'Xtreme Undervolting' the stock conroe

[Lestat]

well with all the hub-bub about overclocking and what volts will get you were i figured i'd give undervolting @ stock a try.

what the heck lets have a little fun.
Stock volts for a 6600 ES is 1.235v
2.4ghz stock for a 6600 ES ram is 888 4-4-4-12 i figured i would let the ram run higher since it im concerned only with the cpu.
1.088v via the bios.
1.056 after droop @ 100% load w/ prime.

i aint looking for 10 hours of prime or anything like that.
undervolting will show its not stable fairly quickly.

so what the heck. if you feel like having a little fun. post some shots of your "Xtreme Undervolt"
and for those who are worried about heat, undervolting will help dramtically.

p.s.
i aint done dropping the voltage yet i just wanted to get the thread started.

cpu-z will show the correct voltages but use whatever you wish to show the voltage.
Speedfan also works fine.

[brandinb]

whats your temps like?

that proc gonna use like 40 watts or something configured like that?

[Lestat]

right now im at 1.000v via bios
and 0.976v after droop.
29c 100% load w/ Prime.
screenshot coming.

what im doing is setting it in the bios. then i will take the voltage of the actual droop. and then after i am done testing that, i go into the bios and set the voltage to whatever the droop was.

what i am seeing however, is that the lower the voltage the less droop its getting. which shows really poor Capacitors and PWM voltage control.

[mk0r]

nice idea, will be handy for people putting these in HTPC or other systems that need low noise or fanless cooling :)

[AndrewZorn]

only .03v droop at 1.0v... yeah definately a lot less at lower volts

[Lestat]

if you use a DMM on the motherboard the board only shows 0.01v droop at any voltage for me.
i am checking on 3 different vcore check points.
even 1.5v droop is still 0.01v when checking the voltage on the motherboard.
so.. take that for what you will.. :\

[Lestat]

here ya go.

time for bed for me so you guys do some benching. have some fun why not!

tommorrow i will do some more and drop the voltage further.
i am going to try for 0.75 or less. so we'll see how it goes.

[Lestat]

3dmark 01 stable

[Btrice]

This is probably a stupid question, but doesn't EIST work by lowering the voltage requirement as well as the multiplier? If so it probably wouldn't be "xtreme" undervolting though.

[PkG.1337]

EIST also downclocks it

[AndrewZorn]

if you use a DMM on the motherboard the board only shows 0.01v droop at any voltage for me.
i am checking on 3 different vcore check points.
even 1.5v droop is still 0.01v when checking the voltage on the motherboard.
so.. take that for what you will.. :\
what does BIOS/speedfan read? is the whole 'vdroop' problem not real?

[pcdoc1]

Absolutley right, so just run stock with EIST enabled and quit wasting time and energy...

[i found nemo]

well, it's his stuff. you should run comparitively temps with stock and temps, with stable undervolt.

[Lestat]

Absolutley right, so just run stock with EIST enabled and quit wasting time and energy...


thats absolutely not gunna work doc you should know that EIST lowers the multi too.

so wtf good does that do.
the freakin point is to run lowest voltage at stock.

and even if EIST lowered only the voltage. it will only do it when the PC is idle not doing full load.

come DOC use your head.. your better than that


and i guess we should all just stock wasting everyone's time and all of us quit overclocking..
yeah i think we should do that.

[Pete]

EIST will not work as efficent as what is being done here!

It lower the volts but it don't under volt!

I think they can go pretty darn low!!, your aiming for 0.75 and i think it's 0.70 they can drop too!

The 29c temps with stock cooler on too?

[K.I.T.T.]

http://i65.photobucket.com/albums/h240/ashish2787/lowvolt-00.jpg

Check out the temps :p:

Will try overclocking it at that vCore later, or may try seeing if it'll run stock at 0.75 v :D

[alpha0ne]

I'll have to give this one a go, see if .75 will work, get back to ya soon :)

[lawrywild]

Man, I wish ASUS supplied us with voltage options below 1.225 or whatever it is that's the lowest..

[K.I.T.T.]

I think it's stable enough

http://i65.photobucket.com/albums/h240/ashish2787/lowvolt-01.jpg

:p:

[|SiLA|]

here's mine E6400

1.15 with +~800mhz

http://img96.imageshack.us/img96/2330/uvolt2ys0.jpg

dunno why i can't give lower voltage :(

[realsmasher]

go away with your damn ES. ;)

no retail c2d can compare with ES results.

i even have problems running 1,2V with stock 6600. (stock is 1,325V)

[Lestat]

0.975 in the bios
23c 100% load via prime.

[FusionZ06]

Hmm this looks like fun I'll definitely be trying this later tonight. I'm down to 1.288 @ 3.0ghz right now :)

[Lestat]

yeah we could add like a 3ghz low voltage barrier too.

Fusion try for stock though also.

[FusionZ06]

yeah we could add like a 3ghz low voltage barrier too.

Fusion try for stock though also.

What's stock again? 1.25???

[Lestat]

no im talking stock mhz.

dont overclock let the cpu run stock and then start dropping the voltage.

so for you it will 2.4ghz

[K.I.T.T.]

Last update before i try lower

http://i65.photobucket.com/albums/h240/ashish2787/lowvolt-02.jpg

:D

[Btrice]

The whole reason I asked about EIST was not to suggest Lestat was wasting his time, but since EIST changes the multipliers 6-9 or whatever for each C2D, if you could read the Vcore with a meter and put the processor under various loads you should be able to get readings for good starting points at each multiplier to do your xtreme undervolting. Who knows you may find you can reduce vcore by a higher percentage at x7 instead of x9.

[Lestat]

well technically since the 7x is running lower mhz then you would be able to run lower volts.
so....?

since eist lowers the multi there isnt a way to judge the proper low voltage end of lets say my 6600 because it would drop the multi clear down to 6x @ 233mhz. so that defeats the point of my undervolting, which is to leave your cpu at stock, whatever that maybe. and drop the voltages.

see what i mean

i didnt have time to test prior to work this morning but i did set 0.975v and she booted just fine.
ran prime for about 5 minutes looked good.
for some reason i was having issues with my nic.
i couldnt get an IP and keep it.
my AP is a bit flakey sometimes and it does this once in a while i just thought it odd that it started doing this at very low vcore.

ill do more testing tonight.

[Btrice]

I said higher percentage. Ok let me spell this out:
(I am sure these numbers are wrong they are just for illustration purpose)
Stock x9 mult @ 1.325 V, Undervolt best stable @ 0.975 V, -26.42% change
Normally x7 mult @ 1.275 V, Undervolt best stable @ 0.875, -31.37% change

I see where you mean it doesn't stay "stock" though because you are forcing a lower multiplier. But aren't you not keeping it stock by disabling EIST in the first place?

[Lestat]

well fellas looks like 0.975 is the best i can do @ stock mhz which again is 2.4ghz for my 6600.
anything less either blue screens or locks windows right when prime starts.

so thats 0.4v lower than what intel specs the cpu at.

makes you wonder wtf they are doing overvolting the cpu so much.
but then again when you take into account all the possible mobo's and the possible vdroops. its smarter to overvolt than undervolt them and have them locking up on various baords from lack of vcore.

still 0.975v thats a nice drop in vcore for stock.

im now getting ready to test lowest vcore for 3ghz.

[Lestat]

I said higher percentage. Ok let me spell this out:
(I am sure these numbers are wrong they are just for illustration purpose)
Stock x9 mult @ 1.325 V, Undervolt best stable @ 0.975 V, -26.42% change
Normally x7 mult @ 1.275 V, Undervolt best stable @ 0.875, -31.37% change

I see where you mean it doesn't stay "stock" though because you are forcing a lower multiplier. But aren't you not keeping it stock by disabling EIST in the first place?


9x multi is 1.23v
7x is i have no idea.... probably the same though.

disabling all throttling functions will make you stay at whatever setting you set it at. for the multi and fsb i mean.
so disable fsb clocking set the default multi and go to town.

set your ram at whever you like i guess. but lower ram. like auto on the ram setting or whatever it is for your system shoul yield the lowest possible vcore.
not sure how much effect higher dram freq will have on vcore. im sure it has some effect but not sure how much.

[zoob]

Mmmm sounds like this will make an awesome low power (ahahaha.. oh boy that made me laugh) rig. So far I've been building very quiet machines using undervolted A64's.. the results here are very encouraging!

Do any of you have Watt meters to measure the decrease in power consumption? I would be quite interested in those figures.

[onewingedangel]

We need some standardised way of testing performance/watt of total systems.

eg.
For 3d performance = 3d mark 05 score/total system power consumption
For general performance = superpi/ total system power consumption

the results could be adjusted to give comparable results for degrees of efficiency.

Just a thought.

[Lestat]

99% of us have no way of measuring power consumption.
doing the math is one way sure, i dont know the equation, but i believe more in the physical measurments of power draw more than something from math only. just the way i am.

standard would be catagories by CPU
6800
6700 (god did anyone even buy one of these things?)
6600
6400
etc.

guidlines are
1)
stock cpu multiplier
2)
Stock fsb ie 233mhz
3)
stock dram (in order to make it equal we must all run at the same speed of ram. i guess we can all set our ram to 667. CAS wont have any effect. frankly i dont think dram freq will effect it at all but to be fair....)
4)
cooling dont matter for undervolting, not that im aware of
5)
3dmark 01 Results (05 isnt the cpu bencher)
6)
must run atleast 30mins of your favorite Prime95 program (from what i found out, on the undervolting, prime fails right away, or locks your system so, 30min's is plenty of stressing)

we could go from there.

but this is just something i wanted to do for fun, if you guys wanna go more official with it we can.
i mean how many of us are running stock ;)

[zoob]

If you guys want to go complicated like that, by all means go ahead :)
I had originally intended someone with a Kill-A-Watt meter to measure the power consumption at stock and undervolt, as more of an indicator of how much an impact the CPU alone makes rather than the entire system.

[Kango]

This could be both fun and informative. Let's follow Lestat's steps and begin to post our results. Knowing the power would be nice but we never know our power for sure when we overvolt/overclock. And don't forget your rig description. :rolleyes:

Kango

[Nullvoid]

Please could you add to that list the motherboard you used? It's nice for people to know which are going to be flexible for options/actually successful as far as undervolting goes (didn't someone already mention their board has no undervolt option at all?).

[sergiu]

Depending on the country you live, almost anyone should be able to measure the power consumtion. First, unplug all the electronics except the computer and go to electronic meter or power counter ( I don't know the accurate translation from my language). Second, counter the time between two light blinks if it is electonic or the time power counter's platen needs to make a full rotation (if it is mechanical). There is a relation between light blinking/platen rotation and power consumtion which is usualy written on the power counter (in my country 1KW is measured at 480 platen rotations or 1000 light blinks). In this way you can measure power consumption of the system quite accurate!

[Lestat]

Depending on the country you live, almost anyone should be able to measure the power consumtion. First, unplug all the electronics except the computer and go to electronic meter or power counter ( I don't know the accurate translation from my language). Second, counter the time between two light blinks if it is electonic or the time power counter's platen needs to make a full rotation (if it is mechanical). There is a relation between light blinking/platen rotation and power consumtion which is usualy written on the power counter (in my country 1KW is measured at 480 platen rotations or 1000 light blinks). In this way you can measure power consumption of the system quite accurate!


ok yeah.
although its a way to to check thats just NUTZ

man,,, you gotta be kidding.

thats a neat idea, im surprised someone actually went so far as to post this method. lol.

but i dont think we are going to be doing this.

technically you would have to go thru and switch off the breaker for every single item in your house except the room your pc is in.
you cant just turn things off. you actually have to cut the breaker. electronic devices still draw power in their off state.
you have frig's and freezers, and A/C and TV's and stereo's and god knows what else.
neat idea,, but no.

[Lestat]

Please could you add to that list the motherboard you used? It's nice for people to know which are going to be flexible for options/actually successful as far as undervolting goes (didn't someone already mention their board has no undervolt option at all?).


haha god i forgot to add that.
so yeah.

ok..
lets get this started.

List Cpu
lowest stable volts
Motherboard Make and Model
PSU
Ram and what Freq you are setting it at
Must include a screenshot showing Prime or some other stressing program and the voltage.
You may include a screenshot of your favorite 3d benching program also. but in my eyes if its prime stable usually 3dmark is also.

[sergiu]

Well, it's the poor man way to check the power consumtion, I tried it several times and it's more accurate than you think (it senses even a small underclocking to a low end graphic card), and it's quite simple when you have a small house with 4-5 electronics. If you have a house with ten rooms and in each room 3-4 electronics, it will just take you several minutes more to unplug all the electionics (unplug them from power, not just turn off). Measuring the power consumtion only takes 1-2 minutes, so I don't think it's a problem if you cut the power of frig's and freezers for such a short period of time. Also, if you're lucky enough, you may have an electronic power counter which shows you directly the power consumtion, it's worth to check out!