For those of you that give a damm, here are a few numbers from a college project I did a while ago. These temperatures were taken using a digital thermometer placed directly onto the die (cap removed) so are probably more accurate than those using mobo LM75's etc which usually read 3-4 degrees too low. Only a standard heatsink/fan was used. I clocked the FSB back to 66Mhz so I could get a wider range of results. At the time, I was running the chip, a K6-2 300, at 380Mhz (95 x 4) on a TMC AI5VG+ with 64MB RAM, ATI All-In-Wonder Pro, Voodoo2 SLI, Maxtor 8.4GB HDD. The chip was put under 100% load by clean booting and running NBench, a simple MOPS-based benchmark in a loop for 5 minutes.

Clock_____FSB__________VCore___Nbench__Temp.
(Mhz)____(MHz)__________(V)__(FPU MOPS)_(C)

133_____(2 x 66)________2.2_____21.2____37.6
166_____(2.5 x 66)______2.2_____26.6____43.3
200_____(3 x 66)________2.2_____31.9____46.3
233_____(3.5 x 66)______2.2_____37.3____49.4
266_____(4 x 66)________2.2_____42.6____52.8
300_____(4.5 x 66)______2.2_____48______57.9
333_____(5 x 66)________2.2_____53.4____61.2
366_____(5.5 x 66)______2.2___LOCKUP IN WIN98
400_____(6 x 66)________2.2_____DID NOT BOOT

200_____(3 x 66)________2.0_____31.9____39.9
200_____(3 x 66)________2.1_____31.9____43.2
200_____(3 x 66)________2.2_____31.9____46.3
200_____(3 x 66)________2.3_____31.9____49.4
200_____(3 x 66)________2.4_____31.9____53
200_____(3 x 66)________2.5_____31.9____57.2
200_____(3 x 66)________2.6_____31.9____60.8
200_____(3 x 66)________2.7_____31.9____65.6
200_____(3 x 66)________2.8___LOCKUP IN WIN98
200_____(3 x 66)________2.9___HA HA! NO WAY.

As you can see, for the range of settings used, temperature scales more or less linearly with both clockspeed and voltage. It's the voltage results that suprised me the most: 0.1V can mean over 5C temperature difference.

For those of you pushing 50-55C with your K6-2's....don't worry, there's quite a bit of headroom left. I think most parts are rated at 60 or 70C by AMD anyway. I'll leave you to draw your own conclusions. I don't know how useful this data actually is, but I just thought that since I took my time over doing it, I might as well share it with you guys.

Mutha Funker

P.S. I've now just finished burning in a K6-2 550. I'll let you know how I get on. Previous attempts have been futile (only 95 x 6 @2.6V). Does anybody get the feeling that the 550 part is just a rebadged 500 rushed onto the market just to eat Celery? Most 500/533's go to 570 easily.



[This message has been edited by Mutha Funker (edited 06-08-2000).]