Choosing a PSU wattage

[enk]

I'm about to invest in a new and more silent PSU for my rig. I know how much power I need now, but since I suffer from chronic upgraditis, chances are I'll prefer extreme performance over silence some day and then it would be sad if my PSU was too weak.

I have an Asus P5K Pro sporting an E8200 and a 8600GT. Here is my power usage measured at the wall plug under various voltages and levels of stress (yes, I used to be all "Who on earth needs 500W?" but then I started overclocking ):

Code: Select all

vcore   idle    orthos  +atitool
1.00    110     133     163
auto*   113     150     175     * cpu-z reports it as a tad above 1.2V
1.28    120     180     200
1.40    121     210     230


What I want to know is this:

How much more will a 8800 use under stress compared to my 8600? (I'm guessing around 100W ?)
Intel's quads have a TDP of 95W while the duos have 65W. Is this the same as saying a quad won't use the double of what a duo uses? What is the max power a current processor has been known to use?

[b.i.o]

Ars Technica has some power stats in their 9600 review--including ones for an 8600 GTS and an 8800 GT:
http://arstechnica.com/reviews/hardware ... view.ars/4
(it's the second chart down on the page)

[Axman]

Which 8800? The G80 GTX consumes, at idle, about 80W, and at full, 160W--from the wall*. With your existing numbers, a 500W PSU will get that just under the wire. Even though that PSU is enough, if you're shopping, I'd get a 550 or 600W power supply just to be inside the efficient part of its curve. That particular GTX is the most power hungry of the lot, and it's factory-overclocked.

*The PSU I used for that test**, at it's worst, is 83% efficient, so the actual wattage is more like 66-136W.

**OCZ/ PC P&C 1000W something or other with unpluggy cables.

[enk]

Thanks for info on the graphics.


Due to the recommendations on silentpcreview.com, it's going to one of the Corsair VXes (450 or 550W). If I am to believe their claims, I'll be getting 84-86% efficiency, so I don't mind only using 20% of 550W with my current system idling.

Spoiler:

The HX520 (modular) costs 25% more than the VX550 here. Is it worth it? Why (not)?

Does anybody know how much a C2Q at, say, 1.5V draws?

[GuitarFreak]

I've got the corsair 620HX and it powers an overclocked e6750 (3.8GHz @ 1.5v) and GTX (630/1720/1030), 4 hard drives, lost of fans, and it's rock solid. I love it.

[enk]

I've got the corsair 620HX and it powers an overclocked e6750 (3.8GHz @ 1.5v) and GTX (630/1720/1030), 4 hard drives, lost of fans, and it's rock solid. I love it.

How do you like the modularity? I'm trying to figure out if I should shell out the extra $30 for the HX520 instead of the VX550.



Btw. I dug out my old 3.0 GHz P4 last night, put it in my new rig and oc'd the beast. It has a TDP of 84W, but 4.4GHz @ 1.575V it made the whole system eat 330W under load. That's more than 200W for a single core alone! And a massive 230W while idle (and the SpeedStep multiplier reduction from 15 to 14 was really amusing )

My puny heatsink could only keep it from throttling for long enough to run SuperPI 1M. The P4 is about three times as slow as the Wolfdale at the same clock: 36 vs 12 seconds @ 3.6GHz!

[GuitarFreak]

Definitely worth it. You can pick what cables you need and it makes wiring a whole lot easier.

[enk]

Got my HX520W today Am going to install it now..


Still wondering, though:

How much power do quads use compared to duals? Many people say a quad generates twice the heat of a dual, but then there's the TDPs 95W vs. 65W. Does it generate the double? Why (not)?

[Axman]

Depending on how you clock it, it will likely consume from ~100W-150W under full load. Intel's TDP is an average, not a maximum. But actually getting it to run at 100% on four cores such that no part of any processor isn't gated is nigh impossible, and Intel automagically drops voltages and frequencies so fast that it's unlikely that you'll really come to depend on a ridiculous heatsink. That isn't to say that they can't haul watt, they just do so rarely--particularly the 45nm CPUs, which idle at almost nothing.

[enk]

Depending on how you clock it, it will likely consume from ~100W-150W under full load. Intel's TDP is an average, not a maximum.

This suggests otherwise:

Thermal Design Power: (Also referred to as Thermal Guideline) The maximum amount of heat which a thermal solution must be able to dissipate from the processor so that the processor will operate under normal operating conditions.

That also agrees with the fact that my E8200 uses only about 40W under load at auto voltage.

But actually getting it to run at 100% on four cores such that no part of any processor isn't gated is nigh impossible, and Intel automagically drops voltages and frequencies so fast that it's unlikely that you'll really come to depend on a ridiculous heatsink. That isn't to say that they can't haul watt, they just do so rarely--particularly the 45nm CPUs, which idle at almost nothing.

Yeah, but that's unrelated to the max heat dissipation. My question still stands. I just forgot to mention I meant while all cores are working at full capacity.

[Axman]

...Under normal operating conditions.

Therein lies the rub. But the way the processor's designed, it realistically won't hit it's full power consumption with existing synthetic tests, like Prime95. The deal is that it shuts off parts of the processor between clocks, so that it remains cool despite operating at full tilt.

Just the same, Intel doesn't count a double-run of Orthos as "normal" and you can realistically pull 125W for the workload with a non-overclocked Q6600.

It's a bit dated, but neither Intel nor AMD has since changed their methods for determining TDP:
http://www.silentpcreview.com/article169-page3.html

Intel is listing TDP numbers that are significantly lower than the actual maximum power draw of their CPUs. They are then relying on the fact that most applications barely use the CPU, assuming that it will remain idle most of the time. In the case that an application does max out the CPU for any period of time, Intel relies on their “Thermal Monitor” to automatically slow down the CPU when it becomes too hot to protect it from overheating.

AMD, on the other hand, lists TDP numbers that are significantly higher than the maximum power draw of their CPUs. They also have listed the SAME TDP for every desktop Athlon 64 so far, and I have little reason to believe that future Athlon 64s will have a higher listed TDP (at least for the near future). We still don't have definitive information about the exact power dissipation of each of the Athlon 64 processors, but it is clear that other than the fastest clock models, it is far below the 89W TDP cited by AMD.

The 45nm parts are significantly improved over the 65nm with regards to idle power and power management (AMD, 'til that point, was still beating Intel over idle power consumption, and that wasn't going to last) and at the low end of the frequency scale, operate well below the TDP prediction (which is based on the highest possible clock speed of the series) like with what you saw with your E8200 (I have an 8400 that maxes out (with SpeedStep enabled) at 47W).

[mosc]

Process technology and idle power have an interesting relationship. If you hold the die area the same, a better process technology will actually INCREASE the idle power. In this case, you do get more transistors to work with however. If you hold the design and transistor count roughly the same when you improve process techs, you essentially shrink the die area which is really what is giving you the power savings. Also, most power savings over the P4 era chips are in architecture changes like turning off parts of the chip you're not using which are independent of process technology.

Leakage power (the biggest chunk of idle power) as a percentage of total power actually increases as you shrink process technologies, I just wanted to point that out.

[enk]

...Under normal operating conditions.

Therein lies the rub. But the way the processor's designed, it realistically won't hit it's full power consumption with existing synthetic tests, like Prime95. The deal is that it shuts off parts of the processor between clocks, so that it remains cool despite operating at full tilt.

Tricksed us! They tricksed us!! My pre..

ok.. but if Prime95 for example doesn't load it fully, what does?

[Amnesiasoft]

ok.. but if Prime95 for example doesn't load it fully, what does?

Crysis...

...just kidding. You probably won't find anything that will realistically provide a full load unless it was artificially designed to do such a thing. For the most part, programs can't be made entirely parallel. And even then, you'd have to make sure it was written in such a way you were using as many parts of the processor at once as possible.

This is all just random speculation from what I do know. I'll wait for Axman to come in and either support me or say I'm completely wrong