Page 5 of 11

Re: [2019-12-31] PrimoCache 3.2.0 and 4.0.0 alpha released!

Posted: Fri Apr 10, 2020 2:26 am
by Support
editor0range wrote: Thu Apr 09, 2020 8:55 pm But is there a way to go even further and let the L1 and L2 cache fill at the same time when I download a game over steam for example? So after the download I can fire it up with RAM speed - if it fits my L1 cache - and after rebooting I still got SSD speed until it´s cached by L1 again?
PrimoCache will fill L1 first and then L2 cache. It does what you want it to do.

Re: [2019-12-31] PrimoCache 3.2.0 and 4.0.0 alpha released!

Posted: Mon Apr 13, 2020 5:35 pm
by Betonf
Logic wrote: Mon Mar 30, 2020 11:48 pm
NB
that the eBoostr guys are able to do most of what I suggest above..! Talk to them!
I haven't heard about eboostr in a long time. It is a pity that the program has no updates.

Re: [2019-12-31] PrimoCache 3.2.0 and 4.0.0 alpha released!

Posted: Tue Apr 14, 2020 6:58 am
by Nick7
Logic wrote: Mon Mar 30, 2020 11:48 pm Feature Requests:
(assuming you want a better, faster product than the competition!?) :)

1:
A 3rd non volatile caching tier:
Cache HDD to SATA SSD to NVME or Optane SSD.
For majority of users (aka 99%+), this is not needed.
You may want this if you like to see huge numbers in synthetic tests.
And for that <1% of users who would even need such feature - they generally will have other means to achieve this - external storage systems, that's what you use if you use it on business scale.
Logic wrote: Mon Mar 30, 2020 11:48 pm 2:
Filtering by size and random/sequential:
Optane is fast at low que depth (QD1) random ~4K, but slower than a MP600 say at large sequential.
So cache the large sequential on MP600 and random 4K on Optane and read/write them simultaneously for an extra boost!
Honestly, don't know of any software that is currently able to do this. Main issue is, you are using a set block size for caching.
Logic wrote: Mon Mar 30, 2020 11:48 pm 3:
Simultaneous reads/writes to take advantage of drive type strengths:
If loading a game; start reading large sequential to small from the HDD (thats good at that)
and simultaneously
Small random to large from from the SSD and/or Optane Drive
and have the reads meat in the middle, or wherever they meet...
It does not work this way. This works with file level cache where you know you are going to read file. Even there it's not 100% sure, as you may not want to read whole file.
Also, reading from HDD - if you have it in cache is silly. HDD still has access time which is 5+ms, unlike SSD.
Logic wrote: Mon Mar 30, 2020 11:48 pm NB
that the eBoostr guys are able to do most of what I suggest above..! Talk to them!
And the main problem - eBoostr works on FILE level. PrimoCache works on BLOCK level.
Logic wrote: Mon Mar 30, 2020 11:48 pm 4: Except for your deferred write RAM cache,
you cannot compete with the speed and 'intelligence' of Windows' Prefetch and Superfetch.
All you are persisting to do is BS the idiots you don't realise they're looking at RAM speed in the Disk benching software... and that said RAM speed is MUCH slower, with less cache hits, than what Pre/Superfetch get.
So stop! You just insulting the intelligence of most users nowadays with this outdated trick that lowers the respect and ups the distrust of influential reviewers and sys admins etc!

Instead:
Make Pre/superfetch controllable/settable as windblows sometime thinks switching them off for SSDs is a good idea.
Use the data they use in the prefetch etc folders to better cache hits in the L2/3 caches too..!
Hell you might even be able to set/control their retardedly set Readyboost app to do #2 and 3 above...
Again, you do not understand difference between super/prefetch vs deferred writes.
Super/prefetch is Windows feature which works on file level....
Logic wrote: Mon Mar 30, 2020 11:48 pm 4:
Add an updated version of MyDefrag (15% faster than Windows' Defrag and anything else) to the package as a means of optimising the last level of the stack to further get optimal perf from PrimoCache...
Here also, you may learn a lot about using data in the prefetch etc folder/s effectively.
?!
You want to cache data to cache to avoid HDD. You want to access most used data from cache. You don't want to use HDD for your data. Main reason for caching.
Logic wrote: Mon Mar 30, 2020 11:48 pm 5:
Running Primocache (or its files) through https://hybrid-analysis.com/
and
https://any.run
etc
makes for some very ...'interesting' reading!!!
Why..!? Thats... really not necessary!??
Rather use that skill for the above '4: Instead' and charge more.
You *do* understand PrimoCache needs to hook to kernel and intercept system calls for it to being able to work?

Re: [2019-12-31] PrimoCache 3.2.0 and 4.0.0 alpha released!

Posted: Sat Apr 18, 2020 3:32 am
by trlkly
Since you are verifying the L2 cache on startup after a dirty shutdown, does this also mean that you can now flush previously deferred writes to disk, making the defered-write option on L2 caches safer?

That is the main feature I am currently looking for. And, honestly, it may be what leads me to decide to actually buy the software after the trial is up (assuming this releases by then). If not, I'll probably bite the bullet and clone my boot partition to my SSD and then boot off of that (while also having to keep a backup of my SSD on the HD in case it fails.) But I would love to have an automated solution like this.

It seems weird to me: Windows itself defers writes to disk with write caching, and that is only in memory, and thus can't be restored on boot I never run into problems with this from improper shutdowns. So why is it so much more risky with Primocache?

Re: [2019-12-31] PrimoCache 3.2.0 and 4.0.0 alpha released!

Posted: Sat Apr 18, 2020 3:46 am
by Jaga
trlkly wrote: Sat Apr 18, 2020 3:32 amIt seems weird to me: Windows itself defers writes to disk with write caching, and that is only in memory, and thus can't be restored on boot I never run into problems with this from improper shutdowns. So why is it so much more risky with Primocache?
I guess the real question is - how do you know an improper shutdown (or crash) with just Windows' cache flushes everything properly?

I haven't had a bad shutdown or even a crash while using Primocache and write caching in years - the software is among the most stable utilities I know of. Come to think of it, I trust it more than Windows 10's disparate parts.

I wouldn't call Primocache more 'risky' in that regard. But Windows 10 may use a different type of writing logic than what Primocache uses, since Primocache is far more geared towards performance than Windows is.

Bottom line from my perspective: If Windows shuts down incorrectly or crashes, uncomitted writes still in memory are going to be lost - there's no way around that. But uncommitted writes on a L2 volume would not be.

Re: [2019-12-31] PrimoCache 3.2.0 and 4.0.0 alpha released!

Posted: Mon Apr 20, 2020 4:06 am
by Support
Windows defers writing with a latency of 2s, while PrimoCache allows users to set the latency freely. Shorter latency, less risk. Besides, Windows caching works at the file system level. Combining with NTFS logging system, Windows is able to fix some errors caused by ungraceful shutdown. However, uncommitted data in memory still will be lost. In rare cases, it may also cause fatal problems.

Re: [2019-12-31] PrimoCache 3.2.0 and 4.0.0 alpha released!

Posted: Sat Jun 20, 2020 9:14 am
by primouser
Does PrimoCache 3.2 support Windows 10 2004 May update?

Re: [2019-12-31] PrimoCache 3.2.0 and 4.0.0 alpha released!

Posted: Mon Jun 22, 2020 10:12 am
by Support
primouser wrote: Sat Jun 20, 2020 9:14 am Does PrimoCache 3.2 support Windows 10 2004 May update?
Yes.

Re: [2019-12-31] PrimoCache 3.2.0 and 4.0.0 alpha released!

Posted: Wed Jul 08, 2020 10:33 am
by RobF99
I am amazed at just how stable the 4.0.0 seems to be. For an alpha.... it runs perfectly! I found that it vastly improves the handling of deferred writing on USB flash drives compared to previous versions. I write a lot of data on a lot of flash drives and with V3 the entire PC would get very slow while handling deferred write to the flash drive and the copy dialog would stay on the screen for a long time and until deferred write caught up. Now it is much smoother and it doesn't slow the PC. I find that at the end of the copy process to L1 or L2 that the OS now closes the copy Window and the files are smoothly written to the flash drive without any significant slowing of the system. Great stuff.

Re: [2019-12-31] PrimoCache 3.2.0 and 4.0.0 alpha released!

Posted: Thu Jul 09, 2020 5:56 am
by Support
@RobF99, Thank you for testing the v4.0 alpha. We're glad that it works well. :)