PSU tier list 2.0

Page 167 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
Okay okay, I agree with you on the definition of holdup time in accordance with the Intel docs.

But anyway, that's besides the point that the EVGA G2 650 most likely still has this problematic issue. I am aware that the voltage at which PWR_OK drops is not affected by the holdup cap.
 


Then you'd be wrong, because according to Aris, it does.

Hi there!

I am afraid that some Super Flower units have this hold-up time issue (besides test results I have also an internal confirmation on this). However in the G2 reviews I didn't have my new scope, so I was only testing the power good signal's time and not the rails' hold-up time, so I don't know if in these units the rails go out of spec before or after the power good signal is de-asserted. This means that the situation can be even worse than it is.

According to my database the only G2 units that didn't meet the ATX spec (regarding the power good signal's hold-up time) are the 550 G2 (12.4 ms) and the 1300 G2 (14.4 ms). The rest had higher than 16 ms but as I mentioned, I didn't check the rails' hold-up time in these reviews due to lack of a high accuracy scope, so I can't be sure if their performance is fully inline with the ATX spec.

If I manage to get my hands on some G2 units I will take full hold-up measurements, but to be frank I don't think that EVGA will be willing to send them.

BR
Aris
Click to expand...
 
What I'm saying is on units like the 550 G2 and 1300 G2, those are the safer units, even though they are out of ATX spec for AC_LOSS to PWR_OK. It's the units that have really high, way above ATX-spec for AC_LOSS to PWR_OK like the 650 G2 and 750 G2 that are less safe, because the means by which they have those higher times is likely from setting the voltage trigger point lower and cheating. The 550 G2 and 1300 G2 most likely don't cheat at all.

All the units have their downsides. If the 550 G2 and 1300 G2 drop the PWR_OK signal before the voltages go out of spec, then it is a good thing, even if the AC_LOSS to PWR_OK time is less than 16ms. The only role that will play really is you'd need a UPS with a lower transfer time if you are purchasing the 550 G2 or 1300 G2. So, let me sum it up, based on my assumptions:

550 G2: below-spec AC_LOSS to PWR_OK time, but has safe voltage.
650 G2: above-spec AC_LOSS to PWR_OK time, but has unsafe voltage.
750 G2: above-spec AC_LOSS to PWR_OK time, but has unsafe voltage.
850 G2: unsure
1000 G2: unsure
1300 G2: below-spec AC_LOSS to PWR_OK time, but has safe voltage.

It's a totally reversed situation where the units that are out of spec are the safer ones, and the units in spec are more likely to be less safe.
 
And I'll bet all of the above have plenty of safety circuitry to circumvent any issues arising from sudden power loss.

I have a 550G2 and I'll say this. It takes a full 2 seconds for power on. I've tested it with plug removal, breaker resets, ups trips etc even cold boot it takes a full 2 seconds to start up once the button is pushed before anything happens, so a specific test of only a part of the whole is kinda misleading to me. The 550G2 may fail ATX standards as far as it's only 12ms not the full 16ms required, but that's only upto pwr_ok, it doesn't take into account the time taken after before startup.

The test might be valid, but you are concerned about one tree when surrounded by a forest.
 


Why would you expect the 650 G2 to have safe circuitry when the AC_LOSS to PWR_OK time is double the 550 G2 while it uses the same bulk cap? All evidence is showing that many of them do not have safe circuitry from circumventing issues from power loss.

If I'm correct, the same circuitry that cuts the PWR_OK signal would be the same as under voltage protection, making the UVP threshold too low. If those are indeed the same circuits.
 
Regardless of what cap is used or whatever circuitry is in place, adding an extremely small timer circuit after the pwr_ok would be child's play. You could easily make that 12ms into 120ms or the 26ms into 260ms. That's what I'd call cheating and with the amount of circuitry in a good psu nowadays, even someone like JG and the other psu gurus would be hard pressed to separate a couple of minor diodes and resistors from the rest.

From everything I can gather on this particular standard, its only reliable verification is in separating the garbage from decent psus, which is easy enough to do with just a visual inspection.

It's not a new standard, in fact it's rather ancient, even if amended periodically. It's honestly almost redundant. Find a single account where ppl have suffered because their G2 turned on too fast and I'll give you it may be valid, but otherwise i see it as inconsequential compared to damage caused from out of spec ripple or lack of protective circuitry.
 
Safety and meeting spec do not line up exactly. There is no guarantee that running at 11.4V for extended periods is 'safe', even though it's within spec.

However, with modern, massively overspecced buck converters, I'd be surprised if running at 10V for short periods caused any damage whatsoever. Running at a lower voltage is going to mean that the input is running at slightly higher current than normal... but that's it. That 15% drop shouldn't cause any more damage than using 15% more power in your CPU by overclocking.

Specs exist to ensure interoperability, not necessarily all safety. Motherboard manufacturers should be expecting that not every product will meet spec every time, and provide adequate protections onboard in the case of e.g. a bad joint somewhere causing excessive voltage drop.
 


See, with your knowledge, you'd be the perfect fit for the JG forums. You could bring some opposition there even.
 
Not everybody is looking to argue, which is mostly what I see over there when somebody has a differing opinion. That, or the tight little "clique" they have just resorts to turning up their noses. Not my cup of tea. Probably not SS's either. Still, I have little doubt that he'd have no problem advocating a more appropriate even view than what I usually see over there.

Mostly, I don't even bother going there anymore except to look at new reviews, which I then take with a grain of salt, for the simple fact that there are some with less than cordial dispositions.
 
It isn't when the PSU is turned on that these delays (or lack thereof) are the problem, it's when it is shut off. Some units don't de-assert PWR_OK before the voltages have dropped out of spec.
IMHO, this is a truly minor glitch, and not the sort that causes equipment damage. PWR_OK dropping is supposed to tell the motherboard to shut down; what's a motherboard going to do in 5,10, or even 20 ms. to "get ready" for power loss? I'd guess nothing, with the possible exception of halting data output through the SATA ports so a disk isn't writing as the power dies. I was under the impression that many hard drives (especially SSDs) had their own protection against this occurrence; is this not the case?
 


Oh believe me I agree 100% with you. At least here on Toms we are one big happy family 😀

But seriously, it's hard for someone like me when I'm told different things from different people. Some people say it's important, others say it's not. I mean, I'm not like an EE who can take a stance on something like this for myself.
 


Are you stupid? What's wrong with you? My AMD FX-9590 runs every game maxed with no noticable fps drop and I heard that every i7 has microstutter because it uses virtual rendering which is very bad for games. Everyone hear moaning about AMD's processors knows nothing, I can show you a benchmark of how it is better, but I'm not even going to bother since you noobs won't understand anyways.
 


And I thought this thread was boring 😛

 
I wonder if they ALL do the same thing, and we simply haven't known about it before because nobody had the equipment or intent to test for it, or they ALL knew, but didn't want to release that information for fear that nobody would send them review samples anymore. We've already seen multiple examples of dodgy reviews where specific shortcomings were sidelined, so this wouldn't surprise me in the least.
 
So, in order to actually have that ATX standard as a viable option...
1. Power must be denied to the psu for a span of no more than 16ms, so basically a brown out dropping voltages below setpoint (hmm genius running an expensive pc during brown out conditions with no ups...) or semi-instantaneous power depletion to the mains from the electric company. Again, with no ups.
2. Serious assumption that windows is indeed still set to autostart after ac loss.

Now I can see the possible relevance when back in the AT days and psu switches were toggle on/off not ATX soft on/ hold off, that that might be a concern, but psu design has come a considerable way since then and I feel like it's nothing more now than an excuse to say something bad about a good psu or something good about a bad one. I personally would much rather have a quality 550G2 psu with its outstanding ripple and great everything else except for its 12ms pwr_ok, than to have spent less for a 550GS unit with mediocre everything, including protections, just because pwr_ok is higher than 16ms. And yet that is exactly what some claim, the GS is a better psu than the G2.

In a pigs eye...
 
UPSs often take 10+ms to switch from AC to internal. If the power goes out and your PSU can't handle 10ms of no power, your PC is shutting down, UPS or not.

Also, none of this is likely to damage hardware. It's more about not losing unsaved work, or OS corruption. Could be a super-cheap office machine, but if it's got your invoice software on it...

UPSs are generally not needed in most of the first world, now. I've had gear that has been up for well over a year with no backup power source. Power cuts in many places are almost always either pre-announced ("We need to replace a transformer, and we're doing it on XXXX-YY-ZZ"), or during a major storm. I don't remember ever having a power cut that wasn't at least somewhat expected, with the exception of stuff like RCDs tripping.

However, there's still loads of things in the power grid that can cause a 5-10ms dip - tap changes for voltage regulation on transformers, for example.

I don't see how windows auto-restarting has anything to do with this.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom