PSU tier list 2.0

Page 187 - 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.
Brief spikes? No. MGS V puts my 970 at 99% period, stays there for hours except when I'm in map or transition. Cpu similar, 50-55%. Hdd is almost constantly spinning, ssd I'm sure is working, because of heat the cpu cooler is bumped up, intake fans are bumped, there's no spikes to it. Even Skyrim puts my cpu to 55ish and cpu to 66ish and no spikes there either.

There's absolutely no way you'll convince me that it'd be acceptable to run any gtx970 system on an overpriced 450w psu. I'll take the cheaper, better built, better output, better warranty 550w psu every day of the week, and that's double on Sunday.
 


Most games cap CPU usage at 55% because they are not programmed well. Programmers hit a wall when it comes to proper threading, as they are engrained with the old style of thinking when it comes to threading..
 


I'm not talking about GPU utilization. You are claiming your GTX 970 is requiring a constant 300W, but that is not true. I don't think you are measuring this stuff at all.

In a GPU review, "Maximum Power" does not mean power at 100% load on the GPU. It means the single highest point in time, which is a brief spike. I'm specifically talking about power spikes, not utilization spikes. Load is at a constant 100% in all of those reviewers tests, yet power fluctuates a lot. This Gigabyte overclocked GTX 970 under a gaming load here is under 100% usage the entire time, yet power is not constant:

h0LToRK.png
 
Overclocked models will reach 240W territory. It all depends on the exact type of model and the exact clock rate. From what I see, Nvidia tends to underexaggerate their power consumption a bit, though. It's typically higher than what their website states.
 
That 75w from pcie, or 75 from 6pin or 150w from 8pin is a generic number. The 18 Guage wire of a pcie on a respectable psu is capable of more. A 6pin has 3 hots, 2 ground and 1 sensor. An 8 pin is 3 hots, 3 grounds and 2 sensors. Power consumption on a asus 97 is 224w stock, on a gigabyte windforce it's 256w stock. 2x 6pin + pcie = 225w. Supposedly. So a 6pin is actually capable of 150w, same as an 8 pin since it carries the same 3x 12v hots. It's why a 6pin-8pin adapter will still work and not melt the conductors. A single 18 Guage conductor of less than 1 ft long is capable of @5A at 12v DC. A 6/8 pin with 3x 12v hots is capable of about 180w total. The only limiting factor of a 6 pin is the 2x grounds, having just the 1 sensor makes the gpu cut back on power draw to @75w, as per spec, 2 sensors allowing full draw of @ 150w.

But none of the specs are written in stone and are stretched on a constant basis. The radeon r9 295x2 had an average draw of 430w and peak of 470w, the hd6990 power draws exceeding the 75w spec of the pcie slot.
 


Where is your source for a Gigabyte Windforce being 256W at stock clocks?
 
I don't know if you're aware, but that's not power consumption of just the card - it's the power draw of the whole setup. Overclockers.com measure the power draw from the wall of the whole computer using a Kill-a-Watt, and that's including the energy losses inside the PSU.

So their whole PC on this chart drew 292 watts AC from the wall in Fire Strike with an OC'd card - that means (assuming that their PSU was 90% efficient at that point) that the total system power consumption was about 262W.
 
If I had a similar setup as them, or a more efficient one (they use an OCed i7-6700K and 3000MHz DDR4 memory - no word on drives) so that I could reliably say my total system power consumption would be similar, then I personally wouldn't have any prejudices against running a 262-300W system on a good 450W PSU.

The Hydro X 450 has two PCIe connectors, an half bridge topology with LLC resonant converter, synchronous rectification with DC-DC converters, very durable caps and a DBB fan, and achieves 1.5% 12V load regulation, 44mV 12 ripple and ~15mv minor rail ripple with small transient response changes - it's a really good unit, and it can output its full 450 watts reliably, cleanly and for a very long time. Out of these 450 watts, 260-300W is a pretty small and sensible portion.
And I certainly wouldn't trade it for a PSU that is 520W, but group regulated!
 
I've seen i5/i7 + GTX970 systems run on 450w units and I've seen them fail with 450w units. I would personally NOT cut it so close whether the build is for myself or a client. There is simply no reason to cut it so close. The price difference is non existent between great 450w, 550w, 650w, and 750w units.
 
Well then, that just begs me to ask. The older 520w etc GB have been an absolutely rock solid unit. Good outputs, very reliable, world wide top sellers for a reason and for the last few years that I can remember, those xfx, seasonic M12 II, antec hcgM have been among the top ranked as far as recommended psu goes. So, apart from lack of compatability with Haswell deeper C-States, what all of a suden makes this newer design so much more attractive that you'd 'certainly not trade'?
 
I agree on this point. I realize those might be good units, but there is no way I'd ever run any system with more than a threshold card, meaning the bottom end of the upper tiered cards like the GTX 750, 750 TI, 950, R7 370, R9 270, etc., on a 450w unit when you can not only stay safely above the probable spike threshold with a 550w unit, but it will certainly be far less inclined to run near it's peak thermal range which means it's going to last longer anyhow.

Also, with a 450w unit you're limited to what you might want to reconfigure later as far as installed hardware, whereas with a slightly higher capacity unit you still have some options. And considering the price of that 450w unit will get you a semi-modular B2 or one of the XFX units, I don't really see how it makes sense if you have a mid to high tiered card regardless of it's actual capacity requirements. Nevermind that you might also want to add another card later, since that's clearly a whole other can of worms.

If I pay 75 bucks for a 450w unit, it better be Platinum or Titanium rated and meant for use with a threshold card.
 


I suppose it's because the 5V and 3.3V rails do practically nothing these days. It's at that point I think where crossload tests are just as important (probably more important) as progressive load tests. As a matter of fact, your computer is probably always crossloading with a heavy 12V and low 3.3V and 5V load. That's why I think DC-DC designs are so much better for a modern day system than a group regulated.

The S12ii and M12ii are so praised because they're surprisingly like the most stable crossloading group regulated designs ever. It's incredible how stable they are. Then you have a PSU like the Be Quiet L9-CM-600W that has horrible crossloading because it's group regulated.

BXsQtJl.png


That is basically your typical group regulated power supply. The S12ii fortunately is much better, but the vast majority are like this. Many people tend to "shun" away crossload tests on Jonnyguru.com because their crossloading is extreme. We are talking almost an unrealistically low load on the 3.3V and 5V rails, as they do have some load. But it's not like those two rails have enough load to really put it in that red area. In the above image, your typical system with a GTX 970 and an overclocked Intel CPU will probably fall someone in the blue or green region on the right.

I mean, if you are running a 300W system on that Be Quiet, there is no way you are going to have 40W of 3.3V and 5V rail load to balance it out into the red zone. At least I can't see that happening. Anyway, Orion basically summed up why I like the FSP Hydro X 450. Not only is it extremely quiet but the double ball bearing fan will have a super long life, and the fan is the first thing to fail in most power supplies. This power supply at full load should be quieter than the EVGA 550 G2 at the same load.

Anyway, a lot of stuff is totally subjective, but I'll have to say that the fan and noise were not things I used to care about much, but after going on Jonnyguru for a while I'm starting to realize that the fan is the most overlooked component on a PSU, and the Hydro X sure as well has a great one. Anyway, I love when we have great discussions like this, I really do.
 
I have never, ever, had a fan bearing in a power supply fail before something else did, at least, not on any quality modern unit from the last fifteen years.

But I don't ever buy junk either. I've had enough failures, here and there, but none of them were fan bearings. I think the last time I ever had a fan bearing fail on a power supply that wasn't already at least ten years old, was back when top mounted power supplies were far more common.
 


Well it's what the experts all day, so I just take their word for it. I mean, nobody wants a crap sleeve bearing on their PSU.
 


Do you have ANY evidence that supports this statement? I'd sure like to see it if you do, especially since it's the first time I've ever heard ANY comment like that at all, much less from any of the experienced reviewers or systems builders.

I agree that ball bearing fans are more desirable, no argument there, but I've never once heard it said that the fans are the most common failure. Pretty sure that's not correct.
 


On cheap PSUs, it is probably true (I have seen it). On a quality PSU, probably not. Hell my Corsair TX850 lasted for ever. Was still good after 8 years.
 


I've had *1*.
Antec VP-450. 2 years old, maybe. It would randomly sound like a shaken coffee can with ball bearings in it.
 
I think really high quality PSUs have not been around long enough to see yet. I mean, the PSUs of these last some years have really exceeded in quality those of the past, and we'll have to wait a solid 10 years or so to see what is failing primarily on these really good units. Do I think the fan could probably fail before the caps on some of these high-end units? We'll have to wait and see.

Jimmysmitty is right, on cheap units the fans probably fail before the capacitors even, which in turn affects cooling and then makes the capacitors die quickly. I mean, you simply don't put sleeve bearing fans in a horizontal mounting position, and nearly every PSU $45 and under has a sleeve bearing fan, and nearly all cases have a PSU configuration that would put that fan in a horizontal configuration. Those sleeve bearings just don't last.

Most cheap PSUs don't use a thermistor for fan control either, so the fan is running at 100% all the time. This is why it often doesn't matter if you have small or heavy load on those budget units, the fan is spinning at a constant 100% speed many times and being sleeve bearing in a horizontal position it just won't last long at all, not to mention the noise and vibration that can result from a sleeve bearing fan in that position.

It was Aris, who does the PSU reviews on Tomshardware and Techpowerup, who said the fan is the first thing to fail most commonly. I'm trying to find the source of where he said it, but I know he did and I know others have, too. I don't just pull these things out of my rear end 😛 now whether they do or don't have the data to back that up, I don't know. But I'm just telling you what I have heard from people with a lot of expertise in this field.

Edit: I found the quote. http://www.tomshardware.com/reviews/power-supplies-101,4193.html

Some of you might not be aware of this, but a PSU's cooling fan is usually the first part to stop working, at least in good-quality PSUs, so we will dedicate some time to discussing cooling fans as well.
Click to expand...

So it's the other way around, in good quality units the fan is the first to fail.
 


I had a VP450 do the same thing. I just ended up replacing the fan.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom