Give us better statistics please Topic is solved

[TheNavigator]

What I need is data to make me able to decide, how large I'm supposed to make my cache.

The data which is needed looks like this:

Theoretical cache size -> % of cache hits

After a while of use, of course.

Two fictional examples:

a)

1024 MB -> 40%
1536 MB -> 41%
2048 MB -> 42%

I would go with a 1024 MB cache.

b)

1024 MB -> 30%
1536 MB -> 50%
2048 MB -> 52%

Of course I would go with a 1536 MB cache.

---

This data can be provided easily, in form of a histogram and even as a graph. Please give it to us.

[InquiringMind]

While it sounds simple, this would be tremendously difficult to do since the "best" cache size is greatly influenced by how people use their PCs. If someone only ever uses 1GB of programs and data, then they'll get no benefit from more than 1GB cache - on the other hand someone else continually loading up different programs and files may (in theory) never get any benefit from caching however much they allocate, since they don't access the same content more than once.

The best advice would be to review your maximum memory usage (via Task Manager in WinXP, Process Explorer or Process Hacker in later versions of Windows), add 20% for a safety margin and allocate any memory beyond that to PrimoCache.

[Bjameson]

The "Cache Free (L1/L2)" reading gives a good indication about possible cache waste.
- If L1 never reaches 32.00 MB or less, you are wasting RAM.
- If L2 never gets full you are wasting L2 disk space.

Decrease RAM until the L1 cache completely fills up, but only just. From there, trim RAM further until your Cache Hit Rate starts dropping. Increase again to get the desired hit rate back. This is the sweet spot; you now have the minimum RAM required to get the best hit rate. As InquiringMind said, this heavily depends on the machine's use.

Adjusting L2 is more difficult because virus scanners and backups can quickly fill up L2, thus giving a false reading. The simplest method is to reboot and load the most used programs. Trim L2 until you get the best balance between L2 size and bootup speed. L2 will (and should) completely fill up so only the bootup/program loading speed counts.