency, fan speed of the PSU in 5 load states: CL1 (approx. 12W, refer to 4.5 Cross-Load section), 30W, 50W, 75W, and 100W.
4-4. Fan Speed, Noise, Temperature
Fan speed is tested over the entire load range.
On June 1, 2020, three noise tests were added, namely 100W desktop application, 400W game mode, corresponding to the full load mode of the power supply. The power supply under test uses a zero-noise resistive load. In a low-noise environment, the specified power is output for more than 10 minutes. After the fan speed stabilizes, the power supply noise is measured at a position 10 cm away from the power supply air inlet.
After generating the FFT spectrogram, you can display the noise value (A-weighted) of the power supply, and you can analyze the high-frequency noise frequency and amplitude between 10kHz and 20kHz, that is, determine whether the power supply has high-frequency howling.
Temperature is captured by thermal camera after the PSU is fully loaded with 230VAC input after 10 minutes.
4-5. 5V StandBy
We test 5VSB according to the requirements of Intel ATX12V specification v2.3 and ErP Lot 6 2013 regulation. Intel ATX12V specification v2.3 requires the efficiency of 5VSB must be greater than or equal to 50%, 60%, and 70% at 100mA, 250mA, and more than 1A loads, respectively. ErP Lot 6 2013 regulation requires the power consumption to be less than 0.5W at 0A output and the efficiency of 5VSB must be greater than 45% at 45mA.
Currently we test 5VSB performance according to the following current load table.
4-6. Cross-Load Test
The cross-load test takes ideas from Intel ATX12V specification v2.4 and SSI EPS12V Power Supply Design Guide v2.92, and combines the concepts of high-end platforms and low-power platforms.
We design seven states for cross-load test:
CL1 – Light Load: Test the voltage stability and efficiency at very low total load.
CL2 – Focus on +5V/+3.3V: Set +5V, +3.3V rails with maximum load and +12V rail with light load, to simulate the simultaneous startup of multiple hard drives.
CL3 – Heavy Load: Set +12V, +5V, +3.3V rails with maximum load, to simulate heavy total load.
CL4 – Focus on +12V: Set +12V rail with maximum load and +5V, +3.3V rails with light load, to simulate overclocking or gaming with single SSD.
CL5 – 12V Max: Test the performance of extreme case where +12V rail is at maximum load and +5V, +3.3V have no loads.
CL6 – 5V Max: Test the performance of extreme case where +5V rail is at maximum load and +12V, +3.3V have no loads.
CL7: 3.3V Max: Test the performance of extreme case where +3.3V rail is at maximum load and +12V, +5V have no loads.
The cross load mainly examines the stability of the output voltage. Within the range of ±5% required by the Intel ATX12V specification, the smaller the voltage deviates and load regulation, the better the accuracy and stability.
4-7. Ripple & Noise
The output voltage fluctuates like the water ripple when observed with an oscilloscope. Ripple and noise is the AC component that lay on the DC output of the PSU. A part of ripple and noise is the residual AC component after rectifying and filtering, another part is the ripple and switching noise generated by the transistor in the circuit. Excessive amounts of ripple can interfere with digital circuits and affect the stability of the circuit.
According to Intel ATX12V specification v2.52, the maximum output ripple and noise (p-p) is 120 mV for +12V, 50mV for +5V, 50mV for +3.3V, 120 mV for -12V, 50mV for +5VSB.
Our ripple test focus on +12V, +5V, +3.3V, and +5VSB, and does not test -12V. We follow the Intel ATX12V specification to connect the decoupling capacitors in parallel to the measuring points on fixture, and use an oscilloscope with 20 MHz of bandwidth to measure.
We test ripple and noise with seven states: 50W to simulate idle with desktop, 100W to simulate daily use, 300W to simulate gaming with single graphic card, maximum total load and maximum load of each rail (+12V, +5V, +3.3V) to test the performance under stresses. Usually the ripple is the largest when the PSU is at maximum load.
4-8. Inrush Current, Hold-Up Time
Inrush current is the instantaneous maximum input current drawn by the PSU at turn-on. Because the PFC capacitor needs to be quickly charged, the inrush current is considerably larger than the normal operating current. The smaller the inrush current, the better the performance. Excessive inrush current can damage components in PSU, such as the fuse, NTC thermistor, bridge rectifier, and AC switch.
Mainly to test whether the T2 time in the Intel ATX12V specification boot sequence is in accordance with the Intel specification is 0.2~20ms. If it is not in this range, the boot may not be lit. Use the full load state to power on, use the oscilloscope to measure the voltage for overshoot, mainly to solve some users’ concerns that the power supply may damage the motherboard, graphics card and other accessories.
Hold-up time is the time that the output voltages stay within required ±5% range after loss of AC input. According to the newest Intel ATX12V specification v2.52, the T5 (AC loss to PWR_OK hold-up time) must be greater than 16ms, in other words the PSU must be able to maintain PWR_OK (Power Good) signal more than 16ms after loss of AC source. Also T6(PWR_OK inactive to DC loss delay) must be greater than 1ms, which means the PSU must be able to keep DC output more 1ms after the PWR_OK signal stopped to keep other hardware running. In short, after loss of AC, PWR_OK signal remains > 16ms, voltage of DC output such as +12V/+5V/+3.3V remains > 17ms.
Sufficient PWR_OK hold-up time (T5) means that loss of AC within 16ms or switching to the uninterruptible power supply (UPS) will not cause the computer to shutdown or restart. Likewise, the DC voltage hold-up time (T5+T6), which is longer than the PWR_OK hold-up time, guarantees other hardware have time to take appropriate procedures, such as retracing the head in hard drives and power loss protection of SSDs.
The test setup for hold-up time is at maximum load with 264VAC input.
4-9. Dynamic Test
The above static test methods are designed to simulate the various conditions of the PSU when the computer power consumption is in a steady state. Assume that a computer steady consume 300W power when it is at maximum load, according to our test result, Corsair RM650x v2018 has 12.038V output for +12V rail, 9.2mV of output ripple, and 0RPM of fan speed at this moment.
However, in reality, the power consumption of the computer changes at every moment. For example, the load of CPU rapidly increases, causing the power limit to change from PL1 to PL2 and keep it for 10ms; When playing game, the power consumption of the graphics card may reach 300W or even higher, for a few milliseconds.
Traditional static test methods do not consider the dynamic changes of power load. Due to factors such as line compensation, resistance of components and input impedance, the output voltage of the PSU generally decreases as the load increases. When the load is removed, there is a process for the output voltage to recovery to steady state.
According to the following graph, when the load shift from I/R-1 to I/R-2 (called “load transient”), the output voltage of PSU drops from Vs-1 to Vs-2, like a step down. It takes time for the PSU to response to change of load, as a result, there actually is a process of “overshoot/undershoot – recovery”. During this process, the peak voltage change is usually larger than the regulated output voltage in magnitude. In other words, when the load increases from I/R-1 to I/R-2, the output voltage first drops to a lower voltage Vpk-1, and then gradually rise up to the steady voltage Vs-2. Conversely, when the load drops from I/R-2 to I/R-1, the output voltage will rise from Vs-2 to Vs-1. This process also has a higher overshoot voltage Vpk-2 than Vs-1.
DC Output Transient Test in ATX12V specification requires the load changes from 50Hz to 10kHz, and the output voltage keeps within ±5%. Currently we only test +12V which’s load transient usually is large in magnitude and high in frequency, and ±5% for 12V is ±600mV.
Our test first verify that the PSU should not shut down, restart, or fail. Then we measure the peak voltages for overshoot and undershoot. Beside that, we also measure the time that the PSU needs to regulate the voltage after load transient occurred Tr-1 and Tr-2.We call them the voltage recovery time (or voltage rebuild time), and we believe voltage recovery time directly reflects the dynamic performance, although Intel’s ATX12V specifications has no requirement of such time.
Based on the current CPU and graphic card in market, our dynamic test stage-2 uses following setups:
12V2: 1A ↔ 9A, step size 8A, rise rate 1A/μs, in equivalent to the mean difference between the continuous current and peak current of the CPU defined in ATX12V 2.52.
12V3: 1A ↔ 13A, step size 12A, rise rate 1A/μs, in equivalent to a 150W graphic card switch cramp between maximum load and idle.
12V3: 1A ↔ 13A, step size 12A, rise rate 1A/μs, in equivalent to a 150W graphic card switch cramp between maximum load and idle.
For small and medium wattage PSU:
12V2 and 12V3 are loaded synchronously, simulating the simultaneous operation of CPU and graphics card. The accumulated basic Load of 12V is 3A and the dynamic Load is 20A/240W, which is equivalent to bringing a mid-end graphics card to play games or run Benchmark.
For 1kw+ wattage PSU:
12V2, 12V3 and 12V4 are loaded synchronously, simulating the simultaneous operation of CPU and graphics card. The basic Load of 12V is 4A and the dynamic Load is 32A/384W, which is equivalent to bringing a high-end flagship graphics card to play games or run Benchmark.
We test dynamic performance with 5 different changing rate: 10Hz, 50Hz, 100Hz, 1kHz, 10kHz. We chose 50Hz and 100Hz as the main test object to show the performance difference in the review. Normally 1kHz and 10kHz will not be mentioned, unless important problem, like the overshoot/undershoot voltage exceeding the requirement, occurs.
Take the Corsair RM650x v2018 as an example for dynamic test:
By default, the included cables are used in the test. The 12V cables come with Corsair RM650x v2018 feature inline capacitors. In dynamic test, neither shut down nor restart has occurred, and the output voltage does not exceed ±5%. The average voltage recovery time is 2.5ms, which is slightly slower than medium level in our current data collection.
Currently, dynamic test result is not being rated.
4-10. Protection Features
Protection function tests currently include Over Power Protection (OPP), Over Current Protection (OCP), and Short Circuit protection (SCP).
Over Power Protection (OPP): The power supply gradually increases the output power from near full load until the power supply fails to work in the protection state. Finally, the power supply must be able to cut into the protection state. If the power supply does not have OPP protection, it may blow up or damage other hardware.
Over Current Protection (OCP): A mandatory requirement for the Intel ATX12V specification. Current protection should be designed to limit the current to operate within safe operating conditions.
Over current protection schemes, where only the voltage output that experiences the over current event is shut off, may be adequate to maintain safe operation of the power supply and the system; however, damage to the motherboard or other system components may occur. The recommended over current protection scheme is for the power supply to latch into the shutdown state. PSU connectors, cables and all other components should not be melted or damaged prior reaching to the OCP trigger.
Short Circuit Protection (SCP): An output short circuit is defined as any output impedance of less than 0.1 ohms. The power supply shall shut down and latch off for shorting the +3.3V DC, +5V DC, or +12V DC rails to return or any other rail. The +12V1 DC and 12V2 DC should have separate short circuit and over current protection. Shorts between main output rails and +5VSB shall not cause any damage to the power supply. The power supply shall either shut down and latch off or fold back for shorting the negative rails.
example: This is the protection features test result of ASUS ROG Strix 750W Gold，No Load Operation and Surge & Inrush Protection can also be judged to be functional according to the above tests and disassembly.
|过流保护， OCP (Over Current Protection)||12V: 86.5A (139.5%) 5V: 28.5A (142.5%) 3.3V: 28A (140%)|
|过功率保护， OPP (Over Power Protection)||1044.53W (139.27%)|
|短路保护， SCP (Short Circuit Protection)||12V: ✓ 5V: ✓ 3.3V: ✓|
|空载保护， NLO (No Load Operation)||✓|
|浪涌保护，SIP (Surge & Inrush Protection)||MOV压敏电阻、NTC热敏电阻、继电器|
This article is our PSU test equipment and methodology. We will continue to upgrade our equipments and improve our methodologies to match the evolving industrial specifications and user demands.
2020-06-01, Version 1.6
- Chapter 4-4, added noise test, and can analyze the frequency and amplitude of high-frequency howling of power supply.
2019-12-20, Version 1.5
- Add 80 PLUS Article “What’s 80 PLUS?” In 4-2. Efficiency.
- Add Another Chroma 8000 SMPS ATS system.
2019-08-14, Version 1.4
- Section 4.8 add the rise time test.
- Added 4.10 Protection Features test, including OPP, OCP and SCP Features test.
- The test method was updated from version 1.2 to version 1.4 on the same day, and the English version was also updated to version 1.4.
2019-08-14, Version 1.3
- NZXT E850 Review began to add the FCP graphics card compatibility test certification Beta (2019-07-25).
- ASUS ROG1200P Thor Review began to adjust the article chapter order (2019-01-25). The product introduction and rating and summary are merged and adjusted to the first page, so that the reader can see the conclusion on the first page.
- Add English Version v1.2.
- Add English Version v1.1.
- Add English Version.
- Add 4-9 Dynamic Test.
- 4-5 “ErP Lot 6 2013 regulation requires … the efficiency must be higher than 45% at 45mA.”
- 4-8 Inrush current use 264Vac input.