EVGA 450 B3 PSU Review
-
Page 1:Features & Specifications
-
Page 2:Packaging, Contents, Exterior & Cabling
-
Page 3:Teardown & Component Analysis
-
Page 4:Load Regulation, Hold-Up Time & Inrush Current
-
Page 5:Efficiency, Temperature & Noise
-
Page 6:Protection Features & The OPP Flaw
-
Page 7:Cross-Load Tests & Infrared Images
-
Page 8:Transient Response Tests
-
Page 9:Ripple Measurements
-
Page 10:Performance, Value, Noise & Efficiency
-
Page 11:Final Analysis
[Update, 10/13/2017: EVGA informed us that as of today all B3 units come from Superflower, not RSY]
EVGA's new B3 power supplies will succeed its B2 family, positioned between the GQ (made by FSP) and BQ (from HEC and Andyson) line-ups. The platform, shared across all B3 models, was designed by Super Flower. According to our sources, though, it's manufactured by a sub-contractor (RSY) since Super Flower's capacity is limited.
It's strange that EVGA didn't improve the B3 platform's efficiency compared to the generation prior; we still see an 80 PLUS Bronze badge. More than likely, the company wanted to keep costs as low as possible and offer something newer-sounding.
Five B3 models cover a range between 450 and 850W. The three lower-capacity members include a couple of PCIe connectors, whereas the 750 B3 has four and the 850 B3 six. None, however, come with two EPS connectors; they each have just one. In our opinion, at least the 850 B3 should have a couple of EPS connectors to support high-end motherboards.
The 450/550/650 B3 models are 150mm deep, and the 750/850 B3 versions are a little larger at 160mm.
The entry-level 450 B3 looks to be a pretty good option for budget-sensitive mainstream PCs. A $50 buy-in gets you a fully modular PSU with decent efficiency and a five-year warranty. That looks like a deal from any angle, so when EVGA was unable to provide us with a review sample, we went out and bought the whole B3 family on our own. That means the PSUs we're testing are the same ones you'll find on store shelves, guaranteed. The question of golden samples is taken care of in this case. Below is a video of our teardown of this PSU.
Specifications
The PSU achieves 80 PLUS Bronze and ETA-S efficiency, while in the noise section is achieves a LAMBDA-A- (25-30 dB[A]) certification. All of its cables are modular, and it is impressive to see a selectable semi-passive mode in this price category. Unfortunately, this unit's temperature rating is lower than what the ATX spec recommends. However, $50 PSUs can get a whole lot worse, so we really can't complain much about the 450 B3's 40°C rating.
A five-year warranty is pretty good for such an affordable power supply. Further, it's nice to see all of the necessary protection features in place (at least, EVGA tells us they're there). Fifteen centimeters of depth make this a fairly compact unit. So all in all, the 450 B3 appears to be one heck of bargain. We're guessing that EVGA sells it at a small profit or none at all in order to put pressure on its competition. Corsair's more expensive CX450, in particular, is in this unit's crosshair.
The following video footage shows our work behind the scenes, demonstrating what we do to break down the PSU. More details can be found on page 3.
Power Specifications
| Rail | 3.3V | 5V | 12V | 5VSB | -12V | |
|---|---|---|---|---|---|---|
| Max. Power | Amps | 20 | 20 | 37.4 | 3 | 0.5 |
| Watts | 110 | 448.8 | 15 | 6 | ||
| Total Max. Power (W) | 450 | |||||
EVGA employs a fairly typical configuration, offering around 100W maximum combined power on the minor rails, 15W at 5VSB, and a +12V rail able to deliver the unit's full power on its own.
Cables & Connectors
| Modular Cables | |||
|---|---|---|---|
| Description | Cable Count | Connector Count (Total) | Gauge |
| ATX connector 20+4 pin (600mm) | 1 | 1 | 18-22AWG |
| 4+4 pin EPS12V (600mm) | 1 | 1 | 18-22AWG |
| 6+2 pin PCIe (550mm+150mm) | 1 | 2 | 18-22AWG |
| SATA (500mm+100mm+100mm) | 2 | 6 | 18-20AWG |
| Four-pin Molex (500mm+100mm+100mm) | 1 | 3 | 18AWG |
| FDD Adapter (+105mm) | 1 | 1 | 24AWG |
The modular cables are long enough. However, the distance between the SATA and four-pin Molex connectors should be greater than 10cm. With such short runs (especially between the four-pin Molex connectors), you might have compatibility issues if you try to use case fans installed far away from each other.
Power Distribution
Since this PSU features a single +12V rail, we do not have anything to say about its power distribution.
MORE: Best Power Supplies
MORE: How We Test Power Supplies
MORE: All Power Supply Content
- Features & Specifications
- Packaging, Contents, Exterior & Cabling
- Teardown & Component Analysis
- Load Regulation, Hold-Up Time & Inrush Current
- Efficiency, Temperature & Noise
- Protection Features & The OPP Flaw
- Cross-Load Tests & Infrared Images
- Transient Response Tests
- Ripple Measurements
- Performance, Value, Noise & Efficiency
- Final Analysis
In this case, it seems like EVGA/Superflower may have incorrectly set the OCP on the 12V rail. I hope they meant to set it to something more conservative like 40A instead of the 50+ it may have been at here.
As for the main fuse not blowing, that is normal: the fuse lives upstream from the APFC choke and most other filtering components while both the APFC and main switching transistors are connected to the APFC-boosted input hold-up capacitor. The transistor's lead blows up from the capacitor dumping its several joules of energy into the lead, the APFC chokes smooth out the current spike and the current seen by the fuse comes nowhere close to reaching its i2t rating. It is the capacitor's charge that blows up the transistor, not AC input current.
Some FETs might have increased Amps on papers, however in real life things can be different especially when the high RDS (on) values increase the operating temperatures. The part though in this PSU which I am not so sure about is the bridge rectifier, since it can only handle 8 Amps at 100C and usually I see more than 100C at the bridge. So if the bridge goes up to 115-120C its max current drops close to 5A according to its spec sheet so we only have 575W handling, which is much below the 700-720W that the PSU draws from the wall with 580W load.
The APFC choke isn't for filtering purposes, but mainly for boosting voltage, this is why it is also called boost inductor. More about how the APFC converter works and an explanation on both most popular types (DCM, CCM) can be found here: http://www.tomshardware.com/reviews/power-supplies-101,4193-10.html. I think it is pointless to explain APFC's operation here since there is an entire section about it. The bulk cap's charge didn't exactly blow the FET, the FET just couldn't handle the increased Amperage that the cap provided because of the high load. Also the bulk cap drew this load from the mains network and didn't make it out of thin air!
Finally please don't lose sight of the forest for the trees and confuse the readers on a subject that is clear. The fuse that protects the socket into which the PSU was connected dropped, meaning that more than 15 Amps passed through it! Given that this PSU's bridge can only handle up to 8 Amps normally its fuse should be less than 8A, however it survived not once, but both times that we tried it! This clearly shows a major flaw and on the same times proves that there is a short on the PSU's primary side, which draws LOTS of current! In any case the main fuse should open the circuit in order to protect the user and the equipment from damage or even fire. The main fuse not blowing is anything else but normal! I just have to point this out
Put a 1mH APFC choke in series with a short circuit and 350V, it'll limit current rise time seen from the input side to 35A/ms, so it does filter regardless of what its other purposes in the circuit is just like any inductor would.
I missed the paragraph about the breaker tripping. If you have RCDs or AFCI breakers, they could be tripping due to faults other than a short-circuit. If you have class-A relays, it could also be tripping faster than the fuse is able to blow.
Edit: missing a 0... 350V/1mH = 350A/ms.
That is deeply troubling, especially coming from a highly reputable hardware vendor like EVGA and who I have nothing but the highest respect for (I've been buying their GPUs solely since the GTX 275).
I sincerely hope EVGA isn't quick to get back because they are trying to find a root cause with Superflower and rule in or out a bad manufacturing batch that got past Superflower, or a bad design. I hope for EVGA it's the former.
@ANDREWJACKSONZA If you mean about the blowing PSU in smokes, nope this is another PSU.
@10TACLE Once we have a response from EVGA we will post it of course. To be frank I waited for a second sample from the time I notified about this problem, however I was not offered one. Also from the start EVGA didn't send any B3 samples and this is why I had to buy them on my own (given the popularity of this line), in order to check them out. With the first chance I will also buy affordable PSUs from other vendors, which usually aren't offered as review samples, in order to evaluate them.
I would buy another 450 B3 but it is out of stock every where I searched in Europe. If I manage to find one I will try the same test again of course. Nonetheless I faced problems with other B3 models as well, which I will mention in the next B3 review.
If what you wanted to simulate is the current rise rate or filtering, no need to: Vinductor = L * di/dt. If the voltage is relatively constant for the period of interest, di/dt = V/L and I just realized that there is a missing 0 in my previous post, should have been 350A/ms for 350V and 1mH.
Personally, I've seen breakers trip faster than fuses can blow quite a few times. On one particularly careless day, I managed to short a PSU to chassis tripping the breaker three times before its 6A input fuse died. Slow-blow fuses don't necessarily die on the first short when there is a faster device upstream.
Or OEM replacements for that matter. I always recommend to people upgrading their OEM PC's crappy PSU. Recent case in point: a friend was in a bind with a dead PC used for his business and had no time to diagnose and/or build another. He rushed out to Best Buy and found an HP OMEN "gaming" PC on sale with a GTX 1070 Founder's Edition (the same thing Nvidia sells direct from their website).
I went over there after he got back and the first thing I did was open the case and look at the PSU: as expected it was a non-branded 500W PSU with terrible max sustained wattage output specs with a low temp threshold. Yeah that's not happening. After he closed up we went to Fry's and picked up an EVGA 550W G2. The 550W G3 was $20 more which is the only reason we got the G2. Thank God we did apparently.
That's not really the situation where overpower protection is needed most of the time, though. If there is a shortage with a yellow wire with a low enough impedance, short-circuit protection will fail to detect it, and then as current rises it's up to OPP to save the day.
Or if companies were smart they'd still be doing multiple 12V rails.
For outputs under 40A, you should be fine with a single rail and working OCP/OPP set to the correct value: even #24 wires can pass safely 40A long enough to trip a working OCP. Not that having multiple output rails would have made any difference in the case of a primary-side catastrophic failure anyway.
Also, having multiple rails does you very little good if their current limits are well beyond what they should have been as it appears to be the case here. A functional 40A OCP limit on the 12V rail here would have prevented the PSU from delivering more than 480W instead of allowing it to self-destruct at 580W.
IMHO, if a PSU represents a potential fire hazard, it seems fitting that it should not receive a better score than a PSU that doesn't. There should be a couple of other "deal-breaker" grade problems that, as I see it, should play into the overall performance rating so the product is not incorrectly portrayed on the comparison charts that you use in the Best PSU.
Examples include:
EVERY PSU or device handling high power is a potential fire hazard and not immune to the occasional material or manufacturing defect that could cause an otherwise good design to fail catastrophically.
With a sample size of one, it is impossible to tell whether this is an isolated defect, a rampant QA problem from a parts supplier or an outright design flaw. The only thing that can be said is that one sample failed. While this may not inspire confidence, it isn't sufficient to definitively condemn the product.
They should be the same platform, so I expect them to be very similar. Internals comparison:
http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story5&reid=500
http://www.tomshardware.com/reviews/evga-supernova-850-g3-psu,4930-3.html
Also, you can see noise ratings at http://cybenetics.com though it seems they don't have the 750 G3 so nevermind.
About noise, checking EVGA site I saw that the fan curve in the 750G3 is much more relaxed then the 850G3 (and 1000G3), so I was wondering how that would translate in the noise output.
But considering the 750 G3 has the same internals and less power then the 850 G3, can that improve the hold up time? I read the 850P2 and 750P2 review here on tomshardware, and the 750 P2 has a much better hold up time. Could that also happen with the G3 series?