My free trial of PrimoCache disk caching software - would buy/recommended

D

Deleted member 259057

Guest
Hey Guys,

After watching an LTT video on Optane vs StoreMI I caught on to his preview of how Primocache actually outperformed them all. So I set about to test it myself.

My System:
Thuban II 1100T
Gigabyte ga78-lmt mATX
16gb Patriot 1333 ddr3
GTX 1060 3gb
240gb Sandisk SSD (SSD Plus model) on Sata2 3gb
2x WD Green 3tb on Sata2 3gb

My Primocache setup:
I gave Primocache 4gb of system managed (shared) memory, typical usage was more like 2gb
40gb of my SSD as non-destructive level 2 cache (99% hit rate after 60 days)
and all of my drives are being accelerated by this setup.

Performance:
What I noted was about a 10% increase in items that weren't cached in the Level 2 cache, such as new updates being applied etc due to the cache now write later feature, but the increases could be dramatic on things like frequently played games, such as League of Legends from my SSD, and steam games from the WD Green.

The seat of the pants feel though is kind of mixed, though I'm not on the greatest hardware what it seemed like to me is game loads, getting into a match, map, or zone for the first time would feel lag, like there was a 100-200 ms delay before the loading began, but then the load itself would be magnificently fast. For instance in League of Legends players I know on faster computers would be at 30-40% loaded before I showed any activity, but I would reach 100% with or before them (getting faster on successive runs as the 'lag' got lessened, as I suspect it would on a more modern CPU and DDR4 RAM).

So, I'm sure some here have more appropriate setups than me, but for my use case at least it seems well worth the $29 investment for the license, especially since my platform doesn't have a similar feature.

I'd also be interested in building a system to use this much like optane is used for Steam gaming quite extensively to good effect, I think pairing primocache with a large 6tb+ main drive and a 128 or 250gb NVME m.2 SSD might be a better use case for the NVME than even installing the OS or games to it.

Anyhow, I'll add screenshots below. Discussion?
 

Attachments

  • primocachesetup1.jpg
    primocachesetup1.jpg
    66.9 KB · Views: 0
  • primocachesetup2.jpg
    primocachesetup2.jpg
    69.4 KB · Views: 0
  • SSD.jpg
    SSD.jpg
    70.3 KB · Views: 0
  • SSD_cachedriot.jpg
    SSD_cachedriot.jpg
    70.5 KB · Views: 0
  • wd3tbgrncachedgame (1).jpg
    wd3tbgrncachedgame (1).jpg
    72.8 KB · Views: 0
  • wdgreen3tb.jpg
    wdgreen3tb.jpg
    69.3 KB · Views: 0
I use it with good results on a box wth a couple spinner notebook drives (4TB total) for Steam games, then half 1TB boot SSD as cache.

It's purely a gaming setup but works great. I need to tinker with it in a more sophisticated setting.

My results with Optane have been excellent, until the pairing crashed and burned.. twice.
 
Yeah, PrimoCache is great software.

I used it years ago when it was called FancyCache. While not an empirical measurement, I got less significantly less LOD pop-in in PlanetSide 2 running FancyCache, even though it was installed on an SSD.



The write buffer feature is great for non-essential work too. Installing huge mods, muxing, copying a ton of small files, etc. It's not something I would run 24x7 though.
 
If you reduce your block size to 4k you'll get even more speedups...
 
I've used Primocache for 8 months straight because of recommendations from others here on this forum. Within the first month of using the program and seeing all the incredible numbers from my benchmarks I was ecstatic with the software and most pleased so I purchased the program for $30.

My rig had 32GB of RAM and had been running 14GBs of it dedicated for Primocache to use.

One day about 1.5 month ago I paused Primocache's cache in order to defragment my hard drives without the write buffer interfering with the process (the developer recommends pausing the cache while defragmenting. Well I actually forgot to ever click to resume the cache and used my PC without Primocache's read and write buffer for a few days doing just productivity work, no benchmarking or gaming at all, just work on the machine with productivity apps.
During this time I noticed that my PC felt considerably faster than it normally did over the past several months so after having this thought, I checked Primocache's hit rate assuming it would be quite high. It had been switched off!
I decided to do a disk intensive real-world subjective test where I opened up a folder with thousands of files in it and mess around with opening/closing the folder. Copying these files back and forth and so on. What I unfortunately discovered is that Primocache was actually slowing my PC down during my real world tasks. I verified and double verified this by enabling the cache, playing with moving files around, disabling it and doing the same. Beyond the shadow of a doubt, in most of the stuff I do on my computer, Primocache absolutely slows things down.
Where I noticed the cache shining the most was installing apps. Applications would install just head-spinningly dramatically fast so it helped there immensely for sure. For everything else it almost always hurt performance.
My advice is to use a stopwatch and test your games and file copies first hand and don't depend on synthetic benchmarks to test this. I'm sure given the right workload situation, the program is great. But for how I use my PC, it hurts more than it helps.
I've since uninstalled the program and don't ever intend to use it again despite owning a license.


Also just wanted to say before someone calls me a hater. I used to tell EVERYONE about this program. Told friends to buy it, went on about how great it was on other forums based off benchmarks. I wanted to love this program.
 
Most of these Cache software cheats benchmark by caching read/write request that are requested to be uncached.
This kind of tricks has been used since the good old dos days.

So you cant use benchmarks that are meant to test the speed of the storage media.
(aka the above benchmark in OP's psot are useless/wrong)

Adding in placebo effect make it hard to take into affect anything about "feels faster working with it"


You need to make some empirical measurement on real workload to avoid both Placebo and benchmark "Cheating"


We saw the same with samsung rapid disk. everyone said it was helpflull eveyone was showing the same flawed/cheated benchmarks
but when we measured realworld applications and usage it was slower than just using windows own cache
 
Last edited:
I gave it a try with 4GB ram allocated to my steam drive for some games i know to have high load times but are too big for my SSD. Figured it can't hurt to give it a test.
 
I've used Primocache for 8 months straight because of recommendations from others here on this forum. Within the first month of using the program and seeing all the incredible numbers from my benchmarks I was ecstatic with the software and most pleased so I purchased the program for $30.

My rig had 32GB of RAM and had been running 14GBs of it dedicated for Primocache to use.

One day about 1.5 month ago I paused Primocache's cache in order to defragment my hard drives without the write buffer interfering with the process (the developer recommends pausing the cache while defragmenting. Well I actually forgot to ever click to resume the cache and used my PC without Primocache's read and write buffer for a few days doing just productivity work, no benchmarking or gaming at all, just work on the machine with productivity apps.
During this time I noticed that my PC felt considerably faster than it normally did over the past several months so after having this thought, I checked Primocache's hit rate assuming it would be quite high. It had been switched off!
I decided to do a disk intensive real-world subjective test where I opened up a folder with thousands of files in it and mess around with opening/closing the folder. Copying these files back and forth and so on. What I unfortunately discovered is that Primocache was actually slowing my PC down during my real world tasks. I verified and double verified this by enabling the cache, playing with moving files around, disabling it and doing the same. Beyond the shadow of a doubt, in most of the stuff I do on my computer, Primocache absolutely slows things down.
Where I noticed the cache shining the most was installing apps. Applications would install just head-spinningly dramatically fast so it helped there immensely for sure. For everything else it almost always hurt performance.
My advice is to use a stopwatch and test your games and file copies first hand and don't depend on synthetic benchmarks to test this. I'm sure given the right workload situation, the program is great. But for how I use my PC, it hurts more than it helps.
I've since uninstalled the program and don't ever intend to use it again despite owning a license.


Also just wanted to say before someone calls me a hater. I used to tell EVERYONE about this program. Told friends to buy it, went on about how great it was on other forums based off benchmarks. I wanted to love this program.

Most of these Cache software cheats benchmark by caching read/write request that are requested to be uncached.
This kind of tricks has been used since the good old dos days.

So you cant use benchmarks that are meant to test the speed of the storage media.
(aka the above benchmark in OP's psot are useless/wrong)

Adding in placebo effect make it hard to take into affect anything about "feels faster working with it"


You need to make some empirical measurement on real workload to avoid both Placebo and benchmark "Cheating"


We saw the same with samsung rapid disk. everyone said it was helpflull eveyone was showing the same flawed/cheated benchmarks
but when we measured realworld applications and usage it was slower than just using windows own cache

Indeed. All the benchmarks are doing is benching the speed of the RAM, so of course you are going to get crazy numbers. Samsung, Crucial, ASUS etc all have some sort of cache or RAMDrive software now. Most of it has really no use, and as yall said, in many of not most cases can result in slowdowns. Some cases however, where you are working for extended periods with the same program or data, putting it into a RAMDrive can really help, big time if you have lots of spare RAM, however right now, with RAM prices still sky high to where they were a few years ago, and SSD prices plummeting it doesn't make sense to use. For that $30 you could have doubled your SSD storage size, as 250GB SSDs are hitting the $30 range. The new SSD also doesn't eat into your RAM, as if you are running many apps or pro programs that need lots of RAM, you will then be at a disadvantage because you will be hitting that RAM limit far faster.
 
I gave it a try with 4GB ram allocated to my steam drive for some games i know to have high load times but are too big for my SSD. Figured it can't hurt to give it a test.

If you cab provide meaningful benchmarks (see my rant about benchmarks "Cheating" above your post)
Please share so we can all get wiser

A simple timer tool like this can do wonders with batch scripts for file copying etc
https://encode.ru/threads/245-timer301-zip
and pelase remember to always works with file in the same physsical place for each test run if we are working on mechanical drives
 
Indeed. All the benchmarks are doing is benching the speed of the RAM, so of course you are going to get crazy numbers. Samsung, Crucial, ASUS etc all have some sort of cache or RAMDrive software now. Most of it has really no use, and as yall said, in many of not most cases can result in slowdowns. Some cases however, where you are working for extended periods with the same program or data, putting it into a RAMDrive can really help, big time if you have lots of spare RAM, however right now, with RAM prices still sky high to where they were a few years ago, and SSD prices plummeting it doesn't make sense to use. For that $30 you could have doubled your SSD storage size, as 250GB SSDs are hitting the $30 range. The new SSD also doesn't eat into your RAM, as if you are running many apps or pro programs that need lots of RAM, you will then be at a disadvantage because you will be hitting that RAM limit far faster.


Its like ppl dont understand that windows already does this dynamically with the available RAM :D
I agree that using 30 bucks for putting a new label on the same thing rather than solving the issue at the core (not having fast storage for the game) seems unoptimal to me as well.

But its like the people that keep advertising for disabling paging file for more performance...No concept of how windows handles memory
 
Its like ppl dont understand that windows already does this dynamically with the available RAM :D
I agree that using 30 bucks for putting a new label on the same thing rather than solving the issue at the core (not having fast storage for the game) seems unoptimal to me as well.

But its like the people that keep advertising for disabling paging file for more performance...No concept of how windows handles memory

Indeed. And most of those "use this", or "disable that" etc etc DO show gains, but often in very, very small specific uses or situations, while everything else suffers, but that almost never comes through in benchmarks.
 
But its like the people that keep advertising for disabling paging file for more performance...No concept of how windows handles memory

Oh god that's definitely the case for Windows 10, just leave the pagefile as it is.
 
Oh god that's definitely the case for Windows 10, just leave the pagefile as it is.

Has been the case for windows for a looong time.
Pagefiles = optimal use of RAM. increased cache size and ensure more relevant data is in RAM

I've sen a person both recommend to disable pagefile and use a memory cleaners. not realizing the total paradox of what he was trying to do
step A: Put all data up in ram
step B: Then but all data down out of ram.
Wait WHAT ...



actually now im currious . ima probabbly run some proper bencmhark on this trials because why the hell not backup my claims up with something :D






Performance:
What I noted was about a 10% increase in items that weren't cached in the Level 2 cache, such as new updates being applied etc due to the cache now write later feature, but the increases could be dramatic on things like frequently played games, such as League of Legends from my SSD, and steam games from the WD Green.


Why is there not data to back this up instead of useless disk benchmarks
?
This would have been such a nice metric to see. Yet without somekind of emperical data all i can do is tuck it up to placebo :(
 
Last edited:
Has been the case for windows for a looong time.
Pagefiles = optimal use of RAM. increased cache size and ensure more relevant data is in RAM

I've sen a person both recommend to disable pagefile and use a memory cleaners. not realizing the total paradox of what he was trying to do
step A: Put all data up in ram
step B: Then but all data down out of ram.
Wait WHAT ...



actually now im currious . ima probabbly run some proper bencmhark on this trials because why the hell not backup my claims up with something :D






Performance:
What I noted was about a 10% increase in items that weren't cached in the Level 2 cache, such as new updates being applied etc due to the cache now write later feature, but the increases could be dramatic on things like frequently played games, such as League of Legends from my SSD, and steam games from the WD Green.


Why is there not data to back this up instead of useless disk benchmarks
?
This would have been such a nice metric to see. Yet without somekind of emperical data all i can do is tuck it up to placebo :(

My favorite are those that clean everything out like that and go "but windows now boots 5 seconds faster!!!" However opening every program or any action etc takes longer because nothing is cached. And you open programs all day, you boot up maybe once a day, or once a week depending on who you are. Now early SSD days, yes, turning off or moving some items to other drives was helpful/needed, but is not the case with newer SSDs for some time now with OSes being SSD aware. Habits die hard though.

As for placebo effect, I am guilty as well, I have done any number of things that felt better at the time at least, only to notice issues over time that just were not worth it.
 
As for placebo effect, I am guilty as well, I have done any number of things that felt better at the time at least, only to notice issues over time that just were not worth it.

i came to hard realaisgin with my own placebo back in my teens after that im anal about testing.
not sure which cpu stress program to use. why not test htem alll (prime95 is king together with linpack. aida64 is only dcent in comparison)
well let do the same with memmory testet ( the built in to windows at boot is amazing horrible and cant even in extend mode not find errors on memory that windows cant boot on)

anyway so far ive test 7zip extration ( probaly CPU bound) and srep extraction (lots i/0)
About to move a gazzillion files around
 
i came to hard realaisgin with my own placebo back in my teens after that im anal about testing.
not sure which cpu stress program to use. why not test htem alll (prime95 is king together with linpack. aida64 is only dcent in comparison)
well let do the same with memmory testet ( the built in to windows at boot is amazing horrible and cant even in extend mode not find errors on memory that windows cant boot on)

anyway so far ive test 7zip extration ( probaly CPU bound) and srep extraction (lots i/0)
About to move a gazzillion files around

Prime95 is not an appropriate program for testing I/O performance.
 
Its like ppl dont understand that windows already does this dynamically with the available RAM :D
I agree that using 30 bucks for putting a new label on the same thing rather than solving the issue at the core (not having fast storage for the game) seems unoptimal to me as well.

But its like the people that keep advertising for disabling paging file for more performance...No concept of how windows handles memory

Windows doesn't buffer writes in RAM to the extent that PrimoCache can. It's waaaay too dangerous to be an default OS feature (if you crash/have a power outage, you'll lose all the data in the buffer), but it's also super useful for certain non-critical things. For example, it sped up demuxing/muxing and transfers with lots of small files immensely for me.

EDIT: Batch demuxing with Staxrip would be a good Windows benchmark, if ya'll are looking for one.

As for the read cache, AFAIK windows will only cache certain types of files, while PrimoCache is totally file agnostic and just caches whatever data is being read. But my knowledge in that area is a little murkey.
 
Windows doesn't buffer writes in RAM to the extent that PrimoCache can. It's waaaay too dangerous to be an default OS feature (if you crash/have a power outage, you'll lose all the data in the buffer), but it's also super useful for certain non-critical things. For example, it sped up demuxing/muxing and transfers with lots of small files immensely for me.

EDIT: Batch demuxing with Staxrip would be a good Windows benchmark, if ya'll are looking for one.

As for the read cache, AFAIK windows will only cache certain types of files, while PrimoCache is totally file agnostic and just caches whatever data is being read. But my knowledge in that area is a little murkey.

Windows cache default all kind of files unless direct no cache read/writes are request or for some specific system files.
this is the non cache read/writes that Primo cache does not adhere too and thereby inflate numbers in benchmarks.
Lazy writes are a part of windows cache as well. you can even select how often you want to clear the write buffers.

If you still do not believe this please link to a proper source of information to backup this claims. I would be interested in it

You do now you can poke around in the Ram and see exactly what files are being cache by windows right?
How did you verify certain files was not cached ?



Prime95 is not an appropriate program for testing I/O performance.
OK... is that just a general information or a specific point that I am missing?
This just seem kinda like a random statement to me.



Anyway som e empirical data on primo cache:

These are 3 diffrent task with scaling from mainly CPU load with light I/O load to medium CPU load/heavy I/O load to almost only I/O load.
Primo cache was no installed in windows standard
Primo cache was then installed using default settings (which was around 6gb of memory cache



*** *** WINDOWS STANDARD *** ***

*** 7zip extraction 2GB file
Global Time = 141.617 = 00:02:21.617 = 100%
Global Time = 140.276 = 00:02:20.276 = 100%
Global Time = 142.023 = 00:02:22.023 = 100%


*** SRep64 decompression 3GB file
Global Time = 27.238 = 00:00:27.238 = 100%
Global Time = 26.817 = 00:00:26.817 = 100%
Global Time = 26.473 = 00:00:26.473 = 100%


*** xCopy 13,698 Files, 414 Folders 4.15GB
Global Time = 196.530 = 00:03:16.530 = 100%
Global Time = 41.434 = 00:00:41.434 = 100%
Global Time = 41.480 = 00:00:41.480 = 100%



*** *** PRIMO CACHE 6GB *** ***

*** 7zip extraction 2GB file
Global Time = 141.586 = 00:02:21.586 = 100%
Global Time = 141.695 = 00:02:21.695 = 100%
Global Time = 141.056 = 00:02:21.056 = 100%

*** SRep64 decompression 3GB file
Global Time = 35.056 = 00:00:35.056 = 100%
Global Time = 16.536 = 00:00:16.536 = 100%
Global Time = 16.006 = 00:00:16.006 = 100%

*** xCopy 13,698 Files, 414 Folders 4.15GB
Global Time = 179.651 = 00:02:59.651 = 100%
Global Time = 108.343 = 00:01:48.343 = 100%
Global Time = 108.343 = 00:01:48.343 = 100%




So in the strong CPU bound case. Nothing is different ( no surprise), so moving on.

On the medium CPU load with HIGH I/O load we see a different.
Windows own cache keeps the times steady at around 26sec to complete the task
Primo cache first show of a slow 35 sec. follow by 2 fast 16 secs

Now this could indicate a a simple greedy FIFO cache.
it slow the first time because it has to empty its cache and fill it with the new data. but it retains data for a long time so the second run is accelerated.
What you have to ask yourself here is how often do you the exact same thing on the exact same data. vs loading new data.
This would be reason for slow load times of games etc etc.

Moving on the the many file strong I/O case primo cache edges in a win on the first run
but then does not compete with windows standard cache at all.
The roles are reversed here.
Maybe Primocache has issues with the tons of small files some kind of overhead. but that would go against it beeing a simple greedy FIFO cache.



I will draw this up as a draw here
More testing is needed to verify marvelous improvements form Primocache.


-- edit ---
Link to RAMmap which can show you cached files from e.g. a copy operation
https://docs.microsoft.com/en-us/sysinternals/downloads/rammap
 
Last edited:
Windows cache default all kind of files unless direct no cache read/writes are request or for some specific system files.
this is the non cache read/writes that Primo cache does not adhere too and thereby inflate numbers in benchmarks.
Lazy writes are a part of windows cache as well. you can even select how often you want to clear the write buffers.

If you still do not believe this please link to a proper source of information to backup this claims. I would be interested in it

You do now you can poke around in the Ram and see exactly what files are being cache by windows right?
How did you verify certain files was not cached ?




OK... is that just a general information or a specific point that I am missing?
This just seem kinda like a random statement to me.



Anyway som e empirical data on primo cache:

These are 3 diffrent task with scaling from mainly CPU load with light I/O load to medium CPU load/heavy I/O load to almost only I/O load.
Primo cache was no installed in windows standard
Primo cache was then installed using default settings (which was around 6gb of memory cache



*** *** WINDOWS STANDARD *** ***

*** 7zip extraction 2GB file
Global Time = 141.617 = 00:02:21.617 = 100%
Global Time = 140.276 = 00:02:20.276 = 100%
Global Time = 142.023 = 00:02:22.023 = 100%


*** SRep64 decompression 3GB file
Global Time = 27.238 = 00:00:27.238 = 100%
Global Time = 26.817 = 00:00:26.817 = 100%
Global Time = 26.473 = 00:00:26.473 = 100%


*** xCopy 13,698 Files, 414 Folders 4.15GB
Global Time = 196.530 = 00:03:16.530 = 100%
Global Time = 41.434 = 00:00:41.434 = 100%
Global Time = 41.480 = 00:00:41.480 = 100%



*** *** PRIMO CACHE 6GB *** ***

*** 7zip extraction 2GB file
Global Time = 141.586 = 00:02:21.586 = 100%
Global Time = 141.695 = 00:02:21.695 = 100%
Global Time = 141.056 = 00:02:21.056 = 100%

*** SRep64 decompression 3GB file
Global Time = 35.056 = 00:00:35.056 = 100%
Global Time = 16.536 = 00:00:16.536 = 100%
Global Time = 16.006 = 00:00:16.006 = 100%

*** xCopy 13,698 Files, 414 Folders 4.15GB
Global Time = 179.651 = 00:02:59.651 = 100%
Global Time = 108.343 = 00:01:48.343 = 100%
Global Time = 108.343 = 00:01:48.343 = 100%




So in the strong CPU bound case. Nothing is different ( no surprise), so moving on.

On the medium CPU load with HIGH I/O load we see a different.
Windows own cache keeps the times steady at around 26sec to complete the task
Primo cache first show of a slow 35 sec. follow by 2 fast 16 secs

Now this could indicate a a simple greedy FIFO cache.
it slow the first time because it has to empty its cache and fill it with the new data. but it retains data for a long time so the second run is accelerated.
What you have to ask yourself here is how often do you the exact same thing on the exact same data. vs loading new data.
This would be reason for slow load times of games etc etc.

Moving on the the many file strong I/O case primo cache edges in a win on the first run
but then does not compete with windows standard cache at all.
The roles are reversed here.
Maybe Primocache has issues with the tons of small files some kind of overhead. but that would go against it beeing a simple greedy FIFO cache.



I will draw this up as a draw here
More testing is needed to verify marvelous improvements form Primocache.


-- edit ---
Link to RAMmap which can show you cached files from e.g. a copy operation
https://docs.microsoft.com/en-us/sysinternals/downloads/rammap

I apologise I missed the part where you shifted the discussion to CPU testing.
 
I must be the only one that just uses this for L2. I've never configured it any other way, just setup to allow my games to be on bulk storage.. while playing from a much smaller SSD.
 
The L2 part akak SDD cache of HDD is definetly a unique point comapred to not installing it.
But hen we are back at just buying a bigger sdd with the money perhaps, im not sure in prices really

also asimilar effect can really be done with simple batch scripts by yourself.
just move af symlink a folder

no need to shell out cash for something you can do yourself with a lot more direct control over it



ZodaEX
Yeah it was just show that I love doing testing of stuff and er very sceptical on anything that might have the slight hint of placebo
 
Last edited:
Back
Top