I have a Presario 906us that is socketed with a Mobilie Athlon XP 1500+. I like to upgrade the processor. The BIOS only supports at best an Athlon XP-M 2200+ T-Bred. TBred however are hard to find these days, however, some folks on the net have successfully upgraded their processor to XP-M 2600+ Barton. Surprisingly, the laptop doesn't POST when you start it up or generate too much heat. Instead, Powernow is disabled because the Powernow is unable to find the stepping table in the BIOS. Windows automatically disables Powernow.

Patching the BIOS is simply out of the question. What I like to know is if I can run RMClock in place of Powernow. Can I get RMClock to transition between different voltage/multipler according to cpu load? Ideally, I would like to see if I can actually run the cpu at a lower voltage and multiper state (you don't need a lot of power to write email or surf the web) than I can with Powernow. This way I have a machine that's actually more energy efficient but have a higher top end in situation when I need it.

Yes, it's perfectly possible if your processor's multiplier is unlocked. Set a minimum voltage/multiplier and a maximum for 100% usage, and it will vary between these as CPU load changes.

I prefer CrystalCPUID as you can specify three load states, rather than using linear interpolation, but it's not as user-friendly as RMClock.

Cool, Mobile Athlon XP are unlocked (at least downward).

1. Does the program need PowerNow to work? I am thinking of trying this in a desktop board first. As far as I can tell, the MB supports Mobile Athlon XP-M.
2. How dangerous is this program, can I accidently set it to a state where I can't boot up?



Why so? I thought the linear interpolation would be closer to how Powernow works and would be friendlier to the system. If I suddenly jump from 500 mhz to 2.0 Ghz, wouldn't I get more timing issues.

Paul

As far as i am aware as long as the BIOS allows the changing of the multi and voltage from windows which it should because PowerNow would never have worked if it didn't.

Since the underclocking doesn't occur until you load the program, your laptop will always start at the default settings in the BIOS.
As StarfishChris says a program like CyrstalCPUID has three power states rather than PowerNow which is based on a table of defined setting. This means it should do what you tell it to change to rather than trying to use a set of incompatible states.

Are you 100% sure it will run in your laptop if it was built before the release of the Barton core? Certainly desktop board were unable to, unless that was because of the 333fsb change rather than the core being the problem? Your laptop would use the Barton at 266fsb rather than 333fsb that a desktop would.

Hmm... Powernow works on the Laptop, but does not work on the PC. This mean RMclock won't work on a machine without powernow?



Ah, I see. I thought states between min and max would be valid. I guess we may high a multipier and voltagae that does not work. So RMclock just goes and adjust the voltage and multipler linearly where powernow and crystalcpuid use a steper table (and crystal has only 3 steps).



No, but I have seen several post from people who tried it. All reported that the cpu works, but powernow fails. The board also toped out at 15x, so the highest you can go is 133x15 or a bit below 2 Ghz (a 2600+).

Not sure what would happen if I plug this into my desktop MB. I have been told by others that it should recognize the mobile XP.

Paul

The multiplier changes might work but i doubt the voltage changes will. What board do you have to test the chip with?



I'm not exactly sure how RMclock works? It has a max setting and min setting and seems to vary speed between them but it must have to put the voltage up to a safe level before making changes the multi?

Having used both programs i would go with crystalcpuid over ANY other speed varying software. When i first used it i found it a bit unstable but the latest builds have been very smooth indeed.



Mobile XP's work in most desktop motherboards but probably won't detect it as a mobile, i believe only DFI motherboards will do that.

Even if it doesn't work i'm sure you would be able to sell the chip on and only lose a small about of money.

The motherboard on the desktop is a Asrock K7-Upgrade-880. It's a strange beast, but it seems to be more stable than the DFI board I used to own (an earlier generation of DFI that were not so reliable as today). Unfortunately, I read the manual today and notice that the timing is set by jumpers. This probably means I can't test Crystal or RMClock on it.

I googled and found a few people that have attached mobile cpu to the machine and said that it works. Asrock flatly denies that it works though. I guess I'll give it a try. If it doesn't work, I'll try it on the laptop. If that doesn't work, I'll just sell the cpu.

Paul

Where can you get CyrstalCPUID?

Thanks,
Chris

Google. It's Japanese (the program will be in English, though) in case you skipped the first link.

Don't forget, to run it at startup, put a shortcut to the program and add the arguments ' /HIDE /CQ' at the end of the target. Copy to Start Menu > Startup.

As a followup, I popped the Mobile Athlon XP 2600+ into the Asrock k7Upgrade-880 desktop. To my disappointment, the MB did not recognize the cpu. It said the cpu was a "Mobile Athlon XP", but not which one. Various people on the net said that the board recognize the cpu when they actually mean that the board recognize the cpu as an Mobile Athlon. Boy, you can't trust information on the net these days, what's this world coming to.

The lack of support left FSB at 100 Mhz and 6x multiper, but at least the computer didn't POST. I attempt to set the FSB and multiper with the jumpers, but apparently Asrock didn't come with jumper caps. Booo. Hsss. There were jumper cap for FSB, so I set it to 133 Mhz and the CPU jumps to 800Mhz. Benchmark indicate that the processor was slower than my Duron.

I ran RmClock and notice that the PowerNow was now active! Previously, I ran RmClock with the desktop processor and it had no options to set. I played around and notice that the highest multipier I can go is 15x, which mean the mobile had a multiplier cap. I bump the multipier to 15x and the processor jumped to 2 Ghz. I ran the benchmark, this time the performance was much better than the Duron. I then set a min and max of 1.45v/3x and 1.45x/15x. The cpu drops to 400 mhz when idle and increase to 2 Ghz under load. Heat output at 2 Ghz seems to be a little bit lower than the Duron.

The only downside is that RmClock only runs in admin mode. I haven't figure out how to run it as admin yet.

I'll experiment with the setup for a few days and transfer the cpu to my laptop, but apparently, you don't need laptop circuit to take advantage of the p-transitions. I am curious to see how low I can go at idle.

Thats great news paulsiu! I don't think it really matters about the name much, you know how fast it is.

I haven't upgrade the cpu on my laptop yet, but I experimented with RmClock on the laptop with the stock cpu. The laptop contains Powernow, I take a reading on the clock, and it read:

Idle Full Power
Athlon XPM 1500+ 1.2v / 530mhz 1.45v / 1325 mhz

I switched off Powernow and run Prime95 and started to reduce the voltage while remaining at the full 10x multipler. When the voltage reached 1.2v, the machine crashed. I rebooted the machine and started it at 1.225v and it ran without any instability.

I started lowering the multipler with Prime95 running to see how low I can go. I got it down to 4x. If I try 3x, the machine would instantly lock. Now that I am at 4x, I started lowering the voltage. I was able to lowering the voltage down to 0.925v (the lowest you can go on this board) and still have a stable system at 4x.

If CPU power is to be believed, an Athlon XP T-bred will generate:

Power
1.45v/1325mhz 37.6W
1.2v/530 10.3W

The above would the be lower and higher power rating from Powernow. With Rmclock, I should have the following power level:

Power
1.225v/1325mhz 26.9W
0.925v/530mhz 6.1W

Assuming that I installed the Athlon XP-M 2600+ Barton into the laptop, I should get the following characteristics.

Power
0.925v/530Mhz 5.9W
1.45v/2000Mhz 55W

I will run some battery test to see if the battery saving is actually working first.

Paul