How much power does it really use?

[Scoob]

Hi all,

I've popped this into the Overclocking sub-forum as I think this will be of interest to us overclockers, who are likely more concerned about power use than most

After some discussions in another thread I thought I’d revisit the whole “how much power does it use” topic. Quite some time back I did some testing using one of those wall socket meters that actually tell you how much power you’re using currently. I’ve not done this for my current rig and today thought I’d do some testing.

It’s in my Sig, but here’s my current gaming PC specification:

CPU: Intel 2500k @ 4.5 (reports 1.35-1.4v depending on load, typically lower, but IBT can see 1.4v)
GPU: Two GTX 570’s in SLI – one Inno3d @ 732 (stock) and one EVGA @ 732 (797 stock for this card)
Motherboard: ASUS P8Z68-V Pro – I use the onboard sound.
Ram: 24gb DDR 3 1600 – 2x4 + 2x8
PSU: Corsair HX750w
Monitor: LG Flatron W2452T – basically your standard older 24” 1920x1200 LCD. No built-in speakers.
Speakers: some old 2.1 Creative Labs jobbies from yesteryear, fairly meaty sub (ON in all tests)
Cooling: Custom External water loop using separate 90w external PSU. D5 Pump.

Note: I am measuring at the socket. Plugged into this socket is my 6-gang power strip that runs: The Main PC, external Cooling Loop, Monitor, Speakers (with Sub) as well as my laptop that’s OFF, but charging.

Test 1: Just the External Loop on, PC is OFF (note: laptop still docked and charging on this circuit)
Power Usage reported: 37w - not too bad, that's a D5 Pump on FULL and 4x 180mm LED fans at 12v.

Test 2: Loop on, PC powered up and sat idle on W7 desktop. Monitor ON.
Power Usage Reported: 230w – note: after a short while this dropped to 205w and stayed there, I assume that’s the various power save features kicking in. So, CPU down to 1.6ghz and GPU’s to their idle state of 50mhz.

Test 3: Loop on, PC powered up and sat at idle on desktop. Monitor OFF as a quick check.
Power Usage Reported: 145w – I guess my old monitor uses 60w then. Sounds about right as it’s a few years old.

For the next few tests, just to be clear, the Loop is ON, the PC is ON, the Monitor is ON and my Speakers are ON. As are my Logitech keyboard (LED’s – keys light up) and my Logitech mouse.

Test 4: Heaven 3.0 – running at 1920x1200, API – DirectX 11, Tessellation – Normal, Shaders – High, Anisotropy – 16x, Stereo 3D – Disabled, Multi monitor – NO, Anti-Aliasing – 8x, Full Screen – Yes, Resolution – System (1920x1200)
Power Usage Reported: 525w Peek – this test pushes the GPU’s nicely, the CPU ain’t so busy.

Test 5: 3dMark Vantage – running at 1920x1200 in Extreme mode. PPU Disabled
Power Usage Reported: 635w Peek – note, I saw no more than 565w peek until the very last feature test.

Test 6: OCCT – CPU Test, Linpack, 25% ram (of my 24gb) AVX enabled.
Power Usage Reported: 228w – that’s just the CPU working hard there, 100% all cores.

Test 7: OCCT – GPU Test, This is Furmark in effect run in full screen mode at maximum complexity.
Power Usage Reported: 674w – we have a new record! Interesting to see, but not really a realistic test in my view.

Test 8: Skyrim gameplay. Note: my Skyrim is beyond Ultra with many many additional graphical enhancements such as textures, meshes and lighting mods. It generally runs at a constant 60fps for the most part, however it does make heavy usage of both GPU’s AND CPU concurrently. This sample was taken as the PEEK value I saw during random gameplay, ALL of it outside, some in open spaces and some in cities & towns. I used god mode along with a conjuration mod to allow me to spawn MANY (30-40) Flame Atronachs and have them fight – this did push the power usage up a little.
Power Usage Reported: 525w Peek – less than I expected. However, I do think I’d be able to beat this figure during regular gameplay potentially, as some areas work the PC that little bit harder. I did do the usual “Go to Dragonsreach and look down into Whiterun” test, as this remains one of the harder scenes in the game from a resource usage standpoint. Add to that the fact I run Improved Whiterun so I have a crap load of extra trees and clutter in that view than standard.

So, in summary, in typical (for me) gaming I never see more than 525w at the wall. Remember, this is power to my PC, my external water loop, my monitor, my speakers as well as my keyboard and mouse of course. Additionally my laptop is on charge in its dock currently. It’s only benchmarks that see usage go over 600w peek. We know 3D Mark Vantage has proven to be a very good test for both CPU and GPU stability, so for that to hit 635w peek is reasonable. OCCT’s own Furmark test however, while good for testing stability, is an unrealistic test in my view, and it did manage to pull the most power at 674w peek – a full 150w more than my Skyrim install that I considered a good real-world test.

I might try Crysis 2 at some point, as that too (if I recall correctly) seems to push both CPU and GPU hard during play.

So, let’s think about this for a moment. I have a good, but fairly moddest 750w PSU. This PSU powers JUST my PC as the monitor, speakers and water cooling loop are externally powered. At the absolute peak power draw we saw of 674w during OCCT’s Furmark test, we can take away the 37w we know my external loop draws and the 60w my monitor draws. That means my PSU is only needing about 580w. Not too bad considering it’s powering an over clocked 2500k and a pair of GTX 570’s! I might bump my GPU clocks up to their 850mhz profile and run some of the tests again if anyone is interested? It’d be interesting to see how much difference it makes – especially in Furmark!

I found this all quite interesting, I wonder if anyone else will? Lol.

Cheers,

Scoob.

 

[Nurf]

Love the fact that you took the time to do these tests. I don't mind those, that go alittle overkill on the psu side, but I hope some of the ppl that think you NEED insane amounts of watts will find this. Don't worry, you have atleast one fan of your work

 

[Scoob]

Heh, thanks Nurf. I thought it relevant to show a real-world example of a gaming PC and what it really uses. Hopefully people will find it interesting and useful.

Get this, I entered my full rig specs into one of those power calculator thingies the other day. It suggested I'd need a bare minimum of a 857w PSU I think it said - an odd number, I thought at first that it meant that's what it expected me to DRAW, but that can't be right.

I forgot to mention that I actually have an additional three 120mm fans in the case, just to ensure that the HDD (it's an old spinner, not SSD) and my motherboard get some airflow at least. I'm running these on a 5v adaptor though, so they are silent.

Scoob.

 

[Nurf]

I might be getting a low watt psu, but things like those calculators did make me rethink if I was underestimating my needs. I would have found this post interesting anyway, but on top of that it showed that I was actually still overestimating a bit.
You could add that you didn't take psu efficiency into account, just to add a bit more shock and awe to the readers that forget about it

 

[JMMP]

Nice work Scoob! This test will elucidate many people with all those doubts like; How much power i need to run my system? And i think that, all you said, people may deduce that if you have a 750w power with 90% of efficency, and a rig similar to yours, will be better than a 1100w with 90% efficency ( like mine) because it will reach that 90% of efficiency sooner! Ah! Just another thing to help understand the power efficency! The efficency doesn´t start to be effective from +/-50%(depending of the brand) of the power supply maximum power delivery!

 

[Scoob]

Hi JMMP,

Yes, from what I understand it's good to "push" (i.e. actually use!) your PSU enough for it to get into its efficient zone so to speak. As such, a 650w PSU running happily at 60% load or so may well use less power than a 1200w running at just 30% load.

The flip side of that is the of running a PSU too close to it's maximum - say a budget 450w at 90% of what it can really do. However, in tests done on my fairly humble 750w PSU, they've pulled well over 900w on proper PSU test rigs before the thing has simply shut down - nice and safe.

This IS encouraging, and it's one of the reasons I like to get a good brand of PSU myself now. Maybe not super high-end, but I simply don't need that as my testing has shown.

I think when things potentially get dangerous - especially on a cheaper PSU - is when you go beyond what the PSU can really do. So, it's above its effecient operating load, so it gets a bit hotter, so it gets a bit less efficient due to the heat, so it pulls a little more...and gets a little hotter... well, you see what I mean. Years ago I had a PSU in a friends machine catch fire due to the very slight extra load of a new GPU psuhing it over the edge. Literally, there were FLAMES out the back. The internals had gotten very hot, the wax-like stuff they use to cover electrical components had melted, caught fire, and was blasted out the back by the fan. All this and the bloody thing NEVER shut its self down! Lesson learnt, quality PSU's bought since that day.

Still, if this thread shows nothing else then it's that PC are getting more efficient. I spoke to my friend earlier, he has a 2600k at 4.8 and two GTX 480's at 800 to 850, and JUST his PC draws far more than my entire setup! So, my PC, External Loop, Monitor etc. pulls less than just his PC. Ok, he is clocking higher than me, so that should be taken into account, but still! Someone running a pair of 670's or 680's or indeed a single 690 would be more efficient still!

I will do some GPU overclocking tests if I get time tomorrow, to see what bumping the cores up to my OC of 850 does to power use.

Anyway, I'm a bit tired and rambling again!

Scoob.

 

[Spiderz]

Thanks for this post, quite a lot of god and helpful information here and makes it easier to understand what sort of PSU is needed for whatever system.

 

[Yeren]

very informative post and something that maybe good for pointing the people that think they need 1500w PSU to run their low end gaming system on air.

very nice read otherwise

 

[iBeInspire]

Really good write up mate, Found it enjoyable and interesting to read
Kinda relates to me, since I'm currently using a borrowed CX430 to run a massively OC'd mid-range system, and will be buying a new PSU soon...

 

[Scoob]

Hi Spiderz, Yeren, iBeInspire,

Thanks for your comments.

I had a little time this evening so thought I'd try some of the tests with a slight GPU overclock.

I have two GTX 570's, one an Inno3D which has a stock clock of 732, and the other an EVGA with a stock clock of 797, though I ran both at 732 in the prior tests.

NV GPU's (not sure about AMD) seem to have no problem in SLI if one card is clocked higher than the other. You'd simply generally see a slightly lower GPU usage % on the faster card. However, I like to keep both of mine at the same speed.

For these next tests I set both GPU's to 797mhz, vCore going from 975mv to 1025mv. Basically the Inno3D is overclocked to the EVGA's stock level. Regardless, both GPU's are now clocked around 9% faster.

I only ran a couple of test, but the results are interesting. I'll list both here so it's clearer. Remember, this is watts as recorded at the WALL socket, so more than just my PC, it includes the Monitor, WC loop etc. as well.

Heaven 3.0

GPU's @ 732 = 525w Max draw
GPU's @ 797 = 617w Max draw

3D Mark Vantage

GPU's @ 732 = 565w, but 635w only in final feature test.
GPU's @ 797 = 607w, but 660w only in final feature test.

That final feature test in Vantage is pretty harsh!

So, with a fairly modest GPU clock - just an extra 65mhz and 50mv - we see a staggering 17.5% higher draw at the wall in Heaven 3.0, but a much lower 7% higher draw at the wall for 3d Mark Vantage.

That's quite a shocking difference really. I'd always considered Vantage to be a fairly harsh test - a good way at identifying a possible instability in the system overall - whereas Heaven is really pure GPU.

So, the "pure" GPU test saw a 17.5% increase in power draw for a 9% overclock. Wheres the "balanced" CPU and GPU test saw a modest 7% increase for a 9% OC. I will say though, regarding the Heaven 3.0 test, that throughout most of the test the draw was in the mid-500's, so only certain scenes pushed those highs. Much like the 3D Mark Vantage final Feature test.

It shows, doesn't it, how different tests stress systems differently.

Oh, as an FYI, I've just put a Windows 8 Pro build on this PC. As I was testing it out and still had the power meter in the wall, I thought I'd run the overclock tests. So, hardware-wise the system is the same apart from the addition of a single additional HDD. I've gone dual boot as not convinced by W8 yet.

What this shows so far is how a GPU overclock has the potential to pull a lot more power - especially if running more than one card of course! Things like this can make the difference between a PSU being "overkill" to hardly adequate. Personally, and I've said this before, I'd be more happy with an 850w PSU knowing what I know. This is purely because I DO like to overclock of course. This would be even more so if I were running my loop (so, pump and fans) from the PC's PSU also.

I'll try some more tests as I get the time.

Scoob.