For as long as I can remember I've always biased my amps around 60% of maximum dissipation. But I just ran across a site that claims 70% is optimal for a Class-AB amp. So I tried it on my go-to Plexi patch and I do believe it sounds better.
Set Power Tube Grid Bias to 0.7 to do this.
The issue with this 70% rule is, that it's based on the old tube production. Newer tube production is no longer made for military usage, therfore the safety operation area of these tubes is not that superior as it was back in the days. The most troubles with newer production tubes were screen voltage issues - for example, EL34s should not take more than 450V on their screens, says the datasheet - in the older days, some EL34s were able to take more than 500V at the screen grids, today we run in various issues with these high voltages.
The 70% rule was just based on fixed bias Class A/B amps. On cathode bias amps (such as the Tweed Deluxe 5E3), the maximum plate dissipation can reach up to 100%....nominal it's between 85 to 95. Some modern fixed biased amps have only 27% stock setting - such as the 5150 or between 30 to 45% on various Mesa/Boogie amps (they said they use special long life tubes, but in reality they just use them with lower bias settings) - I recentely had a a Mesa 395 Simul Stereo amp in, which had 25% stock setting and 35% on their (so called Class A - which isn't Class A just a bit warmer biased Class A/B) tubes......
On the other hand, there were several cathode biased Class A/B designs which go over the 100% mark - such as the Vox AC30 (yes this is no true Class A amp) - I saw 130 to 140%, the newer production Supro amps, some Mesa Boogie amps (Studio Caliber .22/ DC3 etc.) - this wear out the tubes much faster and produce a lot of 120 cycles hum (for US - 100 cycles for EU) - some techs said, this would make them sound better, but that is just stupid......
I normaly bias amps with modern production tubes between 55 to 65% (55% for cold, 60% normal and 65% for hot). I learned the 70% in my EE days, 25 years ago. They also did the biasing with an oscilloscope back then - today I go along with the shunt method - much more accurate.
all the best, cheers
P