14

I have a HDD which has read/write speeds around 100MB/S

Then I found an application called PrimoCache. This software(free to test) claims to increase HDD speed tens of times.

I did run this software and followed instructions. Interestingly CrystalDiskMark utility showed read/write speeds of Gigabytes per seconds. But in actual test of copy a file from windows explorer (from internal SDD) to this HDD(external read/write values remained same as around 100MB/s.

I want to see if anyone have tested this software successfully or this software somehow fools the disk benchmarking tools to show huge numbers. If you have some time to download this software and make it work to increase the HDD speed, please share the procedure here.

enter image description here

6
  • 41
    This is basically cheating the benchmark, I expect. CrystalDIskMark wants to test your actual disk speed so it bypasses caches, but then PrimoCache adds a cache that it doesn't bypass, so you end up testing the cache speed instead of the disk speed. Commented Sep 28, 2022 at 17:44
  • 16
    Without even looking at their advertising copy, best guess is this is some type of RAMdisk. It will, via some algorithm or other, try to guess what you need to read next & copy it into RAM, where - yes - it will be a thousand times faster. Any time what you're reading/writing is not in that cache, you're back to the real speeds. This type of structure can be great if you have an absolute sh*tload of RAM, but miserable if you don't.
    – Tetsujin
    Commented Sep 28, 2022 at 17:44
  • 26
    There is no SSD or mechanical HDD on the market that actually supports 9500+ MB/Sec write speeds except perhaps PCIe Gen 5 and those haven’t been released. The software in question is snake oil
    – Ramhound
    Commented Sep 28, 2022 at 17:52
  • 6
    If it sounds too good to be true, it most likely is. This sounds pretty much like a cache that is placed in between I/O calls and your harddisk.
    – Robidu
    Commented Sep 29, 2022 at 0:14
  • 6
    Note that most OSes in the market already cache both read & write for decades. For the cost of the license, you can just buy a RAM stick which not only adds more cache but is usable for actual works.
    – Martheen
    Commented Sep 29, 2022 at 21:49

7 Answers 7

20

To add to @r2d3's answer, I've ran some tests on my own RAMDisk.

The software I've used is ImDisk, on a 512MB FAT drive (mounted as the R:/ drive).
My system has 2x8GB Kingston KHX3000C15D4/8GX (3000MT/s), on a MSI B450 Tomahawk Max motherboard and an AMD Ryzen 5 3600 CPU.
The RAM usage was about 47-48% at the time of testing.

Doing a 256MiB test, with CrystalDiskMark 8.0.1, I got these results: Read speed - 8577.15MB/s, Write speed - 12726.56MB/s

Without the use of that software, I (kinda) replicated the results you've shown.

In conclusion: that software is using a ramdisk as cache, or some RAM caching.

Your HDD won't magically start working at over 9000MB/s, specially a SATA3 HDD.
And since you've specified a read/write speed of 100MB/s, my guess is that you have a 5400RPM HDD or an half-full HDD.


In fact, quoting the website:

Load Apps and Data Faster

Effectively cache your frequently used applications, documents and other data into faster storage devices, accessing them at up to RAM-like or SSD-like speeds. Make your computer more responsive for creating, gaming and producing, with less boot and load times.

Accelerate Writing

Complete write requests very quickly by temporarily storing incoming data into RAM or SSD storage first and writing them back to target disks later. Enable your computer to handle heavy or stream write IOs, while reducing writes and wear on disks.

It describes exactly what it does, so, it isn't really a scam, as some users have been describing it as.
It has marketing fluff all over it, as many products do.

It is a pricey software, and you might be able to find cheaper/free alternatives.

Also, the website is completely absent of any indication of possible risks, like loss of data.

9
  • 20
    Wait.. doesn't Windows and Linux already do exactly these things since… a long time ago?
    – pipe
    Commented Sep 29, 2022 at 5:10
  • 3
    @pipe yes they have, but it is possible for apps to disable them, for example for ensuring that data is actually written to disk or benchmarking. The best explanation is that PrimoCache just prevents disabling cache for benchmark.
    – ojs
    Commented Sep 29, 2022 at 9:07
  • 4
    Yes, up to a certain point. Linux or BSD, with ZFS, will use as much RAM as possible, for caching. Linux does delay disk writes, by storing the data into RAM. Windows does have a "Cache manager", but I have no idea how/if it works. I do know that Windows has the Prefetch service, but personally, I think it is absolute garbage and will often cause lots of drive accesses for a long time, basically making the system unusable for a while. @ojs I have no idea if it is just messing with benchmarks. It is a big accusation, which I don't think it is the case, based on O.P.'s comments. Commented Sep 29, 2022 at 9:27
  • 8
    @pipe: Yes, but only if you mount with an option like barrier=0 to not force ordering of committing some writes to disk (e.g. of journal vs. non-journal), in order to make things recoverable without a full-disk FSCK on reboot. I assume this software acts like a write cache that reports data as safely on disk even when it isn't, which defeats write barriers and thus is very bad if your system crashes, or is succeptible to power outages. (no UPS or laptop battery.) Also defeats application-level fsync, common in databases (including small ones used by web browsers and so on.) Commented Sep 29, 2022 at 10:04
  • 2
    @Robidu: Yes, exactly my point. That's why real systems trade some performance for crash resiliency, especially filesystem consistency. With a a battery-backed system, you can relax some of the guarantees (e.g. not requiring writes to actually commit to disk, vs. just being sent over SATA or NVMe to the drive's internal write cache, where they will eventually commit as long as power stays on, regardless of the CPU locking up.) But as you say, for safety against software bugs you can't fully relax everything. (Or failing hardware other than RAM, or bumping the power plug...) Commented Sep 29, 2022 at 23:41
31

This application is a scam reminding me of the "RAMDoubler" application. It probably works by hooking the Windows IOControl for disk access, putting the content the data CrystalDiskMark into a cache and returning the call to this system function. The test does not verify if the test data is really located on the disk. As user253751 and Tetsujin already stated this is a kind of RAM test.

You cannot exceed the sustained transfer speed of a disk. Therefore the read speed does not originate from disk access but from previously buffered written test content in RAM.

In order to show that Primocache is scam I recommend you to try out two things:

  1. Increase the test size beyond the amount of available RAM

and/or

  1. Use h2testw to write and read files with pseudorandom content. Look at the speed indication of h2testw.

If you are really paranoid and are operating on file system level you should only use software that uses pseudorandom data for speed tests. Pseudorandom data is near garbage and has the nice property of having high entropy (disorder). That data can hardly be compressed.

Scamware like Primocache could detect regular patterns and use compression techniques to generate an increase in speed.

The best is pseudorandom sequences where you can set a starting value and an algorithm. That is too much too handle for Scamware.

Good luck!

6
  • 7
    Increasing the test size beyond the amount of available RAM doesn't show that Primocache is a "scam", it merely shows that Primocache isn't magic and is instead caching files to RAM (which seems to be what they're claiming to do). A caching application that uses compression would be expected to not perform all that well on pseudorandom data that it cannot compress well, whereas it may very well be able to compress real-world data well, and therefore perform much better for real-world use.
    – NotThatGuy
    Commented Sep 29, 2022 at 13:38
  • 11
    I wouldn't call a program that is has "cache" in its name and describes itself as effectively caching data to be a scam. It's actually very descriptive of what it does.
    – qwr
    Commented Sep 29, 2022 at 20:34
  • 9
    Windows already caches disk reads (see for example memory mapped files), so it is scammish at the very least
    – Yorik
    Commented Sep 29, 2022 at 21:08
  • 6
    @qwr Claims can be both factual and misleading, and these claims are both. It's fleeceware, and the people who buy it will be tech illiterate parents trying to quiet a petulant child complaining about the family computer's slowness, or paycheck-to-paycheck retail employees desperate to keep a 7 year old laptop running while they finish their master's in theology. I do outreach work to help people like that - techies have blind spots around things they think are obvious, and we have to learn to see around them, and maybe donate some low runtime hard drives when we can.
    – user530015
    Commented Sep 30, 2022 at 1:44
  • 2
    PrimoCache does ... pretty much what it says on the tin. The benefits of a third-party RAM-only cache are questionable, but it does also have a SSD cache mode that was traditionally quite useful in the absence of a first-party SSD cache solution. That said, it's less useful now that you can get a reasonably large SSD directly.
    – Bob
    Commented Sep 30, 2022 at 13:44
5

The software does intentionally cheat on a benchmark. The benchmark tries to measure the actual speed of a hard drive or SSD drive, not the speed of any RAM cache, and the software manages to work around that, tricking the benchmark into thinking you have a really fast hard drive.

But it doesn't matter. What matters is how much it helps you with the things you are doing. It will be useless if you copy 10GB video files all the time. But as the benchmark shows, it might be very effective if you use small amounts of data only. Like a few hundred megabytes in small files on a spinning hard drive.

I'd say try it out for free and find out if it helps with your usage. Then see the cost and use it or don't. And you want your data to be backed up: The OS will think that data is written when this software tells it is written. If you shut down your computer, most likely it will actually shut down when it thinks all caches are written to your drives. That could go wrong with this drive.

PS. There are high-end SSD drives nowadays with a small amount of fast and expensive SSD storage, and a huge amount of cheap and slower SSD storage. Small meaning gigabytes. So writing to the SSD will write to the small, expensive part which is fast, and then the SSD drive continuously moves data from small, fast to large, slow, cheap storage. That may actually write a 10GB file fast, and stay fast if you take breaks between large operations.

PS. I learned just recently there are SSD memory cells that can operate either as small, fast or large, slow memory cells, changeable when the drive is running. Right now I have a "2 TB" SSD drive that could operate as a fast 500GB or very slow 2TB drive or anything in between.

1
  • 9
    Your OS already has a disk cache that helps with the kind of cases where using RAM as a disk cache helps. It's hard to imagine this software could do anything to change that significantly. Commented Sep 29, 2022 at 22:23
4

TL; DR: Does Primocache application really work?
Yes it does.


Disclaimer: I'm a happy registered user of PrimoCache. 45 years old, I've had my share of both smartdrv and Host Memory Buffer NVMe SSDs.

I have a HDD which has read/write speeds around 100MB/S

I'm expecting a peak speed here, ie sequential, no multi-threaded.

But in actual test of copy a file from windows explorer (from internal SDD) to this HDD read/write values remained same as around 100MB/s.

Let me sum it up, to be sure not to miss the point. You did not set up a cache on your SSD, you set up a cache on your HDD, and you copied a file from your SSD to your cached HDD. Two things to consider here:

  1. what is the speed of your SSD? Have you tested copying a file from SSD to a RAM Disk?
  2. when you're writing the file, as long as the cache is not full, the data is copied to the cache. Once the cache is full, the file will be copied to the physical disk, getting you back to slow speeds.

I want to see if anyone have tested this software successfully or this software somehow fools the disk benchmarking tools to show huge numbers.

I have tested this software extensively before buying it. I quickly gave up on CrytalMark that only shows irrelevant values. What matters to me is actual file operations. And I can testify that, as long as you keep in the limits of your cache, this software does what it tells it does.


But to answer completely to your questions, let me write a bit more about how I use this software.
I have a SATA cacheless SSD, 16GB of RAM. Software engineering has been my main job for 20+ years.
I have set up a 1.25GB of L1 Write Cache, with defer mode, buffer strategy, flush on sleep.
I don't care that much on prefetching data. My main reads are loading the OS (un-cached), loading Eclipse (once a day, always-on then), launching and closing several Firefox/Chrome windows and tabs (could be cached, but Windows keeps the programs in memory, so irrelevant).
I care a LOT about writing data.

  1. you mentionned speed, and speed IS relevant. Gone are the days where I could stand the hourglass.
  2. looking at the PrimoCache stats, it becomes astounding how much your system writes to your disk, without needing it. It writes data, that wouldn't need to be written, or that would be rewritten later.
  3. by writing less, I make my SSD last longer. I'm not sure the PrimoCache fee is cheaper than a new SSD, but I'll be the one wearing it off. Not an other process.
  4. I've also setup a 512MB RAM disk, for %TEMP% folder, for good measure.

Does Primocache application really work?

Yes Sir, yes it does.
I'm not the only one to say it. See Dave's video, retired Microsoft engineer, and author of an Amiga cache software Supercharging Windows Disk Speeds
Is it the product for you? Well, try the demo. 30 days is a good amount of time to make up your mind. Run your computer as usual (ie, no CrystalMark), and see. Then you could add your answer here ^^

3
  • 1
    Didn't want to be too technical here, but it seems to me that Windows HMB (so, NVMe caching) is limited to 64MB.
    – Olivier
    Commented Sep 30, 2022 at 14:30
  • 1
    The next releveant question, of course, is "is the PrimoCache licence worth a NVMe Cached extra?"
    – Olivier
    Commented Sep 30, 2022 at 14:34
  • Well I do have an SSD with write speed of 2000GB/s. I did added cache using primocache to both my SSD and my 4 TB hard drive(each 8 gigabye). Writing speed in windows explorer showed 110MB/S when tried to copy a folder(10 GB) from SSD to HDD. Am I missing something? Note My HDD is external.
    – S Nash
    Commented Sep 30, 2022 at 18:24
1

PrimoCache - formerly FancyCache - is not cheating software. It's caching software.

The main use of PrimoCache is multi-layered caching, using RAM and SSD as two tiers. Caching improves performance for frequently accessed files, at the expense of spending RAM or SSD space. It does not make your actual HDD faster.

It is highly useful software for knowledgeable power users, and for admins that run low-cost bare metal servers. It takes understanding of cache tiers and manual effort to set up, and is not important to most basic home users.

0

For M2 its does make a difference thought it is small. What primo did which made huge gains for me was for the raid HDD. In my scenario I have 1x2TB M2, 2x4TB HDD and a spare 512GB SSD. You can use a portion of your ram as cache for M2 but that wont make a big difference.

The array is a different story though. By using 2GB of RAM as L1 cache and the whole SSD as L2 Cache for the HDDs its like having an 8TB SSD. This is my setup:

PrimoCache Setup for HDD

0

Scamware? Nonsense! No such thing.

I've been using it for a few years. I/O is dramatically improved especially when adding a SSD for L2 cache.

Real world performance increases are very noticeable. There are several fine tune settings such a deferred write etc, The more RAM the better & faster! Apps start much quicker. Example: Word 2019 up and running in about a second. Excel in under a second, Publisher almost instantaneously. Photoshop ready in a few seconds. File transfers super fast until buffer full.

2
  • 1
    None of this is believable. Please show us some objective tests. Commented Feb 13 at 7:19
  • you know what "caching" means?
    – POB Mail
    Commented Feb 28 at 19:40

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .