UNDERVOLTING FUN! CORE 2 DUO/QUAD

[tenax]

i don't have a quad, yet..but i plan to get a 9450 when they come out...given my htpc case doesn't allow for the biggest and best coolers out there, i have some compromise and i've always felt that my system runs warmer than i like it. i was using a thermaltake typhoon but it had little clearance for airflow to the lid. i am now running an enzotech extreme which has about an inch more clearance, but i like to run the fan between 800 and 900 rpm for an ultra silent htpc so there is some temp sacrifice. taller heatsinks like the thermalrite 120 and such just will not come close to fitting...today i found a thread about undervolting using the 6300 and 6400's as the example (i have a 6400) so i thought i'd put it to the test. in the thread i found it was claiming about the lowest voltage you can go stable on the core 2 duo is about 1.15 or so...and a few mentions that 1.00 should be obtainable..i've never undervolted below 1.3 on my board thinking it would not run at all..who knew? so i did some testing of my own as follows:

6400 overclocked- 8 x 400 (fsb and memory)-1.375 volts needed- cores 50, 51 at idle. 60 to 61C at full load, case temp-39C

6400 overclocked- 8 x 375 (fsb and memory) 1.350 volts needed- cores 49, 50 idle, 58 to 59 at full load. case temp- 37C

6400 at stock- 8 x 266 fsb, 400 memory 1.30 volts- cores 41, 42 idle, 45, 46 full load. case temp- 33C

6400 at stock- 8 x 266 fsb, 400 memory- 1.000 volts (read by 5 monitor programs at .99 to .992 volts!- cores 32, 33 idle. 39, 39 at full load! case temp-26C!

for reference, this was with my gigabyte p35-ds4, 4 sticks of 1 gig ram, overvolted to 2.1 volts as 2 of the sticks are 5300 ram, the other 2 6400 and i wanted to be able to hold my ram speed at 400. mch setting at + .1 volts, fsb at + .1 volts

i should note that with my first try at 1.000 volts, my core temps went as low as 28, 29 idle. i then went drastic and tried .75 volts and it would not boot..matter of fact pulling my power plug didn't work either. i had to reset cmos and lost my saved settings so i might have configured something a bit differently when trying 1.00 the 2nd time. i think it might be the fsb was at stock, not .1 volts over as i know i did have the mch setting with +.1 volts.

regardless, i'm very happy with the undervolt experiment in preparation for the quad to come. while i like to play with overclocking on my system for kicks, it is an htpc that runs 24/7 and everyday computer so the main objective is cool, quiet and stable, yet fast enough for whatever i need it to do. why didn't i try undervolting sooner if cool was one of my main goals? i chalk it up to being used to my last amd processors that required 1.4ish to 1.5 volts as the norm...and little information and discussion available about undervolting on most sites. as of course is common, most people are talking about how much can they overvolt, not undervolt. but, it's an interesting thing to try, especially if compromises need to be made on your cooling options!

ps- the system is running completely stable at idle or load at 1.00 volts! i may try .900 for kicks at some point, but i've had enough fun today!

 

[BonzaiDuck]

Well, I appreciate your approach for slightly different reasons.

I burned out a Crucial Ball'x memory module just by volting the memory at its warranty recommended maximum and tightening down the latencies (which of course, was a reason for requiring that voltage.) I had limited my over-clocking efforts to a 1:1 CPU-to-RAM ratio, so that one finds himself at a limit where the Northbridge core is being stressed and the memory is being pushed to the edge, even when the DDR speed is below spec.

And I had tried using lower CPU multipliers, but wasn't satisfied with that, either. I'm using the Q6600 B3-stepping.

Now I've found a moderate 3 Ghz OC configuration with a 1:1 ratio and decently high memory-cache bandwidth with voltages well below maximum spec for either the CPU or the RAM. I have another configuration at 2.88 Ghz, 4:5 [CPU: RAM], with the memory right at its DDR800 spec, much tighter latencies but lower voltage than with the former setting. And I've scrapped my 3.2 Ghz configuration to replace it with a 3.15 Ghz setting, 4:5 with DDR at 875 Mhz, and a voltage between the first and second settings just mentioned.

Oddly, the bandwidth benchies for the 3.15 Ghz configuration are better than for 3.2 with 1:1.

And I can be happy with 2.88 Ghz for conventional computer work, and then switch to the high setting for gaming. The load temperatures in all cases are so low as to be not worth monitoring anymore. And, for all three settings, the memory voltages have been bumped up just enough to allow me to switch between 1T and 2T command rate without changing anything else.

 

[tenax]

cool! (no pun intended) yeah, it's a good point..nice to have motherboards that you can save various configurations on now..i still like to "play" at overclocking..and will when i get my quad as well, but...for everyday use, there's just no glamour left for me in overclocking..as i've said to many a buddy many a time, once we got past about 1 gig on processor speed, beyond benchies..and of course, now dual and quad cores that have increased our ability to multi task and still have smooth performance, the only really difference we feel is in benchies..i felt the dual cores were the true leap in evolution as an htpc user. for the first time, i was able to watch..and record a show, and burn a dvd and surf and type a document..and not have jerky playback of a recorded show, or have a show record in fits and stutters due to me doing other cpu intensive tasks at the same time.

 

[CallMeJoe]

Search for "undervolt" at the Silent PC Review Forum and you'll get about 80 pages of results. I've been undervolting since I read about the technique for cooling my AMD Athlon Thoroughbred.

edit: I'm not undervolting my current main rig because of the 50% overclock I have on my E6300. As you've discovered, overclocking and undervolting don't mix. Fortunately, with the low heat load of the Core 2 Duo processors and the efficiency of the newer LGA775 coolers, you can have a decently fast running PC with low noise and low temperatures.

 

[tenax]

well said, joe..and that's a great link i'm going to check out!

 

[tenax]

here's the specific link by the way:

http://www.silentpcreview.com/.../viewtopic.php?t=41883

 

[LOUISSSSS]

i wish my ds3P would save my bios configs =(

got reseted once already

 

[tenax]

you mean when you boot, it always restarts, i assume..what have you done with your settings in bios?

 

[Phunk0ne]

Here is my test results, took me two days to test each setting starting from stock volt in BIOS.
I decided to see what my lowest possible VCore was in BIOS with my 3.2GHz overclock (1:1 Ratio), since this speed requires no voltage modification in BIOS.
All other voltages are left @ stock. Flashed BIOS from F6 to latest F10.

I let ORTHOS complete a full cycle before lowering the VCORE.

Lowest stable vcore I could get @ 3.2GHz with my E6750 was 1.2500 in BIOS
Everest shows my cpu @ IDLE @ 1.200V-1.216V / @ LOAD 1.184V

My temp readings C0/C1 in Celsius Degrees
IDLE ~23C/24C
100% Load: 37C/~38C

 

[harpoon84]

Lowest setting on my P5B Deluxe is 1.1V in BIOS (1.07V actual in Windows). My E4400 (L2) does 2.49GHz stable at this voltage.

 

[Thor86]

Originally posted by: harpoon84
Lowest setting on my P5B Deluxe is 1.1V in BIOS (1.07V actual in Windows). My E4400 (L2) does 2.49GHz stable at this voltage.

Pretty much a laptop cpu. Good show. :thumbsup:

 

[tenax]

yep..nice results, gents!

 

[tofumonster]

Err...

i can do 2.66 on my E6400 @ 1.25

pretty good, i think. = D

 

[eelw]

When I replaced my Q6600 B3 to a G0, I decided to undervolt the sucker. It was stable all the way down to 0.96v. Load temps was in the high 30s. With the fan on my ThermalTake V1 disconnected, full loads were in the low 50s. So with a large enough fan, you can indeed run a quad passive.

 

[SpeedEng66]

e6550 (g0 step) @ 2.8ghz (400fsb)
underv to 1.13 - 1.14
foxconn p35a

 

[tenax]

interesting stuff! i did some wiring juggling last night..dropped 2 more C on the cores and 1 on the case..it's still sloppier than i'd like..when i get more core i'm going to do a serious wiring cleanup job..i have more stuff in the way of one intake fan for sure than i like.

 

[lightfoot3b]

E2160 (1.8GHz 1MB cache)
Stock 1.8Ghz @ 1.3X Volts and somewhere around 55-60* Celcius full load (prime 95 x 2).
U-Volt 2.1GHz @ 1.09V idle and 1.07V load, 44*C full load.

Speedstep and EIST on, but I can answer some of your questions.

Gigabyte boards do not seem to change voltages (when speed-stepping) once you manually set voltages. However my Abit does, but my Abit does not allow undervolting! The Gigabyte (IP35-e) will still speed-step but not cut more voltage for idle use. But that is ok, since my undervolt is lower then the speedstep voltage anyway! Plus, simply lowering the clock cycles reduces power use ever so slightly too. I'm going to bust out my kill-a-watt and see what the same clock speeds do with different voltages + idle with a set voltage and idle with speedstepped voltage (on stock settings of course).

I love messing with this stuff... Just need to get my dead 4850 replaced under warranty now.

 

[aigomorla]

http://i125.photobucket.com/al.../p73/aigomorla/OMG.jpg

Lowest i could pull 4ghz windows stable

i have a .97V stock on my Q9650ES. However i didnt take a screenie. I should tho.

 

[VirtualLarry]

Originally posted by: BonzaiDuck
And, for all three settings, the memory voltages have been bumped up just enough to allow me to switch between 1T and 2T command rate without changing anything else.

1T? Really? Have you tested that with CPU-Z to check if you are really running at 1T? And have you Prime95 Blend tested it for 24hr at 1T?

I didn't think any modern chipset would allow running at 1T, nevermind stably.

 

[aigomorla]

Originally posted by: VirtualLarry

Originally posted by: BonzaiDuck
And, for all three settings, the memory voltages have been bumped up just enough to allow me to switch between 1T and 2T command rate without changing anything else.

1T? Really? Have you tested that with CPU-Z to check if you are really running at 1T? And have you Prime95 Blend tested it for 24hr at 1T?

I didn't think any modern chipset would allow running at 1T, nevermind stably.

+1

 

[Philippart]

q9450@stock clock@1.056v rock stable on an Asus Maximus Extreme

 

[suklee]

I just got a Q6600 that I am running at 333x9 at default voltage. CPU-z's voltage fluctuates between 1.136V and 1.248V.

To undervolt on my Gigabyte DS3L mobo, is it simply a matter of lowering the vcore? If it's unstable, I can just return the voltage to normal with no damage done right? Are there any other potential issues I need to watch out for?

 

[BlueAcolyte]

Yea, just lower the VCore. I think it's impossible to actually damage your chip by not feeding it enough power, so don't worry. You can just turn it back to stock if you have to.

 

[suklee]

Thanks. I set my Q6600 @3GHz to 1.15000v in my mobo's BIOS and looks stable so far.

 

[suklee]

Strange, I set my voltage to 1.00 on the BIOS, but HW Monitor and CPU-Z is reporting the lowest as 1.14v. Am I missing something?