Undervolt and underclock quadcore (Q6600), how many W?

Boris Bilinsky · Post by **Boris Bilinsky** » Thu Jun 18, 2009 5:36 pm

I'm thinking of using a Q6600 in a server, but would like to know if there is much energy/heat to save if I do underclocking and undervolting. The TDP should be proportional to the voltage, right? How about the clock frequency?
Maybe the easiest way is to measure the system before and after underclocking and undervolting, but it would be nice to hear from people that already done some testing.

LodeHacker · Post by **LodeHacker** » Thu Jun 18, 2009 11:01 pm

You could try using this calculator: http://extreme.outervision.com/tools.jsp

I'm sure it won't be 100% accurate to tell real world results, but it'll be very close. Remember though that it depends solely on your motherboard how it handles voltages. If you are not familiar with stuff like Vdroop then don't bother doing anything extreme.

Boris Bilinsky · Post by **Boris Bilinsky** » Fri Jun 19, 2009 3:05 pm

Nice tool, seems it will drop quite a lot if I run it at the lowest multiplier and the lowest voltage. From 95 W (2.4GHz 1.25V) to 29 W (1.6GHz 0.85V). I will use a bad axe 2 motherboard so I think the voltage should be stable.

Nighthog · Post by **Nighthog** » Thu Jul 02, 2009 4:51 am

I undervolt/underclock my Phenom X4 9750 2.4Ghz using K10STAT to 1.0Ghz @ 0.816v and according to that calculator my cpu would then only have a wattage off 24W compared to orginal 125W. 1/5th the power

.

erkan · Post by **erkan** » Fri Jul 03, 2009 12:42 am

I undervolted my Q6600 and the saved watt was not so impressive IMHO.

Think I was down to 84 watt iddle from 95-105.

Boris Bilinsky · Post by **Boris Bilinsky** » Fri Jul 03, 2009 8:49 am

erkan wrote:I undervolted my Q6600 and the saved watt was not so impressive IMHO.

Think I was down to 84 watt iddle from 95-105.

Hmm, ok, how much did you decrease the voltage? It seems that I cannot decrease the multiplier in the bios of the bad axe2, so I'll have to do decrease the FSB instead. The minimum is 200 MHz (1.8 GHz @ 9 mult.) which only will yield a 25% underclock. I'll try anyway and post my results, have to get a power meter.

erkan · Post by **erkan** » Fri Jul 03, 2009 11:41 am

I got an 80plus PSU today and iddle in Vista without Aero it stays at 85 watt, occasionaly goes up to 105 w for a second.

Almost the same as with the old Zalman PSU from 2003, I am actually a bit disapointed in this purchase, I thought an 80+ PSU would lower watt more.

Actually I went into bios and can not figure out how to undervolt and underclock, but I still think I am running undervolt.

I have a Asus M-ATX motherboard, 3.5" HD, 4670 GFX board and 4 sticks of 1GB mem, I would love if the machine would iddle at 65 watt.

Boris Bilinsky · Post by **Boris Bilinsky** » Tue Jul 07, 2009 3:34 pm

I tested some different settings in the bios today and i could not underclock a tiny MHz! However, it was no problem overclocking. I don't understand this at all, maybe it's something with my ram or bios version. If it behaves like this my only option is to software underclock by lowering the multiplier.

flapane · Post by **flapane** » Wed Jul 08, 2009 6:55 am

I read here 84, 95, 105w...
Don't trust PSU tools (according to it, my e5200 draws 84w at 3.3ghz lol) nor TDP (which ain't of course the real consumption value).
I don't know q6600, but e5200 indicates how much amperes is drawing, so a program like Everest can tell you an accurate instant power consumption value (@3.3ghz, about 6w in idle, max 27w with intel burn test).
See if it works with an older q6600.

I suppose that a q6600 with Speedstep enabled, could draw something between 15w in idle and 65w in full load, did anybody make some mesaurements?

yuu · Post by **yuu** » Wed Jul 08, 2009 9:37 am

flapane wrote:.....e5200 indicates how much amperes is drawing, so a program like Everest can tell you an accurate instant power consumption value ......?

the motherboard has vrm regulator that indicates amperes with the vrm efficiency included, so the actual cpu power usage is some 10-20-30% less and nobody knows if it is accurate, unless one has amp clamps around the yellow EPS-12v cpu power supply cables and he knows how to go around the efficiency and read real numbers of the actual cpu usage.

my 8400 is 8 Watts actually 1800Mhz 800mv, and probably half of that is lost by 50% vrm efficiency.

http://img39.imageshack.us/img39/8471/1937800.png

and 3600Mhz 1.15V is 42 watts

so the 6600q should be 4 times the 8watts at the same 1800Mhz 800mv ~32watts

flapane · Post by **flapane** » Wed Jul 08, 2009 9:44 am

Why not 16? (2x8w)

Speaking of full load, 42w could let us wonder that a quad (2 dual core packages) could be about 80w at 3.4/3.5ghz due to higher vcore in order to have a stable quad core.

yuu · Post by **yuu** » Wed Jul 08, 2009 11:57 pm

http://xbitlabs.com/articles/cpu/displa ... html#sect0

6550 ~ 40W => 6600 ~ 80W

max VID 1.325
min VID 1.1

1.1^2 / 1.325^2 ~ 30% less heat ~ 55 watts

Boris Bilinsky · Post by **Boris Bilinsky** » Wed Jul 15, 2009 12:36 am

Now I have measured the system. Using a x1550 graphic card and two 3.5" disks:

~100W idle @ 2.4GHz, 1.25V

~158W full load @ 2.4GHz, 1.25V

~145W full load @ 2.4GHz, 1.162V (lowest voltage the cpu allowed to set

)

~100W idle @ 1.6GHz, 1.162V (underclocking with multiplier)

~134W full load @ 1.6GHz, 1.162V

So there is significant power to save, however I think it's physically possible to run the q6600 at lower voltages, especially as it runs fine at 2.4 GHz using 1.162 V and it is spec:d from 0.85-1.5 V, but the cpu (or bios?) won't let me decrease it more than to 1.162 V.

eatbuckshot · Post by **eatbuckshot** » Sun Jul 19, 2009 1:08 pm

Hi, I was searching for posts on calculating power saved from undervolting and I found this thread. Since these posts are fairly recent and I see your having trouble with limited vcore selection I suggest you could try pin modding to undervolt(i'm currently in the process of pinmodding my t8300 on my t61p to bypass the .950v restriction down to maybe .7v).
(SORRY ABOUT THE LINKS, it's not letting me post urls or links)
here is the data sheet for the pinouts download.intel. com/design/mobile/datashts/31891401.pdf
page 46, Socket P (santa rosa platform/ penryn i believe) has a different pin layout from Socket M, even though both have core 2 duos

a guide to pinmodding to undervolt the core 2 duo
nonsensedesign.co.uk/pinmod.pdf

and that came from this
helpful thread on pinmodding the core 2 duo
forum.notebookreview. com/showthread.php?t=365230&page=2

but might anyone have an idea how much power i'd save going from
.950v 600mhz to .7v 600mhz?

smilingcrow · Post by **smilingcrow** » Tue Jul 28, 2009 6:32 pm

I measured power consumption of a Q6600 (G0) at idle and at full load at stock speeds and being able to undervolt in the BIOS made a big difference.
Wattage at idle/Orthos:

Stock 66.5/141
Software 65.5/122.5
BIOS 58.5/107

I used RMClock to undervolt in software and you can also use it to limit the maximum multiplier used. I didnâ€™t measure the power consumption under load at 1.6GHz unfortunately.

Make sure you get the G0 stepping as that has much lower power consumption.