[2017-05-27] PrimoCache 3.0.0 Beta is available now!

First hand news related to PrimoCache
BonzaiDuck
Level 7
Level 7
Posts: 88
Joined: Wed Jan 11, 2017 12:57 am

Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!

Post by BonzaiDuck »

Well, the brief remarks by the last two posters geneo and escaward are the same as mine.

I was thinking to check back with the forums to see what progress had been made, after reading a different forum's discussion on storage performance. As I understood it, version 3 was supposed to make real-time benchmarks feasible for SSD-caching. I'm always tempted to post my Anvil benchmarks on the other forums, just so I can gloat.
User avatar
Jaga
Contributor
Contributor
Posts: 692
Joined: Sat Jan 25, 2014 1:11 am

Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!

Post by Jaga »

BonzaiDuck wrote:Well, the brief remarks by the last two posters geneo and escaward are the same as mine.

I was thinking to check back with the forums to see what progress had been made, after reading a different forum's discussion on storage performance. As I understood it, version 3 was supposed to make real-time benchmarks feasible for SSD-caching. I'm always tempted to post my Anvil benchmarks on the other forums, just so I can gloat.
I've been watching the 2.7.x and 3.x.x performance characteristics, and can say that I think there are some issues there. Speeds in 3.0.0 beta aren't what they were in 2.7.x (which I went back to after a beta test corrupted some drive info while using 3.x).

If you're still using 2.7, do a benchmark and save the results (Anvil is a really good tester). Then test again after upgrading to 3.0 (whenever it comes out). I noticed a significant drop in many individual tests, and in the overall Anvil score when using 3.0.0 beta. Since it's still beta, we have to give it some leeway and understanding. But since the whole idea behind Primocache is speed... 3.0 shouldn't be any slower than 2.7, and right now it is for me.
BonzaiDuck
Level 7
Level 7
Posts: 88
Joined: Wed Jan 11, 2017 12:57 am

Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!

Post by BonzaiDuck »

I guess my primary and only interest in the version 3.0 Beta was similar to that for any piece of software: you would think there would be improvements after ironing out bugs. And I was told that you could see real time performance exclusively for L2 caching with the new version. The only kind of indicator you have now doesn't respond to benchmark programs; you can only see the hit-rate for the SSD-cache, and estimate a performance equivalent.

I could probably turn my perfect Skylake system that took some 8 months to refine and complete into guinea pig for the Beta, and accept the risks. I think I'm more inclined to wait, as these stories about performance shortfalls result in some software improvements toward the certified release. All of my systems perform serious duty in one way or the other, and I'd want a separate system just to test the Beta. A shift in circumstances could change that, but for now, I'll just wait and see . . .
ebfortin
Level 3
Level 3
Posts: 10
Joined: Sun Sep 17, 2017 11:57 am

Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!

Post by ebfortin »

I've been using PrimoCache 3.0 Beta for a month now. No issue. No crash. Speed is as expected, compared to 2.x.
Davey126
Level 7
Level 7
Posts: 99
Joined: Sun Mar 23, 2014 3:40 pm

Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!

Post by Davey126 »

ebfortin wrote:I've been using PrimoCache 3.0 Beta for a month now. No issue. No crash. Speed is as expected, compared to 2.x.
Same - my experience spans several months on two devices.
LuisD
Level 1
Level 1
Posts: 2
Joined: Mon Oct 09, 2017 11:12 pm

Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!

Post by LuisD »

BonzaiDuck wrote:I guess my primary and only interest in the version 3.0 Beta was similar to that for any piece of software: you would think there would be improvements after ironing out bugs. And I was told that you could see real time performance exclusively for L2 caching with the new version. The only kind of indicator you have now doesn't respond to benchmark programs; you can only see the hit-rate for the SSD-cache, and estimate a performance equivalent.

I could probably turn my perfect Skylake system that took some 8 months to refine and complete into guinea pig for the Beta, and accept the risks. I think I'm more inclined to wait, as these stories about performance shortfalls result in some software improvements toward the certified release. All of my systems perform serious duty in one way or the other, and I'd want a separate system just to test the Beta. A shift in circumstances could change that, but for now, I'll just wait and see . . .
Just to stop the speculation, and following on the last 2 replies. I have been using v3.0 since it came out in two pcs without any issue, no crashes or anything. And the performance improvement is just night and day vs v2.7.3 as I only use L2 cache in both computers and v3.0 adds write cache to the L2 cache.

And you can benchmark this performance.

Attached are screenshots of a benchmark with CrystalDiskMark of my main drive (C) which is a 2TB Seagate Firecuda with a 250GB Samsung 960 Evo NVMe SSD running as L2 cache, no L1 cache as that's just wasting your ram with the NVMe cache as read/write cache with defer write enable.

L2 cache volume is max capacity of the NVMe shared for read/write (no individual read/write cache space).

So just go ahead and install Primo Cache V3.0 on that perfect skylake system that took you 8 months to refine...

PD: Advanced Defer-Write Options are set to:
  • - Write Mode: Buffer
    - Free Cache on Written: checked
Attachments
Advanced Level-2 Cache Options.PNG
Advanced Level-2 Cache Options.PNG (9.11 KiB) Viewed 11557 times
Cache Configuration.PNG
Cache Configuration.PNG (22.93 KiB) Viewed 11557 times
Benchmark__PrimoCache_L2_Cache_New.PNG
Benchmark__PrimoCache_L2_Cache_New.PNG (36.65 KiB) Viewed 11557 times
User avatar
Jaga
Contributor
Contributor
Posts: 692
Joined: Sat Jan 25, 2014 1:11 am

Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!

Post by Jaga »

My speed test comparison was done using Anvil and *just* a L1 cache. The L2 caching obfuscates the back-end raw memory caching algorithm (my personal opinion), and since I don't use the L2 on my system it's not a valuable test for me.

Here are the results from 2.7:
Image



And the results from 3.0.0.1 Beta:
Image



Exactly the same cache configuration for both, and yet what I would call a significant difference in results. That's why I indicated people might want to test themselves between 2.7 and 3.0 betas to see what is going on for them. People that actually use the L2 feature will probably experience higher results, though I'm not sure if Anvil will "see through that" and give accurate results anyway, since the copy to L2 for faster access takes a little time, and Anvil doesn't run very long.
ebfortin
Level 3
Level 3
Posts: 10
Joined: Sun Sep 17, 2017 11:57 am

Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!

Post by ebfortin »

PROBLEM WITH L2 CACHE FLUSH:
I just upgraded from 8 GB to 16 GB. When I rebooted the system was awfully slow compared to what it was before. Apparently, the L2 CACHE had been flushed. Why is that? Why flush the L2 CACHE when memory configuration change?
LuisD
Level 1
Level 1
Posts: 2
Joined: Mon Oct 09, 2017 11:12 pm

Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!

Post by LuisD »

ebfortin wrote:PROBLEM WITH L2 CACHE FLUSH:
I just upgraded from 8 GB to 16 GB. When I rebooted the system was awfully slow compared to what it was before. Apparently, the L2 CACHE had been flushed. Why is that? Why flush the L2 CACHE when memory configuration change?
Whenever there's a hardware change the cache is flagged as dirty and PrimoCache has to fill it again. This is standard behavior and it's not new to v3.0.
ebfortin
Level 3
Level 3
Posts: 10
Joined: Sun Sep 17, 2017 11:57 am

Re: [2017-05-27] PrimoCache 3.0.0 Beta is available now!

Post by ebfortin »

Should it check what is the hardware change exactly and act according to that information? Changing RAM will never have an effect on the Cache "dirtiness".
Post Reply