Are all 9000 pieces of hardware actually being added at the top assembly level or are some of them being added in subassemblies. It won't matter much in the overall processing times - I'm just curious.
You could create the assy with 2 configurations - one with hardware, one without. A SpeedPak approach may help with that too. This would allow opening only the non-hardwared version when you don't need to see them all.
I would give serious consideration to looking at upgrading your computer. If your assy already is bogging down, you should get some help for that - it will likely pay for itself in time savings. Then, while your at it, try to plan for the added burden of the hardware.
Ideally the bolts would exist in the assembly level directly above the two sub-assemblies that are connected with the bolts. That means the bolts could be in the top level, or in a sub-assembly, or in a sub-sub-sub-assembly. However, I don't think the location is key at this point. We rarely have any interference or alignment issues due to bolts. The only issue we're focusing on is quantity.
You could create the assy with 2 configurations - one with hardware, one without.
Has anyone done research on performance improvements with large quantities of parts supressed, or congiruations with large quantities of parts suppressed? Does SW actually ignore supressed parts when processing? I only ask because my anecdotal experience with this solution has been varied at best. We've tested SpeedPak but not extensively. It seems like more of a hassle to implement and upkeep than it's worth for the performance improvements we saw.
I may do some testing to see how viable this solution is.
I would give serious consideration to looking at upgrading your computer.
We're actually in the process of upgrading PC's specifically to combat this. Budget and performance met at: i7 3770k (overclocked to 4.5GHz), 16gb of high end RAM, Quadro 4000's, SATA III SSD's, and a nicer ASUS mobo. I completely agree with your suggestion but don't know what more we could do at the current budget.
James, I would be astounded if it took more than 8 hours to regenerate 9000 bolts. I'd suggest inserting the bolts, patterning them wherever possible, then letting the BOM count them for you. It will probably save you a ton of time and be more accurate, too.
I would also add that you will want to make sure the bolts (or at least the configuration used in the assemblies) is as simple as you can get away with.
Jeremy has a really good idea. As a matter of fact your bolts could be as simple an empty part with a construction axis and design table for your part numbers.
You can then take advantage of feature-driven patterns to populate your fasteners and get a BOM count without very much overhead in terms of regeneration or display.