We’re back with another PSU review folks, OCZ sent me a box with a box in it, and I’m pretty sure there’s a PSU in the box somewhere!
This time the lucky unit is one of OCZ’s new ZT Series units, they’re fully modular and come in 550 W, 650 W and 750 W flavors. Being on the benching team my feelings trend towards “go big or go home”, and we got the 750 W unit to play with. Of course, by “play with” I mean test brutally, inspect closely, and dissect.
Specifications & Features
The following is directly from the OCZ product page, my thoughts on each line are in italics:
- Available in 550 W, 650 W, 750 W
- Single +12 V rail for efficient power distribution. (Inexpensive to manufacture? Yes. More efficient? Not really.)
- Heavy-duty protection circuitry. (This is good stuff, we’ll talk about it later.)
- Fully Modular Cabling System. (Very cool indeed, even the main ATX cable disconnects!)
- 80-Plus Bronze Certified – 85% efficient at typical load. (Bronze is good.)
- 140 mm ball-bearing fan.
- Dimensions: 150 x 86 x 175 mm (W x L x H).
- Rated at 45 °C ambient. (Definitely a plus.)
- Premium 105 °C electrolytic capacitors. (105 °C capacitors are good. We’ll check on the “premium” claim later.)
- Erp compliant, low power usage in standby mode. (I like this.)
- Active PFC. (Everybody has it now, but it’s still nice.)
- ATX 12 V 2.2 Compliant
- MTBF: 100,000 hours. (That’s 12.5 years of continuous service, good luck testing that any time soon.)
- 5-year warranty. (Five year warranty is among the best available, good terms too.)
750 W Connectors
- 1 x 20+4-pin ATX
- 2 x 4+4-pin ESP/12V CPU
- 4 x 6+2-pin PCI-E
- 6 x 4-pin Peripheral
- 9 x 5-pin SATA
- 1 x Floppy
The OCZ product page doesn’t have output information so we’ll use a picture of the box for that:
As was stated in the specs, this is a single 12 V rail design. Single rail designs are less expensive to manufacture than multiple rail designs, but these days that’s really the only benefit. If you’re interested in how the single rail craze started keep reading, otherwise skip to the next paragraph now as there is a rough history lesson incoming! Back in the early days of the ATX specification multiple 12 V rails were mandatory, one rail for the CPU and one rail for everything else at bare minimum. All PCIe connectors had to go on the same rail as well. As GPUs started consuming more and more power and people started running two, three, or even four GPUs this became an incredibly large load, and the multi-rail PSUs that followed the ATX spec were hitting OCP (OverCurrent Protection) on the PCIe rail and shutting down. The grand idea was to say “To hell with ATX spec!” and go to a single rail design. Now the PCIe connectors could draw as much as they wanted (up to the limit of the PSU of course) and the problem was solved. The difficulty is that the problem could have been solved simply be spreading the PCIe connectors out over multiple rails and/or raising the OCP point of the rails the PCIe connectors were on, while retaining the safety of having multiple rails with individual OCP. That is exactly what modern multi-rail PSUs do, and it works well. The other reason single rail designs are still around is that every rail needs dedicated hardware to monitor the currant loads for OCP and while that adds to the cost of the PSU. Anyway, to my mind having a single rail isn’t really a selling point, but it isn’t really an issue either.
We also see what almost looks like DC-DC regulation for the 5 volt and 3.3 volt rails, though I would be surprised if that was the case as DC-DC regulation is fairly pricey and is generally only seen when manufacturers are shooting for very high efficiency. This unit doesn’t look to be trying to hit record efficiency, it is trying to bring full modularity and plenty of power to a low price point. We’ll see exactly what the regulation setup is when we crack the unit open in the Dissection Section.
Opening the box, we find a box! Next to it is a bag shaped like a box! There’s a PSU in there somewhere, I hope.
You get a fair number of cables with this unit. Four PCIe 6+2 cables (upper left), two CPU power cables (upper middle), one ATX24p connector (upper right), three SATA power cables (each with three plugs, middle row, left side), two Molex cables (each with three plugs, middle row center), a Molex-FDD adapter (middle row, right), and a nice beefy and fairly long wall power cable (bottom). Also included are a few zip ties and four screws, as well as a manual that has more information than is normal.
If we look closely at the ATX24p cable we see that the 3.3 V sense wire is wired correctly, this wire allows the 3.3 V rail to regulate itself based on the voltage arriving at the motherboard after wire/connector related voltage drops, rather than based on what is inside the unit. We’ll talk a little bit more in the testing section.
Now let’s look at the PSU itself:
Pretty unassuming, its yellow stripe lets you know it exists, but otherwise it is a stealthy flat black. I like flat black, flat black doesn’t show fingerprints very well at all, and that is a definite plus.
Obviously enough, this really is a fully modular unit, even the ATX24p cable comes off. This makes cable routing a snap, and the ribbon style SATA/Molex cables help a lot too. In working with the SATA/Molex cables, I found that their connectors are close enough together that is very hard to remove one without first removing its neighbor on the ATX24p connector side of things first. If you’re trying to remove one by the grey PCIe connectors you may have to remove all the SATA/Molex wires. The PCIe and CPU connectors are actually identical in wiring and shape, despite being different colors. OCZ made the right move given that and set the cables up so that they will work fine plugged into either plug as well. I like this, people get to feel like they’re doing it right plugging everything in right where it goes, but if they accidentally mess it up nothing will die. The SATA/Molex connectors will not plug in anywhere else, nor will anything else plug into them. You can however plug a CPU/PCIe cable into the ATX24p connector, it obviously doesn’t belong there though. Do NOT do this, it will result in dead shorts and maybe dead hardware.
First, a bit about my method in testing power supplies (taken from my previous reviews).
Unlike most computer parts, power supplies require rather specialized equipment to test correctly. Sure, you can plug it into a computer system and see if it can run a 980x and a couple GTX580s, but that doesn’t tell you how much power the PSU is putting out, nor what that output looks like. Worse, if the unit is defective or simply underbuilt/overrated the unit can fail catastrophically and take your computer along with it!
Hence, you really need a load that is strong enough to stand up to a PSU dying while attached to it! Purpose built loads and the testing units to run them cost thousands of dollars. They’re easy to use and very accurate and definitely the ideal way to test power supplies, but also entirely too expensive for me to afford. Instead I have built my own! It’s entirely mechanical and not automated in the slightest, but it can put a serious load on a PSU and will survive the PSU fails in the process. The down side is that it doesn’t have the built-in current sensing that a professional grade unit does, so PSU efficiency is difficult to nail down.
The second part to a good PSU review is an oscilloscope to look at the outputs and check for ripple (I’ll talk about what ripple is in the ripple section), for this purpose I have a BK Precision model 1472B analog scope. It has its pluses and minuses compared to more modern USB scopes. On the minus side, taking pictures of it is a difficult operation at best as you’ll see. On the plus side, it has incredible sensitivity at high frequencies, something that many USB scopes lack.
Lastly, a voltage meter is required, I have a nice cheap unit that I have compared against a highly expensive Fluke 88 and found to match to within 10 mV. That’s good enough for me!
On the procedure end of things, I first check the voltages and ripple with no load on the PSU at all. This is a quite unrealistic test and many PSUs do not appreciate it at all, but I like to test it anyway. If the zero load results are terrible I put a small load on the PSU to simulate idle conditions with a low power computer and test the ripple again. Some PSUs are specifically rated for zero load operation, though this is not one of them.
With that out of the way, I put successively higher loads on the PSU and check the voltages of the three main rails at a SATA connector, all the way up to the PSU’s maximum output.
At the maximum output comes the last test, ripple at full (or very close to it) load. This does assume that the PSU didn’t explode on the way to full load of course! I’d be quite surprised if this OCZ ZT Series 750 W unit did, but it happens on lower end units sometimes.
The 12 V, 5 V and 3.3 V rails need to stay within 5% of their official value to stay within spec, closer is ideal of course. That means 12 V needs to be between 11.40-12.60 V, 5 V needs to be between 4.75-5.25 V and 3.3 V needs to be between 3.135-3.465 V.
If you have survived yet another wall of tests and are still reading, here are the results!
Note that the last two tests are almost entirely 12 volts, this unit is rated for 744 watts of 12v and I want to test that rating, but I can’t leave the 146 watts of 5v/3.3v in place or I’ll be way over the PSU’s total rating. 720 watts of 12v and zero watts of 5v isn’t the most realistic test in the world, but given the independent regulation I suspect this unit uses, it should not pose an issue for it beyond the difficulty of putting out that many watts of 12v.
|Wattages||3.3v rail||5v rail||12v rail|
|0/0/0w||3.44 / 3.44||5.07||12.24|
|48/50/23w||3.46 / 3.44||5.05||12.22|
|96/50/23w||3.46 / 3.44||5.05||12.20|
|144/100/46w||3.48 / 3.40||5.03||12.19|
|192/100/46w||3.48 / 3.40||5.03||12.17|
|288/100/46w||3.48 / 3.39||5.02||12.16|
|384/100/46w||3.48 / 3.39||5.02||12.15|
|480/100/46w||3.47 / 3.39||5.02||12.13|
|576/100/46w||3.47 / 3.39||5.02||12.12|
Two values for 3.3v? Well yes, and here is why. Remember earlier when I pointed out the 3.3v sense wire? It dates back to the earlier days of the ATX specification when the 3.3v rail was used for processor and northbridge power among other things, that was a substantial load and had to be carefully regulated. These days 12v is used for the vast majority of motherboard components. My load tester uses the 3.3v wires in the ATX24p connector for loading the PSU however, and puts a pretty decent load on it! This puts a large draw on those wires and hence on the connectors where they go into the modular connector on the PSU. In some cases this can cause a voltage drop, and this appears to be one of those cases. The right column of 3.3v results are measured at the main ATX24p connector, the left column is measured at a SATA connector. The regulation at the ATX24p connector, with it’s voltage sensing wire, is quite good (though a bit high). The voltage delivered to the SATA connectors however goes out of spec by ~15 mV once the 3.3v load hits 46w. Is this an issue? Is this even really out of spec? No, to the first, and I’m not entirely certain to the second. Not much actually uses 3.3v these days and the motherboard certainly does not use this of it! This circumstance highly unlikely in real life. The rest of the regulation is excellent, and looks like an independently regulated unit. The 12v rail did quite well, even when loaded down quite brutally.
At zero load the fan was definitely noticeable, though far from obnoxious. If the unit is installed in a case I doubt that you would notice it. At 100% load the fan never really sped up much, and didn’t make any more noise than at zero load. This was however in fairly cold testing conditions (it is winter here, after all), given hotter testing the results may be different.
Now for ripple testing! Also, time for another excerpt from my previous reviews explaining ripple testing:
Voltage ripple is a measurement of how far above and below the median voltage the voltage goes. To put it more simply, ripple is how much the voltage bounces around. All power supplies have some ripple due to their design, it is impossible to get rid of all of it, and very expensive to try. The more ripple in the voltage outputs the harder time components have making use of it, this is why almost every computer part out there has capacitors on it’s voltage inputs to smooth the ripple out. In the hunt for the worst of the ripple I check everywhere from 10 milliseconds per divider down to 0.5 microseconds per divider and record the worst that I find. For all of the full load tests I have loaded the entire unitdown to its full rated maximum, not just the rail I am checking.
For the low load results I put a 50w load on the 12v rail, as this unit is not rated for zero load operation and hence testing it there isn’t fair to it.
The ATX specifications have a ripple spec as well of course, for the 12 V rail it is 120 mV, while for 5 v and 3.3 V it is 50 mV. We’ll start with 3.3 volts, head to 5 volts, and then finish off with 12 volts. The low load results are on the left and the full load are on the right, all result photos are taken with the scope set so that each box vertically is 10 mV, and each box horizontally is five microseconds. A microsecond is a thousandth of a millisecond, which is a thousandth of a second. That’s another way of saying that a microsecond is one millionth of a second, not very long!
Ripple comes in two forms, there is the main waveform which is exactly that, and there are transients. Transients are the very brief spikes that show up when the PSU switches it’s mosfets on and off. Transients are also very hard to mop up, the ATX spec calls for a 10 microfarad capacitor and a 100 nanofarad capacitor on the testing rails to mop up transients and serve to emulate the presence of input filter capacitors on the devices plugged into the PSU. My tester has these, and hence the transients are part of the final score of the unit. In reality transients past the ATX spec are far from likely to cause issues, but the spec is the spec.
The main wave forms look excellent for 3.3v and 5v, and good for 12v. The transients on 12v are excellent, but on 3.3v and 5v they exceed the spec by a nose, by 6 mV on 3.3v and by a heartbreaking 2 mV on the 5v rail. While they do violate the spec, it isn’t a huge deal, and not what I would consider a deal breaker. Still, it would be nice to see better results. The 12v rail is the important one, and it did just fine.
Editor’s Note: After this review published, OCZ contacted us about Bobnova’s ripple measurements. After some discussion, we think we’ve come up with a potential reason for slight variation. When he built his PSU tester, ATX12V spec specified a ceramic disk capacitor of 0.1µf and an electrolytic capacitor of 10µf between the test lead and the PSU.
More recent ATX12V spec specifies a low ESR tantalum 10µf capacitor to go with the ceramic 0.1µf, which should do a better job filtering transients. He’s going to order a replacement for future reviews, but in the mean time feel free to check out OCZ’s internal testing of this unit (pdf).
Dissection, Component Specifications
Disclaimer: Power supplies have dangerous voltages inside them, DO NOT OPEN POWER SUPPLIES. It’s just not a good idea, and doing so could very well kill you. Don’t try this at home. Don’t try this at work. Just don’t do it.
With that out of the way, let’s see what is inside the OCZ ZT 750 W PSU! I think this may be my favorite part of reviews.
We’ll start with the fan, it is made by Yate Loon, a well-known company. It’s a 140 mm job rated at 0.7 amps.
With that out of the way (literally!) we’ll take a look at an overhead shot, and then start following the path the electrons take to get through the PSU.
Our tour begins at the AC receptacle that the power cord plugs into, soldered to this are a pair of Y capacitors and an X capacitor, as well as the main switch and a pair of small resistors between the phase pin and the neutral pin. The only purpose I can think of would be to drain the X capacitor when the unit is unplugged and prevent it from delivering a shock to someone who touched the metal plugs. I don’t really see that being an issue though and have to admit to being confused by their presence. The X and Y capacitors are the beginning of the transient (there’s that word again!) filter for the input, then there are two ferrite beads on the leads that go to the main PCB, and the rest of it you can see in the photo of the main PCB, on which there are two more Y caps, another X cap, and two inductors. The PCB’s Y caps are hidden under the X capacitor. Lastly is a fuse to protect the unit, and a thermistor for inrush current limiting. The soldering on the receptacle is a bit sloppy, but mechanically and electrically it is perfectly sound, it just doesn’t look very nice.
Notably lacking is a MOV or TVS diode for surge protection. The aPFC unit can soak up many surges that would kill older designs, but I would still prefer to see some form of surge protection.
Please note that while it looks messy, the white goop on there is fairly important. It helps stabilize the various bits of the PSU during shipping, as well as helping to prevent coil whine under heavy loading.
Next up are the main rectifiers that take the wall plug’s AC and turn it into electronics friendly DC, a pair of T15KB60 units rated at 15 amps and 600 volts are used for this purpose, two more X capacitors surround the rectifiers. Moving along we get to the aPFC section, it takes the 110 V or 220 V coming in to the unit and boosts it substantially before storing it in the large Teapo storage capacitor. Teapo has a fairly bad reputation due to a bad run of capacitors some years ago, as well as Dell misusing some Teapo capacitors in many of their motherboard designs. In PSU land Teapo capacitors work just fine, and are not an issue, though I wouldn’t call them “premium” like the box does. This one is rated at 105 °C, which is nice to see. The aPFC circuit itself has two 24N60C3 mosfets rated at 15 amps at 100 °C and 600 volts, followed by a LQA08TC600 diode rated at 8 amps and 600 volts.
Now the electron’s path splits, as the 12 V, 5 V and 3.3 V are all independently regulated. This is rarer than I would like and refreshing to see, thanks OCZ!
The 12 V is handled by a synchronous layout, with two M6020AP mosfets on the high side and two more on the low side, the M6020AP mosfets are rated at 20 amps at 25 °C and 13 amps at 100 °C, and 600 volts either way. Synchronous regulators are significantly more efficient (and expensive) than normal asynchronous setups (as used for 5 V and 3.3 V), this is how OCZ hit 80+ Bronze with this unit, and likely explains why the fan didn’t speed up during testing.
The 5 V rail is powered by two 20N60C3 mosfets (20 A @ 25 °C, 13 A @ 100 °C, 600 V) and two 30L45CT schottky diodes (30 A, 45 V).
The 3.3 V rail is powered by and identical setup of two 20N60C3 mosfets (20 A @ 25 °C, 13 A @ 100 °C, 600 V) and two 30L45CT schottky diodes (30 A, 45 V).
After all that we reach the output end of things, go through some Rubycon capacitors and head over to the modular output PCB. The output PCB has a few more capacitors on it for final filtering. The 12 V arrives via a nice beefy bus bar, and a similar bar is used for ground. The rest of the rails go via normal wires. Here we run into something annoying, the standoffs that screw the modular board to the PSU’s case are soldered in place without any nuts on them. There is support from the bus bars, but I would much prefer to see nuts on the standoffs anyway. Oh well.
Last, we’ll look at the main PCB’s soldering, by and large it is good. There are a few areas that could use a little bit more solder, but nothing serious. What you can’t see in the picture is a sea of flux over half the board. There is almost always some flux left on the board by the wave soldering process, though I haven’t seen this much before! It’s nothing that will cause an issue though as it is non-corrosive flux.
Final Thoughts and Conclusion
The ZT series is OCZ’s newest, and it is aimed squarely at what is arguably the second most brutal PSU market: the mid-to-high power, good efficiency without a massive price tag market (I would say that the cheap but not dangerous market is the hardest market). OCZ already has a pair of 750 watt PSUs in this market, but the ZT series brings something new to the table: Full modularity at a remarkably low price.
At the current $109 ($99 w/ $10 rebate) price point the only competition worth mentioning comes from OCZ itself and Enermax (plus Rosewill, if you like), and none of them are fully modular. To find another fully modular 750 W PSU you have to spend more money ($130), I am unsure how OCZ managed to keep the price this low but it makes for a good value!
The 3.3 V and 5 V ripple results are unfortunate, but not really what I would call an issue. Better ripple control would be nice though. Voltage regulation was good on the whole, though the unloaded 3.3 V sections violated spec when the ATX24p 3.3 V was faced with a brutal load, much like the ripple results I would be quite surprised to see this cause an issue, but it is annoying. The lack of nuts on the modular output PCB are another possible minor issue, I didn’t have any trouble with them but I have seen reports of similar setups cracking on other PSUs.
Cosmetically the unit is excellent, I like it a lot. Similarly the modular cable setup is fantastic, I love fully modular PSUs as they make customization and cable management a snap. The price is quite good for what you get, too.
To summarize, there are pros:
- Great looks
- Good price
- Fully modular!
- Good 12 V and 5 V regulation
- Plenty of cables
Nothing being perfect in this world, there are cons as well:
- 5 V and 3.3 V rails have spec-violating ripple.
- 3.3 V rail voltage regulation violates spec in certain situations.
- Only solder on the modular output PCB standoffs, no nuts.
- Peripheral cable connectors on the PSU are a bit cramped.
All told, I believe this OCZ ZT Series 750 W PSU to be a good bang for your buck PSU choice, and award it an Overclockers.com Approved stamp!