Why you Should Not Purchase the EVGA G2 Lineup Especially in Homes with Frequent Brownouts

[turkey3_scratch]

(NOTE FROM CREATOR OF THIS POST YEARS LATER: I DON'T SUPPORT THE CONCLUSIONS I DREW ANYMORE)

 

The EVGA G2 lineup is considered by many to be very high quality, the P2 considered fantastic, but the latest testing by professional reviewers shows that some or perhaps all of these units have a serious flaw that won't harm the PSU but could harm your other hardware. It's important to explain the concepts before anything. Very few reviewers test for this stuff - Aris, who does power supply reviews on Tomshardware and Techpowerup, very recently started a new test. This new test shows that many units have newly revealed flaws, and this problem exists with the EVGA G2 and P2 lineup. The EVGA G2 series is a direct copy of the Superflower Leadex Gold series. The EVGA P2 series is a direct copy of the Superflower Leadex Platinum series.

 

Old Hold-Up Time Tests

 

In the past, most power supply reviewers would test "hold-up time" as the amount of time from when there is AC loss (i.e. power outage) to when the unit shuts off, while under full load. For example, if a power supply is under 100% load and there is a power outage, if it takes 20ms for the power supply to shut off, it would be defined as having a hold-up time of 20ms. However, this is not an actual test with what hold-up time really is, and new tests reveal new realities. The old hold-up time tests are not in exact accordance with the definition of hold-up time.

 

Hold-Up Time

 

Hold-up time is defined as the time period from when there is AC loss to when a voltage goes below ATX specification. The ATX specification for the 12V rail is from 11.4V to 12.6V. So if there is AC loss and it takes a power supply 20ms until its 12V voltage gets down to 11.4V, then 20ms is the hold-up time. Some people incorrectly define hold-up time as the time between AC loss and when the power supply shuts off. This is incorrect; a power supply may actually shut itself off after hold-up time. Hold-up time is strictly the amount of time from AC loss until a voltage goes out of specification.

 

AC_LOSS to PWR_OK_OFF

 

When there is AC loss, the voltage outputs of the power supply begin to drop to lower numbers. It is the job of the circuitry of the power supply to detect this and shut off the unit before the voltages go below  the ATX specification. Basically, under-voltage protection. Before any of those voltages go out of specification, the power supply is supposed to cut the PWR_OK signal. The PWR_OK signal is a cable in the main 24-pin ATX cable used to communicate with the motherboard. When on, it's telling the motherboard that its voltages are all safe. When the PWR_OK signal is cut, circuitry is supposed to shut off the power supply as fast as possible so the voltages don't go out of specification.

 

The latest testing methods in renowned power supply reviews are AC_LOSS to PWR_OK tests. This tests the duration of time it takes from there being AC loss to the computer dropping the PWR_OK signal. So, for example, if there is a power outage, your computer is under full load, and your power supply has an AC_LOSS to PWR_OK time of 18ms, then in 18ms your power supply will tell your motherboard, "Hey! Voltages are about to get low. Better shut me off now." And then the unit will be shut down quite rapidly once the PWR_OK signal is dropped.

 

There is a catch though: some power supply units drop the PWR_OK signal after the voltages go out of specification. This means when there is AC_LOSS, the power supply will continue providing the computer with energy as those voltages go deeper and deeper below the ATX specification. Once it gets to a certain low point, then it'll cut the PWR_OK signal, after harm has already been done. In a way, this is a cheating method used by power supply manufacturers to do well on old hold-up time tests that check the duration of AC_LOSS to PSU shutdown time. This cheating method is also dangerous.

 

EVGA G2 550 Flaw

 

The EVGA 550 G2 has this exact flaw. It drops the PWR_OK signal at 10.8V. This information can be found at the Tomshardware Superflower Leadex Gold 550 review here: http://www.tomshardware.com/reviews/super-flower-leadex-gold-550w-power-supply,4416-4.html

 

Quote

The hold-up time tests don't go well. Not only is the measured hold-up less than 16ms (the ATX spec's minimum), but the Power_OK signal drops after, and not before, the PSU's rails go out of spec. This means that your motherboard gets a false power-good signal from the PSU. Indeed, we measured the +12V rail floating at around 10.8V when Power_OK dropped to zero. This is a very low voltage level that applies lots of stress to the voltage regulators of components fed by +12V.

In a high-end PSU like this one, we didn't expect such nasty behavior. We have to admit that we're very disappointed by Super Flower's decision to drop the power-good signal so late, which is probably done to give the false impression that the hold-up time lasts longer. By the end of our review, this is going to cost to this PSU a lot of performance points. Whereas it might have received an award for performance, there's no way it will now.

Some of you may be thinking, "Oh, it's just 10.8V, no big deal." No, this is  a big deal. Power supply experts like Aris (who did the above review) say it is a serious matter, as well as the experts on the Jonnyguru forums. It is a serious issue, and the EVGA 550 G2 should not be purchased because of it, especially in areas where there are frequent power outages.

 

It is important to note that Aris did do a review on the EVGA 550 G2 here, but that was before these new tests came about.

 

EVGA P2 Series Assumed Flaw

 

Most likely, the entire EVGA P2 lineup has the same issue, but right now it is merely speculation (but a pretty good one). The Superflower Leadex Platinum 550 review here: http://www.tomshardware.com/reviews/super-flower-leadex-platinum-550w-power-supply,4281-4.html

 

This review was before the PWR_OK to AC_LOSS tests. Hold-up time was tested to be 13.8ms. That is below the ATX specification. There is no doubt in my mind that it probably drops the PWR_OK signal at 10.8V just like the EVGA 550 G2, in order to "cheat" and get good hold-up times on the old hold-up time tests.

 

EVGA T2 is Safe

 

The EVGA Titanium lineup has been tested and does not have this issue.

 

EVGA G2: Rest of Lineup

 

As of now, the non-550W versions of the EVGA G2 lineup may have these issues, but we cannot know for sure. Power supply experts are pretty confident the rest of the G2 lineup does have this problem, which is very important.

 

What to do Now?

 

Realize that sometimes units we think are incredible really are not. Everybody likes to go around forums talking about how fantastic the EVGA G2 lineup is, but with this problem, how can it be? It shouldn't make any high tier on any list, because Superflower cheats to get good hold-up time on old tests, and sets far too low under-voltage protection values. It is a serious matter, because the VRMs of all your hardware can be seriously affected by such a low voltage, 10.8V. There is a reason the ATX specification exists. 10.8V is two times out of the specification.

[STRMfrmXMN]

Just curious: does, say, a Seasonic G-series do this? I'd assume it has the same issue and it's a non-issue for most people living in western society.

[turkey3_scratch]

I don't think the Seasonic G-Series has been tested for this. Testing this requires very expensive equipment that companies like Jonnyguru cannot afford. Aris is capable of affording it, fortunately. Superflower could easily fix the issue by changing the value to a higher voltage. It'd be a simple manner of changing the data on a chip.

 

I live in western society and I get a power outage probably ever couple of months.

 

There is a Zalman unit here http://www.tomshardware.com/reviews/zalman-zm750-ebt-psu,4373-4.html that drops the PWR_OK signal at 10.24V! That's so dangerous. A lot of Sirfa built units also do this cheating mechanism.

 

I wouldn't expect Seasonic to do it.

[Dark_Fuzzy]

For the people who already have these psus,would something like a battery backup stop this from being a problem?
 

[turkey3_scratch]

6 minutes ago, Dark_Fuzzy said:

For the people who already have these psus,would something like a battery backup stop this from being a problem?
 

Depending on the transfer speed of that battery backup and the load the computer is on. If the computer is at 100% load, with the EVGA 550 G2, it takes only 11ms for the 12V to go below 11.4V. So you'd want a UPS of a transfer time less than 11ms. If your computer is not under 100% load, a UPS with higher transfer time would be fine.

 

I own two EVGA G2 power supplies myself. I own the EVGA 850 G2, and the 550 G2. The 850 G2 was purchased before any of this stuff remotely became known, and the 550 G2 was bought before I was informed of these issues. At this point, I'm not sure what plan of action I might take. In my probably rig I've been thinking of getting a Seasonic Snow Silent 1050W.

 

But yeah, I own two G2 PSUs, so I'm not anywhere near anti-EVGA. Not at all. I'm just a little upset that these power supplies have these issues (I know my 550 G2 does for sure).

[DrM]

Considering this affects the G2 and likely affects the P2 lineup, would it be safe to assume that its also an issue with my B2 750?

 

also i really hope evga or superflower take notice of this, because its such a good psu otherwise and from the op, it sounds like a cheap and simple fix

[Imabigmac]

This is intriguing, will be interested to hear more about this as I'm sure other testers will start to look into this. The G2 550 and 650 are identical or nearly identical, so I would assume my PSU is affected as well. I guess its time to buy a UPS. 

[quan289]

Edit: I had misread...Thought he had mentioned the GS series for some reason.

[turkey3_scratch]

8 minutes ago, DrM said:

also i really hope evga or superflower take notice of this, because its such a good psu otherwise and from the op, it sounds like a cheap and simple fix

Superflower is aware of it because they deliberately did it. They want their PSU to seem like it has better "hold-up time" than it really does. They were the ones who put the 10.8V value in their integrated circuit in the PSU.

[2FA]

38 minutes ago, turkey3_scratch said:

The EVGA 550 G2 has this exact flaw. It drops the PWR_OK signal at 10.8V. This information can be found at the Tomshardware Superflower Leadex Gold 550 review here: http://www.tomshardware.com/reviews/super-flower-leadex-platinum-550w-power-supply,4281-4.html

That link goes to the Leadex Platinum review, not the Gold.

[turkey3_scratch]

1 minute ago, DeadEyePsycho said:

That link goes to the Leadex Platinum review, not the Gold.

Oops, my bad! Pasted the wrong link, fixed now with the correct one. http://www.tomshardware.com/reviews/super-flower-leadex-gold-550w-power-supply,4416-4.html

[DrM]

4 minutes ago, turkey3_scratch said:

Superflower is aware of it because they deliberately did it. They want their PSU to seem like it has better "hold-up time" than it really does. They were the ones who put the 10.8V value in their integrated circuit in the PSU.

I guess i worded that badly.  i meant something more like they realize itlll bring bad publicity/lower sales and change it

[2FA]

Another thing, doesn't the entire G2 lineup vary quite a bit in how well they perform in these analysis tests?

[turkey3_scratch]

On 4/22/2016 at 6:47 PM, DrM said:

I guess i worded that badly.  i meant something more like they realize itlll bring bad publicity/lower sales and change it

I doubt it will, though, because a lot of people love these PSUs (like a lot) and it'll be hard for them to break free of any biases and look at this from an objective standpoint.

 

On 4/22/2016 at 6:48 PM, DeadEyePsycho said:

Another thing, doesn't the entire G2 lineup vary quite a bit in how well they perform in these analysis tests?

Only the 550 G2 has been tested with these tests. It's only a guessing game as to if the rest of the series has this problem, no certified proof, but there is proof the 550 G2 has this flaw.

 

But evidence shows the P2 series might have these issues, but have yet to be tested for sure.

[DrM]

at the end of the linked article, it mentions that the superflower psu uses some capxon caps instad of all jap caps in the g2 550. it even recommends the g2 over the super flower unit, which, i quote: (http://www.tomshardware.com/reviews/super-flower-leadex-gold-550w-power-supply,4416-10.html)

Quote

To conclude, if you want only high-quality caps in your PSU and you also want to avoid the SF-550F14MG's problem with its power-good signal, then stick with EVGA's 550 G2, which uses only Japanese caps and, although it has a lower than 16ms hold-up time, its power-good signal is inline with the actual hold-up time. 

basically, the g2 is safe and has the exact same hold up time as ac loss to pwr_ok holdup time. it seems the superflower one might be more an issue with the lower quality caps and overly aggresive marketing.

[turkey3_scratch]

Firstly, both the 1000W G2 and 1300W G2 use Capxon caps. Also, the EVGA 550 G2 was tested by Aris before he had these new tests. If you look at the article for the 550 G2 review http://www.tomshardware.com/reviews/evga-supernova-550-g2-power-supply,4244-4.html it only tests hold-up time (the right way actually) which is below spec. If they tested AC_LOSS to PWR_OK, it would be at 10.8V.

 

The Capxon caps are no big deal, those caps are dealing with little amount of ripple and virtually no heat. The only reason that the G2 uses more Chemi-Con than the Leadex is for EVGA marketing, because EVGA knows everyone throws a fit if they see Chinese.

 

Aris himself told me that the 550 G2 most likely has this problem. It was in the comments section of the Tomshardware Corsair RM550x review that he said this.

 

Here is what Aris told me:

Quote

As i told you before, the Power-OK test is new, so very few PSUs have been checked so far. The EVGA G2 550, hasn't been tested so far for Power-OK, only for the normal "Hold-UP time" test (failed at this test). But since it is based on the same platform with the SF Leadex 550, it is logical to assume that these 2 units will have similar performance at the Power-OK test as well. 
At this moment the RMx550 is an excellent all around PSU that passed all the tests that was tested. 
(P.S. I'm anxious to see how Seasonic's platforms will perform at this new Power-OK test. )

 

[turkey3_scratch]

Slight update. It seems that the P2 series is most likely safe, as the units tested in it have been fine, though the Leadex Platinum 550 does have the behavior. The T2 as always is safe. Currently the only suspect is the 550 G2.

 

Update. It's all the other way around. The 550 G2 is likely safe, while the 650 G2 and 750 G2 are not. It is very confusing to learn that the old hold-up time tests did not test hold-up time; they tested AC_LOSS to PWR_OK. With the new equipment, they have been renamed to AC_LOSS to PWR_OK. The 550 G2 has an AC_LOSS to PWR_OK time of 11ms. The 650 G2, which uses the same bulk capacitor, has double the time, meaning the 650 G2 is unsafe and the 550 G2 is safe. The 750 G2 is also unsafe, and potentially the 850 G2.

[TheCaptain53]

Will we start seeing re-testing of previously reviewed PSUs? I have an EVGA 650 G2 and would like confirmation that my PSU has this issue, and whether or not to invest in a UPS.

[don_svetlio]

6 hours ago, turkey3_scratch said:

Slight update. It seems that the P2 series is most likely safe, as the units tested in it have been fine, though the Leadex Platinum 550 does have the behavior. The T2 as always is safe. Currently the only suspect is the 550 G2.

 

Update. It's all the other way around. The 550 G2 is likely safe, while the 650 G2 and 750 G2 are not. It is very confusing to learn that the old hold-up time tests did not test hold-up time; they tested AC_LOSS to PWR_OK. With the new equipment, they have been renamed to AC_LOSS to PWR_OK. The 550 G2 has an AC_LOSS to PWR_OK time of 11ms. The 650 G2, which uses the same bulk capacitor, has double the time, meaning the 650 G2 is unsafe and the 550 G2 is safe. The 750 G2 is also unsafe, and potentially the 850 G2.

 

2 minutes ago, TheCaptain53 said:

Will we start seeing re-testing of previously reviewed PSUs? I have an EVGA 650 G2 and would like confirmation that my PSU has this issue, and whether or not to invest in a UPS.

you answer lies above

[TheCaptain53]

1 minute ago, don_svetlio said:

 

you answer lies above

Testing on the 650 G2 was with old testing methodology, meaning that whilst it is likely, the poor AC_LOSS to PWR_OK is speculation. Unless I'm wrong, then someone please correct me.

[turkey3_scratch]

The 650 G2 has really good AC_LOSS to PWR_OK time which is why it is suspicious. The reason it's really good is probably because they set the voltage to 10.8V. It has the same bulk capacitor as the 550 G2, yet it happens to have double the AC_LOSS to PWR_OK time? Not really going to happen unless they lower the trigger voltage.

 

It's a confusing situation where everything below ATX spec for AC_LOSS to PWR_OK time is likely to be a lot safe than everything above and within spec.

[Curufinwe_wins]

Just a thing here but wouldn't newer motherboards with digital delivery have some sort of self shut down when voltages come in out of spec? 

 

Also I am a bit confused with this seeing as I don't see how 10ms of low voltage through a vrm is going to matter realistically... "Stress" sure, but failure from power issues on a very irregular failure seems extremely unlikely. 

 

BTW I would love to see the rest of these lineups tested.

[manikyath]

On 22-4-2016 at 1:07 AM, STRMfrmXMN said:

it's a non-issue for most people living in western society.

it may actually be, because devices like UPSes may base themselves on these numbers, and if they're false, they may not operate properly.

 

that said, a full 50hz sine wave is 20ms, meaning once rectified there's a "spike" every 10ms, so theoretically if you miss out one spike on the sine wave you're already getting dangerously close to spec. (one could argue that before mentioned 13.8ms time is very dangerously close once you start looking at peak to peak times)

[STRMfrmXMN]

7 hours ago, manikyath said:

it may actually be, because devices like UPSes may base themselves on these numbers, and if they're false, they may not operate properly.

 

that said, a full 50hz sine wave is 20ms, meaning once rectified there's a "spike" every 10ms, so theoretically if you miss out one spike on the sine wave you're already getting dangerously close to spec. (one could argue that before mentioned 13.8ms time is very dangerously close once you start looking at peak to peak times)

I'm pretty sure this problem does not apply to the EVGA G2, but to the original unit it was based off of from Super Flower. It's not a direct rebrand as far as I understand.

[Curufinwe_wins]

10 hours ago, STRMfrmXMN said:

I'm pretty sure this problem does not apply to the EVGA G2, but to the original unit it was based off of from Super Flower. It's not a direct rebrand as far as I understand.

I still don't see how this would actually provide enough stress on VRM's, ESP modern motherboards which tend to go insane mode, to cause any real long term damage from isolated incidents.

 

I can imagine this being a problem if it happens on a semi-regular basis, but I think people are massively underestimating the capabilities of modern hardware (not that I think it would matter even on most older systems that were designed with safety factors to account for the much "dirtier" power supply market.)