- Joined
- Feb 4, 2003
- Location
- SF Bay Area
Well, these days I see a lot of people running some very big, expensive PSU's. Sometimes justified, usually not.
I've managed to push the limits of my two wonderful PSU's, namely, Zippy 6701P and Sparkle 550PLG.
The load required is quite shockingly heavy for each.
First the Zippy. 700W, with a +12V rated at 49A with overcurrent protection at 52A.
I don't have any screenies or anything of where things fell apart exactly so I'll just put the scenario into words. X6800 at about 3.8GHz on air, about 1.55v. My rough estimate for that is ~170W. It is just a ballpark figure. All chips are different too, different revisions and even weeks can have significantly different power draw, but I think 170W is a good guess.
I'm going to throw a rough ballpark figure of 14A drawn off the +12v from it. There's inefficiency, but also the cores wouldn't be fully loaded (at least I don't believe they were for me in what I was running yet)
And the video cards. 2x Radeon 2900XT's in CrossFire. The most power hungry video card configuration you possibly have. At stock I'd estimate 420W from both of them combined. So let us say 34A.
Running them at stock with said Conroe on the Zippy resulted in no issues whatsoever. The +12V held steady as I'm used to seeing. 12.05v idle, 12.04v under a load of 3DMark. Food for thought for those of you rushing to buy 1000W+ units.
However I will confess that once I started ramping up voltages and clocks on the video cards, things fell apart quickly. Going up from stock of 743/828 to about 900/1000 remained fine with the voltage up from about 1.15 to 1.3. I think at this point the cards were drawing about 500W, and probably more or less 40A, for a combined +12v loading very very close to the Zippy's overcurrent point of 52A. However, as others have reported, even those it was so close to its maximum rating, the +12v rail remained unflinching. No drop whatsoever. The air blowing from the back was just barely warm, if that.
But that was it. Once I went up to 950MHz core (this required subzero cooling) and bumped the voltage another touch to 1.4, it was over. In Windows everything idled fine, but the second I fired up anything 3D, the rig would reboot immediately. The +12v still showed no sign of being taxed, but the overcurrent protection seemed to be doing its job.
So that was the limit I found for the Zippy. 3.8GHz, and high-clocked, subzero cooled 2900XT's.
After that I began using the Zippy exclusively for the video cards and a hard drive (also served for providing the necessary crossload), and my Sparkle 550W for the CPU/motherboard, and not much else, save for maybe a fan.
For all practical purposes, the only significant load on the Sparkle was the CPU. That's what makes this a rather (extremely) unrealistic situation but an interesting one to note nonetheless.
So, with the cards out of the way this is where I started (very) heavily clocking the CPU up. Everything was looking good until about 5250MHz at about 1.92v on liquid nitrogen. Same wall for two CPU's, one of which was tested by a friend at about 80MHz higher. Random reboots and freezes at inconsistent points.
I didn't have a chance to take rail readings but I noticed that the Sparkle was extremely, extremely hot all over and blowing out extremely hot air. The CPU was likely drawing about 380-400W by itself under such conditions, which was simply too much for the Sparkle to handle, even if it was just one CPU and nothing else. The rails were most likely unstable if I had to take a guess. In any case it had met its match.
So...there you have it. A 5250MHz Conroe with 1.92v can bring down one of the sturdiest 550W PSU's out there.
Two subzero cooled HD 2900XT's along with a high clocked Conroe by air standards can push a Zippy 700W to its maximum rating.
Perhaps if you 1200W PSU guys manage to run 5.3GHz+ and have maybe 3 video cards you might be able to give those a run for their money.
I've managed to push the limits of my two wonderful PSU's, namely, Zippy 6701P and Sparkle 550PLG.
The load required is quite shockingly heavy for each.
First the Zippy. 700W, with a +12V rated at 49A with overcurrent protection at 52A.
I don't have any screenies or anything of where things fell apart exactly so I'll just put the scenario into words. X6800 at about 3.8GHz on air, about 1.55v. My rough estimate for that is ~170W. It is just a ballpark figure. All chips are different too, different revisions and even weeks can have significantly different power draw, but I think 170W is a good guess.
I'm going to throw a rough ballpark figure of 14A drawn off the +12v from it. There's inefficiency, but also the cores wouldn't be fully loaded (at least I don't believe they were for me in what I was running yet)
And the video cards. 2x Radeon 2900XT's in CrossFire. The most power hungry video card configuration you possibly have. At stock I'd estimate 420W from both of them combined. So let us say 34A.
Running them at stock with said Conroe on the Zippy resulted in no issues whatsoever. The +12V held steady as I'm used to seeing. 12.05v idle, 12.04v under a load of 3DMark. Food for thought for those of you rushing to buy 1000W+ units.
However I will confess that once I started ramping up voltages and clocks on the video cards, things fell apart quickly. Going up from stock of 743/828 to about 900/1000 remained fine with the voltage up from about 1.15 to 1.3. I think at this point the cards were drawing about 500W, and probably more or less 40A, for a combined +12v loading very very close to the Zippy's overcurrent point of 52A. However, as others have reported, even those it was so close to its maximum rating, the +12v rail remained unflinching. No drop whatsoever. The air blowing from the back was just barely warm, if that.
But that was it. Once I went up to 950MHz core (this required subzero cooling) and bumped the voltage another touch to 1.4, it was over. In Windows everything idled fine, but the second I fired up anything 3D, the rig would reboot immediately. The +12v still showed no sign of being taxed, but the overcurrent protection seemed to be doing its job.
So that was the limit I found for the Zippy. 3.8GHz, and high-clocked, subzero cooled 2900XT's.
After that I began using the Zippy exclusively for the video cards and a hard drive (also served for providing the necessary crossload), and my Sparkle 550W for the CPU/motherboard, and not much else, save for maybe a fan.
For all practical purposes, the only significant load on the Sparkle was the CPU. That's what makes this a rather (extremely) unrealistic situation but an interesting one to note nonetheless.
So, with the cards out of the way this is where I started (very) heavily clocking the CPU up. Everything was looking good until about 5250MHz at about 1.92v on liquid nitrogen. Same wall for two CPU's, one of which was tested by a friend at about 80MHz higher. Random reboots and freezes at inconsistent points.
I didn't have a chance to take rail readings but I noticed that the Sparkle was extremely, extremely hot all over and blowing out extremely hot air. The CPU was likely drawing about 380-400W by itself under such conditions, which was simply too much for the Sparkle to handle, even if it was just one CPU and nothing else. The rails were most likely unstable if I had to take a guess. In any case it had met its match.
So...there you have it. A 5250MHz Conroe with 1.92v can bring down one of the sturdiest 550W PSU's out there.
Two subzero cooled HD 2900XT's along with a high clocked Conroe by air standards can push a Zippy 700W to its maximum rating.
Perhaps if you 1200W PSU guys manage to run 5.3GHz+ and have maybe 3 video cards you might be able to give those a run for their money.