jmumich


July 08, 2017, 03:29:57 PM 

If you dont care how long the PSU last, then just go nuts on it. No one knows how much voltage and electrical noise you're drawing or how close to capacity you're drawing or the quality of your PSU components etc. This shit makes a difference. 650W rated PSUs are NOT created equally. If you have a G3 series PSU, you can push it a hell of a lot more than a GS or G2 for example. Source: I've blown up 3 G2 PSUs instantly under the same conditions that G3s run smoothly under. I've had one PSU with subpar components blow up at 80% load, while I've ran another PSU with better components at 130% capacity for months.
This is the best answer. It depends heavily on the quality of the PSU. I have a bunch of high quality Corsair and EVGA PSUs that I've run at close to 90% the rated capacity for years without an issue. Before that, I tried to save a few bucks on lowerquality PSUs and learned my lesson quickly  cheaper PSUs are more expensive in the longrun. Keep in mind also that PSUs are more efficient closer to 50% load  that could save some $$ in the long run over running them at 8090% rated capacity (along with less chance of damage and higher resale value). But if OP is going for density  why not contact the manufacturer and ask if you can run at 80% of 120% peak output rating without voiding the warranty? That should tell you all you need to know  if the manufacturer will stand behind it, it should be good. I'd still make sure everything is as fireproof as possible  no matter what load. All it takes is a bad riser and some dust...





Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction. Advertise here.



chiwalfrm
Jr. Member
Offline
Activity: 54


July 08, 2017, 04:07:06 PM 

PSU's ratings are there to show you the limits. You shouldn't be running them at 100% load 24x7. That is why people are telling you 80% of their rating as measured from the wall because while there are different interpretations, this is the safest. Safety is more important than that last GPU you are trying to load.
So what you're saying is because most people don't know what PSU efficiency ratings are and/or basic math, that the blanket 80% rule is safest for them to use? But if we actually look at the numbers. Based on a 750W 87% efficient PSU, 600W at the wall means it is only outputting 522W (600 x 87%) which is 69.6% of its max (522W / 750W). Is that essentially what you are saying? That a 750W PSU can only run safely outputting 522W? Because doesn't that seem wrong? Yes 522W on a 750W PSU, that's already 70%. That 87% rating you quoted is the maximum efficiency, it is not the same from 0 to 100% load. See this article: Quote: "Typically, your power supply's maximum efficiency is between 40% and 60% of it's maximum capability." Source: http://www.corsair.com/enus/blog/2013/october/howtobuildapcthepowersupplyRunning a PSU closer to its limits will also put more wear and tear on the components, it will run hotter for most of its service life, and what's the big deal starting out with a bigger PSU? You pay more upfront but a PSU lasts a long time.




philipma1957
Legendary
Offline
Activity: 1946
A new tool for prediction https://bitvol.info/


July 08, 2017, 04:39:10 PM 

forget

Please support sidehack with his new miner project Send to : 1BURGERAXHH6Yi6LRybRJK7ybEm5m5HwTr I mine alt coins with https://simplemining.net I see BTC as the super highway and alt coins as taxis and trucks needed to move transactions.



thesavoyard


July 08, 2017, 05:29:00 PM 

80% is the law in most countries for datacenter continuous power draw (not peak). If you feel you know better, feel free to max out your PSUs. We'll see you back here soon complaining they are dead.
The 80% rule for data centers that you're referencing is based on the circuit load not the PSU draw. You can't pull more than 80% of the circuit max. And the same should apply to your home. If you have a 15A circuit then don't pull more than 12A continuous, or 1,440W (based on 120V). In THIS case the measured wall draw using a KillAWatt would apply. But that's not the question I'm asking. A 87% efficient Gold rated 750W PSU will pull 862W at the wall while outputting 750W (hence why higher efficency psu's are better, becuse you lose less wattage to heat). The more I research it, the answer is to keep your output below 80% of the OUTPUT (not the input) which would be 600W. And to do so, you should measure your input using a killawatt, then caclulate the actual output based on your efficiency, and not go over 689W (600W / 87%). I'm getting the impression you all don't really know what you're talking about. If anyone who is good at math, and has actual insight into this cares to reply, please do. Well, he is right. The PSU is rated at the output, not input. So I'm running my platinum PSUs at about 50% load, I should be getting the max 93%.




ben8jam


July 08, 2017, 08:03:03 PM 

Yes 522W on a 750W PSU, that's already 70%. That 87% rating you quoted is the maximum efficiency, it is not the same from 0 to 100% load. See this article: Quote: "Typically, your power supply's maximum efficiency is between 40% and 60% of it's maximum capability." Source: http://www.corsair.com/enus/blog/2013/october/howtobuildapcthepowersupplyRunning a PSU closer to its limits will also put more wear and tear on the components, it will run hotter for most of its service life, and what's the big deal starting out with a bigger PSU? You pay more upfront but a PSU lasts a long time. Thank you actually writing something meaningful. Looking at the efficiency graph (see links to chart below) for my PSU, it looks like at 80% of max output (600W output) it is still ~90% efficient. So at 600W it would register 666W on a KillAWatt while 300W (or 40% of max output) it is the most efficient, does that necessarily mean it needs to run that low? I guess my question is, where is there research that states 80% of max INPUT is the magic number? I know you might feel like I'm beating a dead horse here, but there is obviously a lot of misinformation about this. The 80% number is a percentage that datacenters use in terms of the max DRAW on a circuit, not anything that has to do with a PSU. So, I wonder if people have conflated these things, while really they are unrelated. And then people "props" who just calls everyone morons, and has no actual empirical evidence (if you look at his 16 pages of comments, they are 90% calling people morons or dummies or just trolls comments in general. I will call EVGA and get it straight from the horse's mouth tomorrow as to what the best output is for long life while also hard use. And for everyone saying I'm burning my house down, the reason I'm asking this is so I don't do that. But I like science and actual fact over generalized statements. All four of my 750W psus are currently running at max 575W and below. Efficiecny charts for the G2 750W PSU http://www.tomshardware.com/reviews/evgasupernova750p2platinumpsu,45665.htmlhttps://www.techpowerup.com/reviews/EVGA/SuperNOVA_G2_750/6.html Well, he is right. The PSU is rated at the output, not input. So I'm running my platinum PSUs at about 50% load, I should be getting the max 93%.
Thanks. How did you determine your output draw to know you're at 50% and thus the max efficiency? Feel like there needs to be a killawatt type device that is in between the PSU and the cabling. But obviously that would require quite a setup.




Props


July 08, 2017, 11:00:27 PM 

Good luck burning down your house nerd




darkomega69


July 08, 2017, 11:06:23 PM 

Thank you actually writing something meaningful. Looking at the efficiency graph (see links to chart below) for my PSU, it looks like at 80% of max output (600W output) it is still ~90% efficient. So at 600W it would register 666W on a KillAWatt while 300W (or 40% of max output) it is the most efficient, does that necessarily mean it needs to run that low?
Of course you can run it higher. I guess my question is, where is there research that states 80% of max INPUT is the magic number? I know you might feel like I'm beating a dead horse here, but there is obviously a lot of misinformation about this. The 80% number is a percentage that datacenters use in terms of the max DRAW on a circuit, not anything that has to do with a PSU. So, I wonder if people have conflated these things, while really they are unrelated.
I actually looked and couldn't find any research on it, so I agree with you, the 80% for continuous load for a circuit is probably being conflated to "ideal" continuous load on a computer power supply. However, to me, it does make sense that a power supply ran <100% would last longer than a power supply ran at 100%. With that said, getting it around the 80% mark is not a bad idea. I wouldn't sleep over it if I go a little bit higher. I have a 2 power supplies, 900w, with 6x125w load, and a 750w, with 475w load, that comes to a load 83% and 63% respectively. All four of my 750W psus are currently running at max 575W and below.
Awesome.




ben8jam


July 08, 2017, 11:16:10 PM 

However, to me, it does make sense that a power supply ran <100% would last longer than a power supply ran at 100%.
Absolutely. I think the part that is confusing to everyone, is a 750W PSU registering 750W on a KillAWatt is actually only outputting approximately 637W (given say a 85% efficiency rating at that higher draw). To run it at the full 750W output, you would see about 930W on the killawatt (assuming it only runs at a decreased 80% efficiency at max output).




darkomega69


July 08, 2017, 11:32:24 PM 

However, to me, it does make sense that a power supply ran <100% would last longer than a power supply ran at 100%.
Absolutely. I think the part that is confusing to everyone, is a 750W PSU registering 750W on a KillAWatt is actually only outputting approximately 637W (given say a 85% efficiency rating at that higher draw). To run it at the full 750W output, you would see about 930W on the killawatt (assuming it only runs at a decreased 80% efficiency at max output). Maybe it's confusion...or lack of critical thinking.




philipma1957
Legendary
Offline
Activity: 1946
A new tool for prediction https://bitvol.info/


July 09, 2017, 12:29:38 AM 

However, to me, it does make sense that a power supply ran <100% would last longer than a power supply ran at 100%.
Absolutely. I think the part that is confusing to everyone, is a 750W PSU registering 750W on a KillAWatt is actually only outputting approximately 637W (given say a 85% efficiency rating at that higher draw). To run it at the full 750W output, you would see about 930W on the killawatt (assuming it only runs at a decreased 80% efficiency at max output). check out jonnyguru.com he is pretty much the guy with the numbers. http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story&reid=523http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story3&reid=523http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story4&reid=523above is a review of a corsair hx 1200 plat. it has a 100 amp 12 volt rail at hot and cold box tests. 92 amp is 89% hot and 89.4% cold. since we run 24.7 just look at hot numbers 70 amp is 90% hot and 90.6% cold. 46 amp is 92% hot and 91.8% cold. he has numbers for hundreds of psu's I try to pull 70% on my psu's so 1200 watt dc = 90.5 % efficient or 1200/.905 = 1326 max x 70% = 928.2 watts at the kwatt meter look at 1200 x 80% and you get 960 watts very close to the same 960 and 928.2 if the psu was 87.5% eff at a 70% dc load you get 1200/.875 =1371.4285 x .7 = 959.999 which is 960 = 960 So a 1200 watt psu on a 70% dc load with 87.5% eff is the exact same as 80% of 1200 or 960 = 960 so both people would be exactly correct some of the time.

Please support sidehack with his new miner project Send to : 1BURGERAXHH6Yi6LRybRJK7ybEm5m5HwTr I mine alt coins with https://simplemining.net I see BTC as the super highway and alt coins as taxis and trucks needed to move transactions.



Metroid


July 09, 2017, 12:40:53 AM 

Example, take a 1000 watts psu, if you psu is 90% efficiency gold then it means when your computer is at the wall 1100 watts, it means your psu is using its promised 1000 watts but the thing is, as the psu gets close to 1100 watts at wall, 100%, efficiency drops a little, the best efficiency is at 45% offered. If you want the best efficiency at full load then get a platinum or titanium, best brands are corsair, superflower, evga, seasonic, coolermaster.




darkomega69


July 09, 2017, 02:09:15 AM 

if the psu was 87.5% eff at a 70% dc load you get
1200/.875 =1371.4285 x .7 = 959.999
which is 960 = 960
So a 1200 watt psu on a 70% dc load with 87.5% eff is the exact same as 80% of 1200 or 960 = 960
so both people would be exactly correct some of the time.
Correct me if I'm wrong, but 70% dc load is 70% of 1200w, 840w.




philipma1957
Legendary
Offline
Activity: 1946
A new tool for prediction https://bitvol.info/


July 09, 2017, 02:15:20 AM 

if the psu was 87.5% eff at a 70% dc load you get
1200/.875 =1371.4285 x .7 = 959.999
which is 960 = 960
So a 1200 watt psu on a 70% dc load with 87.5% eff is the exact same as 80% of 1200 or 960 = 960
so both people would be exactly correct some of the time.
Correct me if I'm wrong, but 70% dc load is 70% of 1200w, 840w. you are correct. it is a math game if a psu is sending 70% of it 1200 watts in dc it is 840 watts but that is the dc at the end of all the 12 volt cables attached to mobo/cpu/gpu. the ac load at the kwatt meter would be 840/.875 = 960 watts.. or 840 dc/ 960 ac = 87.5% eff .

Please support sidehack with his new miner project Send to : 1BURGERAXHH6Yi6LRybRJK7ybEm5m5HwTr I mine alt coins with https://simplemining.net I see BTC as the super highway and alt coins as taxis and trucks needed to move transactions.



Gotimour


July 23, 2017, 07:06:14 AM 

if the psu was 87.5% eff at a 70% dc load you get
1200/.875 =1371.4285 x .7 = 959.999
which is 960 = 960
So a 1200 watt psu on a 70% dc load with 87.5% eff is the exact same as 80% of 1200 or 960 = 960
so both people would be exactly correct some of the time.
Correct me if I'm wrong, but 70% dc load is 70% of 1200w, 840w. you are correct. it is a math game if a psu is sending 70% of it 1200 watts in dc it is 840 watts but that is the dc at the end of all the 12 volt cables attached to mobo/cpu/gpu. the ac load at the kwatt meter would be 840/.875 = 960 watts.. or 840 dc/ 960 ac = 87.5% eff . I usually run at 80% of the rated power output to be safe and efficient.




