EVGA SuperNova NEX1500 W Classified Power Supply Review

Here we have it folks, EVGA’s entry into the “Ultimate PSU” market. EVGA has gone above and beyond the usual even for this market segment. 1500 W isn’t enough, so they put in eight rails (that can be set to a single huge rail via software or a DIP switch), fan controls, software to monitor and adjust the unit, sixteen PCIe cables. It’s an impressively, excessive PSU, really.

We’ll start with specs and features, even though I know that what you really want is to see the unit.

Features and Specifications

Courtesy EVGA.com:

Far shorter feature list than some manufacturers stick me with, it’s nice. In order: Gold efficiency is good. SuperNOVA software sounds interesting, I’m curious to see how much it resembles reality. Adjusting the 12 V rail voltage is an interesting feature, though probably not actually useful. Single or multiple rail control is also interesting, it definitely beats just being single rail. Overclock mode is an interesting concept that I don’t really like. While I can understand why they probably did it that way (more on that later, in the testing section) overclocking PSUs in general is a bad idea at best. A 10-year warranty is fantastic. Japanese capacitors are nice. Individually sleeved cables look pretty awesome, but can be annoying to work with. Fully modular is great, and Sanyo Denki makes great fans.

It’s worth noting that software control of the 12 V rail voltage means this is a digitally controlled unit, much like others that are being advertised heavily. The big difference is that EVGA isn’t doing a lot to market the digital control aspect of this unit.

That does it for the features sheet, the specs sheet lists protections (it has all of them, OVP/UVP/OCP/PPP/SCP/OTP) as well as an output chart:

That’s a lot of watts. Also a lot of 12 V rails. My load tester only goes up to 116 amps of 12 V, but given that this is almost certainly a DC-DC regulated unit I can grab the extra power needed to cap this unit out at 1500 W from 5 V and 3.3 V and the 12 V rail will still get its full load.

I’m very happy to see multiple rails, a 1500 W single rail unit can do some pretty impressive damage without ever triggering OCP/SCP/OPP. A multiple rail unit will shut down much sooner.

The box has a bit more in the way of specs/features, let’s look at it next.

Photos Part One: The Box

The box is big. Huge, really. It’s the largest PSU box I’ve run across yet. I was very happy to find that it has a carrying handle as it is rather heavy as well.

It’s worth clicking the picture of the rear of the box if you want more specs, it has a number of things that aren’t listed elsewhere. The number of cables, for one. Also worth noting is a goofy looking bar in one of the pictures and a rather industrial looking power cable in another picture. We’ll look at that cable momentarily.

Opening the box, we find boxes!

Photos Part Two: The Box of Shiny Bags

We’ll start with the box on the right, it has shiny things inside it.

Why the cables are inside anti-static bags I have no idea, last time I checked copper wasn’t especially sensitive to static.

We’ll go through the cables, starting with the power cable.

Don’t lose this power cable, it’s a C19 plug on the end and new C19 cables cost a lot. Also interesting is the lump near the PSU end, that’s a ferrite bead and is the first bit of the transient filter. The photo doesn’t really do the thickness of the cable justice, it is quite thick.

All the other cables come in anti-static bags as well, but we’ll skip the in-bag shots. They also come nicely bundled with hook+loop fastener straps. Save the straps, they’re great for cable management.

If you’re asking yourself “What’s a PCIe power plug doing on the CPU power cable?” then you’re thinking what I was thinking at this point. The documentation says it’s for powering a motherboard that accepts additional power. Unlike some motherboard manufacturers that use rather low-rated SATA power connectors for extra PCIe power, EVGA uses nice beefy PCIe power connectors. It’s logical and quite nice. While I like the concept of having a CPU power cable that can also connect to those additional plugs, I would also like to have one that doesn’t have extra stuff dangling off of it. One of the two CPU power cables has two PCIe plugs, one has one. Both are straight 8p connectors, not 4+4.

As a note, there are no ties on the cables and they are very long. The very long is very nice for benching and for big cases. The lack of ties means that they look great as the wires aren’t forced together, but it also makes them somewhat annoying to work with. The ATX24P cable especially.

You get a bunch of accessory power cables too of course, here’s a sample:

We have short and long Molex cables, also one of the many SATA cables, a Molex-2xFDD connector, and a mini-USB to motherboard USB connector. The use of mini-USB (micro-USB? I don’t know. Standard camera type cable anyway) is something that made me very happy. This means that for benching you can plug a normal cable in and monitor/control the PSU from your laptop. I tried it, it works. Compare this to Corsair’s digital unit which has a custom plug on the end of the link cable, you’re stuck with the motherboard on that one. Kudos to EVGA here.

Finally,  cable wise we have the PCIe cables. They’re all red, all 6+2p and come in two different lengths.

These are really annoying to work with, as the +2p connector and wires are completely separate once out of the PSU end plug. This means you can get them all confused and wound together if you’re not careful.

The sleeving on all the cables is nice high density stuff, it’s fairly slick and helps to mitigate the annoyances caused by individual sleeving with no ties, somewhat. Also of note is that the ATX24P cable is not wired for voltage sensing.

Last for this box, you also get a sack to stuff the cables in when you aren’t using them all (good luck using all 8 PCIe cables!), a manual and a cheat-sheet on how to arrange the cables so as to not overload any one rail, as well as explanations of the DIP switches on the back of the unit.

My camera has interesting notions about white balance when faced with a white sheet of paper. It’s really white, not purple.

Now let’s see what’s in the other box…

Photos Part Three: The PSU

Now we can really get down to business, inside the smaller box is this well packaged lump:

I’m impressed by the packaging, EVGA did a good job there. The unit looks pretty cool as well, though the chrome bar is a bit odd. I’ll take some straight on shots first:

EVGA went above and beyond with the labeling here. Not only does the side panel have all the rail wattage information, the other side also has it. One side is “upside down” so that the label is always right side up regardless of what direction the fan is pointing. Also note the very clear labeling of what plug is on what rail, that’s crucial if you want to run in multiple rail mode (which you should, in my opinion). One thing I am very happy to see is that the fan is correctly located on the end furthest from the exhaust grill. This may seem like a no-brainer, but some manufacturers don’t put it there.

Also lurking in one of these boxes are eight rubber stick-on feet and instructions for installing them:

Interestingly there is no warranty sticker as such on this unit. Nothing across any of the seams, nothing over any of the screws. Just a serial number sticker that doesn’t cover anything. It does however peel off very easily. Have fun with that.

I’m pretty sure this is a full retail unit, which raises interesting questions about opening the unit to remove the bar across the back and the warranty.

Last two things of note, the DIP switches on the rear of the unit and the Torx T8 screws that hold it together. The DIP switches are awesome, the Torx screws are very annoying if you’re reviewing this unit and don’t have a Torx set.

And finally, two more angled shots for general glamor purposes. I think I need to water my lawn more often.

Note the red sticker under the power switch, under that sticker is where you find the DIP switches.

Now that we’ve seen it, it’s time to test it. The specs say 50c for 1500 W. We’ll see how that goes, that’s awfully hot and an awfully large load. On the other hand, it’s a gnarly looking fan.

Testing Part One: Regulation

A decent load test of a PSU requires a decent load. Contrary to what some may believe, that means you need a known load that can fully stress the PSU. Computer hardware does not cut it. Worse if the PSU fails during testing it might take out the computer hardware anyway. Commercial load testers cost a lot of money. I do not have a lot of money, so I built my own with juicy power resistors and a Toyota cylinder head. It works great. I’ll be using it to load this thing down fairly severely and will check voltages and ripple (more on that later) at various points. The down side to my tester is that the loads it can put on PSUs are fairly coarse, they go in increments of 48 W for 12 V, 50 W for 5 V and 22 W for 3.3V. Those wattages assume the PSU is putting out exactly the official rail voltage, a PSU putting out 12.24 V rather than 12 V will be at 49.9 W per step rather than 48 W. I file that under the “tough beans” category as I figure if a percent or two of load makes that much of a difference the PSU manufacturer should have hit the voltage regulation more squarely. It does make calculating efficiency difficult at best.  however, given that the input power is read via a Kill-a-watt, the efficiency numbers are dubious to begin with. Kill-a-watts not being known for extreme accuracy on things with automatic power factor correction. For this reason, I am not listing the efficiency.

The ATX spec says that voltage regulation must be within 5% of the rail’s official designation, regardless of load. It doesn’t actually mention that the PSU shouldn’t explode, though I expect they figured it was implied. Exploding is a failure in my book regardless.

It is also worth knowing that I will be testing this PSU at both outdoor ambient temperatures (typically between 15 °C and 20 °C here this time of year) as well as in the Enclosure of Unreasonable Warmth. TEUW is a precision engineered enclosure that I use to route the exhaust air from the PSU right back into the intake fan, it is adjustable to hold the intake air temperature at (almost) any level I want it. This way I can test the PSU’s response to hot conditions as well as cold conditions. For the hot testing I will be running the intake temp as close to the unit’s maximum rated temperature as possible. TEUW, in case you’re curious, is a cardboard box.

I expect the Kill-A-Watt to start screaming at me at the high load end of testing, we’ll see if it shuts down before I can get numbers off of it or not.

Without any more blathering, here are the results!

Sharp eyes will note that the last test is over 1500 W, I forgot to turn the 5 V rail load down to 50 W for that test as I’d planned to do. The NEX1500 really didn’t care, though the Kill-A-Watt probably would have appreciated it.

The first * marks where the fan suddenly woke up, in response to load rather than temperatures. The ** marks where the fan really wound itself up. If you’re used to normal computer fans prepare to be horrified by the noise it makes, it’s a lot louder than most PC/PSU fans. If you’re used to the sort of fans we benching team folks use, this one really isn’t that bad. Quiet, however, it is not. Even at no load and low load the fan makes a fair amount of noise. Low/no load it’s ignorable if it’s in a case, but out in the open you’ll notice it. At full load or close to it, a case isn’t going to help much. Admittedly you’ll have three or four top-end GPUs with fans screaming at you by that point anyway, so who cares?

Lastly, this unit really can cough up 1500 W (and a bit…) with 50 °C intake air temperatures. The exhaust is a toasty 74 °C, but the unit chugs along just fine. Be aware that the grill on the back of the unit and the crossbar will be of a similar temperature to the exhaust, 74 °C / 165 °F in this case. Kind of warm if you try to pick it up. Of course, your ambient temperature is almost certainly lower than 50 °C! I hope it is, anyway.

Regulation wise the 12 V rail comes in at 2.5%, the 5 V rail at 2.7% and the 3.3 V rail at 1.8%, for an average of 2.3%. Not great, not bad, middle of the range really. I was surprised the 12 V rail drooped as much as it did. I didn’t use the software to prop it up, as then it’d likely be too high at idle.

Testing Part Two: Ripple

Ripple is fluctuation of the PSU’s output voltage caused by a variety of factors. It is pretty much impossible to have zero ripple in a SMPS computer power supply because of how a SMPS works, so the question is how much ripple is there? In the regulation testing phase we found out how the PSU does at keeping the average voltage at a set level, now we’re going to see what that voltage is doing on really short time frames. The ATX spec says that the 12 V rail cannot have more than 120 mV peak to peak ripple, the 5 V and 3.3 V rails need to stay under 50 mV.

If that isn’t complicated enough for you, there are three forms of ripple to keep track of as well. Long-term ripple from the PSU’s controller adjusting the output voltage and over/undershooting, correcting, overshooting, etc. Medium-term ripple from the voltage controller charging and discharging the inductor(s) and capacitor(s) that make up the VRM, and very short-term ripple caused by the switching itself. The first and second forms are the most important, if they are out of spec it can cause instability at best or damage in extreme situations. The very short-term (I call it transient ripple) flavor is less crucial, excessive amounts can still cause issues though it takes more of it to do so. The ATX spec does not differentiate, as far as the spec goes 121 mV of transient ripple is just as much of a failure as 121 mV of medium or long term ripple.

I test ripple in a few difference ways, first I test it during the cold load testing. It is tested at zero load and maximum load first. During the hot load testing I test the ripple at maximum load again. I have recently started testing ripple at fairly random loads with the unit still hot, it’s a bit unorthodox (a bit? maybe a lot) but has found issues in the past that did not show up with other test methods.

First up, all rails with zero load on the PSU and with cold intake air temperatures. The scope is set to 10 microseconds and 10 mV. Enjoy these pictures, as the rest of the ripple pictures aren’t nearly as well done!

All three primary rails have about 10 mV of ripple. That’s quite good, really.

Now let’s see how they do with a 1536w load! You also get to see the difference between ripple photos taken at night (above) and during the day (below). You’d be waiting another week for me to have a free night though, so you’ll have to deal with the reflections. The scope continues to be set to 10 microseconds and 10 mV.

The photos aren’t that bad. The ripple is good. I’ve seen better, but not at this load level! Absolutely nothing to be concerned about here. It’s actually better than anything I’ve tested in quite a while.

How will it do at 50 °C you ask? Let’s find out.

Not only can this unit survive the heat, it likes the heat. The filtering capacitors do anyway. Ripple is lower across the board and is straying into “excellent” territory.

One last vaguely interesting thing I ran into. Generally I use a Tt handheld PSU “tester” to fire the unit up the first time and make sure it actually works, as well as to get the PWR_OK delay time. The PWR_OK signal is sent by the PSU to the motherboard once all rail voltages are correct and the APFC bits are fully charged. On most PSUs there is a 300-370 millisecond delay between the motherboard telling the PSU to start and the PSU finishing startup and sending the PWR_OK signal. The ATX specifications call for a 100-500 ms delay. The interesting bit is that the Tt tester reported this PSU as failing that section! I found that somewhat hard to believe and build my own tester centered around an Atmega328 microcontroller (woefully overpowered for this, but that’s the Overclockers way!). This PSU sends the PWR_OK signal in 205ms +/- 5ms. Apparently that’s too fast for the handheld tester! What it does mean is that with this PSU your computer will boot ~150ms faster from the push of the power button. Not exactly a lot, but every bit helps right?

Software Testing

Given new features, we need a new section for testing them! I found the accuracy of the input and output numbers to be dubious at best. The software reads all the rail voltages as lower than they are supposed to be. The 12 V rail adjustment works well, but if you’re going by the software numbers it isn’t hard to take it out of spec while seeing numbers within spec in the software. I always enjoy graphs at least.

Every monitoring option (and there are plenty!) can be graphed. The fan can be set to “Silent”, “Performance”, and “Overdrive” levels, which offer different speed ramps based on the wattage being drawn. Here’s a couple pictures, in case you’re curious what it looks like.

You can set whether the window shows two, four or six different things, I prefer six personally.

Overall it’s interesting and fun to play with, but not really useful.

Dissection

Disclaimer: Power supplies can have dangerous voltages inside them even after being unplugged, DO NOT OPEN POWER SUPPLIES. It’s just not a good idea. Opening a power supply and poking around inside could very well kill you. Don’t try this at home. Don’t try this at work. Just don’t do it.

The bigger the PSU the larger the APFC caps and the more dangerous it is, too. This unit, coming in at 1500 W, is very big. Don’t open it, even if you want to have fun with the Torx T8 and T6 screws. From a discouraging access standpoint they’re a good idea and in theory they’re easier for machines to install, but for me they’re annoying. Oh well.

Popping the top off we find the fan, a San Ace as promised by the specs. (San Ace fans come from Sanyo Denki.) We also get to see the guts of the unit for the first time.

It’s fairly cramped inside this unit, there’s also a plastic baffle to direct air across the heatsinks on various components. This explains why so little air was coming out the back of the unit when the fan was wound all the way up. Less airflow, but far more efficient at getting rid of the heat than if it were simply thrown into the PSU case and allowed to go along its merry way.

Let’s move on to the transient filter, which starts at the receptacle with two Y capacitors, an X capacitor and a ferrite bead.

After the receptacle comes the rest of the filter, on the main PCB. It has a huge (20 A) fuse, a TVS diode, an X capacitor, a relay to short the inrush protection thermistor, a huge inrush protection thermistor (see next pic), two more X capacitors, an inductor, then two more Y capacitors, another inductor, another X capacitor (big), and finally two more Y capacitors. For those keeping score at home, that’s six Y caps, four X caps and two inductors. This is a very solid transient filter, I approve.

Then we have the rectifiers, a pair of GBJ2506 (25 A, 600 V) units. 50 amps ought to do it, don’t you think? That’s a solid 5500-11000 W after all. Following them is the APFC unit, or APFC units I should say. They are two wired in parallel and acting out of phase with each other. This provides very nice smooth power to the main switches. Specs on the APFC bits after the pics.

Each APFC unit has two 47N60C3 (94 A @ 25c, 50 A @ 100c, 600 V) MOSFETs and two C3D10060A (10 A@150c, 600 V). As there are two APFC units, we get four of each all told. A bit of fun math gives us a maximum switched wattage of 22000 W, and diodes for ~15,000 W. In reality I expect something would go BAM much earlier than that (the fuse, for starters), but that gives you an idea of the build level here! Sharing space on the controller board with the APFC controller are a micro-controller to monitor things (for the software, as well as for the benefit of the protections, IC I expect) and the PWM controller that sends the energy on its way towards becoming 12 V.

We can thank Texas Instruments for both the APFC controller (UCC28070) and the PWM controller (UCC28950). They’re both quite advanced pieces and obviously do their job well.

Next up we have four 20N60CFD (20.7 A @ 25 C, 13.1 A @ 100C, 650 v) MOSFETs for primary switches. As a vaguely interesting side note, the Antec HCP-750 uses four of these as well. It’s not often you see a 750 W unit and a 1500 W unit sharing parts! Given that by their spec sheet and the voltage involved, each MOSFET can switch 5,000 W or so, I’d say both units are overbuilt. Don’t get the idea you can pull 20 kW from either unit, though. That would be a bad idea.

Once switched, we head into the very strange looking transformer, to be transformed from 380 V into 12 V. Tacked to the side of it are 12 BSC047N08NS3 (100 A @ 25C, 79 A @ 100C, 80 v) MOSFETs. That’s 1,200 amps of rectification at 25c. That ought to do it, don’t you think?

Under the lid of the transformer is a pair of output filter inductors, getting a picture was not possible.

Now that we have 12 V, a wide swath of Nippon Chemi-Con electrolytic and a couple brands of polymer capacitors clean it up. Then two daughter boards do some DC-DC buck converting to get 5 V and 3.3 V. These boards have heatsinks over the MOSFETs and leave no access to get said heatsinks off, I don’t know what they use.

The 5 V and 3.3 V get a few more of the same capacitors for filtering. Everything is very Japanese in here.

For protections we have another micro-controller, it has a little thermistor hanging off into space for fan control (odd place for it, but whatever), five shunt resistors on the bottom for current sensing, plus three more on top (there really are eight rails), plus another bank of them for overall 12 V OCP and/or single rail mode. It has one cable for the fan, plus two more cables that go to the DIP switches and LED on the back of the unit. There’s a bank of Op-Amps to turn the voltage differential generated by the current sense resistors into something the micro-controller can comprehend, as well as a wide variety of SMD bits and a little trimpot of unknown function.

Now it’s impossible for me to say what exactly the micro-controllers are up to, those are full-on computers. They have a CPU, ram, flash storage, EEPROM storage, and do whatever the code inside them tells them to. The square jobby above has USB support and is also in charge of OCP, both for single rail mode and for multiple rail mode. It also runs the fan and monitors the temperature. Given that the software can adjust the 12 V rail via the USB port, the MCU (micro-controller unit) has something to do with that as well. It’s also in charge of OVP/UVP, almost certainly OTP, and in fact every protection except possibly for OPP. OPP may be there, or it may be in the APFC controller. I’d love to see the code inside this thing, but even if I were able to drag it out it would be in assembly. I don’t speak assembly, so that wouldn’t help much. In any case, all the digital controls live inside those microcontrollers.

Almost done with our tour now. The modular connector board is next up, it’s soldered directly to the main PCB and going absolutely nowhere. It has quite a few electrolytic capacitors as well as a number of polymer capacitors.

That’s some very nice soldering. Let’s look at the main PCB’s soldering too!

There’s some flux that didn’t get washed all the way off, and a few very nit-picky things that could be pointed at. Overall this is the best soldering I’ve seen yet, I call it excellent.

That does it for our tour, we’ve seen what we needed to see. This unit is very nicely built, the specs and quality are all top notch.

In case you’re wondering, all the PCBs have UL number e213009 on them, which belongs to JIANGSU DIFEIDA ELECTRONICS CO LTD. This is either a parent company of Etasis or a sub-company of Etasis, I’m not sure. Either way, Etasis designed this thing, and then EVGA modified it heavily and put their UL number (e355174 ) on the case.

Final Thoughts and Conclusion

To kick this section off, I have to say that I am impressed. I also have to say that for $450 I better be impressed! That’s a lot to pay for a PSU.

To discuss the price more in depth, this is the most expensive of the currently available 1500 W PSUs. Silverstone has a unit that goes for $300, Cooler Master has a unit that goes for $380-$400. At $450 the EVGA SuperNOVA NEX1500 (what a name!) is far from inexpensive. It also however comes with a lot of features that those other units lack. Multiple rails, for starters. Gold efficiency for second. A ten year warranty for third. All together the extra features of the NEX1500 are enough that I’m willing to label the price as decent.

The efficiency of the unit is impressive. Assuming that the Kill-A-Watt is somewhere near accurate, this really is an 80+ gold unit. It clears 87%/90%/87% at ~20%, ~50% and ~100% load levels. Of course the Kill-a-watt might be completely out to lunch, they aren’t known for their accuracy.

The fan is impressive, unfortunately for silence lovers it is anything but silent. At low speed it’s at least decently quiet, at high speed you’ll be impressed by the noise if you don’t work with server fans a lot.

The voltage regulation is good, it’s definitely not the best I’ve seen, but for a monstrosity of a 1500 W unit it’s not bad at all.

The ripple control is quite good. I wouldn’t give it the fantastic stamp, but it’s much better than most units of any wattage.

The NEX1500 looks fantastic, though the chrome bar takes some getting used to. It’d be nice if you could remove it without cracking the unit open.

The cables are very long, which I like. They’re also individually sleeved, which appeals to many. I would vastly prefer that they had some sort of cable tie on them as they can get very unruly if you aren’t paying attention. If you’re planning on using this unit for benching you are definitely going to want to attack it with cable ties to subdue the cables. If it’s going into a case where it will stay, it’s a lot less important.

I do feel obligated to point out that very few people actually need a 1500 W PSU. To justify this you’re going to need four top-end ~250 W TDP GPUs and a SB-E or BD CPU, plus overclocking. Or to be willing to spend an awful lot of money on an awesome looking unit.

The soldering is fantastic, as is the general build quality and component selection.

The 10 year warranty is also fantastic, especially since EVGA has a very good reputation for easy warranty service.

The DIP switches are kind of fun to play with, EVGA says you should use them to run the unit in single rail mode if you’re using multiple GPUs. I say that switching to single rail mode defeats the safety of having multiple rails, and to use multiple rail mode if possible. 1500 W on one rail with a semi-short circuit will set fire to things long before OCP/SCP/OPP triggers.

I found evidence of thought in quite a few places in this units design, starting with the fan placement and continuing all the way through the internals. EVGA and Etasis did a very good job here.

This list is getting out of hand, it’s time for bullet points. There are pros:

• Actually coughs up 1500 W at 50 °C, happily.
• Lots of nice long cables.
• Good ripple control and regulation.
• Great warranty.
• DIP switches and control software are fun.
• Fantastic soldering.

There are a few cons too:

• Non-cable-tied cables may attack you in your sleep.
• Fan is noisy, even at low speed.
• Very few people actually need 1500 W.

Not really much on the cons side, lots on the pro side, this unit gets approved. The only real danger to its approval is that 1500 W really puts it in a niche market, not many people need that. You have to have a LOT of GPUs or be running a couple high-end GPUs on dry ice or liquid nitrogen cooling to actually need that much power. I’m letting it off the hook on that front due to it bringing a lot of features to the table as well as looking awesome, that adds the case modding market.

— Ed Smith / Bobnova