I'd like to read them aswell, cos afaik, connecting power sources in parrallel by shorting the outputs together is a no-no, be it batteries or a psu, because there will be a slight difference in the voltages of the two powersupplies, which means a voltage drop over no resistance = high current, which means high load on the psu/battery. But here, we aren't shorting the outputs of two psu's togther, there is the resistance of the cabling, the pci-e slot, the mobo's filters, the mobo connector, the gpu's fiters, the gpu's connectors, etc. All these resistances are next to nothing, but add all of them up, and there's a little resistance there, so in theory it should be safe, afaik - i haven't actually ran sli using two psu's tho, i've got only a little experience using two psu's. It wouldn't be the first time i'm wrong and cpemma is right tho, so don't do anything yet until we hear from cp
But that's when shorting the outputs of the two psu's, right? That isn't what he'll be doing, or am i missing something?
If he plugs one PSU into the mobo and the other into the GFX card he is essentially connecting the 12V lines of the 2 PSUs together. That is why you can't power one device from separate PSUs.
That^ is true for IDE HDDs, too AFAIK IIRC, as the IDE cable carries some voltages. Not sure about SATA HDDS though.
That's the one! I knew i was missing something So this is how i see it, which afaik is correct, unless i'm missing something else PSU#1 goes to MOBO & GPU#1 PSU#2 goes to GPU#2 & other irrelevant stuff such as fans, hdd, pump, 3.3v load, etc PSU#1 has a low resistance path to the MOBO & GPU#1 compared to PSU#2, so the vast majority of power for the MOBO & GPU#1 is provided by PSU#1 PSU#2 has a low resistance path to GPU#2, compared to PSU#1, so the vast majority of power for the GPU#2 is provided by PSU#2 We may be talking about totals as low as .5 ohm compared to .1 ohm, but that's still a difference of 5x!!! Chances are, the differences will be even greater, further ensuring the correct psu provides the bulk majority of the power. And as for IDE & sata hdd's, both data cables carry voltage, if not, the cables would only be there for good looks. The current flow/resistance/buffers/etc involved with either of these data cables mean that drives are way less of an issue compared to graphics cards. Lets say one of these psu's was to fail of old age or whatever, then that's the same as any pc - the psu can cut out or provide low voltage which means power off or crash/instability etc or the psu could output a high voltage, damaging whatever components it's connected to - so if either psu dies of old age & outputs a high voltage, one or both gpus can die, same for ram, soundcard, cpu, mobo etc - this is the risk in ANY pc, but it's always a very low risk when good quality psu's with more watts then needed are being used. The other thing i haven't mentioned is to ensure both psu's have a low resistance connection for the earths, such as screwing both securely to the same psu case (which is the same as a really thick cable), with conductive screws (standard ones are usually fine) and bare metal on the case for the screws to contact - even better, would be to also attach a thick earth cable shorting the two psu cases together. An easy method is to have eye connectors at each end of this cable, and screw it to each psu case using conductive screws - again the original psu case screws will probably be fine. The harder but better method is of course soldering this earth cable to any of the earth wires, one from each psu, as close to each psu's pcb/case as possible. Even better would be to have multiple earth wires but now i'm rambling and going overboard What i think is unpredicatble & potentially dangerous, is say plugging two different psu's into the same mobo - this would mean a very low resistance connection to both psu outputs.
Very much lower resistances, I hope. Check with Dr Ohm, 10A over a 0.1R lead would lose 1V... The danger is from any difference in the two supplies at source - spec allows ±5% variation from nominal, 11.4V-12.6V on the 12V rail. They're usually better than that, but you can't assume 12.00V from each, nor can you assume because it worked for one person it will work for everybody. There may be other ways to use two supplies - for example, my board won't work without the 4-pin power lead that provides juice to the CPU so that may be isolated from the main board supply. Data cable voltages don't matter, they're relatively high impedance logic signals and don't connect straight to any positive rails.
tbh i would think the the single 620w corsair can handle everything you've got. I recently built a video editing PC at work using one which has been 100% stable and the PSU operates almost silently. It contains the following. 2x Xeon Quadcore Tyan Workstation board 8 GB Ram 2x DVD 9x HDD 1950 Pro Capture card RAID card Sound card 5x case fans It doesnt have the dual GPU's or water cooling gear your running but i would think it would be comparable. Post your specs on the corsair forums and they will tell you if it will handle it.
Getting close to .1ohm aint that hard to achieve, from the gpu board to the psu board, with a pci-e aux power connector & a long cable. If the pci-e aux power connector is a little dented and/or dirty, then a much bigger resistance can had! I gave .5ohm & .1 ohm differences because they are a worst case scenario i think. The mobo power filters can easily have 1ohm total resitance aswell. A difference of .5 to .1 ohm is conservative, a greater difference of say 1 ohm to .05 ohm is very possible, further adding to the proper operation of two psu's when in sli! If there is say an extreme worst case scenario of 1v difference at the source voltages, say PSU#1 = 12.5v and PSU#2 = 11.5v. PSU#1 can give GPU#2 12.5v through a 1 ohm path, PSU#2 can give 11.5v through a .05 ohm path, so PSU#1 can deliver a mximum of 12.5a to GPU#2 but PSU#2 can deliver a maximum of 230a to GPU#2. So PSU#2 will be giving about 19x more power to GPU#2 than PSU#1 would be, talking theoretical limits ofcourse, but the important thing is the 19x power difference, in the way we want it to be. If i go more conservative/worst case scenario with 11.5v & 12.5v voltages, and say PSU#2 has a .1ohm path, then PSU#2 can give GPU#2 115a, so PSU#2 is giving about 9x more power than PSU#1 is, which is still great! And for the absolute extreme unlikely worst case scenario, of .5 ohm to .1 ohm difference with the extremely poor voltages of 12.5v & 11.5v, PSU#1 can give 25a, and PSU#2 can give 115a, which is close to a 5x difference, hence my earlier worst case scenario example of a card getting 20w from one psu, and 100w from the desired psu. The difference of course are much greater in the way we desire for GPU#1!!!!
I just saw rude baboon's post now - where is your vagina Did you ever update that thread? I guess there is another option which rudebab suggests, and that's running the mobo & both cards from the 620w psu, and use the 520w to run hdds, fans, lights, pumps, dvd & a 3.3v load - 2a or 3a dummy load would be fine. Or maybe even swap the 520w with a weaker psu from another system you have/friend/whatever to run all this.
Wow, I see this stirred up a few opinions huh? Thanks for all the input guys I really appreciate it. It looks like the one option I have, that everyone can agree on, is using the 620W for the CPU, MOBO and dual GPUs, leaving the 520W to power the drives, watercooling and lights etc. Can't help but feel like it's a bit of a waste of a PSU now though lol. The 620W is going to be working it's socks off and the 520W is barely going to breaking a sweat. Like I say, I've seen the 620W chew up a GTX SLi setup before and seen a bench test where it comfortably handled a 720W constant load. Thing is, I'm adding a (reasonably heavily) overclocked Q6600 into the mix and volt mods on both GTX's, as well as two Laing DDCs and obviously, a notoriously power hungry SLi Mobo. I also wanted the option to add hard drives etc in the future. My wattage meter (which is pretty cack, so take it with a pinch of salt) reads around 530W under full load atm so I might just be able to get away with the 620W on it's own. The question is: for how long?
If that is 530 at the wall, then the psu will actually be providing a lot less than that to the computer, and afaik that it what is measured in the wattage ratings on the box. Seeing as they are roughly 80% efficient that is between 400 and 450W As for the question of 2 PSUs, have a look at this http://www.bit-tech.net/hardware/2006/08/13/psu_preview_palooza/2 It must be possible therefore... dunno whether the phase issues will be a problems for you, but apparently resistance isn't
That's a very good point cupboard Granted, I'll heed the advice of the more knowledgable on this one but surely the device that cupboard linked to is intended to do exactly what others have said is a no no - provide additional 12v power to the graphics subsystem? Why wouldn't the problems that would be associated with a dual psu setup apply to this device? It is 530W at the wall and I'd read something similar about the "measure" being of that which is actually supplied by the PSU elsewhere, which is encouraging. Also, that's a worse case scenario, with prime95 loading all 4 cores and ati tool stessing the gpu. Whilst I understand it's always best to have ample headroom, nothing in a real world scenario would provide that kind of current draw. I read some threads over at the Corsair forums as well which were encouraging. They suggested that 8800GTX in SLi and 3780X2 in crossfire, along with a quad and some mild overclocking were no problem for the HX 620W. They have even produced new pci-e leads which provide 2 x (pci-e 6 + pci-e 6+2). That'd make setting up SLi a piece of cake. I might chance it...
Thanks cpemma, I knew there must be a logical reason Like I said, I'll listen to those in the know rather than "chancing" it.
Rude babboons suggestion of using the 620w for the mobo & sli, and the 520w for everything else is the simple solution Maybe you have a friend that would like to give you a ~400w psu + something else you want in exchange for your 520w? But just theoretically, even though it's not neccessary in this situation, what do you think of my calculations cpemma? The way i see it, with either a normal secondary psu or one like the fsp booster, the same calculations i provided would still apply, right? After all, by monitoring the primaries output voltage, the fsp booster could either allow it's output to be slightly below the main psu, or above, or hope to match the main psu's voltages, and in all three possible scenarios, the calcs i posted above should still apply! right?
BTW, just like to add some actual practical experience. I have run a dual psu setup with one psu dedicated to running hard drives, optical drives, and fans--for months to years. This was in the days when overclocked Prescotts were frying power supplies. Anyway, never had one of those items fail on me in any way. Did not run a graphics card off the slave psu, for the above mentioned reasons. I have heard of folks running 2 psus across a graphics card for long periods of time without any untoward effects. The 5.25 gpu power supplies do intrigue me. Most power supplies these days--the gpu voltage is independently regulated from the peripheral 12v molex plugs? So you would actually think you might generate the same micro-differential problem. navig
It is possible to run PSUs in parallel, but it takes extra circuitry from what I understand. One example is redundant PSUs in a server. I know in the case of the Intel s7000FC4UR server, it is supposed to be run from a 240v connection. But the system can be run from 120V. The trick is both PSUs need to be connected to provide enough power. In something much older, slot 2 Xeon based server called a 'Megaplex', it advertised 3 redundant 600W PSUs. I ended up encountering one of these and cracked open the PSU to take the rather high powered blower fans from it, and I found it was really 2 off the shelf ATX 300W PSUs joined together through some kind of special circuitry (I think I still have one around somewhere if someone wants to take a look). So it is possible, but probably beyond the scope of the typical modder
