L2 cache almost unused

FAQ, getting help, user experience about PrimoCache
Post Reply
alesf76
Level 1
Level 1
Posts: 1
Joined: Thu Mar 16, 2017 12:22 pm

L2 cache almost unused

Post by alesf76 »

As seen in this picture, almost nothing is written to the L2 cache, and consequently the cache hit ratio is ridiculously low.
I've already rebooted waiting before logging in, and I've already set the gather interval to 1 sec.
Nothing else I can try ?
Attachments
primo.png
primo.png (79.25 KiB) Viewed 5276 times
Babel17
Level 5
Level 5
Posts: 53
Joined: Tue Nov 03, 2015 3:41 pm

Re: L2 cache almost unused

Post by Babel17 »

alesf76 wrote:As seen in this picture, almost nothing is written to the L2 cache, and consequently the cache hit ratio is ridiculously low.
I've already rebooted waiting before logging in, and I've already set the gather interval to 1 sec.
Nothing else I can try ?
Hopefully that changes, but it's important to remember that writes don't get cached to the L2, only reads. That might change with the next update though.
minhgi
Level 10
Level 10
Posts: 267
Joined: Tue May 17, 2011 3:52 pm

Re: L2 cache almost unused

Post by minhgi »

If you enable a small amount of l1 cache, it seem to help cache data to the l2 more often. Just an observation from using Primocache for long time.
User avatar
Support
Support Team
Support Team
Posts: 3731
Joined: Sun Dec 21, 2008 2:42 am

Re: L2 cache almost unused

Post by Support »

How about if you log in and then leave the computer alone for a while?
BonzaiDuck
Level 7
Level 7
Posts: 88
Joined: Wed Jan 11, 2017 12:57 am

Re: L2 cache almost unused

Post by BonzaiDuck »

minhgi wrote:If you enable a small amount of l1 cache, it seem to help cache data to the l2 more often. Just an observation from using Primocache for long time.
For general-purpose desktop workstation, is there an optimal amount of RAM that facilitates that? 1GB? 2GB?

I have had anomalies occur spitting up CHKDSK-fixable disk errors, and tentatively ascribe it to a use of L1 + L2 on an HDD. There is one other cause which could be the correct one. I was in enough of a hurry that I addressed both causes and eliminated the problem. For the time being, since I'm caching a ~140MB/s seq-read HDD to a ~3,000+ MB/s NVMe M.2 drive, the KISS principle suggests to me that I shouldn't need to use any RAM for that HDD -- only L2 on the NVMe.

It would be nice to get the SSD-cache L2 to fill up faster. The fly-in-the-ointment with PrimoCache is the inability to run a benchmark test against the L2-cached HDD and see anything other than spec behavior for it. But you can see the amount of Free Cache available in real-time, and you can see the hit-rate available in real time.

Right now, my hit-rate on the SSD-cached HDD is 89%. That means that 89% of the time the system needs to read files into memory, it finds them on the SSD. And my SSD -- like I said -- has a seq-read spec of 3,500 MB/s, if the full bandwidth of the interface is available through either the mobo M.2 slot or an x4 PCIE. Usually, the drive shows benchies of a top 3,000 MB/s.
Post Reply