drop your chipset temp by 4-5 °C w/ a BIOS setting

[graysky]

I recently started to remove the "auto" setting from some of the user configurable voltages in my P5B-Del and noticed that not only is my system stable to 2x orthos for >2 h (and still counting, I'll probably stop it after 6), but the "system or chipset" temp as measured by speedfan (Winbond W83627DHG sensor array) has dropped as well which to me means the auto settings were very likely over-volting one or more of these numbers just like it did to the vcore on my CPU before I manually set that too.



Anyway, the last time I did a 2x orthos stress test, room temp was about 72 °F (it's 71 now) and the "system" temp as measured by speedfan was 42-43 °C after 2 h. Now it's 37-38 °C which is an easy 4-5 °C cooler. It was 37 for the majority of the test and just recently went up to 38. I'm running a q6600 @ 9x333 with a CPU vcore of 1.3250v. The other settings are set to their minimum values:

Code: Select all

FSB Term. voltage: 1.2v
NB vcore: 1.25v
SB vcore: 1.5v
ICH Chipset voltage: 1.057v

In case you don't know, orthos is a great app to find minimum stable voltage settings since it will stop and report an error if the math doesn't = known values.

Conclusion: don't leave your voltages set to auto, at least with a P5B-Deluxe, if you want to minimize your system temps.

[continuum]

Useful thing that people tend to forget. Although I usually am raising voltages as I overclock.

Although for stability I tend to run 24hr+ of Orthos, at least...

[cmthomson]

P5B and P5W BIOSes silently boost the voltages based on the clock frequencies when they are set to auto. If you care about noise or power consumption, you should always set the CPU and MCH voltages manually on these boards, especially if you are overclocking.

[Matija]

What are the default values? Those mentioned in the code fragment above, or is that undervolted already?

[cmthomson]

The default is set by the VID values that the CPU provides to the motherboard. This can vary by CPU model number, or even lot number.

For most C2D CPUs, the VID is 1.35V.

[Matija]

Yes, but we are talking about chipset voltages here

[graysky]

All I know is that 'auto' mode for chip vcore gives the thing way too much mvolts. I have to assume the same overestimation is true for the other voltages as well.

[derekchinese]

Yup, this pertains to all mobos. Auto will always have to err on the stable side and over volt. It would be cool if auto would automatically do stability testing to find the lowest stable volts so that we don't have to!

On my p5k deluxe, I am testing all voltages at the lowest and it seems to be stable right now with my e6700 running at 3.56ghz (knock on wood that it is stable).

Derek

[Matija]

I just changed those voltages from Auto to the lowest possible ones on my P5B Deluxe, which has an E4300 at 2.4 GHz.

Nothing happened. System temp (idle) is still 42-43 degrees (ambient currently 25-26) - no change.

Weird. Looks like the board increases those voltages only on heavy overclocks...

[Erssa]

Now it's 37-38 °C which is an easy 4-5 °C cooler.

Too bad this has absolutely no effect on noise...

[jackylman]

Now it's 37-38 °C which is an easy 4-5 °C cooler.

Too bad this it has absolutely no effect on noise...

You're probably right, but it's possible that the energy saved will be just enough to keep nearby fans or the one in the PSU from ramping up to the next speed. As i said, it's not likely, but I wouldn't state, "it has no effect on noise" as a fact.

[Erssa]

You're probably right, but it's possible that the energy saved will be just enough to keep nearby fans or the one in the PSU from ramping up to the next speed.

Like you said, I'm probably right . It's possible (but very unlikely), that it could save the psu from ramping up. As for other thermally controlled fans; if they ramp up, their thresholds are too low and need to be adjusted.

I'm curious to know, how much energy chipset undervolting could actually save. So if anyone has any hard data, please share it here. I doubt it's significant, but maybe still worth an article in SPCR/EPCR?

[ronrem]

Any time you get cooler-you can dial down Fan RPMs some. Some folks are thrilled to get a low temp----but there's 2 fans turning 2500 rpm. One fan at 1000 may still give enough cooling -but way less noise. In a Quiet puter---anywhere you can reduce heat produced-get cool air in efficiently-hot air out efficiently--has to be a plus.

[Erssa]

Any time you get cooler-you can dial down Fan RPMs some. Some folks are thrilled to get a low temp----but there's 2 fans turning 2500 rpm. One fan at 1000 may still give enough cooling -but way less noise.

As long as the computer remains stable, you can dial down RPMs, even if you overclock and overvolt. Just look at cmthomsons DIY article.

Running TAT at 100% on both CPUs, the core temperature of the E6600 at stock frequency and voltage (2.400 GHz and 1.35V) goes up to only 68C with the Ninja fan running at just 600 RPM.

I think cmthomson is probably the best example of true silence oriented user here. Few people here would use the word only in front of 68c. But as we know, 68 celsius is perfectly ok and there's really no need to make Nexus slower then 600rpm. Imo aiming for lower temps at the expense of fan speed, is overreaction to temperatures.

Indeed the quest for lower temps has only very little to do with the quest of silence and most coolness oriented people here at spcr seem to cool their pcs at the expense of silence, with unnecessary fans and excess fan speeds. Thermal barrier is not in their computers, it's in their heads. Imo more people should follow cmthomsons example, when it comes to safe temps.

In a Quiet puter---anywhere you can reduce heat produced-get cool air in efficiently-hot air out efficiently--has to be a plus.

If I had an identical computer to graysky, he would end up with a computer that has a 4c cooler chipset then me. I would still dial my fans at the same speed as him (or most likely even slower). We would end up with basicly identical systems, with a very small difference. My system would most likely be just a tad more stable with autovoltage, not that it could be detected, just like the temp difference, unless of course you look at thermal monitors. The only tangible difference would be, that I wouldn't be as cool a person in the internet community, because my computer runs hotter.

Sorry for this rant. I just think most people worry too much about their computer temps. This tip by graysky is a good one of course, because it doens't cost you anything. But seriously, I stand by my original statement running your chipset at 43c instead of 38c has no effect on noise and I'll add that it has no effect on stabilty or MTBF of the hardware either.

[cmthomson]

Most north bridges consume about 20W. Lowering the voltage might save 2 or 3 watts. So watt? The only reason to drop the voltage of a NB is if you have a feeble cooling system (poor heat sink or air flow), perhaps by design (as in my system, which as anemic NB air flow).

Thanks erssa for your kind comments; don't see too many here any more...

One of my pet peeves with postings here in the last year or so is the obsession with CPU idle temperatures. This is completely bogus. The only CPU temperature that matters is the load temperature. The idle temperature will always be low relative to any dangerous level (unless of course the heat sink is incompetently installed) not only because the CPU isn't doing much, but also because unless you've set to the BIOS to all-manual settings, both the voltage and the clock multiplier are reduced automatically by the BIOS when the system is idle.

And, just to state one more time for those who might not know, a Core 2 CPU doesn't throttle until it hits 85C (100C for some of the newer ones). You can run it indefinitely at over 70C. Mine runs at 50% load (folding) 24/7 at about 67C, and has been doing so for over a year now.