It's very simple.The FPB WU give the most points.On Linux 100 or more PPH.The LEDGF=50-60 The Y3B= 25-50

OK so delving deeper confirms RAMMIE's basic premise that different work types produce widely different points per hour....

...but I thought that there was a system in place to iron out such variance and level the playing field for all work types and computers

Sure it is nice to get Great points but delete the others??

I would rather see someone find a way to just give us a steady output on FA@H whatever the wu.

Well, if by "steady" you means same credit regardless the wu, it's very simple to give example 50 credit/wu regardless of FA@H-wu.

If on the other hand you mean constant credit/hour on the same computer regardless of FA@H-wu, things is immediately much harder...

Now, if you're going to calculate something yourself, using only pen & paper, you'll quickly see adding two numbers is faster than multiplying two numbers, especially if each number includes many digits. Let's say you're using 1 second/digit to add two numbers and 3 seconds/digit for multiplying two numbers, meaning you'll use 3 seconds to calculate 123 + 456 but use 9 seconds to calculate 123 * 456.

For computers it's similar, some calculations are harder to do than others. It doesn't neccessarily need to be addition and multiplications, but it can example be additions compared to calculating the square-root of a number. So, let's say you've got a computer using 1 ns for doing one type of calculations and 3 ns for doing another. If you've got a wu with a 50/50-mix of calculation-types, a computer will example use 1 hour on the fast calculations and 3 hours on the slow calculations.

But, is different times for different calculation-types a problem? Not if all FA@H-wu is always a 50/50-mix, since it doesn't matter if a computer does one wu with 1 hour fast + 3 hours slow compared to another with 2 hours fast and 6 hours slow, you can still get the same credit/hour.

The problem is, all wu's very often isn't always a 50/50-mix, you'll have some wu's being a 25/75-mix and others a 75/25-mix. In the 1st. instance the computer will use example 0.5 hours on the fast calculations and 4.5 hours on the slow calculations while in the 75/25-mix the computer will use 1.5 hours on the fast and 1.5 hours on the slow.

To solve this latest problem, one suggestion could be "count how many fast calculations are done and count how many slow calculations are done". By doing this it's still possible to get a constant credit/hour on a particular computer.

Unfortunate where's an addition problem to this. While computer A uses example 1 ns on the fast calculations and 3 ns on the slow calculations, computer B uses maybe still 1 ns on the fast but only 2 ns on the slow.

Meaning, for computer A you have:
25/75-mix: 5 hours
50/50-mix: 4 hours
75/25-mix: 3 hours

For computer B you have:
25/75-mix: 0.5 hour + 3 hours = 3.5 hours
50/50-mix: 1 hour + 2 hours = 3 hours
75/25-mix: 1.5 hours + 1 hour = 2.5 hours.

If example computer A for 50/50-mix is worth 60 credits, and you "count fast & slow flops" for A, you'll have for A:
25/75-mix: 75 credits ==> 15 credits/hour
50/50-mix: 60 credits ==> 15 credits/hour
75/25-mix: 45 credits ==> 15 credits/hour

For B you'll get, if assumes A is the wing-man on each wu-type:
25/75-mix: 75 credits ==> 21.42 credits/hour
50/50-mix: 60 credits ==> 20 credits/hour
75/25-mix: 45 credits ==> 18 credits/hour

Things isn't much better if B decides the crediting, in this instance it's A getting variable credits/hour.

So, how do you want to fix this problem?

I'll give you two additional tips on the quest of solving the credit-problem:
1: While it's of course possible to say computer A gets 15 credits/hour regardless of wu and computer B gets 20 credits/hour regardless of wu, this clearly breaks the "equal pay for equal work".
2: One idea could be to split wu's in two, having a wu only containing the "fast" calculations and another wu only the "slow" calculations. Since B is significantly faster on the slow than A, only issuing the slow to B and all fast to A would be a "solution". Unfortunately a wu can include many types of calculations where different computers can have both advantages and disadvantages depending on calculations. Additionally other things like memory-speed also can play a role, so where's no way to handle all the differences between computers, even in the unlikely instance it's possible to split a wu into the different computational parts.