Rico has his third installment on LINQ to SQL performance up on his site and he finally lets us in on what he thinks the problems are/were. I'm not sure how many more posts he's got up his sleeve before he delivers the punch-line, but he's certainly building up the momentum.

Roger Jennings is keeping tabs on the whole process over at Oakleaf, complete with his own tests.  He's getting somewhat different values, but that's to be expected since there are quite a number of different variables that can affect the result.  Roger thinks Rico's machine is faster than his.  Rico's thinks Roger has a faster disk.  Someone needs to come up with a LINQ to SQL benchmark and start selling computers based on 'real-world' performance. :-)

Everyone is anxiously anticipating Beta 2 where supposedly every thing is now fast and furious. Does it end up with 4x overhead, 3x, 2.5x, x.x?  I'd really like to say, but I promised Rico he could reveal the numbers.

However, I did not promise not to reveal the non-numbers!  Err, whatever that meant.  It's definitely not 10x or 8x or whatever was happening back with the May 2006 CTP.That was before it was 'Orcas' when the focus was on just getting the feature set right, before everything went into 'production' mode. Now we really mean it. Now the performance is definitely better than even with the introduction of the compiled query feature. (Since some of you peeked ahead and took a look at your Beta 1 perf using compiled queries.)

Most people will be pleasantly surprised.

And look, I didn't digress one bit.