I'm sure a very few of you may remember back a couple months ago when I was going to try to get 200MHz out of a 486 chip (AMD X5-133) when I was hindered by a poor motherboard (50MHz bus was only 40MHz). So I only got to 160MHz. Well here's the next one!...
Getting an OEM Intel Pentium II 233MHz SL264 Malaysia processor to 500MHz!
Abit BE6-II motherboard based on the i440BX (I may have perfected my RAID mod on this board finally!)
Intel Pentium II 233MHz (3.5x66, 2.8v) SL264 Malaysia (does 370 or so on 2.9/3.0v)
nVidia Riva128 4MB AGP1x video card
One 128MB stick of generic 7ns PC133 (CL2 @ 100, CL3 @ 133).
Enlight 250w PSU
Huge Gateway2000 OEM black heatsink
Even huger aluminum heatsink off eBay
One 118w peltier
One 30w peltier
Two 36w peltiers
As many 80mm fans I need to keep those pelts cool!
I am still waiting for the aluminum heatsink and the two 36w peltiers to arrive (I just ordered them today). I have been experimenting with the removing the case from the processor the last couple days. I've figured out how to remove the case and heatsink without doing any damage (after 3 attempts where I fecked at least one thing each time).
The first time, I got the casing off but the HS was still attached, and there were brackets as on the P3 chips. While removing them with a screwdriver, the screwdriver slipped once, damaging a really curvy twisty trace to a cache chip. I tried using conductive paint to fix it at first to no avail. So then I'm sitting up at 5am this morning soldering a single thread of a PSU molex cable between the two points. I think I did a damn good job considering it was 5am and the chip actually worked upon booting again! Yay fixed it!
Then I got everything together and later this afternoon I decided to mess around a bit more. Since I put the casing back on earlier this morning, I had to remove it again. This time I busted off a tiny capacitor. I decided I didn't really want to fix a SMD cap, so I put it in to test it and it worked. Then when I finally got the casing completely off again I saw that I had busted off quite a large SMD cap, so I just threw it back in the baord to test it and it worked. That's when I decided to try extreme overclocking this thing. Why the hell not? I have really nothing to lose! This chip fries, all the more reason to head out and buy a Cel-T.
I did some temp probing today with the casing off and I found that of course, the core was the hottest (this core is frickin huge! Probably 4-6x as large as my XP). The other hotspot was the cache controller, which is directly behind the core on the reverse of the PCB. I figured most of that heat was coming from the Klamath core itself. I also probed the 4 128k cache chips, all of which read 37*C each, a bit above the ambient temp of 32*C today. These don't normally have heatsinks (same as cache controller). I checked the cache chips, and they are rated as "A7", which I guessed means 7ns, or 143MHz. That means I have a pretty much guaranteed OC to 280MHz core speed, but I already knew that. Since I've had this chip running at about 370MHz, that puts the cache running at 1/2, or 185MHz...must be pretty close to the limits without heatsinks. I'm banking on supercooling them to provide me with an additional 75MHz, pretty unheard of, but hopefully its gonna work.
Anyways, my plan is to put the 118w peltier on the core and clamp it down with the heatsink (I think the clamps will JUST be long enough). Then the 30w will probably go on the backside of the core, on the cache controller. The two 36w peltiers will go on the back side of the cartridge as well, on the cache chips. Hopefully this will cool the cache chips on the front enough to clock a bit as well. If needed, I can always back off the L2 latency (or completely disable it, but I wouldn't want to).
I know about condensation and will worry about that when the time comes. If I find that the heatsinks just can't keep up with the large amounts of heat being output, I have been thinking about possibly welding one of the heatsinks to create channels between the fins (fairly large spaces between fins) and turn it into a waterblock, but I'll do that only if needed.
Do I think this is possible? Not really. I can't find a datasheet on the cache chips, made by NEC, so does anyone know that if I increase core voltage, that the cache will get an increase as well? I think that's how it works on all chips up to the Cel-T/P3-T, but I am not sure of myself.
This project will not be taken under way until later next week when I get the other peltiers. I will be posting pictures later next week as well.
Basically, just something to keep me busy
Intel Pentium 4 2.53B C1 @2.85GHz (stock volts), watercooled @ 27*C load
MSI 865PE Neo2-P (i865PE + ICH5), 1GB PC3200 DDR RAM
ATi Radeon X700Pro 256MB AGP8x
Seagate Barracuda 7200.9 160GB SATA, Maxtor DiamondMax 9 Plus 80GB
NEC DVD-RW (ND-3520A), Lite-On 52/32/52x CD-RW (LTR-52327S)
Microsoft Windows XP Professional SP2 My Heatware | My Beerology | My eBay rating
About the L2 cache chips, I don't think the voltage to them will increase as you increase the core voltage. I know this definitely wasn't the case with my old Slot A Athlon - increasing the voltage only increased the core voltage and not the L2 cache voltage there, as my cache always crapped out above 350MHz, no matter the voltage, but the chip worked like a charm as soon as I increased the L2 cache divider.
I hope you succeed, it looks like something for the record books if it works. Just be careful with that screwdriver near the CPU! Its a miracle its still working