Dual Rail 12V and PSUs

[Oklahoma Wolf]

True indeed.

The thing with the OCZ 600w I'm still puzzled over, is how the rating is done on them. At Topower's website, the P6 600w with the single 12v has it rated at 35A. The dual rail OCZ, as well as the triple 600w Topower, add up to 38A. The quad goes up from there to 41A total, and the triple and quad have peak ratings to boot. I have to wonder why so much difference in 4 units based on one 600w design. Maybe they're all only capable of 35A draw combined 12v - I'd love to see that put to the test.

 

[larva]

Yeah, I distrust the OCZ rating also, it looks like a self-serving bit a wishful thinking created soley to try to allay the fears of those who know that a 18A cpu line is weak for a supply saying 600 all over it and costing what the OCZ 600 does.

I don't really trust the 26A rating on the Zippy that much more, see the comments I added to the above post on that one.

 

[Oklahoma Wolf]

Interesting about the Zippy - was wondering if there wasn't a limit there somewhere. At any rate, IIRC it almost costs as much as their 700w monster with the 45A 12v rail - I think that would be the one I'd go for had I that kind of money or that kind of power needs. And the FSP 550w 36A unit is about a third the price of it - you sure gotta pay for those extra current numbers.

 

[AlabamaCajun]

How many watts does the fattest Prescott takes. Remember to back-calc the amperage to 12V. If you have CPUs pulling 200+watts then I say we are introuble. Whats really needed is for Mobo makers to start really beefing up the caps around the CPU to handle to draw a little better. I've seen two Mobos with a larger cap farm at the socket lately, BFG is one, I don't remember the other. I'm sticking with Duals, I know they work. I'll see If I can get the peak wattage list on these hot-a$$ cpus today.

 

[larva]

The whole point is this is too back-calced already. The way you know you are in trouble is that you can overload the 20A rail on an OCZ600 with a Prescott at 5GHz. It doesn't really matter what numbers you put on it, the fact is that the OCZ600 chokes and you put the OCZ520 on the same load, and it doesn't. The numbers behind it are academic, not easily assesed, and beside the real point, anyway. That being which supply works, and which doesn't.

You have to understand, the numbers commonly tossed about don't mean a whole lot. All these power supply numbers are ratings, not measurments of output. And the numbers bandied about for cpu draw are uniformly wrong, and ignore the complex nature of the load as well as its variablity with time. It's better in most cases not to confuse the issue with too much theory, but rather to try the supplies and see which works better. This is the end result people are after, regardless of how it discects numerically.

Most systems out there, yours (and mine) included, run comfortably off a dual rail supply. Don't try to extend this to a rule, though, because in this case it simply leads you to the wrong conclusion (and recommendation, which is why this is not stuck).

 

[AlabamaCajun]

What we have to do is line the chips up. I got frosted by the 800 series chips. The Intel data shows them running at 125A and from the calcs I get about 150 or so watts. But with a little digging I find theses hungry mosters can wolf down more like about 200 watts pushing you over the 16-18A rails of these dual PSUs. The data presented here is not theory but information gathered from PSU makers and peoples experiances backed by knowledge of regulation. But we all know good science can get beaten up by something left out. I did check out the AMDs and all the consumer Pentiums. I know they will run on a CoolerMaster 450 or PC-P&C and should run on your Fortron 500 are better (EXCLUDING the 800s & Extremes) . What stands now is to get PSU makers to start pushing up the 12V rails to 22 and 24amps on the PSU side. IMO, 30 wold be better but we don't have the choice that we had back in the linear regulation days of doubling the wattage needed. I will update the Header in this thread to disclaim the 800s and extremes.

 

[larva]

Again, whatever calcution that is taking place here is on top of a foundation of assumption. There are cases of both dual Xeons and a 5GHz Prescott amongst the forum members that cannot be adequately powered by the OCZ600. And if you think that a 4.5GHz Prescott wouldn't make the 12A rail of a Coolermaster 450 walk funny, think again.

I agree that roughly 24A ratings will indicate rails that are strong enough. But that is only strong enough for today's power consumption, and of course that will have risen by the time such ratings occur. What is innarguable is that the 34-36A 12V output of the OCZ PS520, Sparkle FSP550-60PLG, and PC Power&Cooling 510 is no overkill for agressive rigs. Dual rail supplies with the same output level cannot power the most hungry cpus well. This leaves the inescapable conclusion that, at present, dual rail supplies are limiting regardless of the degree to which you analyze the other electrical implications of their topology.

 

[AlabamaCajun]

Thanks Larva, from what I'm reading it won't matter how much we write here since the industry seems to be out of sync. Every day I read a new post, My PSU let off a loud bang, My HD just crashed, My video card artifact, My system won't post or reboots. I'm thinking of writing to PSU makers to see if we can get more power for these massive power sucking chips that are hitting the market. Can we get more wattage on the 12V rail(s) away from the 5V rail.
If we took it another way do we pressure Mobo makers to beef up the CPU power ring aroung the socket?

 

[larva]

I think we just have to hope that truly elegant solutions like the Pentium-M proliferate. These concerns are exactly why the Prescott's evolution was cut short. Power draw (and in turn heat production) were and are out of hand. What's mysterious to me is why the dual core Intel cpus have two Prescott cores. Obviously if a single Prescott is spiraling out of control for power and heat reasons two of these same cores on a single package does nothing to ease the situation, and in fact worsens it.

Were Intel to use the currently technology to its fullest, we would have dual core P-Ms that use less power than a single Prescott of high clock rate. And the performance would be awesome. I guess Intel didn't decide to change course soon enough, and the remnants of the power-hungry evolution aren't feasibly exorcised yet.

I do think dual rail supplies are a great idea, for the many varied technical reasons you detailed. I studied electrical engineering for some time, and have a degree in it. I understand the wisdom of Intel's move, and feel that over time it will be the dominant topology, and for good reasons. It's just that present, the size constraint of the ATX power supply has limited total 12V current output to less than 40A, and massing this on a single rail has proven to drive heavier loads.

Stock rigs have hard time pressing the limits of good dual line supplies, and this is the usage Intel envisioned when moving in this direction. We as overclockers have different priorities, though, and at present seem to be living well outside Intel's design envelope where power consumption is concerned. It's good to understand why dual rail supplies have significant merit from a technical perspective, but this notion must be tempered with the overiding concern of cpu power draw. Quite simply, until the disadvantage of 18A cpu current limits can be overcome, the other positive aspects of the dual line topology must be viewed as secondary.