Bitcoin Forum

So my Ultra 1050w PSU died and I swapped in my old corsair 550w. The question is, do you think a 550w will hold up on a system with a core 2 duo running at 2.4ghz, 4gb ram, 2x radeon 5830's, a 10,000 rpm hdd, and a 7200rpm hdd?

I don't want to kill the machine but I do want to try mining.

According to the calculator here (http://educations.newegg.com/tool/psucalc/index.html) I'd be running at 659 W.

What do you think?

You want to keep the PSU running at about 50-60% load.  5830s pull about 150w from a piece.  You're cutting it close =/

Err -- why's that? Fr0m what I've read, they typically reach peak efficiency ar0und ninety percent l0ad and are 0ften able t0 reliably 0utput 10-20% 0ver what they're rated.

I'm running three rigs with dual 0C'd 5850s 0n 450W PSUs. Have been since May, 2011 with0ut a hitch.

Might want to re-read.

http://www.corsair.com/media//tx850w_efficiency.jpg

At least in 0ne case, that seems t0 be c0ntradicted here: http://www.codinghorror.com/blog/2007/05/when-hardware-is-free-power-is-expensive.html -- it's aks0 c0ntradicted here: http://www.silentpcreview.com/article263-page4.html

I'm n0t sure which c0mp0nent ch0ices are causing the efficiency curve t0 shift, but even am0ng Active PFC PSUs, each unit has its 0wn curve. I c0uld be misreading 0r missing s0mething big. I d0n't mean t0 c0me 0ff as a c0ntradict0ry assh0le -- just curi0us.

Can I donate you a non-broken keyboard?  Reading l33t speak, intentional or not, is just too painful.

enigma

If you'd like. Alternately, I could stop being lazy and thoughtless - just start using a fancier clipboard app [I think I actually made a specific keyboard layout for this problem years ago if I can find it...], saving us all trouble, and encouraging me not to open my mouth to ramble on about stuff I don't know about. ;)

When I run both 5830's, the PSU fan speeds up so it's audible. Running only one does not cause any issues. I'll just run one card for now. Yay ultra.

First "blog" shows a Corsair HX520W Power Supply.

"It's a decent result; efficiency increases by more than 10 percent across the board. But there's a catch: the power supply efficiency curve peaks at around 250 watts. "

250/520 = 48% :)

Second one shows a 300W with peaking efficiency at 300W...no fucking idea.

oh this is a gem. I got the Ultra 1050w PSU for their lifetime warranty. I guess TigerDirect neglected to tell me that their open box policy says they come with a 3 month warranty.

http://www.tigerdirect.ca/applications/searchtools/item-details.asp?EdpNo=5466881&SRCCODE=CANBP236&cm_mmc=EML-_-Weblet-_-CANBP236-_-feedback

I count 3 (THREE) places where it says it's a lifetime warranty if I register. Now Ultra won't go through with the RMA. Wow.

This was bothering me, so I asked an IT director. He replied that many lower-end consumer-grade PSUs (even if they're marketed to "enthusiasts") have an efficiency curve and peak out at about half their rated output. There are two PSUs which frequently have a linear increasing curve - those being server-grade PSUs and PSUs designed in partnership with a mass producer for producers of PCs with a small form-factor. PSUs rated 80+ are also more likely to have the linear increasing efficiency due to how stringent the requirements are for higher loads. Marketers generally don't give specifics on their efficiency curve (or lack thereof), and just because a PSU is labeled "Active PFC" doesn't mean it will operate under the preferable efficiency curve -- it could just be the higher-end PSUs are under-rated. Just as with cars and batteries, every PSU is different and will have its own efficiency curve.

Those quotes don't contradict anything.

The first blog says power supplies can have an efficiency of up to 90%.  It doesn't mean AT 90% load.

The second article had a correction from their obviously wrong data
http://www.silentpcreview.com/article263-page5.html

In the correction the peak efficiency is @ 150W (50% load) for 84.8% efficiency.

While the curves do vary and 80-Plus Gold PSU tend to have flatter curves the peak efficiency is in the 50%-60% range.  If you had a PSU which was 90% efficient at 1000W you would sell it as a 1200W (or maybe 1350W) PSU.  So PSU are never efficient at max load.

Check the PSUs they checked in the first blog -- specifically the 480W Antec PSU, where peak efficiency is between 390W & 515W.

Yeah and it is horribly inefficient and low powered.  The more efficient unit was more efficient across its entire curve and peaked at 50%.

While there is an exception to every rule generally high efficient power supplies are most efficient at 50% to 60% of rated load.

Generally miners should be looking for high efficiency power supplies and they will never have peak efficient at 90%+ load.  The rating requirements and the marketing mean peak efficiency is going to be 50% to 60%. 

Take a unit which maxes out at 90% efficiency at 800W.  which is easier to sell, as a 90% efficient 1000W unit or as a 90% efficient 1350W unit for the same price?

1. That blog post is from 2007. The review it gets its data for the 480 from is from 2004.
2. Checked what site reviewed that Antec 480... and when? Starting to see a link here?
3. Who the FUCK graphs efficiency vs. *input* power?

Appeal to authority.
Also, *all* psus have an efficiency curve. By definition. what moron would claim otherwise?

Server PSUs? Nope. They usually peak at about 70% load, partly because they're underrated to increase reliability, partly because servers statistically spend more % of time at high utilization = load.
SFF PSUs? nope again, generally at 50-60% load, some a bit lower, some a bit higher.

Link? Efficiency curves for all real 80+ units are freely available from ecova, so finding a few examples should be trivial, right?
And... err... yes, a PSU being labeled "Active PFC" usually means it has... active PFC. Nothing else. Duh.

So PLEASE inform yourself before making up claims in the name of some "IT director" k?

Old corsair 550W? What model exactly?
If it's a VX it *should* work fine at least for a while, just follow the usual guidelines (underclock GPU ram if possible, make sure everything gets enough fresh air, don't start overvolting GPUs with a wimpy PSU, ...). Might last for years, might decide to go tits-up after a few months at continuous high load.

Or you could buy a keyboard using the profit you made from mining bitcoin.

In case there are still doubts about PSU efficiency:

Customers prefer buying 80+ certified PSUs than non-certified "junk".
Therefore, manufacturing and selling 80+ certified units is more lucrative.
The manufacturers are interested first and foremost in maximizing profits.
As a result, only as much R&D budget is being allocated to PSU engineering as is necessary to ensure passing the 80+ testing procedures and receiving certification.
Failed designs unable to receive the certification due to design errors are sold to manufacturers building cheap units often bundled with PC cases.

All 80+ certification levels define the 50% load as the one where efficiency must be the highest.
Therefore, it is reasonable to expect that any given PSU, regardless of whether it has or has not been 80+ certified, will work at peak efficiency somewhere around the 50% load mark.