C2D Undervolting experience?

Any people out there undervolting their C2D laptops?
Wondering what voltage limits I should be able to reach safely.

Mike

 

Here's my own Core 2 Duo undervolting experience:

Model: XPS M1210
Processor: Core 2 Duo T7200 (2.00GHz, 4MB L2 Cache)
Processor sSpec: SL9SL ( link to Intel's sSpec page)

(Click on the image for a full sized version)



So far, I was able to go from 1.200V, down to 1.150V fully tested and stable. That's a decrease of 4 steps, or 0.050V / 4.17%.

What does it mean? Well, Intel's Core 2 Duo Mobile Processor specifications sheet (.pdf format) comes in handy again.

Under Section 3.10 - Processor DC Specifications sheet, the various operating parameters are listed out:
Vcc @ HFM (Supply voltage at Highest Frequency Mode) = 1.0375V - 1.3000V
Icc @ Vcc HFM (Nominal supply current at Vcc HFM) = 41A

Since Power = Voltage x Current, the absolute maximum power consumption for my processor at the highest frequency mode would be:
Before: 1.200V x 41A = 49W
After: 1.150V x 41A = 47W
Theoretical Scenario: 1.0375 x 41A = 43W (assuming best case undervolting)
(All rounded to 2 significant figures)

In my case, I saved 2W when maxing out the processor. With a 85WHr extended 9-cell battery, that means:
85WHr / 49W = 1.7 Hours
85WHr / 47W = 1.8 Hours
85WHr / 43W = 2.0 Hours (assuming best case undervolting scenario)
(That is before taking into account other parts of the system which uses power as well, but it should be constant between the 2 so we are only looking at the influence of the processor being undervolted)
So I gained about 0.1 hour/6 minutes of extra battery life due to the processor running at a lower voltage.
Under the assumed best case scenario, you will gain about 0.3 hour/18 minutes of extra battery life.

In reality, since the processor will be switching between the power states quickly, it will transition down to the lowest frequency mode/power state for as much as possible (which is 1GHz @ 0.95V nominal), averaging/cancelling out the benefit gained from saving 2W. I failed to notice any gain in battery life either running the processor at a lower voltage so far, further suggesting that undervolting does not have any tangible benefits in increasing your battery runtime.

As for operating temperatures, I've noticed the following:
Before: Idle - 45°C-55°C, Maximum load (passive cooled before fan kicks in) - ~82°C
After: Idle - 45°C-55°C, Maximum load (passive cooled before fan kicks in) - ~78°C

Since at idle it will run at the lowest frequency mode and power state, there will be no difference in idle temperature. At maximum load, I notice that the processor does run about 3°C-4°C cooler undervolted passively cooled, however, once the fan kicks in there is virtually no difference in the system's running temperature.

In my opinion, undervolting provides no significant tangible difference in making your system run significantly cooler, or a longer battery runtime. You'd be better off/gain much more by simply capping your processor speeds to run at its lowest frequency state while on battery power.

I guess the whole undervolting thing is to provide bragging rights more than anything at this stage.

 

Thanks again for your insight. However, I think the difference you can achieve in temperature is a worthwhile effort. I went from a max temp of 78 degrees to 72 degrees C. This is with a 100% load where the fan is running constantly running prime95.

Additionally, I think your battery consumption calculations may be off. If you calculate it using the minimum voltage .95V you'd get

.95V X 41A = 39W

85Whr/39W = 2.18 hrs

I think with a 9 cell battery, the M1210 can go 4:00 to 4:30, so something isn't right.

Mike

 

I HIGHLY doubt you can get any Core 2 Duo to run stable at 0.950V while running it at its High Frequency Mode (which would be 2.00GHz in the T7200).

In theory, making the switch from 1.200V down to 0.950V will, as you said, provide a noticable improvment to the battery life and temperature (as that's a 10W, or 20% saving in power compared to the 49W maximum value).

However, Intel's own specification sheet listed that:
Vcc @ HFM (Supply voltage at Highest Frequency Mode) = 1.0375V - 1.3000V
With the majority of Core 2 Duos shipped at 1.200V Vcc @ HFM.

Therefore in my calculations below, I only used 1.0375V as the "assumed best case scenario". In fact, I think the majority of Core 2 Duos out there will only go at best, 10% lower (~1.0750V-1.0875V) while still running completely stable.

If you can get your Core 2 Duo to run 100% stable at 0.950V at its maximum frequency all loaded out please let me know and provide us with some screenshots and details


Still, these savings are only realised, assuming that you let your system run 100% fully loaded and running at the highest frequency mode. The fact is, most of the time the system is sitting idle (web browsing, watching movies, using Office apps etc.), and therefore the processor will kick down to 1GHz using as little juice as possible anyways. So, on average, one will not see much difference, if at all (and that's my point of the original post).

 

I have mine running at 1.0125v at 12x. Been running for a week now fine. Temp after 4 hours of continuous gaming is 62c peak. Idles around 40c. Seemed to help with the GPU temp a bit too, maybe just not as hot in there. GPU temp went down by 3c. Now maxes at 68c. Harddrive temp maxes at 32c.

my specs are

XPS M1210
Core 2 Duo T7200
Gefore 7400go
80Gb 7200RPM HDD
2GB RAM

Havnt tested battery life. dont think it will change at all since 6x is always .95v

hope this helps

 

Screenshot (this was taken right after gaming for a while)

 

Thanks georgel12.

Wow, I stand corrected, I guess Intel's yield and quality on their Core 2 Duo silicon must be really good in order for you to get 1.0125V @ 2.00GHz and still run stable. That's a 0.1875V/16% decrease which in my opinion, would be a worthwhile endevaour.

I could try and experiment/tweak mine a bit more, however, I've been running Vista RC2 for the past month and unfortunately, NHC wouldn't work with it at this stage. Just hope they'll release a version compatible with Vista by next month (seeing it's gone gold/RTM this week already and launch is imminent).

 

Angrymob,

I wasn't implying that I could run at .95v, I was only trying to figure why the calculated power consumption at .95v would only yield 2.18 hours? Maybe the 41A value changes at a different multiplier?

What would the calculation look like to get to 4:00 hrs of battery life in power saving mode.

 

yea, i had RC2 for a little while also. the temps were at least 5c higher in RC2 than in XP. also, NHC just didnt work for me at all in RC2. not sure wat i did wrong. now if only there was a way to make the GPU run cooler...

 

PanamaMike:

Well, the current does change depending on what frequency, and power (or "C" state) the processor is in. So, you are correct, at different multiplier settings, the processor will run at a different frequency and thus the current draw will be different.

Looking at the datasheet again, here are all the possible current ratings:

Icc @ Vcc HFM (Highest Frequency Mode) = 41A
Icc @ Vcc LFM (Lowest Frequency Mode) = 27.3A
I AH (C1 state, Auto-Halt/Stop grant) @ Vcc LFM = 18A
I SLP (C2 state, Sleep) @ Vcc LFM = 17.8A
I DSLP (C3 state, Deep Sleep) @ Vcc LFM = 17A
I DPRSLP (C4 state, Deeper Sleep) = 12.1A
I CCDC4 (C4 state, Intel Enhanced Deeper Sleep) = 9.9A

We have the following combinations for all Core 2 Duo processors:
- Maximum workload @ maximum clockspeed: 1.200 x 41 = 49W
- Maximum workload @ minimum clockspeed (1.00GHz): 0.950 x 27.3 = 26W
- Minimum absolute power consumption @ C4 state (machine idle, ~0% load): 0.950 x 9.9 = 9.5W

However, if you run the processor at maximum clockspeed, the current draw will still be at 41A, hence why even if you set the voltage to 0.950V, at 12x multiplier setting running at max clockspeed it'll still consume up to the maximum of 41A.

Most of the time, your processor will cycle through the various power states/frequency mode, therefore, it is very rare that you will have to draw a constant 49W (maximum possible) all the time. In fact, if you look at my XPS M1210 review:

Worst Case scenario

Screen to maximum brightness (7/7), CPU and GPU at maximum speed, WiFi and Bluetooth turned on.
3DMark06 in looped demo mode

1 Hour 45 Minutes (105 Minutes)
System power usage: 48.5 Watts

Video watching via WiFi

Screen to medium brightness (3/7), CPU with SpeedStep management, GPU at minimum (PowerMizer set to max battery), WiFi on, Bluetooth off.
Streaming videos off WiFi network from media server continuously

3 Hours 40 Minutes (220 Minutes)
System power usage: 23 Watts

General web surfing and light workload

Screen to medium brightness (3/7), CPU with SpeedStep management, GPU at minimum (PowerMizer set to max battery), WiFi on, Bluetooth off.
Surfing the internet, chatting online, using MS Word/Excel etc.

4 Hours 30 Minutes (270 Minutes)
System power usage: 19 Watts


Hence why undervolting the CPU is not the only answer to try stretch out your battery life, as you cannot change the current draw and the processor itself will try to switch to a lower power state as much as possible as well.