Table of Contents
● Introduction
● Measurements
● ASSY 250407
● ASSY 250425
● ASSY 250466
● ASSY 250469
● Conclusion
● Heatsinks
● RF shield

Introduction
In forums, Youtube and social media, every now and then failing "hot chips" are mentioned. In the C64, some chips are surpriosingly hot, but still in working working condition. So, how hot is hot? Time to get the meter on the bench and find out.

I do have a (thermal) FLIR camera with a nice resolution, but for the measurement, I prefer my dual channel thermometer (UNI-T UT320D) and a surface proble (a K-type thermocouple). Measuring accurate temperatures with an FLIR camera requires setting the exact emissivity of the component to be measured, which depends on the material. So FLIR cameras are good for finding hotspots, even unexpected ones. The accuracy of the thermometer is +/-(0.5% + 1°C).

Before the measurements, I have run the C64 for at least one hour with the dead test cartridge to allow the temperatures to settle and to provide equal conditions.

The interesting value is actually the temperature rise, which means how much warmer the IC is than the ambient. Hence, the first value taken was the ambient temperature.

I have measured the temperature at the center of the top surface of the chip, waiting until the displayed value didn't change anymore, which took maybe 30 seconds for each chip.

I have performed this measurement for every IC and relevant semiconductor in all mainboard versions, that I own. The mainboards were unmodified in the original state and no heatsink was attached to the VIC-II or any other IC.

The power supply used in all tests was one of my C64 PSU Combi. The transformer was a 10VA model (that is relevant for the temperatures of the internal 12V and 5V regulators - 7812 and 7805).

Measurements

ASSY 250407, PAL, S/N WG A 22471
The ambient temperature was 22,3°C

Component	Type	Note	Temperature
U1	6526	CIA 01/83	37,2 °C
U2	6526	CIA 50/87	38,3 °C
U3	2364	ROM (BASIC) 901226-01	47,3 °C
U4	2364	ROM (KERNAL) 901227-02	47,2 °C
U5	2332	ROM (Character) 901225-01,12/83	39,3 °C
U6	2114L-20	1kx4 Bit (Color-RAM)	38,3 °C
U7	6510	CPU 13/83	48,8 °C
U8	7406	Hex Inverter (o.c.)	33,9 °C
U9	4164	DRAM 64Kx1 Bit (200 ns)	31,8 °C
U10	4164	DRAM 64Kx1 Bit (200 ns)	32,6 °C
U11	4164	DRAM 64Kx1 Bit (200 ns)	32,5 °C
U12	4164	DRAM 64Kx1 Bit (200 ns)	32,6 °C
U13	74LS257	Multiplexer	30,8 °C
U14	74LS258	Multiplexer	28,8 °C
U15	74LS139	Demultiplexer	29,5 °C
U16	4066	Quad Analog Switch	27,1 °C
U17	N82S100N	PLA "C64A", 33/82	38,6 °C
U18	6581	SID 49/82	53,9 °C
U19	6569R1	VIC II (no heat sink) 10/83	58,9 °C
U20	LM556	Dual Timer	29,7 °C
U21	4164	DRAM 64Kx1 Bit (200 ns)	31,5 °C
U22	4164	DRAM 64Kx1 Bit (200 ns)	32,6 °C
U23	4164	DRAM 64Kx1 Bit (200 ns)	32,5 °C
U24	4164	DRAM 64Kx1 Bit (200 ns)	32,0 °C
U25	74LS257	Multiplexer	29,6 °C
U26	74LS373	Octal Latch	31,3 °C
U27	74LS08	Quad 2-input AND-Gate	30,6 °C
U28	4066	Quad Analog Switch	26,8 °C
U29	74LS74	D-Flip-flop	31,1 °C
U30	74LS193	Binary Counter	33,8 °C
U31	74LS629	Dual VCO	38,1 °C
U32	MC4044	Phase-frequency Detector	33,8 °C
VR1	7812	12V Voltage regulator	45,3 °C
VR2	7805	5V Voltage regulator	45,3 °C
CR4		Bridge Rectifier	50,6 °C
Q3		Transistor	27,3 °C

ASSY 250425, PAL, S/N WG A 367728
The ambient temperature was 21,9°C

Component	Type	Note	Temperature
U1	6526	CIA, 05/86	39,5 °C
U2	6526	CIA, 05/86	39,3 °C
U3	2364	ROM (BASIC) 901226-01, 03/86	42,2 °C
U4	2364	ROM (KERNAL) 901227-03, 02/86	40,0 °C
U5	2332	ROM (Character) 901225-01, 02/86	38,6 °C
U6	2114L-20	1kx4 Bit (Color-RAM)	33,5 °C
U7	8500R3	CPU, 05/86	42,9 °C
U8	7406	Hex Inverter (o.c.)	31,7 °C
U9	4164	DRAM 64Kx1 Bit (200 ns)	29,4 °C
U10	4164	DRAM 64Kx1 Bit (200 ns)	30,7 °C
U11	4164	DRAM 64Kx1 Bit (200 ns)	30,6 °C
U12	4164	DRAM 64Kx1 Bit (200 ns)	30,0 °C
U13	74LS257	Multiplexer	29,1 °C
U14	74LS258	Multiplexer	28,7 °C
U15	74LS139	Demultiplexer	28,4 °C
U16	4066	Quad Analog Switch	26,2 °C
U17	906114-0	PLA 906114-0, 03/86	41,0 °C
U18	6581R3	SID, 06/86	46,4 °C
U19	6569R3	VIC II (no heat sink), 05/86	58,2 °C
U20	LM556	Dual Timer	28,8 °C
U21	4164	DRAM 64Kx1 Bit (200 ns)	29,3 °C
U22	4164	DRAM 64Kx1 Bit (200 ns)	29,4 °C
U23	4164	DRAM 64Kx1 Bit (200 ns)	30,0 °C
U24	4164	DRAM 64Kx1 Bit (200 ns)	30,1 °C
U25	74LS257	Multiplexer	29,3 °C
U26	74LS373	Octal Latch	30,7 °C
U27	74LS08	Quad 2-input AND-Gate	27,9 °C
U28	4066	Quad Analog Switch	25,7 °C
U31	8701	Clock Generator	38,1 °C
VR1	7812	12V Voltage regulator	41,5 °C
VR2	7805	5V Voltage regulator	39,0 °C
CR4	DBA20B	Bridge Rectifier	42,0 °C
Q1	D880	Transistor	30,1 °C

ASSY 250466, PAL, S/N DA4 162554 B1
The ambient temperature was 22,0°C

Component	Type	Note	Temperature
U1	6526	CIA 49/86	38,3 °C
U2	8521	CIA 48/86	40,4 °C
U3	2364	ROM (BASIC) 901226-01, 50/86	43,0 °C
U4	2364	ROM (KERNAL) 901227-03, 49/86	43,0 °C
U5	2332	ROM (Character) 901225-01, 34/86	42,3 °C
U6	2114L-20	1kx4 Bit (Color-RAM)	37,3 °C
U7	6510	CPU 50/86	45,1 °C
U8	7406	Hex Inverter (o.c.)	32,9 °C
U9	41464	DRAM 64Kx4 Bit (200 ns)	31,0 °C
U10	41464	DRAM 64Kx4 Bit (200 ns)	32,2 °C
U13	74LS257	Multiplexer	29,1 °C
U14	74LS258	Multiplexer	31,4 °C
U15	74LS139	Demultiplexer	30,0 °C
U16	4066	Quad Analog Switch	28,1 °C
U17	906114-0	PLA 906114-0, 50/86	43,8 °C
U18	6581	SID 37/84	55,7 °C
U19	6569R5	VIC II (no heat sink) 52/86	58,2 °C
U20	LM556	Dual Timer	29,5 °C
U25	74LS257	Multiplexer	28,9 °C
U26	74LS373	Octal Latch	33,8 °C
U27	74LS08	Quad 2-input AND-Gate	29,8 °C
U28	4066	Quad Analog Switch	26,8 °C
U31	8701	Clock Generator	41,3 °C
VR1	7812	12V Voltage regulator	43,2 °C
VR2	7805	5V Voltage regulator	41,3 °C
CR4	DBA20B	Bridge Rectifier	38,3 °C
Q1	D880	Transistor	31,0 °C

ASSY 250469, PAL, C64G, S/N DA4 362455
The ambient temperature was 22,0°C

Component	Type	Note	Temperature
U1	6526A	CIA 14/90	31,3 °C
U2	6526A	CIA 14/90	33,1 °C
U3	74LS08	Quad 2-input AND-Gate	28,6 °C
U4	23128	ROM (KERNAL/BASIC) 251913-01, 15/90	34,9 °C
U5	2332	ROM (Character) 901225-01, 12/90	34,6 °C
U6	8500	CPU 14/90	36,3 °C
U7	8565R2	VIC-II (no heatsink) 13/90	43,0 °C
U8	252535-01	Super PLA, 09/90	27,6 °C
U9	8580R5	SID 11/90	42,0 °C
U10	41464	DRAM 64Kx4 Bit (200 ns)	33,0 °C
U11	41464	DRAM 64Kx4 Bit (200 ns)	31,5 °C
U18	4066	Quad Analog Switch	26,5 °C
U19	2114L-20	1kx4 Bit (Color-RAM, other PCB revision)	29,2 °C
U20	8701	Clock Generator	40,3 °C
U21	4066	Quad Analog Switch	27,5 °C
U22	7406	Hex Inverter (o.c.)	31,4 °C
U23	74LS14	Hex Inverter	30,9 °C
CR2	DBA20B	Bridge Rectifier	26,8 °C
Q5	D789C	Transistor	36,0 °C
Q1	D880	Transistor	31,8 °C

Conclusion
1. VIC-II
is the hottest chip on board, the latest version on ASSY250469 is running much (~15°C) cooler, though.

2. SID
It is a hot chip, as well. R3 runs ~9.4°C cooler, than the first model. Again, the short board/ASSY250469 SID is the coolest.

3. ROMs (KERNAL, BASIC, CHAR)
The ROM chips are surprisingly hot, especially in ASSY250407. The short board ROMs are much cooler.

4. CIA
There are three different types of CIAs, the 6526, the 8521 (as user port CIA in some ASSY250466, causes diagnostics to fail) and the 6526A in the short boards. The temperature of the first and second are approximately the same, the 6526A runs about 7°C cooler.

5. CPU
There are three types of CPUs again. the 6510 runs pretty hot, the 8500R3, which I have found in ASSY250425 is about 5°C cooler, than the 6510. The 8500 on the short board is again the coolest CPU.

6. PLA
The early PLA N82S100N runs significantly cooler than the Commodore 906114-0, which is another warm chip. The short board Super PLA is coolest.

7. Voltage regulators 7812 and 7805
The voltage regulators are supplied directly from the 9VAC of the power supply, which is rectified and conditioned in a way to serve as an input voltage for those regulators. The 9VAC transformer is not a regulated device. It will output 9VAC only at nominal load. Below this nominal current, the voltage is higher. This means, the voltage regulators have a higher voltage drop to their output voltage, which means, the dissipated power is also higher and as a result, the regulators get warmer. It is a known fact, that the old boards, which don't have a heatsink for the 7805 will crash after a while (the 7805 is shutting down) with some modern PSUs. Thus, it is strictly recommended to use power supples with a transformer not "stronger" than 10VA. More isn't better here.

8. ASSY250469
The short boards are known for their durability. Beside being younger, the fact, that they run much cooler is responsible for that.

Download my Excel Data.

Heatsinks
It is a known fact, that cooler components have a longer life. According to the application report "Calculating Useful Lifetimes of Embedded Processors" by Allan Webber, an increase by 10°C will approximately half the lifetime of a semiconductor. Hence a decrease of the inner (junction) temperature by 10°C will approximately double the lifetime.

The goal of the next measurements was to find out the effect of heatsinking the VIC-II and other ICs.

A VIC-II (6569R3) on an ASSY250425 board was prepared with a thermocouple on the center of the bottom surface. It was secured in the described location with kapton tape (a heat resistant tape., which is often used for solder masking etc. in electronics manufacturing).

As always, it is interesting to determin the temperature increase above ambient temperature. Hence, a 2nd K-type thermocouple was connected to the thermometer to allow sim,ultaneaous measurment of both temperatures.

It seems to be pretty popular to glue several small and cheap heatsinks to the ICs. I tend to doubt the benefit of doing so. There is a bad thermal connection between adjacent heatsinks, so the middle heatsink is cooling the hottest part (the center of the IC, where the die is situated) and the two other heatsinks are cooling the ends of the IC, which are less warm. The thermal coupling between the heatsinks can be assumed as poor.

To prove this method, I have first attached such a small heatsink to the center of the IC.

The temperature rise with one small heatsink is 39.2°C. The small heatsink in the center is "worth" 1.3°C (40.5°C - 39.2°C).

The temperature rise with three small heatsinks is 37.6°C. This method is worth 2.9°C (40.5°C - 37.6°C). Since the central heatsink provides 1.3°C, each of the additional heatsinks is worth only (2.9°C-1.3°C)/2 = 0.8°C. This shows, that this way of attaching heatsinks is not really effective.

The better way is to attach one DIP40 sized heatsink, since it will spread the temperature of the hot center of the IC over a bigger area.

The temperature rise with one DIP40 sized heatsink (Fischer ICK40B, RTH=15.8K/W) is 35.5°C. This method is worth 5°C (40.5°C - 35.5°C).

Fimally, the efficiency of the original cooling solution, the shield/cover should be determined. The original RF-shield was attached, adding some fresh thermal paste to the surface of the VIC-II.

The RF-shield (this kind) is worth 40.5°C - 38.9°C = 1.6°C, which is even worse than three little heatsinks. Originally, there wasn't even thermal paste on the VIC-II.

There are different ways to attach heat sinks to an IC or transistor. a thin layer of thermal paste and screwing or clamping might not work with DIP ICs. So, it needs to ble glued in some way. There is special thermal conductive heatsink glue, which is probably very effective, but permanent. A heatsink tape (doublesided adhesive heat conductive tape) is ok, but usually less effective than the glue. The thermal conductivity of the glue is approximately 3 times higher than of the tape. The tape can be removed if desired, though.

Attaching the heat sink with any other glue or double sided tape is not a good idea, since those materials are not made for this purpose and less heat conductive.

Actually, calculating a heatsink and its effect on the junction temperature is simple math. You need to know the power dissipation inside the chip and the thermal resistances junction-case (RTH J-C) and junction-ambient (RTH J-A). You could find those parameters usually in the data sheets of the ICs. The MOS Technology INC. data sheets are lacking that information.

just multiply the dissipated power (Pd) with the total thermal resistance (Rth tot). The thermal resistance has the unit K/W (Kelvin per Watt). a differnce of 1K equals the differnce of 1°C, so no conversion required. In case a heatsink is attached, add the thermal resistance of the heatsink to RTH J-C (junction case). It is really that simple!

The thermal resistance of that DIP40 heatsink is RTH=15.8K/W. A better heatsink would have a lower value. The problem is, that those heatsinks might be higher, which is definitely a problem in a C64C case. A wider heatsink with more fins might be good, it would fit on ASSY numbers newer than ASSY250407, because they don't have a cage around the VIC-II circuit.

The short board VIC-II (8565) and the short board SID (8580) are running much cooler, because they dissipate way less power (Pd). As a result, a heatsink will make much less difference for the junction temperature. It is in my opinion not benifical to put a heatsink on those.

We have found a temperture decrease on the bottom of the VIC-II (6569R3) of 5°C. This will also be the temperature decrease of the junction. We also learned that a rule of thumb says, that a decrease of 10°C will approximately double the lifetime of a semiconductor.

The factor for the lifetime can be calculated like this:

For a decrease of 10°C, the factor is 2 (of course), for 5°C it is 1.4. The DIP40 heatsink we have attached to the VIC-II is adding 40% to the lifetime.

I think, that attaching a heatsink to an IC, that is 50°C or hotter is helpful. Maybe even the hot old ROMs and the CPU (older date codes, like in the ASSY250407) could be benefical. Attaching one to a less warm IC (like the CIAs) is rather decorative, than helpful. It does no harm, though.

What helps most is decreasing the ambient temperature of the ICs. The easiest thing is to remove those ugly cardboard RF shield. Moving the hor air out of the C64 is a good thing, too. A fan produces interferences, so the 5V for driving the fan should be filtered and protected with a diode (to prevent voltage spikes while switching the power on and off). Nevertheless, I think, not producing fan noise is part of the charm of the C64. A small fan, that fits the case will always be noisy.

RF shield
The RF shield has always been suspicious to produce excessive heat. After once again removing the heat sink from the VIC-II (6569R3) in ASSY250425 (S/N WG A 367728), I have attached a thermocouple (K-type) to said IC (in the middle of the top surface) and fixated it with kapton tape. The leads went out through the expansion port. a second, identical thermocouple was left outside the case, far enough away from the C64 todisplay the ambient temperature. After running dead test for about an hour, I have measured the temperature of the chip with an open top half of the case.

The ambient temperature was 21.8°C, which means a temperature rise of 40.6°C.

Now, only the top case with keyboard was put on top of the case.

Finally, the RF shield was installed and the case was closed.

Description	VIC-II temp	Ambient temp	Temperature rise	Delta T	Lifetime factor
open lid	62.4°C	21.8°C	40.6°C	-12.2°C	233%
closed lid, no RF shield	68.6°C	22.4°C	46.2°C	-6.6°C	158%
closed lid, RF shield	75.1°C	22.3°C	52.8°C	0.0°C	100%

The temperature rise is the IC temperature minus the ambient temperature. Now, this value is compared to the temperature rise ("Delta T"). The lifetime factor is calculated according to the previously shown temperature.

Compared to the original state ("closed lid, RF-shield installed"), removing the RF-shield is a huge plus (58%) for the VIC-II. "open lid" is not really an option. This calculation is for the VIC-II only, since all ICs are roasted under the RF shield, the actual gain in lifetime is higher.

Removing the RF shield is a very effective method to increase the lifetime of a C64! It is recommended to always get it out. The C64 does not meet any modern EMC rules, anyway.

[...]