I ran my LSP editing on Win7 x64, the fastest icore7 proc that I could get from Fry's, 32GB, and 4 fast SSDs in RAID 0... and STILL my 13,000 channels were slower to sequence than I wanted.
Agreed, I'm running close to 5,000 channels and as the sequence is programmed, it gets slower and slower to optimize each sequence. They do run ok in the scheduler, it's just slow to optimize in the sequencer.
I don't believe that LSP takes advantage of the multiple cores, memory or the dedicated graphics co-processor, so I'm hoping the re-write David has done on LSP, will take advantage of those with the next release.