We have a server with 1500 users databases and an awful adhoc workload. We are on 2008 and have 10 logical cores per socket/numa node. We see a large percentage of CMEMTHREAD waits (75%). BUT as far as spinlock stats - seeing SOS_TLIST as the largest spinner. This is a delta on a 10 minute sample:
Spinlock Collisions Spins SpinsPerCollision SleepTime Backoffs
SOS_TLIST 4673403 2274201787 486 2877 36028
LOCK_HASH 7391 6823378 923 0 893
BUF_FREE_LIST 5409 2324733 429 7 154
SOS_OBJECT_STORE 3036 66379 21 0 3
SOS_SUSPEND_QUEUE 1801 588115 326 1 42
Reading this article
http://blogs.msdn.com/b/psssql/archive/2011/09/01/sql-server-2008-2008-r2-on-newer-machines-with-more-than-8-cpus-presented-per-numa-node-may-need-trace-flag-8048.aspx
How do I determine if this trace flag is appropriate?