Jan Kivar's report on CrystalCPUID, a user-configurable Cool 'n' Quiet that now works not only with Athlon 64s, but also Intel 600 series and the Pentium M processor.

Great review by Jan Kivar. I am too much of a newbie to play around with this tool but I look forward to reading of experiences by those more experienced and/or adventurous.

Great review Jan,

Just 3 quick points:

#1: It might be a good idea to include the registry entry that allows it to start at boot

#2: CnQ does NOT have to be enabled in BIOS for CrystalCPUID to work - at least not on my Abit VIA board.

#3: Benchmark programs usually report decreased performance due to startup delay. For the benefit of newcomers it might be beneficial to note that this is purely synthetic and that there's no performance loss.

Great review. I had a question, though. Jan mentions in the article:

"it's good to know that similar utilities are also available for the Linux community."

Is this a reference to RMclock (which is discussed right before this statement)? I went to the homepage and it looks like RMclock is a windows program. What, then, are the linux programs that perform similar functions as CrystalCPUID?

I have been using this on Windows since shortly after getting my amd64 machine. I have a Abit AV8 which is a VIA board. Have not had any problems with and I like being able to undervolt.


On the Linux side with later kernels Q'n'Q fuctionality is built right into the kernel. No need for a third party app or driver. This is not very "user" configurable at his point. But I have modified my kernel code to get it to work the way I want. I don't advise this unless you know what you are doing. On the other hand the kernel code to handle this is fairly simple so if you are a programmer you can do some tweaks on this without too much difficulty.

It does on my board, though. Better safe than sorry IMO.

Will it work with MSI Mother Board with ATI chipset?
Is it compatible with MSI: RS480M2 (MS-7093)?

Changing multipliers on the fly works fine, but voltages remain
at ridicules high levels 1.65V!!! max/ 1.35V min. I have 'enabled voltages'
but nothing happens. Must be a BIOS issue (A8V rev2, 1009.007).
I have to disable CnQ to be able to control the voltages at all.

Anyway, it's only an 8W(AC) difference between idling at 2.4GHz @1.5V
versus 1.1GHz @1.35V. That's why I don't use CnQ on this board.

The article clearly points out that CnQ has to be disabled in Windows in order for CrystalCPUID to work. On page 1:

"2. Configure Windows NOT to use Cool 'n' Quiet. Go to Control Panel -> Power Options, and ensure that "Minimal Power Management" is not selected."

Mike, Burcakb was referring to the BIOS instructions right above the line you quoted, where it instructs the user: "1. Check if "Cool 'n' Quiet" option is found in BIOS, and turn it on. " On some boards at least, it doesn't matter if BIOS CnQ option is on or off. Which implies that CrystalCnQ may work even on some boards where CnQ is known to be permanantly disabled in the BIOS.

And for us cool cats with Athlon XP systems , you can use 8rdavcore. It allows controlling your pwm speeds and a whole lot more. Good job Jan!!

Nice write up Jan. This will be a lot more helpfull for a lot of people than the "help" included with the program itself.


I played some with CrystalCPUID 4.1.4.226 and that motherboard a few weeks ago. It worked. Don't ask me details, as I am not using it now. My PSU on it's lowest speed is still the loudest component in my setup, so it didn't help with the noise.

Very nice, now I don't have to care that the normal CnQ stopped working after installing SP2.

Finding the lowest voltages will have to wait untill I feel I have more time tho. My fan still turns off at 1.1V after a while

CnQ + ATITool underclooking the graphic card meant that case temp took a real plunge.

Many s for you.

Excellent work, Jan!



CrustalCPUID works with AXP if:

-The CPU is somehow unlocked, either pre 0339, post 0339 with bridges 5:1 and 5:2 closed (I think), or a mobile CPU

-A mobo that supports multiplier changes on the fly (not nVidia)

I used it with my locked 2500+, unfortunately the voltage change didn't work on my Abit KV7.

I noticed that when I used this software it disturbed the mp3 player, the sound disappeared for a fraction of a second. I think it happened sometimes (but not always) when it switched speed. This is an issue that I guess is gone on A64 since it's made for this from the start.
A64 owners, have you ever had this problem when switching speed automatically with C&Q or CrystalCPUID?

One thing that I would like to see is a program that lower the multiplier when undervolting, and raises the "FSB" when overclocking (AUTOMATICALLY). The two reasons for this is that I don't use unlocked CPU's, and I don't want to run the FSB att 300 MHz at idle.
How does A64 work with a FSB lower than 200 MHz? Running at a lower HT multiplier all the time is obviously no problem, but I don't know about the actual bus speed. Any help would be appreciated!

Nice article Jan

I still don't totally get 'up threshold' and 'down threshold' but it seems to be working fine so I am not going to worry about it.

If you're folding, you'll never enter CnQ and thus never miss a beat so to speak.



On my system, which is normally at 240x10, I can drop down to 160 but below that I lock up. Perhaps this has something to do with the memory timings I have optimized for 240.

BTW, CnQ with AMD's driver and minimal power management works just fine for me. I don't have a need for this application. Then again my CPU hasn't been below 100% for weeks.

Great! Now why did we even bother using CrystalCPUID in the first place? Can't remember...

Great article, made me try the program - in a little way, since I'm already using RMClock to undervolt below that allowed by CnQ, this is a little one-person's opinion using both RMClock and CrystalCPUID

1) At the end of the article, you mentioned RMClock was buggy. I've been using this for the past week without problems so far.

2) in some ways RMClock is easier to manipulate than Crystal with same functions. Both allow changes on the fly, both allow 0.025V changes, but RMClock blocks you from making some multiplier changes (eg. my Winnie 3000+ on Asus A8V only allows 4x, 5x, 8x, and 9x).

3) I'm currently running at x5 0.9V at minimal and x9 1.15V at maximal; it loads at startup and been stable with Prime95 for at least 12 hrs. I haven't yet tested the limits of undervolting, these numbers were used based on another person's experience with same mobo

other comments:
1) at least on my board (Asus A8V) Crystal and RMClock works whether I turn on CnQ at BIOS level (for the earlier poster with the same board, I've got the same BIOS revision 1009 dated jan 2005). However, the programmer at RMClock also recommends turning this feature on

what are other ppl's experiences?

I have been using RmClock on a Mobile Athlon XP with a VIA KT880 chipset under Windows 2000 and later Windows XP Pro. Unlike the editor's experience, I did not encountered any problems with RmClock and found the interface much easier to use than Crystal. Some of the other utilities like 8AVcore doesn't work with my board.

Actually, I used it mostly to get around the problem of my MB not being able to set the multipler of the Mobile Athlon XP from the BIOS. The MB boots with the cpu locked at 6x. Using a Athlon XP-M 2600, I was able to change the multipler anywhere from 3x to 15x, and voltage from 0.925v to 1.45v (though I think the actual voltage is probably higher, since the BIOS report 1.65v). I could even use the RmClock to occassionally overclock. Note that RmClock will not work if you're not using a mobile processor. The previous process was a Duron and RmClock could not make any adjustments. I set it up to dynamically go up in clock speed when there is a load.

I also tried RmClock on my laptop and it worked without issue, except that I could not go below 4x for some reason (cause crashes).

Another problem is that the utilities can only run by an account with admin. I normally set up user accounts with no admin permissions for security. Normally, I am trying to figure out how to run RmClock as a service under the admin user account.

Paul

Having a shortcut in the Startup-dir makes it easier to temporarily disable the automatic starting -- just drag the shortcut out from the Startup-dir. A registry key requires more tinkering IMO. [One can disable the excecution of Startup-dir shortcuts by keeping either Alt, Shift or Ctrl depressed when Windows is starting up. Don't remember which keys (one or more) exactly, though.]

That's true. I've read from some other forums that some (older) games don't work properly if the low state Interval Time is too long. The game measures the CPU speed when starting, and when the state is eventually changed to max., the game runs too fast.

Cheers,

Jan

Unfortunately I have little experience with Linux in general. The claim for Linux support is based on this post, in which one forum member uses a "hacked cool 'n' quiet kernel driver". I don't know how one should hack the driver, and it's IMO even more dangerous than using CrystalCPUID (stability issues if one messes up with the kernel).

But, as hvengel already mentioned in this thread, it surely is possible. Maybe he could post some more specific info in this thread.

Cheers,

Jan

It's Shift (you can do that when putting in CDs to stop autorun, too).


Re Linux - I believe there's a program called 'speedfreq' that lets you change frequency to whatever you want without the powernow daemon. I don't use it because of folding, but there should be an option (or another program) that undervolts as well.

wowie fred!

Great article!

I was wondering if anyone has tried CrystalCPUID with a P-M on either the Aopen or DFI 855 board, and if so, was it stable?

Is there any advantage to use it with a Winchester 3000+, which has a relatively low Thermal Design Power of 60 Watts?

You could turn it into less than 10 watts when it's idle.

I messed with it a tiny bit yesterday, but couldn't get it change the Vcore and Multiplier dynamically. I'll continue playing with it because it looks like it would be a nice replacement for the unstable (on my P-M systems) RMClock.

Your note suggests a number of utilities are available for the GNU/Linux community. Would you care to elaborate, an update to the article would be nice.

So far in this thread I have seen kernel hacks mentioned and only one program SpeedFreq which looks to be alpha. Maybe were not so well covered...

Anyone know a micro-ATX board that this would work on?

A little higher up in this same thread:
http://forums.silentpcreview.com/viewtopic.php?p=172439#172439
The MSI RS480M2-IL is a micro-ATX board.