PSU Design

Message boards : Number crunching : PSU Design
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Profile -= Vyper =-
Volunteer tester
Avatar

Send message
Joined: 5 Sep 99
Posts: 1652
Credit: 1,065,191,981
RAC: 2,537
Sweden
Message 1975417 - Posted: 16 Jan 2019, 14:58:26 UTC - in response to Message 1975409.  

Is there a definitive answer to which is better and why, single rail or multiple rail PSU's. The dearer up-market ones seen to have single rail designs.


https://www.youtube.com/watch?v=rC-ImSeYn2A

_________________________________________________________________________
Addicted to SETI crunching!
Founder of GPU Users Group
ID: 1975417 · Report as offensive
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22161
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1975446 - Posted: 16 Jan 2019, 17:55:47 UTC

And to keep it simple.
A single rail PSU should be more stable, but may be less efficient at part load. But like many things it depends on the exact design.....
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1975446 · Report as offensive
W-K 666 Project Donor
Volunteer tester

Send message
Joined: 18 May 99
Posts: 19014
Credit: 40,757,560
RAC: 67
United Kingdom
Message 1975862 - Posted: 19 Jan 2019, 1:35:49 UTC - in response to Message 1975572.  

Many thanks for the replies. Not quite sure I'm any the wiser though. It would seem to me that splitting the load 2 or 3 ways would balance out the power load for stability rather than shoving it all down 1 rail, but I appear to be wrong. Anyhow the low end 750W max PSU's that I would ever need come with multiple rails anyway.

I'm simply not in the market for 3 x GTX 1080 in a single machine like some people can apparently afford. Distributed computing is simply a hobby to me, not a way of life, and is funded accordingly on a pension. My low end machine runs a 750Ti quite happily on a 550W PSU, I only upgraded that because it was an office Dell machine with a 350W PSU meant for 9-5 buisness use, not 24/7 crunching. My top ma/chine is designed as a server anyway, and has 2 x GTX960 cards in it and it came with an 850W PSU as standard, more than adequate.

But the replies are appreciated, thank you.

The problem I see with multiple rails is, do you know how to balance the loads, or would you, as most do, run the psu output cables as neatly as possible. An example, of neatly, being to take one output and run it near the front of the case and power all the hdd's, the cd/dvd drives and the front panel fans. Not realising that the start/stall current of a motor is, as a rule of thumb, about 3.5 times the nominal running current.
ID: 1975862 · Report as offensive
Profile ML1
Volunteer moderator
Volunteer tester

Send message
Joined: 25 Nov 01
Posts: 20147
Credit: 7,508,002
RAC: 20
United Kingdom
Message 1975867 - Posted: 19 Jan 2019, 1:58:17 UTC - in response to Message 1975862.  
Last modified: 19 Jan 2019, 2:03:45 UTC

The problem I see with multiple rails is, do you know how to balance the loads...

Therein lies the design compromise and the usage and utilization problem...

Single rail design is the obvious way to go. Better lower supply ripple and lower noise. Far more responsive to load. And no concerns for the user about needing to load balance on separate less powerful supply rails.

Additionally: Cheaper to make due to fewer supply ripple capacitors needed. The single rail controller should be no more expensive than for multiple rails, and you can get a cost reduction for spreading the load over more phased power transistors... Less supply ripple for the supply capacitors also so a further cost reduction there also.

No brain needed: Go single rail.


Shame that Marketing and profits artificially bump up the price...

Happy cool crunchin',
Martin
See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)
ID: 1975867 · Report as offensive
Profile Wiggo
Avatar

Send message
Joined: 24 Jan 00
Posts: 34744
Credit: 261,360,520
RAC: 489
Australia
Message 1975868 - Posted: 19 Jan 2019, 2:03:38 UTC

Where dual rail PSU's really fall down is in systems with GPU's that pull all their power from the motherboard which usually leads to the motherboard rail becoming unstable and overloaded while there is almost no load on the other rail.

With multiple rails you must work a balance out between those rails while good single rail PSU's don't require that at all and why I prefer to use good single rail PSU's wherever possible over multi-rail jobs.

Cheers.
ID: 1975868 · Report as offensive
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22161
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1975910 - Posted: 19 Jan 2019, 8:44:30 UTC

Single rail is certainly easier to design and manufacture, the component count is lower, and not just capacitors but power devices, resistors, would components etc..
Internal cabling is very easy to manage in both single- and multi- rail designs, and external, is up to the user. I don't know of a single-rail computer PSU that doesn't have a whole variety of cables and or sockets sprouting from it. Indeed being presented with a totally de-labeled PSU without taking the case off it is all but impossible to tell what type it is.
Properly designed a multi-rail should be more efficient across a wider range of loads, but that isn't always so, but most consumer grade PSUs are a long way down on the best industrial design devices, but then with industrial devices they are designed against a known load whereas consumer devices have to cope with a much wider range of loads.
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1975910 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13722
Credit: 208,696,464
RAC: 304
Australia
Message 1975917 - Posted: 19 Jan 2019, 11:03:05 UTC
Last modified: 19 Jan 2019, 11:05:57 UTC

Looking at the Wiki on computer PSUs, the rise and fall of multi-rail PSUs was a result of regulatory requirements, and was brought in as part of the ATX 12V v2.2 spec.
3.2.4. Power Limit / Hazardous Energy Levels
Under normal or overload conditions, no output shall continuously provide more than 240 VA under any conditions of load including output short circuit, per the requirement of UL 1950 / CSA 950 / EN 60950 / IEC 950.

The requirement was later deleted from the ATX 12V v2.3 spec.
Single vs. multiple +12 V rail


{I'd always thought it was a cost thing- lots of lower current rails would be cheaper to build than one honking great rail for a high power PSU.}
Grant
Darwin NT
ID: 1975917 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13722
Credit: 208,696,464
RAC: 304
Australia
Message 1975956 - Posted: 19 Jan 2019, 20:17:53 UTC

Modular or non-modular is more of an issue IMHO.
Having another connector in the wires just adds another point of potential failure. The actual risk is extremely low, but it is doubled by having 2 connectors instead of just one.

When repairing electronics for a living, connectors were a bigger source of faults than component failure, or dry joints (faulty soldering). Devices that run continuously tend to be at less of a risk than those that turn off & on due to the effects of thermal cycling (heating & cooling resulting in expansion & contraction). And the higher the voltage and or current (particularly current), the greater the risk of issues even without the effects of thermal cycling.

Semi-modular PSUs are the best compromise IMHO. The high current Main & Aux supplies are hard wired, and the low current supplies (storage, PCIe external power etc) are removeable.
Grant
Darwin NT
ID: 1975956 · Report as offensive
Profile Wiggo
Avatar

Send message
Joined: 24 Jan 00
Posts: 34744
Credit: 261,360,520
RAC: 489
Australia
Message 1976029 - Posted: 20 Jan 2019, 7:49:17 UTC
Last modified: 20 Jan 2019, 7:50:46 UTC

Antec use to be my preferred PSU's until they went dual rail years ago so I went onto Corsair instead (all problems were solved).

Cheers.
ID: 1976029 · Report as offensive
Profile Wiggo
Avatar

Send message
Joined: 24 Jan 00
Posts: 34744
Credit: 261,360,520
RAC: 489
Australia
Message 1976035 - Posted: 20 Jan 2019, 8:42:00 UTC
Last modified: 20 Jan 2019, 8:58:17 UTC

Now when you get into 1000W+ PSU's you're starting to look at a different kind of beast as almost all of them are 2 smaller PSU's built into 1 which is why they are longer than their standard counterparts and are purposed especially for multi highend GPU's so they do carry 2 or more 12V rails (while1 PSU will provide the 5V rail power the other 1 will supply the 3.3V rail).

Most 1000W PSU's are actually 2x cut back 550W PSU's grafted together (and no, two 500W grafted together will not give an actual 1000W output, 950W maybe) and that increases as the ratings goes up (most 1600W PSU's are two 850W PSU's grafted together where both 5V and 3.3V rails are also grafted together into single rails) and then there are those overly long PSU's of 1800W+ that are actually 3 normal PSU's combined.

You really need to get into their internals to work them out.

Cheers.
ID: 1976035 · Report as offensive
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22161
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1976036 - Posted: 20 Jan 2019, 8:49:31 UTC

Why go for a Platinum rated PSU over a Bronze rated one?
a) Marketing
b) Energy saving
c) Marketing

Its a fact that all "consumer" grade PSUs are pretty inefficient over a wide range of loads and hit their marketing badge efficiency over a narrow range (despite claims to the contrary in very many adverts).
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1976036 · Report as offensive
juan BFP Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 16 Mar 07
Posts: 9786
Credit: 572,710,851
RAC: 3,799
Panama
Message 1976051 - Posted: 20 Jan 2019, 12:17:56 UTC - in response to Message 1976036.  
Last modified: 20 Jan 2019, 12:30:42 UTC

Why go for a Platinum rated PSU over a Bronze rated one?
a) Marketing
b) Energy saving
c) Marketing

Its a fact that all "consumer" grade PSUs are pretty inefficient over a wide range of loads and hit their marketing badge efficiency over a narrow range (despite claims to the contrary in very many adverts).

Sure is marketing, most of the users not even know the real difference from Bronze, Gold, Platinum or Titanium PSU's

But certainly use a platinum PSU makes a difference in a high cost electric rates places like i live, not forgetting the efficiency, hot is an issue here too. The external temp is already high all the year. Each watt of heat i could spare is a gain. LOL

Single or Multi rail? If you don't know what it means go for a single rail, especially if you have a power hungry GPU.

my 2 cents.
ID: 1976051 · Report as offensive
Profile zoom3+1=4
Volunteer tester
Avatar

Send message
Joined: 30 Nov 03
Posts: 65709
Credit: 55,293,173
RAC: 49
United States
Message 1976121 - Posted: 20 Jan 2019, 20:40:00 UTC

I live in a hot environment at least in our summer, so the a/c has to fight the heat outside as well as whatever is made in here, a new dual pane vinyl window helps keep it warmer in the winter and cooler in summer, since metal framed windows use the same metal that cpu heatsinks do, aluminum.

And a psu that makes less heat is desirable, as do led lights, cfl lights get too warm to touch after a few minutes, yet an led light barely registers any heat after a few hours and can be touched without fear of burning ones hand, sure there is some heat in an led light, but it's constant and very low, and unlike cfl's led's do not use or contain mercury.

I have a platinum 1050w psu, as soon as I have some more memory for My PC and some energy of My own, I intend to upgrade My PC, most of the time I'm constantly in physical pain and that saps a lot of energy even with two types of pain pills, 8hr 650mg Tylenol and Naproxen Sodium, the one time I am not in pain?

When I'm asleep as it's very tiring.
The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's
ID: 1976121 · Report as offensive
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22161
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1976210 - Posted: 21 Jan 2019, 11:17:47 UTC

As Chris suggests the first thing to do is to work out what the power demand is going to be, add a sensible amount to that for start-up surges and the like (there are some pretty good power demand calculators out there to help).
Then take a very careful look at the manufacturers data - walk away from any that don't publish load/draw figures (scary how many manufacturers don't), obviously considering purchase price.
Most of us who are running multiple GPUs are running fairly symmetrical loads (all the same generic type e.g. 3 off GTX1080), so a single-rail PSU may well be better, but if the load is highly asymmetric (e.g. one GTX690 and one GTX760) a multi-rail PSU might be a better bet - provided each rail is separately regulated (which they should be....).

One thing that does surprise me is how many folks trip up when looking at 110 vs 230 v mains - a Watt is a Watt, so the power (Watts) drawn from the wall should be more or less the same for a given load, but the current drawn from a 110v supply will be about double that from 230v supply. Potentially there will be a difference due to different frequencies (60Hz vs 50Hz), but in general iron-losses are quite small with the modern switched-mode PSUs we all use, so have little impact on power from the wall.
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1976210 · Report as offensive
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 1976211 - Posted: 21 Jan 2019, 11:23:29 UTC - in response to Message 1975409.  

Is there a definitive answer to which is better and why, single rail or multiple rail PSU's. The dearer up-market ones seen to have single rail designs.


. . I believe the current European standards specify the single rail PSU design.

Stephen

. .
ID: 1976211 · Report as offensive
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 1976212 - Posted: 21 Jan 2019, 11:35:02 UTC - in response to Message 1975572.  

I'm simply not in the market for 3 x GTX 1080 in a single machine like some people can apparently afford. Distributed computing is simply a hobby to me, not a way of life, and is funded accordingly on a pension. My low end machine runs a 750Ti quite happily on a 550W PSU, I only upgraded that because it was an office Dell machine with a 350W PSU meant for 9-5 buisness use, not 24/7 crunching.
But the replies are appreciated, thank you.


. . FWIW, my SFF HP desktop with a 240W PSU is happily running a GTX1050ti. I have a power meter on each of my PCs and this one at full crunch is drawing less than 110W from the wall. No drama there ...

. . Most PSUs achieve their highest efficiency at around the 50% load point. ( from about 40% to 65% then downhill either side of that)

Stephen

:)
ID: 1976212 · Report as offensive
juan BFP Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 16 Mar 07
Posts: 9786
Credit: 572,710,851
RAC: 3,799
Panama
Message 1976213 - Posted: 21 Jan 2019, 11:37:17 UTC - in response to Message 1976200.  
Last modified: 21 Jan 2019, 11:47:12 UTC

But going back to my original question, it seems that the answer is that single rail PSU's are better at dealing with high loads, and they are mostly in the top PSU ranges which also have more efficient power outputs as well.

May i suggest read something like: https://en.wikipedia.org/wiki/80_Plus

That article clearly explain the difference from 80 plus to 80 plus titanium... but in simple words...

A 80 plus Bronze in 115 V has an efficiency of 85% at 50% of the rated load a platinum has 92%.

And where the difference goes? Most to HEAT simply that.

In simple terms, others factor may interfere, using a 1000W PSU running at 50% of the load as an example the gold PSU uses 7% more electric power to supply the same 500W than the platinum, that difference makes you paid an additional of 35 W in your electric bill. Now multiply this for your electric power rate (kw/hr) and you will see how much you will going to paid more at the end of the month. Using an electric rate of 0.20 US$ per kw/hr that gives about 5$ per month if you run 24/7.

So besides the environmental problem (more heat generated bad for global warming) and the need to use more efficient tools, what you paid at the end of the month to your local electric company is what you "feel" in your pocket.

That's is why hungry host normally runs on high efficiency PSUs to save few watts and big money at the end of the month, but is not mandatory.

Hopes this answer your question
ID: 1976213 · Report as offensive
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22161
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1976216 - Posted: 21 Jan 2019, 11:51:08 UTC
Last modified: 21 Jan 2019, 11:52:03 UTC

...provided the manufacture has actually had the PSU verified correctly.
I recently had three failed PSU on the bench, all were the same manufacturer (once one got into the beasts), all rated at the same "650W" output, but with different efficiency ratings, but all had exactly the same internal layout and components (and all had failed in the same manner - I had hoped to make one good out of the three....) It really is a case of "buyer beware".
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1976216 · Report as offensive
juan BFP Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 16 Mar 07
Posts: 9786
Credit: 572,710,851
RAC: 3,799
Panama
Message 1976217 - Posted: 21 Jan 2019, 12:04:56 UTC - in response to Message 1976216.  
Last modified: 21 Jan 2019, 12:19:57 UTC

...provided the manufacture has actually had the PSU verified correctly.
I recently had three failed PSU on the bench, all were the same manufacturer (once one got into the beasts), all rated at the same "650W" output, but with different efficiency ratings, but all had exactly the same internal layout and components (and all had failed in the same manner - I had hoped to make one good out of the three....) It really is a case of "buyer beware".

Obviously we are talking about high quality, well tested & verified PSU's. Anyway who put a low quality high power PSU on a high cost host? Makes little sense to do that. Provided the PSU is one of the less expensive items on a high cost host you expect it coming from a high quality manufacturer. A bad "cheap" PSU could easily burn your expensive GPU or MB/CPU components. LOL
ID: 1976217 · Report as offensive
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22161
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1976220 - Posted: 21 Jan 2019, 13:05:33 UTC

These weren't "cheap" ones, but sensible middle of the road prices, and well known brands.
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1976220 · Report as offensive
1 · 2 · Next

Message boards : Number crunching : PSU Design


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.