So instead of all the "How far can I OC a Mobile?" posts

[MercenaryForHire]

I'm trying to build an XP-based system that can be passively cooled. There's a direct 92mm exhaust that can pull off the Zalman flower-sink, so I'm wondering what I need to go for here.

The 35W Mobiles seem to be good candidates, as they're already way down the ladder in heat output. Any other contendors? Speed isn't a huge priority as much as cool-running and balance.

I'm also even considering an A64, but doubt that the 185W PSU in the system could handle it.

- M4H

 

[Tiamat]

Any idea what mobo you are looking at? I guess the obvious requirements would be at least 1.0V CPU Voltage offering. and AGP/PCI lock so you can use low FSB after you have lowered the multiplier down as far as it operates at. (Do the mobiles have some low multipliers that they will not function at?)

As for semi-passive setup, A quiet PSU that has a 120mm fan (Seasonic tornado comes to mind) and a 120mm case exhaust (antec fan or similar, or better), and a decently strong 120mm front fan (low volt Evercool 120mm) might be able to cut the cake for ya. Just speculation, but i dont see why not.
133fsb x 8 gives just over 1ghz. I dont know if you want to go below that, I personally havent seen much below 1 ghz.

Another thought, maybe those huge huge tower heatsinks will cool better since the fins are parallel to the air flow?

 

[Dman877]

Downclock a 35 watter to 1.0 ghz and see how low you can take the voltage with full stability. I'm guessing an average 2400+ 35 watt will run at 1 volt at 1 gig.

 

[MDE]

Google for the Socket A wire mod, it'll let you set the multiplier as high or low as you want to, as well as the voltage.

 

[MercenaryForHire]

I've got the PSU and case already set - I've hacked up an old IBM Netvista case and upgraded the PSU to a rubber-mounted 92mm. It's so quiet that a Seagate CudaIV sounds deafening.

I'll google the SocketA mod.

- M4H

 

[Sonic587]

This is an interesting question...I'll try it with my 35W 2400+...I'll post back with results later.

 

[trexpesto]

Originally posted by: Sonic587
This is an interesting question...I'll try it with my 35W 2400+...I'll post back with results later.

Yeah I was hoping for some resluts!

Great thread idea, alot of people will probably want to do this when upgrading to A64, turning their SocketA into the HTPC....

I did the wire trick on my TBred to unlock the low multis so I could UC it. I used frazzled end of my stereo antenna wire, nice small diameter. Now though, I wish I had made it flatter with a hammer, or carved a little relief for it in the socket plastic (not as good because of crumbs), because I think it is tilting the CPU die just a little. Which is booty. When I pulled the HS off to try a new one, and cleaned off the grease, the base had an area burnished a bit, corresponding to the high end of the CPU. This was one of those tough clip-ons though.

 

[Sonic587]

Ok, just tried it for about 15 minutes! Not long, I know. I'll explain why. Here are my results even though they may be pretty much useless:

Hardware:
Thermalright ALX-800 w/AS5
AMD Athlon XP-M 2400+ 35W I stepping.
Albatron KX18D PROII

My goal was 1GHz@lowest voltage possible(1.1V for my motherboard). To my surprise, it booted up and I proceeded to stress it with P95 along with multitasking. I am still running P95 at my underclocked voltages right now. It's on the 7th test currently.

Pic of prime and MBM5 after testing

This was done with absolutely NO airflow to the CPU AT ALL while running the programs. I took off the side of my case to insure that the heatsink/CPU would get the worst conditions I could throw at it (I have better temps with the case side on than off). MBM5 was set to take readings at intervals of 1 second.

As you can see, in the space of about 13 minutes, the CPU core rose 12C to a max of 38C. Very comfortably within limits, but a little too fast. It's well known that the ALX-800 chokes under low-airflow situations, however.

Sadly, I had to stop testing when I decided my passive cooled northbridge was getting too hot for my liking. With the fans started up again, it didn't take long for everything to cool down. On the plus side, the CPU seems to be stable.


I will probably reattatch my NB fan tomorrow and retest. So, take this quick run at face value.

 

[Lonyo]

I tried 10x100 @ 1.1v on my NF7 v2 and Barton M 2500+

Wouldn't boot.

 

[Sonic587]

Ok, I am back with a more more thorough testing :beer:

This time, P95 ran for a full hour.

Pic near the end of testing.

As you can see, the max temp recorded was 53C for the core. It hovered around 51-52C for the majority of testing. Remember, this is in a NO airflow situation with a heatsink that is pretty bad with low speed fans, let alone no fan. I think it did well. :thumbsup:

With some sort of duct that you are talking about, I think you could drop those temps by as much as 10C. That is, if your heatsink is good enough.

Hope this helps.

 

[trexpesto]

Yeah and the passive chiset sink from Zalman sounds good