C2D, low power at idle & high O/C with new BIOS.

[smilingcrow]

I’ve previously posted here that my experience with C2D motherboards is that once you change the VCore in the BIOS you lose the ability to change VCore with RMClock and the like.
It was recently brought to my attention that the more recent BIOS releases for the ASUS P5W DH Deluxe now remove this limitation, at least to a degree. I have tested this on a P5B deluxe and found it to be true for this Asus board also.

What you actually get is two fixed VCore values that you can switch between using Speedstep. The lower value is currently fixed at 1.05V with both boards, whilst the higher value is dependent on the voltage that you set in the BIOS. Typically, for the P5B at least, the VCore reported by Speedfan/CPU-Z is less than what you set in the BIOS.
To configure my system to enable this feature I did the following:

Set a manual VCore in the BIOS.
Using RMClock, I set at least two multipliers that would be used ‘on demand’ as the system load varied.
I set the VCore for the lowest multiplier, which was 6, to the lowest setting available, which was 1.05V in this system.
I set the VCore for the highest multiplier that I wanted to use, which was 9, to the highest VCore available, which was 1.212V.

This also worked with CrystalCPUID. Although the range of voltage settings is different, the principle is the same; you set a high voltage for the higher multiplier and a low voltage for the lower one.
The voltage settings work in such a way that there is a cut-off point for each state. You have to set the VCore at or below a certain value to trigger the lower VCore state and at or above a certain other value to trigger the higher VCore state.
In other words, just setting the values to the highest and lowest that the utility you are using allows should do the trick.

It’s a shame that the lower VCore value is only 1.05V, as that limits the usefulness of this feature.
With an E6600 over-clocked to 3.2GHz, the idle CPU clock will be 2.13GHz. This may well be stable at 1.05V.
With an E6300 over-clocked to only 3.0GHz, the idle CPU clock will be a high 2.57GHz. It seems unlikely that this will be stable at 1.05V

[argusarray]

How does this impact idle power at 1.05V on the P5VD2-MX? I noticed in your previous post that it idles at 58W @ 1.136V. Can you use this to underclock?

Thanks,

-Argus

[smilingcrow]

How does this impact idle power at 1.05V on the P5VD2-MX? I noticed in your previous post that it idles at 58W @ 1.136V. Can you use this to underclock?

The BIOS that I used with this board didn’t allow under-volting, so this technique isn’t relevant. This class of boards typically doesn’t allow you to change VCore in the BIOS, so I doubt that a newer BIOS revision would help.

I can’t recommend this board as it has a very poor IGP and I found it uncompetitive against the Conroe945G-DVI. I didn’t try over-clocking it as I wasn’t interested in it enough to warrant spending the time.
I have been meaning to write up some subjective notes on the boards I have tested in case they are useful to anyone.

[plympton]

I can’t recommend this board as it has a very poor IGP and I found it uncompetitive against the Conroe945G-DVI. I didn’t try over-clocking it as I wasn’t interested in it enough to warrant spending the time.
I have been meaning to write up some subjective notes on the boards I have tested in case they are useful to anyone.

What kind of controls does the Asrock Conroe945G-DVI board have for voltage/underclocking in BIOS? I'm getting one to build this week and my OS won't have any software controls for SpeedStep, etc, so whatever the BIOS can provide is the best I will have.

I'm going to pair it with an e6600 and a Zalman 9000Al-Cu that I had kicking around (7000? Dunno the number.. I needed an adaptor to get it to fit the 775, apparently).

-Dan

[line]

What kind of controls does the Asrock Conroe945G-DVI board have for voltage/underclocking in BIOS?

It can underclock but it can't undervolt.

You can read more here: ASRock ConRoe945G-DVI: Core 2 Duo goes mATX @ AnandTech

I'm going to pair it with an e6600 and a Zalman 9000Al-Cu that I had kicking around (7000? Dunno the number.. I needed an adaptor to get it to fit the 775, apparently).

It is the CNPS7000B-(Al)Cu that requires a separate adapter (ZM-CS1) for socket 775.

[smilingcrow]

What kind of controls does the Asrock Conroe945G-DVI board have for voltage/underclocking in BIOS? I'm getting one to build this week and my OS won't have any software controls for SpeedStep, etc, so whatever the BIOS can provide is the best I will have.

The BIOS goes as low as 140 for the FSB, no VCore adjustment.
I’m writing this with the FSB at 166 for an idle clock speed of 1GHz. I just disabled RMClock and noticed that the VCore is oscillating between 1.1x & 1.3x volts every second or two. Strange!
If you will be running Linux then I’m surprised that there isn’t a utility similar to RMClock for it. I thought there was such a thing for Cool n Quiet, so I’m surprised there isn’t something for Speedstep. Unless Linux people don’t like Intel!

[puppydg68]

Hi smilingcrow, I'm extremely new to Overclocking, hoping you can help.

I'm going to purchase an e6300 and conroe to be used in a silent PC.. either Zalman 7700 or XP-120+Nexus120 for cooling.

I leave my machine on always, so I want a stable OC with an asus P5B. I'm not greedy, looking to get somewhere in the range of 2.4-2.8Ghz extremely stable and quiet.

I've never done it before, so I think I'll just go with what is available in the bios. I'm not sure i understand what speedstepping is it the Asus advertises AI NOS, which will dynamically boost the CPU into an overclock only when required?

I'm not even sure where to start, i would really appreciate any help to point in my in the right direction. I visited anandtech, but it seems everyone is invested in getting the highest speed out of their processor, none of the OC'ing is with keeping the "quietness" in mind.

[paapaa]

If you will be running Linux then I’m surprised that there isn’t a utility similar to RMClock for it. I thought there was such a thing for Cool n Quiet, so I’m surprised there isn’t something for Speedstep. Unless Linux people don’t like Intel!

There is the cpufreq subsystem:

http://www.kernel.org/pub/linux/utils/k ... ufreq.html

Don't know about voltage settings but the above supports Speedstep fully. AFAIK.

[smilingcrow]

Hi smilingcrow, I'm extremely new to Overclocking, hoping you can help.

I’ve just determined a better method for over-clocking C2D when using stock voltages. It’s a particularly good method for using with the E6300, which otherwise is hampered by its low multiplier. I’ll post details on this tomorrow.

[puppydg68]

Hi smilingcrow, I'm extremely new to Overclocking, hoping you can help.

I’ve just determined a better method for over-clocking C2D when using stock voltages. It’s a particularly good method for using with the E6300, which otherwise is hampered by its low multiplier. I’ll post details on this tomorrow.

Thanks so much, I'm looking forward to it. Thanks for helping out someone new to this.

[smilingcrow]

I'm going to purchase an e6300 and conroe to be used in a silent PC.. either Zalman 7700 or XP-120+Nexus120 for cooling.
I leave my machine on always, so I want a stable OC with an asus P5B. I'm not greedy, looking to get somewhere in the range of 2.4-2.8Ghz extremely stable and quiet.
I've never done it before, so I think I'll just go with what is available in the bios. I'm not sure i understand what speedstepping is it the Asus advertises AI NOS, which will dynamically boost the CPU into an overclock only when required?

Speedstep allows a CPU to alter its clock speed and voltage dynamically based on the CPU load. It changes clock speed by altering the multiplier but keeps the FSB constant. The lowest multiplier for all C2Ds is currently 6 and for the E6300 the highest multiplier is 7; E6400 = 8, E6600 = 9 etc.
Using the E6300 that you want to purchase as an example we have the following if the CPU is not overclocked:

Lowest Speedstep clock speed = 6 * 266 = 1.6 GHz
Highest Speedstep clock speed = 7 * 266 = 1.86 GHz

To achieve 2.8GHz with the E6300 will require a FSB of 400, which gives:

Lowest Speedstep clock speed = 6 * 400 = 2.4 GHz
Highest Speedstep clock speed = 7 * 400 = 2.8 GHz

This should be possible to achieve with almost any E6300 and using the stock voltages. You will need to use DDR2-800 as the FSB of 400 will mean that the RAM will be running at an affective minimum of 800 MHz (2 * 400) with the P965 chipset. Although many DDR2-667 modules will run at DDR2-800 when over-volted, this is not guaranteed and is dependent on the BIOS being able to over-volt the RAM. I’m not sure if the P5B offers this feature; the deluxe version does.
At this speed and voltage a C2D is easy to silence quietly with a good heatsink/fan.

The other part of the Speedstep equation is CPU voltage control. This is of less importance with the E6300 as the difference between its two speeds is quite small when overclocked to 2.8GHz; 2.4 – 2.8 = 400 MHz
As a comparison an E6600 overclocked to 2.8GHz would have a lower Speedstep speed of 1.87GHz, which gives a difference of 1.87 – 2.8 = 930MHz.
So there is much more scope to run a significantly lower VCore setting at idle with the E6600 than there is with the E6300. Since many motherboards don’t allow you to lower the VCore below ~1.15V, this isn’t as big a deal as it sounds. An E6300 at 2.4GHz might only need 1.2-1.25V, so the difference at the voltage level isn’t dramatic.

I have a P5B Deluxe which does allow me to use as low as 1.05V when over-clocking, even when the VCore is set to auto in the BIOS. This is useful as when used with an E6600 I get these settings:

FSB = 360
Idle: 2.16 GHz, 1.05V – tested Prime95 stable for 2 hours.
Load: 3.24 GHz, 1.376V – tested Prime95 stable for 6 hours.

[Edit] These VCore values as reported by CPU-Z are incorrect. The power consumption stays constant regardless of the VCore that I select with RMClock, which strongly suggests that the VCore is always at a high state, even when CPU-Z says otherwise. No other utility shows a vaguely meaningful value for VCore with this board and BIOS version. [Edit]

I’m not sure if the non-deluxe version works in the same way though!
A new BIOS for both the P5B boards has just been released, so who knows how these boards behave now anyway!

[smilingcrow]

For this whole issue regarding the newer Asus BIOS’s allowing a fixed low VCore value at idle after you manually set a VCore in the BIOS, I have been relying purely on CPU-Z to measure voltage. Well it turns out that the CPU-Z data is incorrect, as when I took power measurements they didn’t back up the readings. The power consumption at load remains static regardless of the VCore that I set with RMClock and is reported by CPU-Z. I had the same result when using CrystalCPUID also.
Ironically, when setting the voltage to the lowest available, 1.05V, there seems to be a slight increase in power consumption of 1 – 2W.

I usually like to measure things with multiple tools, but in this case no other utility would read the VCore of the P5B Deluxe; which should have triggered a warning bell. Well I finally got around today to doing a bit of testing on the P5B Deluxe and monitoring power consumption when this error reared its ugly head.
Sorry to have built up any hopes and maybe the BIOS just released has changed this!

It’s possible that the P5W DH Deluxe also now suffers from this malady, as recent BIOS revisions ‘seem’ to have given it the same behaviour. Only testing with a power meter will verify this.
I witnessed fairly similar behaviour in a Gigabyte GA-965P-DS3 with BIOS F4, although it wasn’t there with BIOS F3. In this case though CPU-Z showed that the VCore was fixed at ~1.42V regardless of what Speedstep and RMClock were doing. Power consumption measurements seemed to back this up, although I didn’t make many notes on this issue, so I’m not sure how thoroughly I tested it. It’s still open to question in my mind.

It’s turning into a bit of a minefield with regard to overclocking C2D whilst keeping low power consumption in mind. Let’s hope that motherboard manufacturers don’t standardise on this practice, or idle power consumption for an over-clocked C2D will seriously suffer.

[puppydg68]

Thank you so much for your help, it's all starting to make sense to me. Those anadtech articles make my head spin, So you would recommend using the speedstepping to keep idle consumption low, thereby keeping the fans quiet most of the time?

After reading your below post, are you saying at this point there is no advantage to undervolting the C2D on p5b boards? So you recommendation is only touching the FSB and overclocking from there?

One last question, am I correct to assume that taking the FSB beyond the 400 will most likely result in requiring more voltage to the CPU in order to sustain a stable overclock at that rate, which would then in turn cause the chip to run hotter and less quiet?

[smilingcrow]

After reading your below post, are you saying at this point there is no advantage to undervolting the C2D on p5b boards? So you recommendation is only touching the FSB and overclocking from there?

Experiment with it; you will gain something from under-volting, but an overclocked E6300 has less to gain than just about any C2D CPU.

So you would recommend using the speedstepping to keep idle consumption low, thereby keeping the fans quiet most of the time?

You can have the fans quiet ALL the time with a C2D at 2.8GHz, provided you have the right components. In this scenario you are using Speedstep purely to lower power consumption and heat levels, not noise. If your heatsink/fans aren’t top notch, then Speedstep will also help to lower noise levels if your fans are temperature controlled.
BTW, Speedfan doesn’t currently work with the P5B, so you’ll need to determine how you will regulate your fan speeds.

am I correct to assume that taking the FSB beyond the 400 will most likely result in requiring more voltage to the CPU in order to sustain a stable overclock at that rate, which would then in turn cause the chip to run hotter and less quiet?

It will vary a lot depending on the CPU, heatsink, fan, case, ambient temps etc. I estimate that around 3.2 GHz is the limit for silent cooling of a C2D. You are correct in that once you have to start increasing the voltage that cooling quietly becomes more difficult.

P.S.
There’s no need to quote the whole of a previous post, it can have the affect of cluttering the thread unnecessarily, especially if it was a long post that you have quoted. One alternative is to direct your reply to the person in question.

[smilingcrow]

I tested the P5B Deluxe with the recent BIOS version 0614 and it exhibits the same issue as previously noted; if you overclock in the BIOS, the VCore becomes fixed even though CPU-Z reports it as varying depending on what you set it to with RMClock.
Fortunately, if you under-clock in the BIOS this problem doesn’t occur.
Also, if you leave the FSB as AUTO in the BIOS and increase the FSB with Clockgen, the VCore can be changed with RMClock or Speedstep for that matter.
I think I was using a beta of Clockgen and you need to set the clock generator to the Cypress CY28551. It’s been stable in my experience.

[marcob]

Just a short report, the BIOS 0804 behaves the same, and somehow I doubt it's a matter of BIOS.

[NeilBlanchard]

Welcome to SPCR! And thanks for the info.

[marcob]

Thanks! I've been lurking around here for over a year, so I just decided to register and have a say or two

I'm in a lot of trouble right now, this P5B(deluxe, bleah) really s*** when it comes to fan/voltage control. Via speedfan, there are only two fan ports that can be controlled, one is the CPU should you be that lucky and have a decent 4-pin fan and the other one is for all chassis fans(yes, one control for all fans together, no individual setting).

But, the biggest problem of them all is the voltage control, by using RMClock I am limited to 1.168V - 1.275V, and I need 1.325 in order to run stable(actually 1.305V measured with the DMM, this mainboard has a huge VDrop). So that means I can't set the voltage @ auto and control it with RMclock.
Does any of you know whether it's possible to unlock the VID of an E6600? I guess it is programmed inside the processor, so almost no chance, but I thought I'll ask anyway.
The only solution to this problem would be a piece of software that can change the CPU voltage via the mainboard not by using the VID. Does anyone know of any such software?

PS: I hate asus for this
PS: My next mainboard will be either Abit or DFI(Abit is the king of the fan/voltage control, it's amazing what it can do, while DFI also offers 3 controllable fan channels, but marginally better performance/support)