Volts and Watts

Printable View

08-11-2007, 04:40 PM
Danger30Q

Volts and Watts

I thought this post would be appropriate here in WCG since most of us are concerned with wattage consumption of our crunchers.

I know the equation is (amps x volts = watts). I tried during some research to figure out how much wattage changes when you change the cpu volts but I realized it doesn't scale linearly.

Basically I'm curious to know how many more watts are consumed when increasing cpu volts from 1.35 to say 1.45 volts.

My Windows Home Server crunches 24/7 with a 65W Brisbane 3600+ X2. The stock volts I believe are 1.35V. Is there anyway to estimate the increase in wattage consumption from 1.35V to 1.45V?

How would this change when calculating it for a Core 2 Duo?

Thanks a ton!
08-11-2007, 05:34 PM
[XC] riptide

Quote:

Originally Posted by Danger30Q

I thought this post would be appropriate here in WCG since most of us are concerned with wattage consumption of our crunchers.

I know the equation is (amps x volts = watts). I tried during some research to figure out how much wattage changes when you change the cpu volts but I realized it doesn't scale linearly.

Basically I'm curious to know how many more watts are consumed when increasing cpu volts from 1.35 to say 1.45 volts.

My Windows Home Server crunches 24/7 with a 65W Brisbane 3600+ X2. The stock volts I believe are 1.35V. Is there anyway to estimate the increase in wattage consumption from 1.35V to 1.45V?

How would this change when calculating it for a Core 2 Duo?

Thanks a ton!

Right. I run a QX6700 on this rig. I'm gonna set its clocks back and change from 1.35V to 1.45V. Its always in a Kill=a=watt. BRB. :) Give me 2 minutes.
08-11-2007, 05:39 PM
[XC] riptide

From 1.35-1.45 = 30Watts change. 100% loaded.
From 1.45-1.56 = 40Watts change. 100% loaded.

Now this is set via Windows app... not actual after droop.
08-11-2007, 05:46 PM
alexio

Power draw increases +- exponentially with vcore. So: 1.45 / 1.35 = 1.074074....
1.074 ^2 = 1.154
The increase in power draw is 15.4%
Power draw increases a little worse than linearly with clockspeed.

Example: 65W TDP, 10% increase in voltage + 15% increase in clockspeed =
1.1^2 X 1.15 X 65W = +-90W.

It's a very rough calculation though....
08-11-2007, 05:47 PM
fornowagain

Exponential? No, power comes from the square of the voltage.

I've simplified this and ignored static power leakage. But its pretty close in practical terms. A common expression for semiconductor power P, P=kLV²F

Where:
L = Load (from software)
V = Core voltage, default vcore V1 is VID.
F = Chip frequency
k = Empirical constant for the chip.
Pd = Total cpu power
TDP = Thermal Design Power

Therefore delta P, i.e. change in power Pd-TDP. Rearrange for Pd, the constants cancel.

Pd = TDP x (F2/F1) x (V2/V1)²

Or a little easier to understand

OCed Watts = TDP x (OCed Mhz / Default Mhz) x (OCed Vcore / Default Vcore)²

online version
08-11-2007, 05:51 PM
alexio

Quote:

Originally Posted by fornowagain

Exponential? No, power comes from the square of the voltage.

Yeah, I used the wrong therm, but meant the same thing. It's almost 4.00 AM here now :p:
08-11-2007, 05:59 PM
fornowagain

Quote:

Originally Posted by alexio

Yeah, I used the wrong therm, but meant the same thing. It's almost 4.00 AM here now :p:

Yeah, I saw your squared term after looking, 3.00AM here:p:
08-11-2007, 06:06 PM
fornowagain

Quote:

Originally Posted by [XC] riptide

From 1.35-1.45 = 30Watts change. 100% loaded.
From 1.45-1.56 = 40Watts change. 100% loaded.

Now this is set via Windows app... not actual after droop.

Also note that, that includes the PSU efficiency losses.
08-11-2007, 06:16 PM
alexio

Quote:

Originally Posted by fornowagain

Also note that, that includes the PSU efficiency losses.

And don't forget CPU power circuity losses. Mofsets get warmer for a reason when the power draw of a CPU increases. It may just be a few watts normally, but it may be a bit higher when they are being pushed to the max.
08-11-2007, 06:21 PM
[XC] riptide

Quote:

Originally Posted by fornowagain

Also note that, that includes the PSU efficiency losses.

I haven't. And I haven't forgot the extra heat load drop in efficiency, nor the PSU effeiciency... or the variable vdroop effect.

However, my simple measurements are real world for the purposes of the OT.
08-11-2007, 06:26 PM
Danger30Q

Thanks a lot for the info guys. I'm thinking some others here would be interested in this as well. Some of you really know your physics calcs. I have a few more that you can hopefully answer.

Is there an easy way to factor in a very rough estimate of the total power losses from all circuitry and PSU efficiency?

Would the same formula work for an overclocked video card with increased voltage?

How do you calculate increased power consumption from adding volts to your overclocked RAM?
08-11-2007, 06:28 PM
[XC] riptide

Quote:

Originally Posted by Danger30Q

Thanks a lot for the info guys. I'm thinking some others here would be interested in this as well. Some of you really know your physics calcs. I have a few more that you can hopefully answer.

Is there an easy way to factor in a very rough estimate of the total power losses from all circuitry and PSU efficiency?

Would the same formula work for an overclocked video card with increased voltage?

How do you calculate increased power consumption from adding volts to your overclocked RAM?

Well lets see then..... BRB

2 x 512 Sticks Corsair XMS @ 395Mhz.

Set at 2.2 in Wondows. Change to 2.3 = 3W increase.

Change from 2.3 - 2.1 = 8W decrease. Hmm interesting. Full load of course running a VERY RAM intensive APP 900megs used.

I switched my X1900XT from OC'd 3d mode (vcore at 1.575 mem at 2.1 clocks at 756/820 core/mem) to 2d mode using atitool (vcore at 1.175, mem 2.1v , clocks at 500/600 core/mem). Kill-o-watt dropped by 25 Watts. Now GFX was not loaded at all.
08-11-2007, 06:42 PM
alexio

Quote:

Originally Posted by Danger30Q

Thanks a lot for the info guys. I'm thinking some others here would be interested in this as well. Some of you really know your physics calcs. I have a few more that you can hopefully answer.

Np ;)

Quote:

Is there an easy way to factor in a very rough estimate of the total power losses from all circuitry and PSU efficiency?

Efficiency of a good PSU at load should be +-80%. Efficiency is rather consisted on most PSUs across different loads unless you put a very small load on it or you max it out. In those cases it can drop below 70%. The power circuity for the CPU should have 75-90% efficiency unless you overstress it. It's hard to find more accurate numbers than these.

Example: If the efficiency of the PSU is 80% and the efficiency of the power circuity is 85% then the real efficiency (from wall outlet to CPU) would be:
0.80 X 0.85 X 100 = 68%

Quote:

Would the same formula work for an overclocked video card with increased voltage?

For the GPU alone it would. The extra stress on the vRAM from the faster running GPU is neglectable. What you do have to keep in mind is that a GFXs power circuity isn't build to run at higher voltage in most cases. You may be pushing the components which can add quite a few watts.

Quote:

How do you calculate increased power consumption from adding volts to your overclocked RAM?

Exactly the same as with the CPU and GPU. The temperature factor may become very important if you don't actively cool your RAM while running high volts trough them. This can significantly increase the power the chips consume.
08-11-2007, 07:03 PM
Danger30Q

Thank you riptide for testing this out - I really should buy myself the kill-a-watt.

Is my math correct using a 6700 with stock of 2.66ghz and 1.35V stock and OC'd to 3.4ghz and 1.45V:

65 TDP x (3.4/2.66) x (1.45/1.35)^2 = 95 watt increase due to OC

The default watt consumption at stock everything is 65 TDP x 1.35 = 88W
08-11-2007, 07:12 PM
alexio

Quote:

Originally Posted by Danger30Q

Thank you riptide for testing this out - I really should buy myself the kill-a-watt.

Is my math correct using a 6700 with stock of 2.66ghz and 1.35V stock and OC'd to 3.4ghz and 1.45V:

65 TDP x (3.4/2.66) x (1.45/1.35)^2 = 95 watt increase due to OC

The default watt consumption at stock everything is 65 TDP x 1.35 = 88W

Default wattage is 65W (1.35V X 48A), no need to multiply it by the vcore, because that is included in the wattage already. The new power draw is +- 125W, so the increase is 60W. 65 x (3400/2667) x (1.45/1.35)^2 = 125W
I think you forgot the "^2" part.
08-11-2007, 09:51 PM
Danger30Q

Now I'm confused with my math...I get the same as my old calculation:

(1.45/1.35)=1.07407 and 1.07407^2=1.1536

65 x 1.275 x 1.1536 = 95.6 W?????
08-11-2007, 09:57 PM
STEvil

you're leaving out the mhz increase.
08-11-2007, 11:58 PM
Okda

isn't this calculator is accurate enough ?

http://www.extreme.outervision.com/tools.jsp#cpuoc
08-12-2007, 12:09 AM
Mad_Man

Quote:

Originally Posted by Danger30Q

Now I'm confused with my math...I get the same as my old calculation:

(1.45/1.35)=1.07407 and 1.07407^2=1.1536

65 x 1.275 x 1.1536 = 95.6 W?????

one thing to remember, 65W tdp cpu doesnt use 65W, 95W tdp cpu doesnt use 95W at load. and so on.
If cpu is flagged as 95W tdp, it simply means it is under that. But above next lower tdp rating. you've got the point.
Actually, without knowing the stock full load power consumption, you can calculate only theoretical increase in percentage.
in other words, 1.47*65W is the same 47% increase of 1.47*125W.
but if your cpu uses 64W this increase of 47% will translate to 30W increase
and if your cpu uses 48W (the same 65W category) this same increase of 47% will yield in ~22W increase.
and the diff is even larger with higher tdp ratings.
08-12-2007, 12:10 AM
littleowl

Quote:

Originally Posted by SaFrOuT

isn't this calculator is accurate enough ?

http://www.extreme.outervision.com/tools.jsp#cpuoc

Hey SaFrOuT Thanks!!! :clap:
08-12-2007, 12:15 AM
Mad_Man

Quote:

Originally Posted by SaFrOuT

isn't this calculator is accurate enough ?

http://www.extreme.outervision.com/tools.jsp#cpuoc

hell no... imho :P
it says my 4000 brisbane has stock Vcore 1.35V and 2000Mhz
sad its 2100mhz and mine was 1.3V
and it says its 65W stock.
well, if you read my previous post you got the point...
08-12-2007, 12:24 AM
littleowl

hum ok so I am stupid please explain all that in english! :D
08-12-2007, 12:36 AM
STEvil

Quote:

Originally Posted by Mad_Man

one thing to remember, 65W tdp cpu doesnt use 65W, 95W tdp cpu doesnt use 95W at load. and so on.
If cpu is flagged as 95W tdp, it simply means it is under that. But above next lower tdp rating. you've got the point.
Actually, without knowing the stock full load power consumption, you can calculate only theoretical increase in percentage.
in other words, 1.47*65W is the same 47% increase of 1.47*125W.
but if your cpu uses 64W this increase of 47% will translate to 30W increase
and if your cpu uses 48W (the same 65W category) this same increase of 47% will yield in ~22W increase.
and the diff is even larger with higher tdp ratings.

AMD rates that way, yes. Intel rates on average processor useage... so "65w" processors can actually put out more if you push them to full potential, which is why Intel laptops run so hot a lot of the time.
08-12-2007, 01:10 AM
Mad_Man

Quote:

Originally Posted by littleowl

hum ok so I am stupid please explain all that in english! :D

thats what im trying to do all the time lol... english is sometimes very hard, especialy if i try to explain smth more... educated rofl

Quote:

Originally Posted by STEvil

AMD rates that way, yes. Intel rates on average processor useage... so "65w" processors can actually put out more if you push them to full potential, which is why Intel laptops run so hot a lot of the time.

my bad, i didnt know that.
maybe its long time since i had intel setup.
if i had spare $425 ill buy new Q and mobo :(
08-12-2007, 01:16 AM
littleowl

Quote:

Originally Posted by Mad_Man

thats what im trying to do all the time lol... english is sometimes very hard, especialy if i try to explain smth more... educated rofl

my bad, i didnt know that.
maybe its long time since i had intel setup.
if i had spare $425 ill buy new Q and mobo :(

now see that is just bull! I asked you to come here and crunch! :shakes: and then you say I am stupid...... OH wait I said that! :rofl: Thanks that gave me a nice lol at 515 am!
08-12-2007, 01:25 AM
Mad_Man

anything for you to wake you up at 5.15 :p:
nothing can be nicer than a nice lol in the morning... well, there's something nicer :P
mmm, to much "nice" in one post...

next time you should think twice before asking someone to join, especialy if he have "mad" in nick :rofl:
08-12-2007, 01:30 AM
littleowl

Quote:

Originally Posted by Mad_Man

anything for you to wake you up at 5.15 :p:
nothing can be nicer than a nice lol in the morning... well, there's something nicer :P
mmm, to much "nice" in one post...

next time you should think twice before asking someone to join, especialy if he have "mad" in nick :rofl:

Are you sure your not a rip tide JR? :D

I am not waking up I am getting ready for bed at 8am

:D I will try and remember that mad part! :p:
08-12-2007, 01:43 AM
Mad_Man

Quote:

Originally Posted by littleowl

Are you sure your not a rip tide JR? :D

I am not waking up I am getting ready for bed at 8am

:D I will try and remember that mad part! :p:

then you have seriously switched daytime lol
i remember, mom told me that some babies have switched theyr daytime and they are woken up in night and sleeping during day. but she said "babies" not adults :ROTF: :ROTF: :p:

and yes, im sure who am i :)
08-12-2007, 01:47 AM
littleowl

:rofl: well if I had my way about it I would be home sleeping! but no I am stuck here making money :p:
08-12-2007, 02:11 AM
Mad_Man

then i hope the money will be used for good things. another Quad, or two ;)
and then you'r stuck in work for eternity just to pay that power-bill
08-12-2007, 02:41 AM
fornowagain

Quote:

Originally Posted by [XC] riptide

I haven't. And I haven't forgot the extra heat load drop in efficiency, nor the PSU effeiciency... or the variable vdroop effect.

However, my simple measurements are real world for the purposes of the OT.

You misunderstand my point, I wasn't nitpicking so chill out. The OP wanted change in Wattage from increases in Vcore. I'm merely pointing out, that your measurement point is after the AC-DC conversion and includes an approximate 20% loss from the PSU. I pointed this out so that you can more easily compare numbers against theoretical increases. I didn't include a half dozen other reasons for slight error in numbers, anything from power unity to heat load. Just the biggest contributer thats,

a. Verifiable from psu efficiency curves
b. Easy to include.

Quote:

Originally Posted by [XC] riptide

From 1.35-1.45 = 30Watts change. 100% loaded.
From 1.45-1.56 = 40Watts change. 100% loaded.

Now this is set via Windows app... not actual after droop.

ALL I was saying is apply the losses from the psu to the figures to make them more accurate.
08-12-2007, 03:18 AM
[XC] riptide

Quote:

Originally Posted by Mad_Man

and yes, im sure who am i :)

Son, I thought I told you to get to bed earlier. C'mon you have school in the morning :rolleyes:

Quote:

Originally Posted by fornowagain

You misunderstand my point, I wasn't nitpicking so chill out. ///snip

Hey. I am chilled. I'm Mr. Ice. Wasn't getting excited at all. Just saying that AFTER all the PSU drops, and circuit realities thats what he should expect. Hell even the Kill-o-watt has got margins of error.:)
08-12-2007, 03:29 AM
[XC] mysticmerlin

Quote:

Originally Posted by Mad_Man

then i hope the money will be used for good things. another Quad, or two ;)
and then you'r stuck in work for eternity just to pay that power-bill

Nahh, he leaves that to me to get the quads. He just spanks it to them :ROTF:
08-12-2007, 03:31 AM
fallwind

1 Attachment(s)

Thought you guys might like to see this result from my crunching farm! That's 776w to power:

-4 Q6600's B3 3.2GHz 1.45 vcore (BIOS) fully loaded with BOINC+WCG.
-4 motherboards
-6 sticks DDR2
-10 120mm fans
-2 MCP655 pumps
-4 HD's
-1 wireless bridge
-4-400 Sparkle power supplies at 80%?? efficiency.
-zero video cards

That calculates out to 620w being used by the components after figuring in PSU efficiency. I don't know how much power the MB's, pumps, HD's, RAM and wireless bridge would use but I'm guessing maybe 200w? That leaves 105w for each CPU or 26w per core.

That's sounds pretty efficient to me. I mean that's less watts than a vacuum cleaner for over 50GHz of crunching power! If I had G0's the figure would probably drop some more. Things could get pretty interesting in a few years if/when 8-16 core CPU's arrive! :up:
08-12-2007, 04:46 AM
Mad_Man

Quote:

Originally Posted by [XC] riptide

Son, I thought I told you to get to bed earlier. C'mon you have school in the morning :rolleyes:

:sofa: :eek:
sorry dad, im guilty
:ROTF:
well, its 2.45pm here, so i have some time left till dusk

Quote:

Originally Posted by fallwind

Thought you guys might like to see this result from my crunching farm! That's 776w to power:

-4 Q6600's B3 3.2GHz 1.45 vcore (BIOS) fully loaded with BOINC+WCG.
-4 motherboards
-6 sticks DDR2
-10 120mm fans
-2 MCP655 pumps
-4 HD's
-1 wireless bridge
-4-400 Sparkle power supplies at 80%?? efficiency.
-zero video cards

That calculates out to 620w being used by the components after figuring in PSU efficiency. I don't know how much power the MB's, pumps, HD's, RAM and wireless bridge would use but I'm guessing maybe 200w? That leaves 105w for each CPU or 26w per core.

That's sounds pretty efficient to me. I mean that's less watts than a vacuum cleaner for over 50GHz of crunching power! If I had G0's the figure would probably drop some more. Things could get pretty interesting in a few years if/when 8-16 core CPU's arrive! :up:

some serious power there... how much do you pay for power-bill?
08-12-2007, 08:06 AM
fallwind

$150/month roughly, power is relatively cheap here. This is my first summer with the farm though so I'm not looking forward to the bills that are coming because of central a/c use.
08-13-2007, 10:46 AM
v0dka

That's not a bad powerconsumption at all fallwind! Most gaming rigs that half of that easily. And then youre only powering a dualcore, GPU and a fancy monitor. You got 25Ghz crunching for the same amount of watts. :eek:
08-13-2007, 01:43 PM
fallwind

Actually it's 51.2GHz now. All 4 racks of the farm now sport Q6600's running at 3.2GHz. Less than 200w each including efficiency losses isn't bad at all! This was the first time I hooked up the Kill-a-Watt since upgrading to all quads and I was quite surprised at the result. I was thinking atleast 1kw. I guess these cheap Sparkle PSU's are pretty efficient and the Q6600's don't use as much power as you would think.
08-13-2007, 02:46 PM
mike047

Quote:

Originally Posted by fallwind

Thought you guys might like to see this result from my crunching farm! That's 776w to power:

That's sounds pretty efficient to me. I mean that's less watts than a vacuum cleaner for over 50GHz of crunching power! If I had G0's the figure would probably drop some more. Things could get pretty interesting in a few years if/when 8-16 core CPU's arrive! :up:

That is very good for 52Ghz:up:

I use 70A[8.4kw] dead pull 24/7 for my house. Thats 32 dual cores, 14 single cores, water heater, well pump, microwave, small appliances and 16 various size fans @.08us per Kw.
08-13-2007, 03:52 PM
[XC] riptide

Quote:

Originally Posted by mike047

That is very good for 52Ghz:up:

I use 70A[8.4kw] dead pull 24/7 for my house. Thats 32 dual cores, 14 single cores, water heater, well pump, microwave, small appliances and 16 various size fans @.08us per Kw.

Damn! 70Amps! Did you need special wiring for that load?
08-14-2007, 01:05 PM
mike047

Quote:

Originally Posted by [XC] riptide

Damn! 70Amps! Did you need special wiring for that load?

100A service to the house. I have dedicated circuits for each 4 boxes and monitor sets[ some have more on them]. I balanced the load, it's 34a/36a on each leg, when the well or water heater comes on it stays balanced as they are 220v and pull equally from each leg.

All times are GMT -8. The time now is 04:54 PM.

XtremeSystems