• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

CPU energy usage at idle..

loaky

loaky

loaky

Associate
Joined
6 Feb 2010
Posts
138
I've got a little home server running Windows Server 2012. It spends a good chunk of it's life (over night, in the day) sitting at idle when i'm not using it. It's currently got a i3 2120T which copes ok, one of the the lower power i3's. Although I've recently come across an i5 3470 for free, a chip quicker by quite a stretch. What i'm looking for is a idea what the energy usage would be just sitting at idle of the two chips. As I know most modern chips really bring the clock etc down when idle, and the i5 being newer maybe more. The TDP figures are 35W for the i3 and 77W for the i5. Now I realise if they were running at 80-100% utilisation their whole life i'd be saving a chunk, though that's not the case. Any views everyone?
 
If it helps, the system idle difference between an i5-3470 and an i7-3770K is very small:

http://hothardware.com/reviews/intel-core-i53470-quadcore-cpu-review?page=8
http://www.anandtech.com/show/5871/intel-core-i5-3470-review-hd-2500-graphics-tested/5
https://www.pcper.com/reviews/Proce...sor-and-HD-2500-Graphics-Review/Overclocking-

The final link shows the idle power of the i5-3470 is about the same as the i3-2105. I doubt you'll find a better direct comparison than that.

Ivy bridge is also quite a bit more power efficient than sandy bridge. I'd definitely do the swap.

If you're very keen on reducing the power you would do better to focus on the PSU and motherboard I think, as well as reducing the number of HDDs.
 
The actual amount of money you'll save by sticking with the i3 is probably going to be quite small. As you've said, it'll down-clock at idle to use even less power. If it's free, I'd get the i5 and forget about the energy savings.

EDIT: looks like I'm wrong and the i5 is possibly more efficient. Things like a more efficient PSU and fewer HDDs will save more energy.
 
Last edited:
Yeah, TDP is the max they could use, though you can exceed this with overclocking. Idles between any of the CPUs is going to be negligible, dont forget at idle your system probably falls out of the load bracket that PSUs are certified 80plus for - as in, at idle, your power from the wall is likely to be less than 100W, even on a huge 6-8 core monster. PSUs are rated 80 plus at 3 load levels, 20/50/100%, with the most efficient load for a PSU being somewhere between 50-75%.

But an 80plus PSU at only ~5-10% load can be in the 70s or even 60s % efficiency. So if you have a 850 watt supply, only drawing 50w at idle, you're actually drawing 95w from the wall. If you got a new CPU that used more power, so at idle you're drawing 60w, you may find you only increase your wall draw by 5w, not the total difference of the CPU, because you're getting into a more efficient load for your PSU.
 
Some good links there thanks guys! and having more of a search myself the general consensus was a 2120 and a 2120t at idle there was minimal difference, so based on that i'm thinking there wont be much in it. I changed them over lastnight and been running prime95 etc. to make sure it's stable and seems fine.

I took a few readings from HWInfo before and after, sitting at idle and under load.

i3 2120t @ idle.

Temperature 0 52°C (125°F) [0x32] (Core #0)
Temperature 1 35°C (95°F) [0x32] (Package)
Power 0 4.92 W (Package)
Power 1 0.80 W (IA Cores)
Power 2 0.11 W (GT)
Power 3 4.01 W (Uncore)
Voltage 0 0.99 Volts (VID)

i3 2120t @ full load.

Temperature 0 59°C (138°F) [0x2B] (Core #0)
Temperature 1 42°C (107°F) [0x2B] (Package)
Power 0 25.19 W (Package)
Power 1 20.99 W (IA Cores)
Power 2 0.10 W (GT)
Power 3 4.10 W (Uncore)
Voltage 0 0.97 Volts (VID)



i5 3470 @ idle.

Temperature 0 31°C (87°F) [0x47] (Core #0)
Temperature 1 34°C (93°F) [0x47] (Package)
Power 0 5.37 W (Package)
Power 1 1.15 W (IA Cores)
Power 2 0.03 W (GT)
Power 3 4.19 W (Uncore)
Voltage 0 0.88 Volts (VID)

i5 3470 @ full load.

Temperature 0 54°C (129°F) [0x30] (Core #0)
Temperature 1 58°C (136°F) [0x2F] (Package)
Power 0 38.68 W (Package)
Power 1 33.12 W (IA Cores)
Power 2 0.07 W (GT)
Power 3 5.50 W (Uncore)
Voltage 0 1.14 Volts (VID)

I don't know the accuracy of those but even just used as a number to compare there's really not a lot in it. For the boost in speed when needed too it's definitely worth it!

EDIT: I wouldn't rely on the temps too much some of the them were still cooling down or just after being powered on.
 
Last edited:
MajorMaccas said:
Yeah, TDP is the max they could use, though you can exceed this with overclocking. Idles between any of the CPUs is going to be negligible, dont forget at idle your system probably falls out of the load bracket that PSUs are certified 80plus for - as in, at idle, your power from the wall is likely to be less than 100W, even on a huge 6-8 core monster. PSUs are rated 80 plus at 3 load levels, 20/50/100%, with the most efficient load for a PSU being somewhere between 50-75%.

But an 80plus PSU at only ~5-10% load can be in the 70s or even 60s % efficiency. So if you have a 850 watt supply, only drawing 50w at idle, you're actually drawing 95w from the wall. If you got a new CPU that used more power, so at idle you're drawing 60w, you may find you only increase your wall draw by 5w, not the total difference of the CPU, because you're getting into a more efficient load for your PSU.
Click to expand...

I started writing my reply before your post and got sidetracked, some good info there thanks, never have really looked into PSU efficiency.

I think if I was going all out for efficiency i'd look at replacing the PSU, I think it's a 350w unit, not massively cheap though at the same not as efficient as modern specced units. Though until it's failing I imagine the cost / savings make it not really worth it.

I could do with getting a power measuring plug/socket one day as it's difficult basing it on a Wattson reading as fridge / freezer / other gubbins always click on and off making it difficult to get an accurate reading.
 
The other thing possible is undervolting. Everything will undervolt including memory, CPU and chipset.

To give idea my i7 4770k voltage is offset by -0.090 in bios and this is with an overclock to 4Ghz. I run my 1600 memory at 1.35v also.

It helps to have a more stable PSU, the better regulated the more room for undervolting.
 
JasonM said:
The other thing possible is undervolting. Everything will undervolt including memory, CPU and chipset.

To give idea my i7 4770k voltage is offset by -0.090 in bios and this is with an overclock to 4Ghz. I run my 1600 memory at 1.35v also.

It helps to have a more stable PSU, the better regulated the more room for undervolting.
Click to expand...

Yeah I have undervolted chips in the past, that through your bios or software?

I used to use a program called RMClock on an old build, you could specify the voltages at all the multipliers. I had a T7200 processor on a mini-itx board that just didn't like keeping cool, did the trick perfectly.
 
You must log in or register to reply here.
Back
Top Bottom