Romex Software Forum

Hello, i've spent all my day searching on how to parameter Primocache, but i think i'm still doing it wrong.

- 2Tb HDD (Segate Barracuda) [OS + Games]
- 250 Gb SSD (Crucial MX500 SATA/m.2) [caching 2TB HDD] both drives new

I just want faster booting, and better gaming performances.
However, i don't see any upgrades. My booting time is slightly increased (but far from ssd boot speed), and no changes for gaming, apart some games are freezing for 2 to 10 sec.

I created a level-2 storage with all my SSD, and i assigned read/write cache task to all HDD partitions.

Does it just takes more time to be effective ?
Is Read/Write good option for my case ?
Are my settings good ?

Thank you !

---------------------My Config---------------------
CPU: Ryzen 5 2600
Motherboard: ASRock AB350M PRO4
RAM: 16GB G.Skill Aegis 3000Mhz
GPU: RTX 2070 Super Windforce OC
(i don't think the rest is necessary)

Suggestions:

1) Remove all volumes from the cache task -except- the "C:" volume.
2) Turn on the "Start at Windows Boot" so it will prefetch when Windows starts up.
3) Change your Block Size from 4k to 16k (your memory overhead right now is huge at 5.25GB)

You said you assigned the entire SSD to L2 and assigned it " read/write cache task to all HDD partitions", but then you forced the L2 to be a read-only using individual read/write cache space.

Your boot speed will probably be the same with or without Primocache, since it has to prefetch at boot time while Windows is starting (or just after depending on settings).

You might want to disable write caching on the L1 (set it to 100% read), and then disable individual read/write cache space on the L2 (so it handles deferred writes instead). Your total RAM isn't large enough to effectively use the L1 as a good sized write cache, so stick with reads only on it.

Since most of your caching is going to happen in the L2, and that needs to be populated from a conventional spinner HDD, it is going to take a while to get the L2 cache built up so that you see speed improvements.

Alright, thanks a lot for your time and advices !
You're a hero

What exactly does memory overhead ?

I've read the doc, but still don't understand

The larger a cache is, and the smaller the block size, the more memory the cache task will consume tracking it all. By either reducing cache size or raising block size you reduce overall memory overhead required. That in turn increases memory available for either the OS, or apps, or to put back into the cache (increasing it's size).

What you had going reduced your available RAM to a mere ~1.3GB, which meant highly probable swapping and in turn, slower performance overall.

Memory overhead is extra memory used to maintain the cache tasks for cache indexing database etc. Smaller block size, bigger memory overhead. In your current setting, overhead is too much. Your system is only about 1GB memory left. This is not enough for games. As Jaga suggested, you may change the block size to 16KB or 32KB to reduce memory overhead. I would suggest reducing L1 cache size to 4GB, thus system will have more free RAM for games. Some games require lots of free RAM to run smoothly.
I agree what Jaga suggested except that I prefer L1 using shared R/W setting for Defer-Write and L2 is 100% read. Usually with 10s latency 4GB L1 is big enough to cache all writing requests. Unless there are lots of writing requests and you see some "Urgent Writes", there is no need to using L2 write caching.
L2 will speed up your system boot time and game loading. The reason why you haven't seen significant improvement is that L2 is populating the data and haven't provide much cached data. Usually read cache will reach 80%~ 99% with L2. You may check "L2Storage Read" and "L2Storage Write". See https://www.romexsoftware.com/en-us/pri ... cache.html.

Going to agree with Support - dropping the L1 cache size a bit will help boot times if that's one of your interests. It will force more of the Windows boot files into the L2 instead of the L1, and those don't need to prefetch at boot. You'll have to see if 16KB or 32KB is a better block size - the practical difference will be minimal in use and probably help significantly with reducing overhead. But the difference between 16KB and 64KB (if you decide to try it) will be noticeable, though 64KB blocks free up a lot of overhead.

I did not noticed you responded me, so i'm late but thanks a lot guys !

Recent Seagate Barracuda drives are using Shingled Magnetic Recording (see ArsTechnica: Buyer beware... and this related Slashdot discussion.

If this applies to you, then you want to avoid small writes to that Seagate Barracuda at all costs.

As such it may be worth using only part (say 50%) of your SSD for caching, and the rest for your OS and most frequently used programs.