Question Having trouble picking a PSU for uber-build

[MoreMoneyThanSense]

I don't like buying twice; last time I bought a PSU was over a decade ago and it lasted me nicely, but I am looking to make a new computer and I want a PSU that will cover all my bases and make sense given the rest of the build.

My intended build: 3950X CPU, x570 Aorus Xtreme, at least one 2080 ti, 64 GB Trident Z Neo RAM, custom watercooling loop (mostly Heatkiller), D5 pump, >= 10 fans, and a gen 4 NVMe SSD (and possibly a couple backup drives). I would be pushing OC wherever I can, e.g. the 2080 ti can pull 380W with the right BIOS flash. I'm estimating the 3950X to pull > 200W.

1. I plan to be using this thing long-term and with the system running 24/7
2. Would love for the PSU to be quiet even when pushing the system... the longer it can handle things quietly, coolly, and stably without the fans making noise, the better. I'm already planning to use three 360 rads in my watercooling loop to help keep things quiet, but I want the PSU quiet too.
3. Not sure yet if I'll eventually want to add a second video card or not, but I'd like the PSU to be able to handle extra hardware if I were to add it.
4. Should be good-quality parts, good protections, low temperatures, etc.

I've mainly been eyeing Corsair, EVGA, and Seasonic brand PSU's in the 1000W+ Platinum/Titanium range but I'm finding it very hard to pick one that is right for me because I am not as familiar with their sound profiles, nor am I confident in what level of wattage would be appropriate. It seems to be that 1000W "should be enough" but I wouldn't want to add more hardware down the line only to find that I am stretching the limits of my PSU.

Any input appreciated, thanks!

 

[jonnyguru]

I'm finding it very hard to pick one that is right for me because I am not as familiar with their sound profiles

https://cybenetics.com/index.php?option=power-supplies

 

[MoreMoneyThanSense]

https://cybenetics.com/index.php?option=power-supplies

Thank you for the link. I am a little surprised it doesn't have the EVGA Supernova 1000 T2 but I tried putting the rest I was looking at into a table as reference. Looks like the EVGA PSU's tend to be louder than the Corsairs which are in turn louder than the Seasonics, hmm.

MODEL	FORM FACTOR	WATTAGE	AVG	AVG EFFICIENCY	VAMPIRE	AVG PF	AVG NOISE	EFFICIENCY	NOISE RATING	DATE
EFFICIENCY (%)	5VSB (%)	POWER (W)	OUTPUT [DB(A)]	RATING
SSR-1000TR Ultra (Sample #2)	ATX12V	1000​	91.644​	79.663​	0.0532916​	0.992​	13.16​	ETA-A+	LAMBDA-A++	2/26/18​
SSR-1000TR Ultra	ATX12V	1000​	91.703​	79.825​	0.0548672​	0.992​	12.08​	ETA-A+	LAMBDA-A++	1/16/18​
AX1000	ATX12V	1000​	92.107​	79.719​	0.0470888​	0.99​	22.85​	ETA-A+	LAMBDA-A	11/22/18​
AX1200i	ATX12V	1200​	89.489​	82.769​	0.0539247​	0.995​	22.64​	ETA-A	LAMBDA-A	5/3/18​
AX1500i	ATX12V	1500​	91.264​	80.974​	0.0564061​	0.993​	27.55​	ETA-A+	LAMBDA-A-	5/2/18​
AX1600i	ATX12V	1600​	92.478​	82.158​	0.0467618​	0.992​	23.29​	ETA-A+	LAMBDA-A	12/19/18​
AX1600i (Sample #2)	ATX12V	1600​	92.288​	81.676​	0.0481594​	0.992​	23.7​	ETA-A+	LAMBDA-A	1/2/18​
SuperNOVA 1600 T2	ATX12V	1600​	91.769​	78.664​	0.111363​	0.992​	34.51​	ETA-A+	LAMBDA-S++	10/20/18​
SuperNOVA 1600 T2 (Sample #2)	ATX12V	1600​	91.823​	78.814​	0.12028​	0.994​	34.05​	ETA-A+	LAMBDA-S++	11/22/18​
SuperNOVA 1000 G3	ATX12V	1000​	89.652​	76.748​	0.13439​	0.987​	33.81​	ETA-A	LAMBDA-S++	11/13/18​

 

[hftvhftv]

You're looking at top tier equipment, in the ballpark of 1200-1600 watts, you probably could get away with 1000 watts if you go single GPU, but if you ever consider adding another power hungry card like the 2080 Ti you'd definitely need the 1200 watt.

 

[MoreMoneyThanSense]

1200W seemed like a good target but I couldn't find any titanium-level offerings from the three brands I had been looking into.

AX1200i (which is Platinum) seemed like a good pick but I had also seen some iffy reviews on Amazon that the units die after a few years and the warranties don't get respected. https://www.amazon.com/CORSAIR-AX1200i-Platinum-Certified-Modular/dp/B008Q7HUR0#customerReviews

 

[jonnyguru]

Thank you for the link. I am a little surprised it doesn't have the EVGA Supernova 1000 T2 but I tried putting the rest I was looking at into a table as reference. Looks like the EVGA PSU's tend to be louder than the Corsairs which are in turn louder than the Seasonics, hmm.

MODEL	FORM FACTOR	WATTAGE	AVG	AVG EFFICIENCY	VAMPIRE	AVG PF	AVG NOISE	EFFICIENCY	NOISE RATING	DATE
EFFICIENCY (%)	5VSB (%)	POWER (W)	OUTPUT [DB(A)]	RATING
SSR-1000TR Ultra (Sample #2)	ATX12V	1000​	91.644​	79.663​	0.0532916​	0.992​	13.16​	ETA-A+	LAMBDA-A++	2/26/18​
SSR-1000TR Ultra	ATX12V	1000​	91.703​	79.825​	0.0548672​	0.992​	12.08​	ETA-A+	LAMBDA-A++	1/16/18​
AX1000	ATX12V	1000​	92.107​	79.719​	0.0470888​	0.99​	22.85​	ETA-A+	LAMBDA-A	11/22/18​
AX1200i	ATX12V	1200​	89.489​	82.769​	0.0539247​	0.995​	22.64​	ETA-A	LAMBDA-A	5/3/18​
AX1500i	ATX12V	1500​	91.264​	80.974​	0.0564061​	0.993​	27.55​	ETA-A+	LAMBDA-A-	5/2/18​
AX1600i	ATX12V	1600​	92.478​	82.158​	0.0467618​	0.992​	23.29​	ETA-A+	LAMBDA-A	12/19/18​
AX1600i (Sample #2)	ATX12V	1600​	92.288​	81.676​	0.0481594​	0.992​	23.7​	ETA-A+	LAMBDA-A	1/2/18​
SuperNOVA 1600 T2	ATX12V	1600​	91.769​	78.664​	0.111363​	0.992​	34.51​	ETA-A+	LAMBDA-S++	10/20/18​
SuperNOVA 1600 T2 (Sample #2)	ATX12V	1600​	91.823​	78.814​	0.12028​	0.994​	34.05​	ETA-A+	LAMBDA-S++	11/22/18​
SuperNOVA 1000 G3	ATX12V	1000​	89.652​	76.748​	0.13439​	0.987​	33.81​	ETA-A	LAMBDA-S++	11/13/18​

That's actually generalizing it quite a bit. If you look solely at the ratings, I'm sure it's easy to come to that conclusion.

You need to look at each report and see WHERE the fan noise is more (low loads, high loads) to know why the PSU got the "average noise" score it got.

I personally would prefer a PSU that was quieter at low loads. If a high fan RPM at higher loads is hurting the overall score, I don't care because the PC overall should be much louder than the PSU since it's using so much power.

 

[MoreMoneyThanSense]

That's actually generalizing it quite a bit. If you look solely at the ratings, I'm sure it's easy to come to that conclusion.

You need to look at each report and see WHERE the fan noise is more (low loads, high loads) to know why the PSU got the "average noise" score it got.

I personally would prefer a PSU that was quieter at low loads. If a high fan RPM at higher loads is hurting the overall score, I don't care because the PC overall should be much louder than the PSU since it's using so much power.

Is it always a tradeoff like that? I'd want it quiet at both low and high loads however possible -- I find it frustrating when things get noisy just because I am opening up a big program or game or crunching a bunch of math in a script or whatever. I know it's "tougher to do" but that's why I am willing to pay more for it, unless I'm SOL and loud noise is just inevitable somewhere.

 

[jonnyguru]

1200W seemed like a good target but I couldn't find any titanium-level offerings from the three brands I had been looking into.

AX1200i (which is Platinum) seemed like a good pick but I had also seen some iffy reviews on Amazon that the units die after a few years and the warranties don't get respected. https://www.amazon.com/CORSAIR-AX1200i-Platinum-Certified-Modular/dp/B008Q7HUR0#customerReviews

503 reviews with an average of 3.5 stars and you're going to focus on the handful of one stars?

Not sure what "not respecting the warranty" means. It has a 10 year warranty. If it dies. You send it back. That's it.

But IMHO, Titanium is over rated. The efficiency is so close to Platinum it's almost immeasurable. and you're at a point of diminished returns. Furthermore, the complication of the design to squeeze out that extra bit of efficiency is potentially cause for higher failure rates.

Unless I had the budget for an AX1600i (even if I didn't need 1600W), I would opt for an HX1200i any day of the week.

 

[MoreMoneyThanSense]

503 reviews with an average of 3.5 stars and you're going to focus on the handful of one stars?

Not sure what "not respecting the warranty" means. It has a 10 year warranty. If it dies. You send it back. That's it.

But IMHO, Titanium is over rated. The efficiency is so close to Platinum it's almost immeasurable. and you're at a point of diminished returns. Furthermore, the complication of the design to squeeze out that extra bit of efficiency is potentially cause for higher failure rates.

Unless I had the budget for an AX1600i (even if I didn't need 1600W), I would opt for an HX1200i any day of the week.

For the build I am making I would want something with a higher average, yes. I run my systems pretty hard and in my experience fans are usually some of the first things to go. I definitely feel more comfortable with units that don't have occasional scatterings of "used it for a few months / couple years and it just randomly died." Maybe it's user error, but without knowing anything else to go off of, I can't ignore it either.

Since I am running the systems longer-term I whipped up some basic math in a spreadsheet to try to estimate things like cost savings, and admittedly it's pretty small and takes a number of years for it to make any sense on that front, but it is there. Though I don't have the context yet for assessing expected costs of other error/failure points by upgrading to titanium but you're right, to be rigorous I'd have to take that into account too (I'm not aware of how to model those error rates though), but I do care a lot about keeping temps and noise levels low, high-quality build, good voltage reg, ripple, efficiency, spikes, good cabling, modular, etc etc.

The AX1600i looked great... definitely overkill for my build but I haven't taken it off the table either.

Do you generally think platinums are a better pick than the titaniums for most people? Are those other error points enough of a concern / do they occur often enough to make them less worthwhile?

 

[jonnyguru]

Do you generally think platinums are a better pick than the titaniums for most people? Are those other error points enough of a concern / do they occur often enough to make them less worthwhile?

Let's just say I am director of a PSU R&D team and I see the struggles behind trying to squeeze an extra 1% efficiency out of a PSU topology and at the end of the day it's just not worth it. I mean, the end product "works", but you're working outside of a lot of comfort zones to make it work.

The reason I suggested the HX1200i is because it is HELLA mature. It reviewed well, has been around forever, and everyone that has one seems to have no issues with it.

Look at it this way: Do you want a DB9 because it looks cool and it's fast but is prone to mechanical and electrical problems or do you want a Ford F150 that can take out a brick wall and keep on going, even if you haven't changed the oil in 100,000 miles?

 

[MoreMoneyThanSense]

Let's just say I am director of a PSU R&D team and I see the struggles behind trying to squeeze an extra 1% efficiency out of a PSU topology and at the end of the day it's just not worth it. I mean, the end product "works", but you're working outside of a lot of comfort zones to make it work.

The reason I suggested the HX1200i is because it is HELLA mature. It reviewed well, has been around forever, and everyone that has one seems to have no issues with it.

Look at it this way: Do you want a DB9 because it looks cool and it's fast but is prone to mechanical and electrical problems or do you want a Ford F150 that can take out a brick wall and keep on going, even if you haven't changed the oil in 100,000 miles?

Definitely want the the thing that takes out the brick wall, perhaps the paradox I am after is that I want it to take out that wall quietly, haha. I did not realize there were major comfort-zone issues with reaching that extra level of efficiency, not worth the tradeoff to me.

 

[jonnyguru]

Definitely want the the thing that takes out the brick wall, perhaps the paradox I am after is that I want it to take out that wall quietly, haha. I did not realize there were major comfort-zone issues with reaching that extra level of efficiency, not worth the tradeoff to me.

Go ahead and look at the complete Cybenetics report for the HX1200i. You'll see that it's very quiet. On top of that, it uses an authentic Matsushita/Panasonic fluid dynamic bearing fan and not one of these fake "hydro bearings" everyone tries to pawn off as an FDB.

 

[MoreMoneyThanSense]

It indeed looks to be very quiet (I'm slapping a bunch of graphs together to see them all at once) -- random side question but I found it interesting that for the HX1200i there are these lines/pockets at regular intervals under 600W where it suddenly gets even quieter < 17 db for very brief ranges of wattage. More of a curiosity but what's happening here?

There's also a similar (but different) "striping" with the EVGA SuperNOVA 1600 T2, and the graph for the AX1600i is very interesting as well.

 

[jonnyguru]

It indeed looks to be very quiet (I'm slapping a bunch of graphs together to see them all at once) -- random side question but I found it interesting that for the HX1200i there are these lines/pockets at regular intervals under 600W where it suddenly gets even quieter < 17 db for very brief ranges of wattage. More of a curiosity but what's happening here?

If the load is only on the +12V versus the +12V and +3.3V and/or +5V. Since +3.3V and +5V is derived from the +12V, there is always a load on the +12V. And the Corsair PSUs not only control fan speed by temperature, but also load (using an op-amp to measure).

 

[FALC0N]

The reason I suggested the HX1200i is because it is HELLA mature. It reviewed well, has been around forever, and everyone that has one seems to have no issues with it.

I think this an excellent point. There is something to be said for having a mature and reliable platform. Especially when you are tossing $200 at it.