Quad Q8400S undervolt 0.925V 24W one fan need to be removed

Ecological issues around computing. This is an experimental forum.

Moderators: Ralf Hutter, Lawrence Lee

Ksanderash
Posts: 353
Joined: Sat Oct 18, 2008 6:30 am
Location: Moldova, exUSSR

Post by Ksanderash » Mon Sep 21, 2009 11:44 pm

electrodacus
Damn, I'm so busy these days, I even don't surf the net ) Sorry, I'll rework all calculations some time later, including at-the-wall one. My compie is lying disassembled on the floor up to now.

electrodacus
-- Vendor --
Posts: 372
Joined: Sun Jul 12, 2009 10:30 am
Location: Canada
Contact:

Post by electrodacus » Tue Sep 22, 2009 9:43 am

I hope I will find a solution to eliminate the only fan in my system (it is not noisy since is only at 5V under 800rpm) but it get a lot of dust in my system I will probably be able to use a filter but I will prefer to remove this completely even if I need to change the motherboard.
If you know a motherboard with LGA775 and integrated video that will be able to run completely passive please let me know.

electrodacus
-- Vendor --
Posts: 372
Joined: Sun Jul 12, 2009 10:30 am
Location: Canada
Contact:

Post by electrodacus » Sun Oct 18, 2009 4:33 pm

I found this board AOpen nMCP7ASt-V and opened a new topic about this under CPU and Motherboards Link but no feedback for now is hard to find any info or review about this board .
I need to know if I can undervolt the CPU and if is really the mobile version of 9300/9400 chipset this will make a huge difference since is half the TDP of the desktop version.

ntavlas
Posts: 811
Joined: Mon Jul 16, 2007 2:35 pm
Location: Greece
Contact:

Post by ntavlas » Mon Oct 19, 2009 4:39 am

Have you though about using a thermalright hr05 on the northbidge? The sli version should be able to clear the cpu heatsink even if it takes a bit of bending.

electrodacus
-- Vendor --
Posts: 372
Joined: Sun Jul 12, 2009 10:30 am
Location: Canada
Contact:

Post by electrodacus » Mon Oct 19, 2009 9:29 pm

ntavlas wrote:Have you though about using a thermalright hr05 on the northbidge? The sli version should be able to clear the cpu heatsink even if it takes a bit of bending.
Is really tight there but I guess it will still not be able to cool 15.5W the XP140 is much larger with 6 heat pipes and at 20W I will get 65C.
Even if North bridge can handle more than 65C I will like to have al components in my PC under 60C.
I'm trying to find a better motherboard for LGA-775 hope to find one with low TDP chipset and CPU voltage control.

ntavlas
Posts: 811
Joined: Mon Jul 16, 2007 2:35 pm
Location: Greece
Contact:

Post by ntavlas » Tue Oct 20, 2009 12:38 am

It might be larger but it`s not really designed for passive operation when the hr5 is. It`s also not certain that the igp dissipates 15 watts all the time, it`s the max rating but the typical use could/should be lower. Besides, even if you find a better motherboard, a heatpipe chipset cooler will still be very useful.

Ksanderash
Posts: 353
Joined: Sat Oct 18, 2008 6:30 am
Location: Moldova, exUSSR

Post by Ksanderash » Tue Oct 20, 2009 12:59 am

electrodacus
Now I'm doing some experiments on undervolting the Biostar G31D-M7 (Flex-ATX formfactor, that is between mini-ITX and micro-ATX in size). I roasted to death two Biostar G31 boards already (due to inaccuracy though) to figure out how this can be done at ease. There are five regulated voltage sources on the board that have a direct impact on consumption: 1.07V + 1.5V ICH, 1.2875V CPU, 1.3V Vtt, 1.8V RAM, 1.3V G31. The only thing that I couldn't get control over it is RT8841, power PWM ocntroller of CPU's VRM. It has a lot of protection, and official means described in the datasheet to introduce a negative Vout' offset are very limited - I could drop down 1.2875V only to 1.23V or something (a tried 3 methods). Got to play with VIDs, and forget about lazy chip's feedback tweaking.

...

I'm now on about 20W DC in idle (E6300+G31+2x2GB), excluding HDD/ODD/USBFLASH. I'm awaiting ordered K-type thermocouple for my brand-new Agilent multimeter, so I'll be able to measure the temperature not only by finger :) And I'll tell you then whether it is possible to run G31 w/o active cooling, or not. It depends on air convection around the board, btw. Fastening the board verticaly helps a lot for cooling. Desktoped (horizontaly lying, I mean) mainboard will get overheat at the same conditions.

...

P.S. As for 5V to 4.5V reduce in order to get lower consumption -- I did a recheck, and I beg you pardon for given misinformation. No reduce. My old multimeter readings were wrong, and I'll investigate why.

electrodacus
-- Vendor --
Posts: 372
Joined: Sun Jul 12, 2009 10:30 am
Location: Canada
Contact:

Post by electrodacus » Wed Oct 21, 2009 9:32 am

ntavlas wrote:It might be larger but it`s not really designed for passive operation when the hr5 is. It`s also not certain that the igp dissipates 15 watts all the time, it`s the max rating but the typical use could/should be lower. Besides, even if you find a better motherboard, a heatpipe chipset cooler will still be very useful.
In my opinion there is not much difference (if any) between a normal heatsink and a heat pipe one in passive cooling (passive cooling not using a case fan so no air circulation in the case) natural convection only.
Currently I use a Intel heatsink glued on top of chipset heatsink the Intel heatsink is about 120g aluminum on top of the 40g original northbridge heatsink but it still getting to hot.
I still tiring to find a better motherboard with a mobile chipset for LGA 775.

electrodacus
-- Vendor --
Posts: 372
Joined: Sun Jul 12, 2009 10:30 am
Location: Canada
Contact:

Post by electrodacus » Wed Oct 21, 2009 9:40 am

Ksanderash wrote: I'm now on about 20W DC in idle (E6300+G31+2x2GB), excluding HDD/ODD/USBFLASH. I'm awaiting ordered K-type thermocouple for my brand-new Agilent multimeter, so I'll be able to measure the temperature not only by finger :) And I'll tell you then whether it is possible to run G31 w/o active cooling, or not. It depends on air convection around the board, btw. Fastening the board verticaly helps a lot for cooling. Desktoped (horizontaly lying, I mean) mainboard will get overheat at the same conditions.
It probably depend on the manufacturer of the board but 20W seems low to me I have the same G31 + undevolted and underclocked Q8400s (only about 4W at idle) + 2x2GB RAM and an SSD that is only 0.4W max so is not important and I get a DC of about 25W to 27W. How do you measure the DC ? and where ?.

Ksanderash
Posts: 353
Joined: Sat Oct 18, 2008 6:30 am
Location: Moldova, exUSSR

Post by Ksanderash » Wed Oct 21, 2009 2:57 pm

electrodacus
I connect the ampermeter in the break of +5VDC / +12VDC / +3.3VDC supply chain. Similar wires are grouped together. The pic (sorry for the quality, it's celluar-phone made)
Image

And my preferable multimeter, with data aweraging function, TrueRMS (so my AC readings are very accurate, but ...so high, that I'm ashamed to bring them up, about 65W in idle :shock: )
Image

And the current data. First table is default board+CPU+RAM DC wattage, no voltmod at all. The second is a completely shoveled over board. And you can see that not so much can be economized. Additional minus 2W from +5V if you go 166 instead of 266FSB. Khe-khe, Biostar has very convenient DC-DC converters -- all on +5V :) And there is a thought in my head -- what if I give them supply from +12V, chips do support this. But doing so I'm sticky with this board forever, until S775 disappears, cause too much attention to this piece of silicon and tin :)
Image

electrodacus
-- Vendor --
Posts: 372
Joined: Sun Jul 12, 2009 10:30 am
Location: Canada
Contact:

Post by electrodacus » Wed Oct 21, 2009 10:40 pm

Ksanderash
Thanks for your detailed description the DEF number seems about right I will expect to see a difference in 2D and 3D load on 5V line.
I get similar numbers so they are probably right.
I see in the first photo probably the current measurement for the 5V line you need to have 4 wires that go to ATX connector, I only see 3 but probably because of the picture.
And it seems that your board is not using the 12V line only the CPU since you only have 4W at idle (I get about 5W for the CPU and 3.4W for the board at idle on the 12V line).
I do not understand how you reduced the power by 7W did you also checked at the AC you will also see at least 7W there. I have an option to reduce the Northbridge voltage but I see no effects in power consumption so I guess is not working on my board.

Ksanderash
Posts: 353
Joined: Sat Oct 18, 2008 6:30 am
Location: Moldova, exUSSR

Post by Ksanderash » Thu Oct 22, 2009 11:38 am

electrodacus wrote: And it seems that your board is not using the 12V line only the CPU since you only have 4W at idle (I get about 5W for the CPU and 3.4W for the board at idle on the 12V line).
I think the most of these 4W uses CPU VRM, cause both 2 x FP6321 PWM controllers of NB/SB/DRAM are sitting on +5V, and they are the major power drainers. Btw, schemes based on FP6321(and many many analogues) have a very simple design. I even think about a homebrew picoPSU making. Can try to make(w/o oscilloscope this will be tough) +5V and +3.3V easily, up to 20A, and up to 93% efficiency. Not necessary FP6321, it can be APW7120, or RT9214, or other clones. Don't sure about -12V source... It needs some additional IC, like MC34063. And mentioned PWMs can be found on every half dead motherboard.
I do not understand how you reduced the power by 7W
Voltmoded every power related IC by pencil (and then solder in right resistor).
did you also checked at the AC you will also see at least 7W there.
Not yet, but sure there will be a positive change :)
I have an option to reduce the Northbridge voltage but I see no effects in power consumption so I guess is not working on my board.
I can tell you how the voltages are controlled by BIOS on my board (via multi I/O controller IT8712F), but more easy way is just checking voltage on NB (on a ferrite coil, or mosfet, or el-cap, or ceramics, whatever), if there is a change.

electrodacus
-- Vendor --
Posts: 372
Joined: Sun Jul 12, 2009 10:30 am
Location: Canada
Contact:

Post by electrodacus » Thu Oct 22, 2009 7:50 pm

To much work and money to make your one PSU The one I sell uses the APW7120 for the 5V and 3.3V and IRFR3711Z .
And you can get the green board for 19.95 + 9.99 shipping if you only need the board.
I will need to check if the bios settings affect the voltage on northbridge but is probably not.

Ksanderash
Posts: 353
Joined: Sat Oct 18, 2008 6:30 am
Location: Moldova, exUSSR

Post by Ksanderash » Fri Oct 23, 2009 8:24 am

electrodacus wrote:To much work and money to make your one PSU The one I sell uses the APW7120 for the 5V and 3.3V and IRFR3711Z .
And you can get the green board for 19.95 + 9.99 shipping if you only need the board.
Huh, the problem is that shipping tax from Canada to Eastern Europe will be about 40-50$, afaik. Yeah, I'd like to, but that's quite rejecting factor.

electrodacus
-- Vendor --
Posts: 372
Joined: Sun Jul 12, 2009 10:30 am
Location: Canada
Contact:

Post by electrodacus » Fri Oct 23, 2009 9:24 am

Ksanderash wrote:
electrodacus wrote:To much work and money to make your one PSU The one I sell uses the APW7120 for the 5V and 3.3V and IRFR3711Z .
And you can get the green board for 19.95 + 9.99 shipping if you only need the board.
Huh, the problem is that shipping tax from Canada to Eastern Europe will be about 40-50$, afaik. Yeah, I'd like to, but that's quite rejecting factor.
if you only need the PSU board the weight is under 250 gram and shipping to Moldova by air is 9.99$ is true that if the weight is over one kg then the shipping is 51.99$ this is the reason I do not recommend buying the brick (AC adapter) that is heavy and the shipping is more expensive than the product.
Shipping from Canada to Europe including most of East Europe

up to 250g 9.99$
up to 500g 18.95$
up to 1000g 34.95$
up to 2000g 51.99$

Ksanderash
Posts: 353
Joined: Sat Oct 18, 2008 6:30 am
Location: Moldova, exUSSR

Post by Ksanderash » Mon Oct 26, 2009 10:14 am

electrodacus
Thanks, got it. Now searching for a man who will buy it for me, as I haven't the first notion about e-payment :)

...

I got a good haul recently -- EA10953 by EDACPOWER Electronics Co. It is 90W 20V adapter, pretty regular brick. I measured AC 5.8W consumption when no load, and 68.34% efficiency at 31.3W DC load (AC 45.8W). Haven't tried on a real system -- just attached a low Ohm power resistor as a load to calculate efficiency. The datasheet claims 85% typical, and I see only 79.2% efficiency at 61.8W DC load (AC 77.9W)

Klusu
Posts: 196
Joined: Thu Jul 24, 2008 10:57 am
Location: Riga

Post by Klusu » Mon Oct 26, 2009 12:03 pm

Ksanderash wrote: (so my AC readings are very accurate, but ...so high, that I'm ashamed to bring them up, about 65W in idle :shock: )
Wrong. You can not measure W. Your real W must be perhaps 42W. If you are lucky, you can get 30W from the wall with a cheap ATX PSU (which may make you re-evaluate the need for a pico-like PSU).

Ksanderash
Posts: 353
Joined: Sat Oct 18, 2008 6:30 am
Location: Moldova, exUSSR

Post by Ksanderash » Mon Oct 26, 2009 1:27 pm

Klusu wrote:Wrong. You can not measure W. Your real W must be perhaps 42W.
Hi, Klusu, nice to meet you again. Could you please explain, why? My multimeter calculates effective AC power irregardless of the wave form. Plus both PSU (ATX and brick) have a APFC scheme.

Klusu
Posts: 196
Joined: Thu Jul 24, 2008 10:57 am
Location: Riga

Post by Klusu » Mon Oct 26, 2009 11:04 pm

Does it measure power? At all?
TrueRMS is one thing, power factor is another. Are you sure your PSU has active PFC (have you opened it)?

Ksanderash
Posts: 353
Joined: Sat Oct 18, 2008 6:30 am
Location: Moldova, exUSSR

Post by Ksanderash » Tue Oct 27, 2009 5:26 am

Klusu wrote:Does it measure power? At all?
Of course not! Power quality analyzer costs a lot of bucks :) But it measures AC current, and can handle waveforms up to 3 crest factor, so I can calculate the power by myself, simply multiply V on A.

Here is an explanatory document, Can you trust your meter reading?

Image
Are you sure your PSU has active PFC (have you opened it)?
O-ops :oops: Sorry for missinforming, I confuse Inwin IP-S350Q2-0 with my old Delta GPS-400AA-100A. No, my current ATX PSU hasn't APFC. And EDACPOWER brick has. I disassembled it to see what is inside. APFC is built on a STm chip.

Klusu
Posts: 196
Joined: Thu Jul 24, 2008 10:57 am
Location: Riga

Post by Klusu » Tue Oct 27, 2009 7:13 am

Multiplying you get VA, not W.
The document is about RMS. You need one on PF.

electrodacus
-- Vendor --
Posts: 372
Joined: Sun Jul 12, 2009 10:30 am
Location: Canada
Contact:

Post by electrodacus » Tue Oct 27, 2009 1:28 pm

Klusu wrote:Multiplying you get VA, not W.
The document is about RMS. You need one on PF.
You are right he is getting the apparent power (VA) so he will pay that if PF is 1 or less is PF less than 1. Probably less :).

I will try to explain as simple as possible with my limited English :)

You have 3 type of power real power P measured in Watt , apparent power S measured in (VA) and reactive power Q measured in (var)
The PF power factor is a number between 0 and 1 .
PF is 1 for a pure resistive load and 0 for a pure inductive or capacitive load.
P = S x PF.
Your home power meter is measuring P and is what you pay :) only the big companies pay S and they have special meters to measure this.
For example if you have a PF of 0.5 then P = S x 0.5

Let's get real numbers and say S= 100 VA and PF is 0.5 then P = 50W.
You will pay for 50W ... but the power lines between you and power plant will see 100VA so 50 or 60 times a second you take more than 50W but the you give back some power.
Is like having a capacitor that is charging at your home with energy provided by the power plant generator but then the capacitor is discharging back so you send energy back to the power plant.
The problem is that you lose on the lines more power than if you only take 50W with a resistive load so you do not take more and then send back power to the power plant generator.
Conclusion is that if you have electrical equipment with PF less than one then your loss is in the wiring inside the house but is small compared to the power plant company losses on the power grid so they pay much more than you for the loss.
In the end they will still charge you with the higher energy costs :)

Ksanderash
Posts: 353
Joined: Sat Oct 18, 2008 6:30 am
Location: Moldova, exUSSR

Post by Ksanderash » Thu Oct 29, 2009 4:12 am

Klusu
electrodacus
Well, exactly so, I do measure the apparent power, that is multiplication of RMS values A and V.

But how can we segregate the active power from the reactive in the apparent one? That involves tricky calculations, I suppose. I wonder, what type of power shows Kill-a-Watt? Active, apparent? There is a switch PF on the control panel -- so we can calculate the reactive component by ourselves?

Klusu
Posts: 196
Joined: Thu Jul 24, 2008 10:57 am
Location: Riga

Post by Klusu » Thu Oct 29, 2009 7:45 am

You can not "segregate". You can only use a proper meter. Some meters show PF also (better or worse), then you can "calculate".

electrodacus
-- Vendor --
Posts: 372
Joined: Sun Jul 12, 2009 10:30 am
Location: Canada
Contact:

Post by electrodacus » Thu Oct 29, 2009 3:54 pm

Ksanderash wrote:Klusu
electrodacus
I wonder, what type of power shows Kill-a-Watt? Active, apparent? There is a switch PF on the control panel -- so we can calculate the reactive component by ourselves?
The Kill-a-Watt will measure real power and will also display PF so you can calculate apparent power or maybe some models calculate that for you.

Ksanderash
Posts: 353
Joined: Sat Oct 18, 2008 6:30 am
Location: Moldova, exUSSR

Post by Ksanderash » Thu Nov 26, 2009 11:41 am

electrodacus

At last my tests with DC-DC board were performed. Don't know about disabling CPU fan cause it blows on NB heatsink too.

MtnHermit
Posts: 99
Joined: Fri Jan 01, 2010 9:25 am
Location: Colorado

Post by MtnHermit » Fri Jan 01, 2010 9:50 am

Reply 1

MtnHermit
Posts: 99
Joined: Fri Jan 01, 2010 9:25 am
Location: Colorado

Post by MtnHermit » Fri Jan 01, 2010 9:50 am

Reply 2

MtnHermit
Posts: 99
Joined: Fri Jan 01, 2010 9:25 am
Location: Colorado

Post by MtnHermit » Fri Jan 01, 2010 9:51 am

Reply 3

MtnHermit
Posts: 99
Joined: Fri Jan 01, 2010 9:25 am
Location: Colorado

Post by MtnHermit » Fri Jan 01, 2010 9:54 am

electrodacus wrote:
Image
The thing that strikes me about these results are the diminishing returns and/or the Sweet Spot. To me the 2GHz at 0.925V is the best compromise between power and performance. From my perspective the load power is less meaningful because we seldom run PC's at 100%, therefore how can I get the idle power down is of most interest.

You may find this thread interesting: Quad-core system under 25W idle.

In any case, I'm very interested in these low power topics.

Post Reply