"I was
wondering if it was a larger area of silicon so the heat can migrate out
of the die quicker."

That's not the crux of the matter. Power dissipation is power dissipation.
The suitability for audio use is the same as the suitability for any linear use.

Look at the spec sheet carefully, or actually link me to one. Better yet, I got a random one right here :

http://www.classiccmp.org/rtellason/...ata/irf430.pdf

Let that load in another tab for now. Power dissipation is power dissipation. Most of the time it is largely dependent upon the case style but there are other factors. The sheet in the link lists all three in the family at 75 watts.

I would like to see a spec sheet that gives a different Pd for pulsed and linear operation because if it does (not that I have ever seen one), it miust be figured a different way. For example RF power transistors are sometimes rated at their maximum capability in class C operation. This figure is useless to you for audio, or any linear application.

Now on that datasheet, look at figure 3 on the PDF page 4. Looing at the lines and the increments, it is a linear graph, not logarythmic. The lines drawn in the chart indicate the gain of the device. Ideally, those lines would be perfectly straight for linear use, but that is not achievable in real devices.
Even with the line not starting at zero, absolutely straight would be ideal, it doesn't matter where it starts.

As you can see, the line that represents 125C is the straightest, but you can't just run it it at 125C becasue you would have to derate for power dissipation so much that a pair of these 75 watt devices might get you 10 watts or some ridiculous low power like that. It would be very inefficient to say the least, and it would take class A to keep them that hot anyway.

Now, if you look at a MOSFET that is specifically only designed for switching, the "knee" in that curve is likely to be alot more pronounced, and that is a desirable characteristoic for a switcher because the idea is for it to spend as little time as possible in the linear region. No time at all would yield the most efficiency for a switcher. The only dissipation would be the leakage current when turned off times the voltage applied, and the voltage drop across the drain amd source times the current carried. One number of each is very low, therefore the power dissipation is very low at those times.

Then we have the time actually spent between states. That is where the real power dissipation comes in usually. What exacebates the situation is the fact that these devices frequently turn on faster than they turn off. This can cause problems in a totem pole arraingement obviously, or an inductive load as the voltage wants to go go go but the currentis not yet completely shut off off off. Thus the device can be and is frequantly optimized for switching.

This will make it perform very poorly in a linear application. In fact a long time ago I worked on a TV in which someone had replaced a video amplifier with an RF amp transistor that was designed to be a switcher. The result was a picture that had good blacks and whites, but almost no shades of gray, as if it was clipped. However, the actual video level was about right. Do you understand why ?

On a bipolar spec sheet you will find a similar curve but usually it looks upside down sort of from the gain curve on the FET data sheet here. It will have the current gain plotted against collector current. In that case you want the curve to be horizonatally as flat as possible for linear use. For switching, it is a whole different story.

At any rate, minding the divisions on the graph of course, of the three curves there in figure 3, the 125C line is best and the 25C line is worst. You will not get a perfectly straight line, but the closer you come the better.. Ironically the -55C curve is not better, but that is just how these things are.

Now, if you want to parallel fets in a linear application, I can do that, but I need a case of beer. It is not as simple as just using source resistors because for that to work effectively the resistance value would be too high. I have a neat little drive circuit that not only accomplishes the current sharing accurately, but since it effectively shunts drive voltage, eliminates problems with storage time (?), the turning off slower than turning on thing. That can cause alot of problems.

Even in switching circuits, if you are heading for the high frequencies you have to optimize the drive to make sure the damn thing turns off in time. This requires pretty much shorting out the charge on the gate to source capacitance. Driving it linear is not exactly the same, but some of the same principles apply. In more modern devices you don't have to jump through as many hoops. Semiconductor manufacturers do improve their products and lessening drive requirements in all ways is considered an improvement as it is attractive to engineers who have alot to say about which devices are chosen and therefore bought.

This is long enough. I probably will never catch all the typos...