- Jun 30, 2004
- 16,176
- 1,771
- 126
I'm posting this as a caution to better understanding for anyone who read my thoughts about using an M.2 NVMe as a caching drive with PrimoCache.
Yes, it can be done.
Yes, you can use any SATA SSD or M.2 SSD of any size as the caching drive. Or, yes -- Primo allows for drives of any size as caching SSDs.
Here's the fly in the ointment. The larger the SSD cache, the larger the memory overhead in addition to RAM caching which Primo also does.
For a 256GB caching volume on an SSD (my experiment used a Crucial MX100 500GB total), PrimoCache needs nearly 6GB of RAM without even allocating RAM for RAM caching. By comparison, a 64GB caching SSD needs something more than 1 and less than 2GB.
So I have no doubt that all this would work, for some number of enthusiasts who want a tiered storage scheme that includes the fastest (RAM and M.2) to the slowest of HDD spinners.
But if you have 16GB of RAM, you'd best add 8 more or simply double it.
So with the cost of the M.2 NVMe, another $100 on additional RAM, the Primo software or $30, well -- you have to ask yourself . . . .
Yes, it can be done.
Yes, you can use any SATA SSD or M.2 SSD of any size as the caching drive. Or, yes -- Primo allows for drives of any size as caching SSDs.
Here's the fly in the ointment. The larger the SSD cache, the larger the memory overhead in addition to RAM caching which Primo also does.
For a 256GB caching volume on an SSD (my experiment used a Crucial MX100 500GB total), PrimoCache needs nearly 6GB of RAM without even allocating RAM for RAM caching. By comparison, a 64GB caching SSD needs something more than 1 and less than 2GB.
So I have no doubt that all this would work, for some number of enthusiasts who want a tiered storage scheme that includes the fastest (RAM and M.2) to the slowest of HDD spinners.
But if you have 16GB of RAM, you'd best add 8 more or simply double it.
So with the cost of the M.2 NVMe, another $100 on additional RAM, the Primo software or $30, well -- you have to ask yourself . . . .