• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Radical UNDERclocking

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Xymox1

Registered
Joined
Aug 9, 2011
I have a ASUS Rampage 3 Extreme. Ive got it nicely stable with a i7 950 at 4.3Ghz. Ive also got a EVGA GTX580 FTW. Im using a Cosair AX1200 power supply.

At 100% CPU and 100% Graphics card it draws 588 Watts. At idle its 229 Watts. That works out to 63.7 cents a day on my power plan and so $19.11 per month if I leave it on 24/7 at idle..

With this MB I can store and recall BIOS profiles apparently. Not sure how easy it is during boot to select profiles tho..

So I was thinking I could make a really seriously low power under clock..

Anyone with experience doing this ? This is a weird thought for me. Just how low can I go in all the settings ? Where do I get the best power savings ? Can I clock a i7 950 at 1Ghz ? Whats the best voltages for my super underclock ?

Hahaha.. Fun thought. Super underclocking ! How slow can you go ? hahaha

Can you reach 1hz ? heheheh...
 
588W at load? Wherever you are reading that, I dont believe it to be correct. Way too high.

Anyway here is where you will get into problems, the lower you go the lower the efficiency on that already overkill PSU. So if you are sitting at say 125W idle, your efficiency may only be at 70% as opposied to the 90% in that PSU's sweet spot.

Why dont you just have all power saving features enabled? Did you over extend your self in the purchase? Its like you bought a Ferrari but cant afford the gas...LOL!
 
Hahaha.... I can afford the gas... hehehe... But I was just thinking if it was easy to setup a low power profile I might ! It would be a fun challange.. 1Hz.. haha..

I am a electronics engineer and I have serious test equipment handy. So I measured the RMS current being drawn from the wall directly using a Fluke Scopemeter. It can measure the line voltage at the same time as it measures the current being drawn and do the math on the waveforms to calculate a accurate RMS power..

Thats a correct value.

Well.. I think its right. I suppose I mighta got something wrong but its a pretty simple measurement.

I have a tektronix scope and some other gear I could triple check my measurements I suppose..
 
A heavily OCed Bloomfield and a GTX580 would probably be 600W. Assuming he's measuring at the wall, the components are drawing less then that depending on the PSU efficiency. It seems right to me.
 
OoOO... With 3DMark 11 doing the Xtreme test with running Prime95 at the same time I hit 675 Watts RMS.

This sounds reasonable with my MB / CPU / GPU and factoring in the loss in the power supply as you correctly point out.

Its mostly GPU.. Idle current draw at 121Vac is 1.99A RMS. With 100% CPU it draws 3.2A RMS.. Thats a difference of 146 Watts RMS being drawn by the 100% load on the CPU alone... That sounds about right considering the power loss in the power supply.

So the 2A at idle must be the CPU + all the stuff on the MB + I have 2 Raptors in a raid 0 and 2 big drives in a raid 1. My OS is on a SSD.. There is also a Adaptec RAID controller card for my RAID 1.

According to ASUS RCTweakit my CPU pulls 18.4 watts at idle. ( 13.3 A at 1.375vdc Vcore )..

Hmmm.... Yea must be my HD's and Adaptec RAID card, fans, water pump and IOH ICH all drawing the rest of the 200 Watts at idle.. That does seem a bit high tho..

Interesting tho... Yea using different gear I measured about the same power draw from the wall. I think its pretty accurate.. Maybe off by 1%..

If the CPU is only 1/4 of the total power draw at idle tho im not sure underclocking and power saving is really gonna do me a lotta good to my idle power draw.. Sure having it go into standby will seriously help me tho...
 
Last edited:
A heavily OCed Bloomfield and a GTX580 would probably be 600W. Assuming he's measuring at the wall, the components are drawing less then that depending on the PSU efficiency. It seems right to me.
I barely broke 400W at the wall with a 580 and i7 2600k (look at my 580 review). Now, that was with a killawatt but still. 588W should not be right with 580 at stock. Another thing, Im running a 580 and 2600k on a seasonic 560W psu at the moment, the fan barely spins upon load...

TDP on the 580 is 244W by they way. 950 at 4.3v isnt more 200-225W. ;)

Anyway, what was the question? :p
 
Last edited:
Ive never heard of that equation before...sorry.

TDP is a general idea of power. Thats what ALL reviews sites base their power ratings off of when they test. It seems to be a direct correlation at minimum. ;)
 
Well TDP is Thermal Design Power, it's how much heat (in watts) the device puts out, not how much it's consuming. A perfectly efficient device wouldn't put out heat, it would use all consumed energy to do work. Since it isn't perfectly efficient, some of the energy is used producing heat, some is used to do work. I'm pretty sure that the majority of the energy is dissipated as heat, as work is force over a distance, and it's not exactly moving electrons very far.
 
To be frank its above my head. I just know, from reading so many reviews they use it as a direct correlation to power consumption. Maybe all those review sites are wrong, I dont know...

Im going to try to run how he has it (which is an incredibly unrealistic load unless folding maybe and see what my KAW says....hang tight and look for an edit here in a few. :)
 
Hehehe... Yes THE QUESTION... We did stray from the topic...

Underclocking.. I guess I will rephrase my question in the light of my new measurments :)

Is it possible to underclock and undervolt maybe to achieve a lower then normal power consumption ? Also is it worth it ?

On the off-topic discussion, im sure of my power measurements from the wall. But yea I think I agree that they seem high for idle. It seems like I have a extra 75-100W in there I dont understand.

Maybe your right and somehow I mismeasured. Its a pretty easy measurement tho. I will use a better method. I was using a http://www.fluke.com/Fluke/usen/Accessories/Current-Clamps/i30s.htm?PID=56297 current clamp around the hot lead of a power cord. But I will cut the hot lead and measure with a direct current measurement. I have a meter here some place that will go to 20A using a current shunt.. I also need to look at the current waveform and see what that looks like. I will check these things and report back.

I used a Tek MSO3034 scope plus the above current clamp for my previous measurements. http://www.tek.com/products/oscilloscopes/mso3000/

Earthdog: thanks dude ! I am interested to see what you get... Im interested in idle mostly..
 
Last edited:
It is possible to do so, sure, just the opposite method of overclocking. Is it worth it? Well, see my other statement. :)

___________

OK... here goes... obviously not apples to apples since I have a 2600k so in order to compensate a bit I used way more voltage than I needed to and a higher clockspeed

i7 2600k @ 4.5Ghz 1.48v loaded
Asus GTX 580 Direct Cu II stock

(all measurements at the wall with ~%90 efficient PSU)

Idle @ stock all the way around = 90W
Load @ stock (CPU ONLY) - 155W
Idle @ 4.5Ghz = 116W
Load @ 4.5Ghz (CPU ONLY) - 235W

Prime95 (4.5Ghz CPU) + 3D11 (Stock 580) = 445W Peak...~425W average.

Your equipment is WAY more accurate than mine, so... I dont know. Mine about match every review I have ever seen, thats all I know. No clue man! :)
 
Hmmm..... well thats a big difference...

I forgot to mention I am overclocking my 580.. But just looking at idle is where the difference seems to be.. Hmmmmm.....

I just kinda jury rigged up a way to measure current with a meter directly. Using a meter with a current shunt in it is way more accurate then a current clamp. Even tho its a much less expensive way to measure it.. It agreed with both other ways I measured so its gotta be correct..

How are you measuring current / watts ? What device is doing the measurement ?

This really isn't a hard measurement technically. The only complication could be that the meter is not doing a good RMS measurement. Is your reading RMS ?

Im uploading a youtube of my last test using the meter with a current shunt in it. Jury rigged but it does the job.. Gimme a few.. I will post the vid..

I do have a lot of other *** in my computer. 4 drives a high power Adaptec RAID card that gets hot not to mention fans and a water pump.. I just dont see all that drawing 100W tho...

Hmmm.... This is a mystery... I could start pulling stuff from the computer..

I appreciate your time :) Its interesting...
 
Sadly, a Kill A Watt... like I said your equipment is by far more accurate.

Did you adjust voltages too or just overclock the 580? What are your settings? I can see a 580 at 1.15 and 900+ clocks getting close to that number... Lets see the difference in my 580 at stock vs 900+ core and 1.15v...

Edit... now that the whole story has been told (overclocked GPU) I get a lot closer...

1.15v 920 core 2100 memory

~330W with no P95 and 3d11 (peak)
~430W with no P95 and 3d11 (peak)
~530W with P95 and 3d11 (peak)

Idle at 4.5Ghz and o/c GPU - 122w
 
Last edited:
Hehehe... Be easy on me.. Im new to this :) I didnt even see voltage settings on the 580 ! But Ooooo that sounds interesting :)

Lets see.. Im using EVGA Precision and its set at Core clock 930 shader 1860 memory clock 2310. Its stable.

Its a EVGA GTX 580 FTW Hydro Copper 2

My vid is like 7 more minutes from uploaded..

There will be a break in my posting after the vid. I have to go check on some stuff. Be back in like 1hr..
 
Most 4xx and 5xx cards will downclock alot at idle.

Also, at one time my wifes rig and mine were nearly identical save for PSU's. I had a 1200w Corsair and she had a 600w Antec. With the same OC on both rigs her rig pulled around 30-40w less than mine at idle. So I would think that is the efficency coming into play since at idle our rigs draw around 200w.

Instead of downclocking the computer how about replacing all your incandecent bulbs in your house with CFL or LED bulbs? I recently replaced ~15 60w incandecent for 13w CLF's. Thats a 705w savings, assuming all the lights are on at the same time. :thup:
 
See my edit... Im betting with pump and more fans than me and such its not out of the realm of possibilty. You just didnt say you overclocking the GPU or had water and such in the first post when this started! :)
 
Again this is jury rigged but using a meter with a current shunt is more accurate then using a current clamp and a scope..

To get power you multiply the current times the voltage. My line voltage is 120Vac.

Measuring current and power.
.
 
Yea sorry for multiple posts.. Yea sorry for not mentioning all the other stuff.. You do have me curious tho and I will explore it more shortly...

Yea not hitting the sweetspot on my power supply indeed must be less efficient.. I wanted this supply tho for its dual 8 pin ATX connectors. I like a good connection to the MB :)

onefstsnake: Indeed I do already use lots of CFL's. I also use Crestron automation gear to do lots of savings. I even wrote my own code for a thermostat because I wanted to control duty cycle and do other tricks.. So im already saving a lot in other ways...

MOSTLY I just liked the fun idea of seeing how slow a CPU could clock ! Thats a fun exercise for absolutely no pourpose at all !..
 
I believe you, just the whole story wasnt told up front. I sure believe that a raid card, pump and more fans than me ( I have two) can be in the ballpark of the difference between our results.
 
Back