XPG Fusion 1600W Titanium Power Supply Review

The word extreme is thrown around in PC gaming hardware quite often, but rarely does it apply as much as it should to the XPG Fusion 1600 Titanium power supply. This is the most powerful and expensive power supply we've ever had for review. It's so powerful, that it would have blown up our old PSU load tester, but lucky for us, our new one can test up to 3800 watts (it's a beast)! XPG Fusion 1600W We first saw this power supply at Computex 2023, and we gave it an award for being one of the most impressive products we saw at the whole show. There it was being demonstrated running four RTX 4090 graphics cards on a heavily overclocked system, all day long, in a warm trade-show hall, pulling upwards of 2000 watts at anyone time. With 80 Plus Titanium efficiency, and rated for 1600W+, with the latest ATX 3.0 and PCIe 5.0 standards, it's ticking all the right boxes for those wanting to build something extreme. Gaming PCs are one thing, but this PSU is moving into the class of rendering, mining, AI, and other extreme PC tasks for those running multiple graphics cards, multiple processors and big overclocks 24/7. It features a Planar Transformer, GaN FET, a fully digital design with three MCUs to manage and monitor the digital LLC, there are multiple 12VHPWR connectors, and it has a full software suite to manage temperatures, monitor voltages, efficiency and performance history, and it's equipped with 100% Japanese capacitors rated at 105c, and will deliver 80 Plus Titanium efficiency, and is also Cybernetic Titanium certified too, comes with 8 major industrial protections, and a class-leading 12-year warranty. This isn't just looking like a PSU, this is looking like the King of the PSUs. Packaging & Accessories The box is big, really big, it may not look like it, but it is. I can't even lift it with one hand, it's stupidly heavy. It stands out too, it's like Clifford the Big Red Box. Around the back, there are various photos of the PSU, as well as a rundown of all the cable lengths, the rated watts/amps are well-detailed too. Down the side, you can see many of the key features, such as the customisable fan curve, ATX 3.0, and heavy-duty components. It's a thick box too, with double layers of cardboard on the outside, and then an inner foam layer for the PSU and a box for the cables. There are a lot of cables here, so the cable box alone is pretty huge, but given the size of the PSU, that's hardly surprising. All the cables are just crammed in the bag, but they are all cable-tied to keep them as neat as possible. There are two 12VHPWR cables included. As well as a pair of 12VHWR (from the PSU) to two 6+2 (for the GPU) cables. There are six more PCIe cables, as well as two sets of 4+4 CPU power cables. As well as a plethora of SATA and Molex connections for your peripheral components. A Closer Look This PSU is so big, that its 140mm cooling fan only covers two-thirds of the top of the unit! It's like someone built an extension on the back, and well... I guess they actually did. It's not the prettiest-looking power supply, which is fine, but it's not exactly ugly either, with some cool Fusion graphics on the side, and some lines pressed into the metal that add strength and a bit of style too. It's the same on the other side, albeit with the label inverted. This PSU uses a higher-rated C19 power cable, a larger square header, as C13 cables top out at around 1600W, so a thicker gauge of wire is required to deal with the higher sustained power delivery, which could easily exceed 2000W from the wall when pushing this PSU. Got connectors? Sure do, and plenty of them! It's all fairly standard stuff, except the Prime Port, which is a data cable to connect to your motherboard for software control. On the base of the PSU, you'll find the rated wattages and amps, and boy are they are some scary big numbers, with each 12V rail able to pull 50A each, and there are SIX of them; you can pull the full 1600W on the 12V rails alone. A Closer Look - Interior The fan is a HA13525H12SB-Z, one of the most common fans on many high-end PSUs on the market today, I see it in most of my PSU reviews. There's a large shield on here to prevent debris from getting into some of the higher voltage components, not uncommon on PSUs, but they're usually transparent not black. With the case opened up, it does look like there are two power supplies smashed into one here, and that actually isn't far from the truth, as I can see two main bulk capacitors, two transformers, two 5v and 3.3v converters, and well, to get to 1600-2000W, it does take a lot of hardware. There are a lot of additional boards in here, such as this active APFC modular which has a finned heatsink that's pressed up against the side of the PSU housing. That would restrict airflow and be of concern, but it's such an efficient PFC unit that cooling just isn't an issue. There are two transformers, again with another control board between them. Where possible the PSU appears to use solid-state connections or fixed control boards direct to the main PCB rather than use wires. Everything is far more shielded than usual too, with protective film between most components, albeit that's hard to see in some of these pictures. Where big voltages are concerned, pretty much everything appears to be doubled up on, from capacitors to converters. The AC Line-In uses thick wired, two shielding blocks, and again, it's just very robust looking. What's missing are big heatsinks, you usually see large fins on the bridge rectifiers, but these Infineon IPDQ60R010S7 units are so efficient, that they don't need cooling! There's a lot of that marshmallow-looking material too, it's some kind of thermal compound, but it's hard to the touch, acting like a bonding agent too. At the heart of the power supply are these two enormous bulk capacitors, each rated for 450V and 820 uF, making 900V and 1640 uF in total, with additional capacitors littered around the housing that are frankly too numerous to count without a full disassembly of the unit. There's a wall of PCBs stacked in rows, with banks of capacitors and chokes, ensuring the PSU can deliver both high power and low ripple on all the rails. There are hard metal interconnections rather than cables, and there are numerous black cables for voltage monitors and temperature sensors too, there's just so much going on here! Methodology & Testing At eTeknix we take the power supply testing procedure very seriously and have invested a lot of resources into acquiring the appropriate testing equipment. For all power supply reviews, we test the power supplies with dedicated power supply testing equipment. This means we can get the most accurate results from our testing as opposed to using software benchmarks (such as OCCT) or multi-meter readouts which are broadly inaccurate. Testing Hardware Sunmoon SM-8800 Active Load Tester Stingray DS1M12 USB Oscilloscope Voltcraft DT-10L laser tachometer What the eTeknix Test Procedure Involves Testing each power supply at 20/40/60/80/100% load (with load balanced across all rails) and measuring PFC (power factor correction), efficiency (actual power divided by power "pulled at the wall") which also uses a 50% load test, and voltage regulation (deviance from expected voltages of 3.3/5/12 on the main rails). Measuring ripple with an oscilloscope at 20/40/60/80/100% load. Transient loads taken with an oscilloscope are taken from 0% to 100% on the 12v raid, 3.3v rail, and then both combined. Measuring fan speed after a stabilisation period of five minutes at each loading scenario using the Voltcraft DT-10L laser tachometer and a reflective strip on the fan. Testing each power supply's OPP (Over Power Protection) mechanism and seeing how many watts each power supply can deliver before shutting down. Any deviation from this will be explained per review as appropriate on the respective testing pages, but this methodology page will remain unchanged. Other Things to Consider We recognise that a single 12-volt cable can provide only 6 Amps before overheating (which corrupts voltage regulation and efficiency) so we used an adequate number of cables for each power supply to ensure there is no efficiency loss from poor cable selection. If there is more than one 12VHPWR on a PSU, all will be used to test their combined loads. We use the same time scale and horizontal millivolt scale on our oscilloscope for all ripple tests, that is a 20ms T/DIV (horizontal) and a 0.02 V/DIV (vertical) meaning the scale is from -80mV to +80mV, ATX spec dictates that the 12v rail must fall within 150mv of ripple and the 3.3/5 within 50mv so that scale allows us to include both 150 and 50mV peaks. (Some older PSU reviews use different scales which were later ditched as the visual representation they give is inadequate, in these reviews written measurements are provided only). Deviance is the terminology used to represent the way voltages diverge from the expected values Voltage Regulation To test voltage regulation we load the power supply to five different load scenarios that give an equal spread of load across every single rail. So that means 20% on all rails, 40% on all rails and so on. We then calculate the average deviation of each rail from its expected voltage. Ideally, anything of +/- 2% regulation is seen as a very good result. Power Efficiency Power efficiency is measured by calculating the supplied wattage divided by the wattage drawn at the wall/plug, multiplied by 100 to give a percentage. We then compare that to the particular 80 Plus certification the company claims to see if it meets that. You can see the 80 Plus certifications below, we always test power supplies at 230v. Transient Loads Power Factor Correction Power Factor Correction is the ratio of the real power flowing to the load, to the apparent power in the circuit. PFC aims to make the load circuitry that is power factor corrected appear purely resistive (apparent power equal to real power). In this case, the voltage and current are in phase and the reactive power consumption is zero. The closer the number to one the better as this allows the most efficient delivery of electrical power (Source - Wikipedia). Ripple An oscilloscope can easily measure noise and Ripple. These show how much voltage fluctuation there is on a particular rail. We tested the rail stability of the 3.3-volt, 5-volt and 12-volt rails using an identical time and millivolt scale for all graphs. millivolt ripple is measured by the peak-to-peak size of the voltage curve. The latest ATX 12-volt version 2.3 specifications state that ripple from peak to peak must be no higher than 50 millivolts for the 3.3-volt and 5-volt rails, while the 12-volt rail is allowed up to 120 millivolts peak to peak to stay within specifications. Millivolt figures are stated to the closest increment of 5 given their variability. OPP Power supplies are often quoted as having various protection mechanisms and the most important of these is Over Power Protection. In our testing, we crank up the power draw until the power supply either shuts down (meaning the OPP mechanism is present and working) or blows up (meaning it is either not present or not working). We then note the maximum power consumption before the power supply shut down, or before it blew up. Fan Speed When testing in a power supply laboratory it is difficult to take fan noise readings as the noise from the Sunmoon test equipment and air conditioning corrupts everything. The next best thing in our circumstances was reading off the fan speed with a tachometer to get an idea of the noise. The ambient temperature during testing is held constant at 22 degrees, with 1 degree of variation. Each power supply had a consistent period of 5 minutes to stabilise between each load scenario. In my experience, the following general relationships apply between noise levels and fan speeds, though it can vary greatly between the type of fan used. Below 800 RPM - Inaudible/Silent 800 to 1000 RPM - Barely audible 1000 - 1200 RPM - Audible but still quiet 1200 - 1400 RPM - Moderately noisy 1400 - 1800 RPM - Noisy 1800 - 2000 RPM or above - Intolerable Internal Components This is just an example of what you can expect to see on a PSU interior. Efficiency I knew this power supply was going to be good, but I'm still absolutely blown away by how well it performed. Hitting the 80 Plus Titanium and Cybernetics Titanium rating for efficiency is exceptional, meaning the power drawn from the wall is remarkably close to the actual power output. This is also reflected in the Power Factor Correction, and the closer to 1.0, meaning no correction is needed, the better. This is the first time I've ever seen a PSU score anywhere near this high, and the first time I've seen one hit exactly 1.0 at any load. PFC (Power Factor Correction) Voltage Regulation Voltage regulation is extremely consistent at any load across all rails, with some of the best tolerances we've seen to date from any PSU. Ripple Unsurprisingly, the ripple is extremely low, given we need to be under 50mv at 12V and under 150mv at the remaining voltages, it's giving a very clean power delivery. 3.3V 5V 12V 5VSB Over Power Protection and Max Wattage What, 1600W isn't enough for you? Don't worry, we knew this PSU could do more, as it was demonstrated at Computex running four RTX 4090s and drawing just shy of 2000W consistently, and even in our own testing, we managed to push it a little harder still, reaching just over 2100W. Transient Loads These results measure the power supply from 0% load to 100% load on various rails to simulate a power spike from your system. Typically, we would like to see as small a spike as possible relative to the requested voltage. With modern GPUs being notorious for power spikes, this is becoming more relevant in modern PSU design. The specifications for ATX 3.0 were released in February 2022. It includes the new 16-Pin 12vHPWR connector that can deliver up to 600 W to graphics cards. These incorporate data lines for components to negotiate power capabilities with the PSU so they do not draw more power than the PSU is capable of delivering. The specification also has more strict requirements for handling spikes, two times the nominal output for 100 microseconds. Linear, clean and accurate transients are what we like to see, and again, this PSU just hit its mark every time in the transient testing, with no over or undershoot and a very quick load time response. 0-100% 3V 0-100% +5VSB 0-100% 12V Cross-Loads Just like we saw in all other testing, the performance is consistent from 10% to 100% with little to no deviation. The 12v rail drops a little when under load, but is still well within the 3% target we like to see. Fan Speed There's no switch for Hybrid fan mode on this PSU, despite it being capable of drawing over 2000W of power, it's so freaking efficient, that the fan is ALWAYS in hybrid mode. It put out so little heat in our testing, that the fan didn't even spin at 20 or 40% load, even when loaded for 20 minutes, no fan was needed. The fan didn't even bother reaching its full 1500 RPM till we reached 100% load, and even then, the actual PSU housing was barely warm to the touch. How Much Does it Cost? The XPG Fusion 1600W Titanium Power Supply is a monster, completely overbuilt and over-engineered for a very specific and small set of use cases that simply makes it pointless for virtually all PC gaming builds. Few things make that truer than the eye-watering price, which what little stock we can find has a price slightly above £550, making it £100 more than rival models from the likes of be quiet!, Seasonic, Corsair, and amusingly, it's still £100 cheaper than the utterly overpriced ASUS ROG THOR 1600W. Overview Would it be worth buying one of these for a gaming PC? Well, yes and no, it has some advantages, such as the truly exceptional efficiency, the best power factor correction I've ever seen, as it was near perfect, and it runs remarkably cool as a result, so if efficiency and heat are important to you, it's hard to argue with the results of our testing today. The power delivery is just fantastic, with very low ripple, consistent voltages at all loads and crossload, with good transients, making this a PSU that would be very well suited to extreme overclocking, where voltage tolerances can make, or literally break your hardware. With a pair of 12VHPWR connectors, it's capable of running any of the latest Nvidia graphics cards, such as the RTX 4090. However, that also means you could run two of them, or have one spare for future-proofing should next-gen cards require more than one. Or if you're running a rendering rig or workstation, you can very easily power four Nvidia RTX A6000 or AMD Radeon PRO W7900 cards and still have enough PSU juice left to run the new series of Threadripper CPUs. I would have thought a PSU of this size would be scary to test, putting it on our load tester I half expected something to blow up, or for some coil whine, or a fan to be spinning like heck to maintain its performance. But no, it was one of the coolest, quietest and most stable PSUs I've ever tested, and certainly the first time I've ever pulled 2000W+ from one. Should I Buy One? The conclusion of this review could not be much easier for me, as the exceeded my expectations in every regard. It pulls more power than it's rated for, consistently and easily, it's cooler and quieter, more efficient and more stable than just about anything else I've ever tested too. Sure, it's also the most expensive I've tested too, but if you need massive power with no compromise, then that's exactly what you're going to get.