Undervolting: how low can you go?

fwki · Post by **fwki** » Thu Apr 10, 2008 4:35 am

I was surprised myself, especially since the Q9450 defaulted to 1.2375v or thereabouts out of the box. Here's my setup experience: viewtopic.php?t=47335&highlight=

verix · Post by **verix** » Sun Apr 13, 2008 3:01 am

I have some slightly better results for the Athlon 64 X2 4400+ Brisbane (G2) than Mikael: 2.3 GHz @ 0.960v!

Ran Prime95 Small FFT's for 16 hours, now doing Blend. If that goes successfully, OCCT and then done. I may try lower than 0.95 in BIOS but I'm already pretty happy to run stock at less than 1v.

If anything goes wrong, I'll update this post.

Palindroman · Post by **Palindroman** » Sun Apr 13, 2008 9:09 am

I have entered all the entries of the past 2-3 months and added this one myself:

AMD X2 4450e (AM2) stock - 1.25V

Minimum: 1.0 GHz @ 0.850 V
Stock: 2.3 GHz @ 1.100 V

Results of over 60 CPUs have been posted this far. Thanks, everyone!

k23o · Post by **k23o** » Sun Apr 13, 2008 10:09 am

AMD Athlon X2 4200+ Socket 939

Minimum: [email protected]

ST · Post by ST » Sun Apr 13, 2008 10:15 am

ST wrote:s939 Opty 165

P-Min State = 800Mhz @ .850V
P-Mid State = 2.0GHz @ .995V
P-Max State = 2.4GHz @ 1.125V

I can overclock to 2.8GHz @ stock 1.35V and even 3.0GHz Dual Prime 95 Stable, but that requires ~ 1.475V. Since I don't play FPS anymore, the above P-States are my typical setups.

fyi - i had a write up about overlclocking / undervolting about a year back :

viewtopic.php?t=30703&highlight=rmclock

Changed my system to a Q9450 and can undervolt as follows:

1.00v @ 2.7GHz
1.08v @ 3.0GHz

My mobo doesn't give me much options to undervolt further

bhvm · Post by **bhvm** » Tue Apr 22, 2008 7:01 pm

Bah! Mine is much better!

Tool- RMClock

Scanario 1-
Account - web surfing.
Backgroud processes-
Antivirus, Transparency, Dock with big PNG Icons, 3D effects (WinFlip), Yahoo widget engine with 10 widgets (Including a Big Animated Analog Clock)

Foreground tasks- Opera, Yahoo messenger, MSN, Rediff, Skype.

RMClock profile-performance on demand.
All Mulitpliers at 0.95V (the minumum option) 6X~10X
Systmem Rock Solid!
Usually runs at 6X (1ghz)
Core temps- 23.C ~ 33.C

Fan/Noise -Absent.
Graphics temp- ~40.C

Scenario 2-
Plain undervolting

Background tasks- Minimum/Absent
Foreground - Game (NFSMW,DOOM, Call of duty etc)

Multiplier 10X 1.667 Ghz at 0.96 (Just a Notch higher Than Min)
Lower Mulitpliers at 0.95V

Aquamark score - 43K
Temps 50~60.C
Fan at low sppeds
Graphics temp 50~60.C

Before Undervolting-
Multiplier - Switches 10X 1.667Hgz at 1.26V and 1Ghz At 0.95V
Websurfing Temp- 40~58.C
Gaming temp- 80~90.c (Ouch! My pants!)
FAN- Huh! Its better called a Hairdryer!

Config-
1>Custom HP DV 8/9XXX series (Import
from Singapore)

2>17.2 in Widescreen, 1400X900
Dual lamp,Ultra bright (OMG!!!!!
Gorgeous!!!) 5:3 Aspect ratio

3>Intel Core QUAD, 1677Mhz,667
Mhz FSB , 2MB L2 cache, 31W
socket 479

4>Intel 945PM Chipset,Pcie X16
support,Sata Support,Upto 2G RAM
support.

5> 1 GB DDr2 667 mhz RAM.

6>Nvidia GF 7400Go OEM SLI!!!!
- 2X nvidia G72 cores at 400Mhz
each + 512 MB RAM (700Mhz)each
OC'd to 564/864 For gaming Profile

7>Intel HD audio

8> 2X Seagate Momentus 7200Rpm
Sata2 HDDs (80GB X2)

9>HP DVD Writer/RAM with
Lightscribe

10>Inbuilt Bluetooth and Wi Fi, +

Dial-up Modem +100mb/s lan
11>Ports-
4X USB2
1X S-video and VGA
1X Firewire
1X PCM cia and Express card slot
5 iN 1 Card reader

Misc-

1>HP Quick play Buttons
and remote control

2>Copper Heatpipe cooling

3>Fullsize + Numeric keyboard

fwki · Post by **fwki** » Wed Apr 23, 2008 5:37 am

How about testing some of these "stable" UV's at load with a long-term stress test? As overclockers know, many systems appear stable and one could go weeks under normal use before demonstrative errors. Overclocks aren't accepted as legitimate without rigorous testing, nor should undervolts. Anything else is simply speculation.

To quote Palindroman, the OP, "Although the procedure is quite simple, finding out the minimum voltage values for your CPU takes a bit of time testing". Also for those who haven't taken the time, his link to the undervolting tutorial is still active and it does explain how to find a vaild, stable UV.

Nighthog · Post by **Nighthog** » Wed Apr 23, 2008 6:50 am

How low do you guys think I could undervolt a Phenom X4 9750 "Agena" cpu?

Orginal seems to be 1.3volt. Well that was the volt I got when I had installed windows and checked cpu-z. In bios you only get "NORMAL" volt(or +0.125 or whatever increment you wanna overvolt)... then there is a 1.2500 note below, but you won't get that in "normal" whit the gigabyte 780G motherboard,

"Normal" Â´volt gives you 1.3volt under idle and 1.4volt under 100% load.

I've undervolted the cpu in AMD OverDrive to 1.2volt(1.3v @ 100%load) but I don't know how low I should go. don't want to risk it to much as you can't undervolt in bios but have to do it in Windows through AMD OverDrive.

ryboto · Post by **ryboto** » Wed Apr 23, 2008 9:09 am

Nighthog wrote:How low do you guys think I could undervolt a Phenom X4 9750 "Agena" cpu?

Orginal seems to be 1.3volt. Well that was the volt I got when I had installed windows and checked cpu-z. In bios you only get "NORMAL" volt(or +0.125 or whatever increment you wanna overvolt)... then there is a 1.2500 note below, but you won't get that in "normal" whit the gigabyte 780G motherboard,

"Normal" Â´volt gives you 1.3volt under idle and 1.4volt under 100% load.

I've undervolted the cpu in AMD OverDrive to 1.2volt(1.3v @ 100%load) but I don't know how low I should go. don't want to risk it to much as you can't undervolt in bios but have to do it in Windows through AMD OverDrive.

Does CrystalCPUID support the Phenoms? If so, that's a great way to do slight undervolts. Just try it in steps. 1.2v, stress test, 1.175v, stress test, repeat. If it doesn't crash right away, and you want to be sure it's stable, stress test for longer. If it's completely stable, push down another notch.

bhvm · Post by **bhvm** » Wed Apr 23, 2008 6:55 pm

fwki wrote:How about testing some of these "stable" UV's at load with a long-term stress test? As overclockers know, many systems appear stable and one could go weeks under normal use before demonstrative errors. Overclocks aren't accepted as legitimate without rigorous testing, nor should undervolts. Anything else is simply speculation.

To quote Palindroman, the OP, "Although the procedure is quite simple, finding out the minimum voltage values for your CPU takes a bit of time testing". Also for those who haven't taken the time, his link to the undervolting tutorial is still active and it does explain how to find a vaild, stable UV.

Hmmm....

Speaking of stability, How does One year of Rock solid performance sound to you?
I don't believe Much is synthetic tests as they are far from real scenarios.
On the other hand, I Keep encoding Videos and MP3s for my portable player often, and thats the best test I think.

Upto full speed 1.66G at 0.95V, stability confirmed

fwki · Post by **fwki** » Thu Apr 24, 2008 5:25 am

bhvm wrote: Hmmm....
Speaking of stability, How does One year of Rock solid performance sound to you?
I don't believe Much is synthetic tests as they are far from real scenarios.
On the other hand, I Keep encoding Videos and MP3s for my portable player often, and thats the best test I think.

Upto full speed 1.66G at 0.95V, stability confirmed

I'm glad that works for you, I just don't have time for a year of testing between BIOS tweaks.

bhvm wrote: 3>Intel Core QUAD, 1677Mhz,667
Mhz FSB , 2MB L2 cache, 31W
socket 479

That's an interesting cpu. I didn't know Intel made a socket 479 QUAD.

bhvm · Post by **bhvm** » Thu Apr 24, 2008 6:35 am

fwki wrote:
bhvm wrote: Hmmm....
Speaking of stability, How does One year of Rock solid performance sound to you?
I don't believe Much is synthetic tests as they are far from real scenarios.
On the other hand, I Keep encoding Videos and MP3s for my portable player often, and thats the best test I think.

Upto full speed 1.66G at 0.95V, stability confirmed
I'm glad that works for you, I just don't have time for a year of testing between BIOS tweaks.

bhvm wrote: 3>Intel Core QUAD, 1677Mhz,667
Mhz FSB , 2MB L2 cache, 31W
socket 479
That's an interesting cpu. I didn't know Intel made a socket 479 QUAD.

Hello dear,
This is not a BIOS tweak or something. Being A Laptop (DID I mention that this is an lappy??) bios enhancements are simply absent! I've used RMClock to reduce Voltage.

As for tha CPU, Even i am puzzled.
This system is made is singapore, and was gifted to me.
In the specs, it does not mention QUAD, but I mentions TWO Dualcores (Thats 2x2=4)
Other thing that might have caught your attention is that it has got TWO nvidia G72 cores in tandem,with 512MB RAM each!

What do you say?
The system is 2 years old, and QUADs were absolute rarity even in desktops.And laptop SLI was out of question.

maybe this some RARE engineering sample that HP made to test Quadcore implementation,and it somehow slipped out to me?

By the way, temps are amazing!
It TOPS out at 33.c when websurfing(and loads of background stuff)
I live in tropical climate with No Air conditioning

and Gaming temp maxes out at 44.C

fwki · Post by **fwki** » Thu Apr 24, 2008 7:14 am

bhvm wrote: As for tha CPU, Even i am puzzled.
This system is made is singapore, and was gifted to me.
In the specs, it does not mention QUAD, but I mentions TWO Dualcores (Thats 2x2=4)
Other thing that might have caught your attention is that it has got TWO nvidia G72 cores in tandem,with 512MB RAM each!
What do you say?

My brother got me a Rolex from Singapore.

bhvm wrote:This is not a BIOS tweak or something. Being A Laptop (DID I mention that this is an lappy??) bios enhancements are simply absent! I've used RMClock to reduce Voltage.

RMClock cannot change voltage directly. All it can do is request a VID change from the MB BIOS. If the MB BIOS does not have the capability, the request is ignored. To determine if the VID request is implemented, use cpu-z which reports Vcore after Vdroop.

bhvm · Post by **bhvm** » Thu Apr 24, 2008 7:18 am

fwki wrote:
bhvm wrote: As for tha CPU, Even i am puzzled.
This system is made is singapore, and was gifted to me.
In the specs, it does not mention QUAD, but I mentions TWO Dualcores (Thats 2x2=4)
Other thing that might have caught your attention is that it has got TWO nvidia G72 cores in tandem,with 512MB RAM each!
What do you say?
My brother got me a Rolex from Singapore.

bhvm wrote:This is not a BIOS tweak or something. Being A Laptop (DID I mention that this is an lappy??) bios enhancements are simply absent! I've used RMClock to reduce Voltage.
RMClock cannot change voltage directly. All it can do is request a VID change from the MB BIOS. If the MB BIOS does not have the capability, the request is ignored. To determine if the VID request is implemented, use cpu-z which reports Vcore after Vdroop.

Temps and Fans have come down a lot, which means Voltage has indeed changed. I tried even gaming at bare minimum voltage (0.95) But it hung after 2 Hrs.

Then I set it up to (0.96V) the just next step, and is stable even after 4 Hrs of DoomIng!

Cistron · Post by **Cistron** » Fri Apr 25, 2008 3:51 am

Hmm, my E2160 currently runs with 1.285V core voltage at 3GHz, but that value was set by default. So I'm wondering how far I can go down. I'll experiment on the weekend.

SKeptical Thinker · Post by **SKeptical Thinker** » Sat Apr 26, 2008 5:15 am

I have a BE-2350 in a Gigabyte ga-ma78gm-s2h. I have set the CPU voltage in the BIOS at 0.975. Speedfan reports 1.0-1.04 depending on load. It's running with 10x multiplier and 210 clock.

BE-2350 2.1GHz 1.04V both cores at 100%
2.1GHz 1.00V idle
2.1GHz 1.3V idle 1.36 both cores at 100%

Modo · Post by **Modo** » Sat Apr 26, 2008 8:23 am

Intel Core 2 Duo E6550 (stock: 2.33 GHz @ 1.35V) on a Gigabyte GA-G33M-S2H (Intel G33 chipset) motherboard

Minimum: 1.6GHz @ 0.9V
Stock: 2.33 GHz @ 1.1V

These are BIOS settings. The actual voltages fluctuate slightly below that (no more than 0.05V, though).

smilingcrow · Post by **smilingcrow** » Sat Apr 26, 2008 12:30 pm

It would be very useful if people would specify whether the voltages they are listing are as set in the BIOS or through software or as reported by CPU-Z/Speedfan etc and whether they are reported at idle or under load. Otherwise itâ€™s just too vague.

Q9300 at stock speed - 2.5GHz
VCore set in BIOS (or Easytune) - 1.037V
Idle â€“ 1.008V
Full load (Prime95 x4) â€“ 0.976V
Both as reported by CPU-Z.
Stability testing â€“ 1 hour

Q6600 (G0) at stock speed - 2.4GHz
VCore set in BIOS (or Easytune) - 1.05V
Idle â€“ unknown
Full load (Prime95 x4) â€“ 0.976V
As reported by CPU-Z.
Stability testing â€“ 2:20 hours

bhvm · Post by **bhvm** » Sat Apr 26, 2008 6:36 pm

Nice ones!

amyhughes · Post by **amyhughes** » Sat Apr 26, 2008 11:56 pm

AMD X2 4450e (AM2) stock - 1.25V

Minimum: 1.0 GHz @ 0.825 V set in BIOS, reported by cpu-z: 0.832 idle, 0.832 load
Stock: 2.3 GHz @ 1.025 V set in BIOS, reported by cpu-z: 1.04 idle, 1.056 load
Prefered: 2.0 GHz (200x10) @ 0.95 V set in BIOS, reported by cpu-z: 0.96 idle, 0.976 load

ryboto · Post by **ryboto** » Sun Apr 27, 2008 3:39 pm

I guess I'll amend mine with load results. The processor folds 22/7, plays games sometimes.

AMD X2 4000+ G1 Brisbane @ 2.41ghz, 1.125v as set by CrystalCPUID

Idle vcore: 1.128-1.136
Load vcore: 1.136

edit: reported the same by both Everest and CPU-z, Abit EQ doesn't report beyond 0.01v.

gvego · Post by **gvego** » Mon Apr 28, 2008 2:31 am

I didn't try to find the lowest, but my sweetspot with X2 BE-2350 G2 is 250MHz x 10 = 2,5GHz @ 1.05V - reported by CPU-z 1.077V

Rock solid for everything I've tried (ORTHOS, games...)

Memory is Supertalent DDR2 800CL4 @ 1000MHz 5-5-5-15 2,0V on GA-MA78GM-S2H, IGP > 900MHz without overvolting

Question: what is the difference between BE-2350 and "normal" brisbane @ equal freq. and vcore regarding power consumption?

ryboto · Post by **ryboto** » Mon Apr 28, 2008 7:56 am

gvego wrote: Question: what is the difference between BE-2350 and "normal" brisbane @ equal freq. and vcore regarding power consumption?

Is the 2350 a G2? I think they're just binned to operate at a lower vcore.

davidh44 · Post by **davidh44** » Thu May 01, 2008 12:46 am

ryboto wrote:
gvego wrote: Question: what is the difference between BE-2350 and "normal" brisbane @ equal freq. and vcore regarding power consumption?
Is the 2350 a G2? I think they're just binned to operate at a lower vcore.

Both G1 and G2. And yup, just binned to operate at lower vcore.

At stock settings:
My G1 runs at 1.2v, and 1.1v in Cool 'N Quiet idle mode.
My G2 runs at 1.25v, and 1.0v in Cool 'N Quiet idle mode.

Whereas my "normal" X2 4200+ Brisbane G2 runs at 1.3v/1.1v

fwki · Post by **fwki** » Thu May 01, 2008 6:17 am

Palindroman wrote:If I may ask, how come the Meroms don't go below a 0.950v (T5600) and 0.975 (T7400) respectively. According to this site it should be able to vary between 0.75~1.3V.

AFAIK, the motherboard limits the lowest VID setting and and this low level setting can override the range set in cpu registers when the MB minimum is below the cpu minimum. When the MB minimum is above the cpu minimum, you won't be able to reach the cpu minimum. Under software control (e.g. RMClock), the minimum VID is determined by the range in the cpu registers as long as it's within the MB limits.

Unfortunately most BIOS's for laptops and mobile cpu desktops are fairly limited, so it's not surprising that we are running into VID minimums significantly within cpu capability. It's just another blow to the MoDt market...crappy, expensive mobile cpu MB's and now very efficient 45nm desktop cpu's.

leifeinar · Post by **leifeinar** » Sat May 03, 2008 4:02 am

got my self an qx9650.... so i tried different voltages. i guess there will be a lover clocked 45nm quad than the q9300. either 2,0ghz for htpc og a 2,3ghz. so i simulated this and got some really nice voltages

6x333 = 0,8V
7x333 = 0,864V
8x333 = 0,944V
9x333= 1,04V

and OC. 3,66 @ 1,19V4, and 4,0 @ 1,31V

http://www.diskusjon.no/index.php?showtopic=947482&hl=

gvego · Post by **gvego** » Sat May 03, 2008 7:10 am

gvego wrote:I didn't try to find the lowest, but my sweetspot with X2 BE-2350 G2 is 250MHz x 10 = 2,5GHz @ 1.05V - reported by CPU-z 1.077V

Now I know the minimum:
260x5.5=1430MHz on 0.8V (0.816 in CPU-z) minimum setting in CrystalCPU (up on 80%, 100ms refresh)
260x9.5=2470MHz on 0.95V middle setting (up on 80% down on 20% 1000ms refresh)
260x10.5=2730MHz on 1.20V (i guess this could be lower, as it was ok above 2800MHz on the same voltage (down on 70%, 10sec refresh)

Now... what would be the best low profile cooler if I wanted to put this in some low HTPC case?

I am considering buying a ninja mini, and using it without vent.
It should be OK for 0.8V, right?

ryboto · Post by **ryboto** » Thu May 08, 2008 7:58 am

Just updated my overclock. The Ballistix in my system are EPP ready, so I thought I'd try them out at 1000mhz.

AMD X2 4000+ G1

2.5ghz @ 1.15v Set by CrystalCPUID, monitored by CPU-Z under load as 1.17v

kinda wish I had a G2, but I can't justify any upgrades right now, so I guess I'll just have to look on and be patient.

chahahc · Post by **chahahc** » Tue May 13, 2008 6:11 pm

Intel Core Duo T2400

Stock voltage: 1.8ghz @ 1.256v
Undervolt: 1.8ghz @ 1.025v

Seems stable enough. Very similar to the T2500 undervolt.

Strid · Post by **Strid** » Sat May 17, 2008 3:27 am

Xeon E3110 @ stock 3 GHz ~ 1.056 V load, as reported by CPU-Z 1.45.
Specs in sig.

Undervolting: how low can you go?

Another BE-2350

Intel Core 2 Duo E6550