David Alvarado's Dual C366 BP6 Adventures
Agg Note: This is a collection of 4 historical articles from 1999, which I have converted to the new article system and placed together. Some pics would have been nice, but not everyone had a digicam in 1999. :)
31-Aug-99 - David Alvarado shared his experiences with C366's (originally posted to alt.comp.hardware.overclocking):
If anyone cares, here's how I'm doing with 9 untested Cel 366 week23 Malays I've been messing with.
GW 25603-32 fansinks
bh6 and bp6 boards
inwin a500 cases with 80mm intake fans
prime95 torture for testing (couple hours)
ambient temp 84-89F (garage)
cpu id# speed Vbefore sanding Vafter sanding
...70 550 2.1 2.1
...71 FAILED at 550 at up to 2.4V with 100% copper/alum sanding both cpu
...72 550 2.0 2.0
...83 550 2.1 2.0
...85 550 2.0 2.0
...86 550 2.2 2.1
...87 550 dunno 2.1
...88 550 2.2 2.1
...89 550 dunno 2.1 (did 2.0 for a long
Most all of these were tested in the bh6 systems. It looks like sanding helped bring down the voltage in about 3 of 4 cases where I actually checked for it. All cpus except the one that totally flailed where sanded approx 25% copper. Interesting to note the serial number directly before and after the "failed at 550" chip did 550 at 2.0 and 2.1V. Overall unscientific success rate at 550 = 88%, with sanding. Avg case temps in the bh6 systems were around 36-38C in this hot room. The 2.0 cpus are all going in the bp6 boxes.
These machines seem to require a little more effort that the bh6 machines.
I'd appreciate any input on the bp6 stuff. Thanks.
- First I used nylon washers as spacers to get around the mobo grounding problem with the inwin a500 case. This seems to work okay.
- I noticed that the Fan1 and Fan2 headers are located such that the Fan1 connector is closest to the CPU2 socket and visa-versa. I kinda wish the fan connectors were closest to thier respective CPU sockets, but whatever.
- The bios I have doesn't have a RPM readout for the case fan. I couldn't figure out how to get the FAN3 rpm in MoboMonitor, either. Anyone know if a bios fix is coming?
- In MBM4, It took me a while to figure out that the sensors need to be Winbond2, Winbond1, and then Winbond3 to get the app to display CPU1,CPU2, and Case Temp in that order.
- MBM seems to report the wrong Vcore for CPU2 (1.5 Volts??).
- In bios, my CPU2 temps are consistently higher that CPU1. It looks like others are reporting this "problem" regardless of single/dual cpu utilization. Hmmm.
- Like other people, My case temps are *way* too high. I guess the case temp is actually the temp of the green sink on the bx chipset part. I used a large house fan which seems to help a lot. I'm considering more aggressive cooling in these bp6 machines. The air exhasting from the p/s fan seem
noticeable warmer- even with 1 cpu installed, than the bh6 machines. Has anyone replaced the green sink with a larger heatsink/fan?
On 02-Sep-99 he sent us some followup info:
Last night I yanked the green sink on the first bp6. I used needlenose pliers to gently twist and pull the white push tabs out from the top side of the mobo. They came out undamaged without too much trouble. I put my finger on the naked chip and fired her up. The chip got pretty damn hot within just a few seconds. It was hot enough that it hurt to touch it for more that a few seconds at a time. I wasn't expecting that kind of heat. I sprayed some canned air directly on the chip at close range. The air came out all frosty and I observed a distinct melting pattern in the shape of a square about 10mm on a side, directly in the center of the chip. I also noticed that the sink is totally not flat (no I didn't sand it). I fixed a 40mm x 10mm 486 fan to the sink, which fit and screwed in perfectly. I applied heatsink compound to the chip and reattached the green sink/fan combo to the board using the original spring-loaded pushtabs. I pressed down on the fansink to squish out excess compound. I connected the power cable and fired her up.
MBM 4.10 reported about the same case temp as before ~45C once everything warmed up with the cover on. The other bp6 machine (unmodifed) was reporting the same temp +/- 1C in the same room. What *did* change was that the machine appeared to me more stable. At 2.1V/2.1V, this machine would get rounding errors in prime95 torture, usually within an hour. Last night after the mod, it ran all night for the first time with no errors at the same voltage. This AM temps were: CPU1: 39, CPU2: 45, Case: 45.
I do seem to recall my Case temps going as high as 50 something a few days ago, but it was much warmer weather at that time.
Next, I'm going to try to modify the inwin a500 case to increase intake fan air flow to bring down the temp. Also, I'm thinking of modifying the powersupply cover to reorient the air holes to the bottom instead of on the side where they are now. I suspect that the air flow caused by the exhaust fan in the ps is competeing with one of the cpu fans. It would be nice if I could rig a 80 or 90 mm fan horizontally with baffles directly above the ps and have it exhaust air out of the holes at the top of the case on the back side. In retrospect, another case with provisions for a second exhaust fan might have been a better choice for this space heater.
On 03-Sep-99, he sent a further followup:
I decided to ditch the inwin a500 midtower for a Q500 fulltower. This will allow me to put an addtitional 92mm exhaust fan up above the ps and should also improve cooling to cpu2 since this case has the ps oriented horizontally up above and not "hovering over" the atx mobo (as it should be). I noticed that the intake holes on the front panel are larger in the q500. The vertically mounted ps in the a500 midtower had its intake vents
only 20mm from the cpu fan and directly opposing the cpu fan air flow.
Removing the ps and checking cpu rpms seems to confirm my concerns of poor air flow at cpu2. The new case should improve all this and with the additional fans, improve my cpu2 and case temps.
I also confirmed the exact location of the "case temp" sensor. The picture of the mobo here calls out the "Themal sensor" on the left. You can't really see it in the pic, but if you have a look on your bp6 at the area there directly between the two square groups of 4 caps you'll see a tiny little surface mounted device labeled "RT3". I pointed the little red straw tube on a can of compressed air directly at that puppy and gave her a couple good frosty blasts while the machine was running. The case temp reported by MBM4.10 immediately fell down from 45ish to 16C! Looking at the location, its not hard to figure why this device doesn't normally get much airflow. I'm putting a fan on the area in
addition to the fan on already on the green sink.
I get the feeling that other bp6 owners already figured this out, but it seems that I wasn't the only one who wasn't clear on where the sensor was. I'll let you know how the case swap goes. I'll try to get photos too.
Peace from Newport Beach, CA.
The final update came on 08-Sep-99:
Since that post I've done some work on the system. The machine is now housed in a modified Inwin q500 full tower. I've cut the sheet metal "grille" out in the front of the case, sealed the intake area with weather stripping and fitted a 92mm Sunon 3-wire intake fan. I have another 92mm on the exhaust vent above the p/s. The exit vent is not modified - the inefficiency of the sheet metal holes doesn't seem to be as crucial here. I have a third 92mm mounted parallel to the mobo about 4 inches above the surface. blowing on the board between the green chipset sink and the case temp sensor RT3 (which I've precisely located now). This fan is simply circulating air in the case to prevent hot/cold spots and to even the case temps out to provide a more "meaningful" Case temp RT3 reading. Testing showed that the area around the RT3 sensor was stagnant air-wise before the changes. Not difficult to see why - the RT3 sensor gets hot exhaust air from the CPU2 sink, but it is kinda blocked from cold intake air by the AGP card. I actually use a business card to shield the RT3 sensor from the cpu exhaust and the case temp dropped a few degrees almost immediately. I still have the 40x10 mm 486 fan on the green sink, which along with heatsink paste, did help stability, although it may not be necessary now that the larger fan is nearby. The p/s fan is the stock 80mm which is exhasting along with the 92mm above it. One thing of interest is that the p/s fan exhaust air is considerably warmer that the exhaust air coming from the 92mm fan. The power supply heat is surely a factor as is proximity of the cpus to the power supply air intake vents. It might be a useful to remove the 80mm power supply fan and fit a(nother) 92mm fan on the power supply intake vents above the cpus for increased cooling. My gateway PC here at work has its power supply fan on the outside of the actually power supply but inside the pc case, sucking air into the power supply in this fashion.
My temps running two instances of prime95 under win2k, one pegged to each processor, are much better now. Case temps with the covers on in my warm garage, are below 40. Typically 36C. CPU temps are fairly close to each other at around 38 when warmed up. The machine is now totally stable at 550+550/2.1V, 2.1V. Its actually not that loud, thanks the the Sunons. This machine is actually for my brother. I have a 2nd machine that I'll finish after this one that will be for me.
Peace from Newport Beach, CA.
Of course, it has to be said - cooling the temperature sensor can lead you to falsely believe you have lowered your in-case temperature. However I think David's got the right idea - certainly bad airflow can be a killer on a highly overclocked dual system. Thanks again for the info, David! - Agg
These articles originally appeared here, here, here and here.
All original content copyright James Rolfe.
All rights reserved. No reproduction allowed without written permission.
Interested in advertising on OCAU? Contact us for info.