Solved - overclocking the HP 2140

[Twain]

Hi there

Finally I got my HP 2140 HD running with higher clock frequencies 8) 8) .

As already figured out by other people in the Internet (many thanks here to Abo from SetFSB ) the built in clock generator is hardware locked not to allow software overclocking in default configuration.

I disected the netbook and prepared it for some measurements in the lab:



For configuration of modification rights the clock generator reads in the state at pin 4 during start-up of the netbook. As can be seen on following signal graph over subsequent time a high voltage level (clock modification not allowed) is applied to the pin till the pin begins to act as a PCI clock output with 33 MHz frequency:



I switched off the HP and measured the resistance being present on the pin 4 of clock generator versus GND and VDD. The resistance was completely given by the resistor in the red box on following picture pulling the pin softly to VDD (sitting on the bottom side of the PCA).



I desoldered the resistor and moved it to the green boxed location. After I started the machine and took the measurement again:



As you can see, the configuration level is low now (clock modification allowed) . Thus I put the stuff together and booted the netbook into Windows XP. I configured SetFSB to a reasonable level and performed a first Passmark 7 benchmarking. See the results:



I'm going to test out the potential on both my 2140 HD netbooks during the next days. So far the situation looks promising, with displayed overclocking setting the netbook play 1080p almost well.

Cheers from Switzerland
Twain

[Warder]

Awesome, thanks for sharing Twain!

So does that jumper you de-soldered basically lock the mini to a clock speed, while the other one unlocks it?

[Twain]

Hi again

The 10k resistor does maintain the access rights within the clock generator chip. If tied to VDD then the chip won't allow overwriting its internal clock registers. If tied to GND the clock frequency can bed adjusted with software tools as SetFSB etc.

Please be aware that the resistor is really tiny (0402 form factor) and it requires a certain degree of experience with handling of microelectronics. You need a supersmall solder tip and some very thin solder wire. Otherwise the mission will fail...

I also experienced that shifting the gears down from high clock frequencies will lead sometimes to a system crash if both FSB and PCI clock was modified. This can be prevented by adjusting only the FSB what is adequate for all the relevant applications.

Today I performed a Passmark 7 benchmark on the second machine with impressive result:



I also tested the 1080p capability. With a little more throttle (approx. 1.9 GHz) the 1080p footage is played with no stuttering at all!

Enjoy!

Twain

[mikez]

Often times it is easier to have a brand new resistor on hand - -
then the old one can just be clipped rather than removed and reused.

Also note that the new one does not have to be a surface mount package - there looks
to be plenty of board room for a tiny, leaded, resistor - with a spot of hot glue for body support.
Just don't go wild with the lead length, there is a high frequency signal at this point.

Even so, having the proper sized tools is a "must have" for this job.
(Plus great eyesight, a steady hand, nerves of steel if this is your first job . . . . )

[Twain]

Using a brandnew part is definitely a good idea even if I'd have serious doubts about wired parts since in any case the wire legs would apply a certain amount of mechanical force to the tiny tiny solder pads (dimension of one pad: 0.5x0.3 mm or about 0.02x0.01 inch!). To clip the old resistor is also not a good idea due to same circumstances...

If you're used to work with SMT electronics then it's not a big deal. Otherwise go and find someone being used with.

Twain

[gregw]

Wow... That's a really cool hack! How was the CPU temperature when clocked at 1.9 Ghz? Even at just 1.6Ghz, the CPU temperature on my 2140 already heats up pass 60 Celsius (monitored using Everest) when working hard..

[mikez]

Should not be a problem - the usual design range for these cpus is 60 - 95 degrees core temp.
When the coffee you spill starts to boil on the NetBook - then you have problems.
- - - -
Personally, I like to see my equipment running with a core temp from 40 to 50 degrees - -
Every 10 degrees hotter approximately halves the life of a silicon device, ratings be damned.
- - - -
A bit of trivia for anyone thinking to finger their heatsink (degrees C):
42 - Uncomfortably warm
44 - Painful
54 - 30 seconds to burn
72 - 1 second to burn
88 - Burn on contact
Human skin is a bit more sensitive than silicon (beach sand) to heat.
Not a surprise to anyone who has walked barefoot on a tropical beach.

[Twain]

Well, here come the requested temperature readings. The ambient temperature during the tests was quite high (about 28 °C).

With normal clock speed (1.6GHz):



With overclock speed (1.85GHz):



Imho the CPU temperature and voltage is not the bottleneck for instability, it's rather the RAM. At the high temperatures now the upper speed limit is @ 1.85GHz for stable operation.

Twain

[Twain]

Just to give you an idea how 1080p looks on the HP 2140 with 1.85 GHz:

http://www.twains.ch/Dolphins_1080P.mpg

It's the dolphins 1080p video from the Microsoft HD swowcase. It's 1080p but anamorphic (only 1440 pix wide). It stutters once or twice very shortly over the whole length of the movie. With higher clock frequency and/or a better codec I see really no problems to achieve a seamless playback. Every kind of 720p works pefectly anyway.

Twain

[reded23]

Thanks for the great instructions ! I can't get much past 1800 with my machine...screen blanks out...
1800 is fine though ;-)