Scorecard is updated

It finally happened! Almost 9 months has passed since last scorecard update, and now ~~the baby is born :)~~ you can see the newest results.

Most important changes:

Test suite was migrated to .NET 4.0.
All tested ORM tools are updated to their most recent versions. New versions are mentioned on their description pages.
There is a new player, Devart LinqConnect ORM, showing perfect results right on the first round.
LightSpeed ORM is removed from this test suite by decision of Mindscape.
Finally, all the tests were performed on 64-bit Windows 7.

Brief overview of new results:

Most of players show nearly perfect LINQ test results, so I suspect they're quite comparable from this point now. All the typical cases are fully covered. I hope ORMBattle LINQ test played a noticeable role in this process.
There are no changes in LINQ test pass results for Subsonic (that's predictable - the project frozen for at least one year) and NHibernate. The case with NHibernate is a bit more complex: currently we test its latest stable version (v2.1.2 GA), although NHibernate 3.0.0 alpha1 is already available. But we were unable to update our tests for it (see the commit comment here). Thus any help in resolving this issue is appreciated.
BLToolkit is still definitive leader in performance tests; all the other commercial players are quite comparable there now. As I hope, that's partly because of this test. There are no changes in NHibernate and Subsonic results here.
Lots of people were interested in EF 4.0 results: EF team has improved LINQ support (they got additional 5% here), but its performance results are noticeable worse - from 10 to 20% in most of the cases, and nearly twice - on materialization test. So if earlier it was one of leaders on this test, now it's closer to losers. If you know any possible reasons of this, please describe it in comments to this post.
New LINQ to SQL shows much worse performance on CRUD operations. Please let know us, if you know the reason \ possible fix - the difference looks quite suspicious; probably, there is some mistake in test (this could appear e.g. because of change in default behavior).
Our own ORM, DataObjects.Net, also shows some decrease in materialization performance. The only reason I can imagine here is lack of almost any profiling of this part during last year. So hopefully, we'll improve our own results soon.