Feedback

Does StoreMI Bring AMD on Par with Intel Optane Memory Caching?

Author:
Subject: Storage
Manufacturer: Various

We aim to find out

Back in April of this year we first took a look at the storage performance of the then-new X470 chipset for the 2nd generation of Ryzen processors. Allyn dove into NVMe RAID performance and also a new offering called StoreMI. Based on a software tiered storage solution from Enmotus, StoreMI was a way for AMD to offer storage features and capabilities matching or exceeding that of Intel’s mainstream consumer platforms without the need for extensive in-house development.

Allyn described the technology well:

AMD has also launched their answer to Intel RST caching. StoreMI is actually a more flexible solution that offers some unique advantages over Intel. Instead of copying a section of HDD data to the SSD cache, StoreMI combines the total available storage space of both the HDD and SSD, and is able to seamlessly shuffle the more active data blocks to the SSD. StoreMI also offers more cache capacity than Intel - up to 512 256GB SSD caches are possible (60GB limit on Intel). Lastly, the user can opt to donate 2GB of RAM as an additional caching layer.

View Full Size

We recently did some testing with StoreMI after the release of the 2nd generation Threadripper processor evaluation was out of the way, just to get a feel for the current state of the software offering and whether or not it could really close the gap with the Optane caching solutions that Intel was putting forward for enthusiasts.

Continue reading our look at StoreMI and Optane Memory Caching!

The Pricing Question

Before diving into the results of our testing, let’s talk about the cost question. AMD is adamant that StoreMI is a “free” feature while the Intel Optane solution requires the consumer to buy special hardware they might not otherwise be purchasing. The premise is as follows:

  • Most modern PC enthusiasts that are building a new system today will buy a large hard drive for mass storage, in the 4TB to 12TB range.
  • Most modern PC enthusiasts will also buy a single SATA or NVMe SSD ranging from 250-500GB in capacity to act as the primary OS and active applications drive.

The AMD StoreMI solution will work and accelerate your hard drive in this situation, combining the SSD (up to 256GB of it) and HDD into a single, tiered storage solution. For Intel consumers, you don’t really need to accelerate the SSD but you DO want to accelerate the storage on the HDD so that any games you have stored there continue to start up quickly. To do that, you’ll need to purchase a 32-64GB Optane Memory drive from $60-120 and configure it.

So, there is truth to the claim of cost efficiency on the side of AMD. All else being equal, including the processor, motherboard, hard drive, and SSD, you are going to need to buy one additional component to accelerate a separate hard drive for Intel platforms. Going with the AMD Ryzen processor and StoreMI solution saves you up to $120+.

Performance Testing StoreMI vs Optane Memory Caching

Now let’s dive into the performance we measured on our two systems. Configurations were as follows:

Intel

  • Intel Core i7-8700K
  • ASUS ROG Strix Z370-F Gaming (BIOS 0805)
  • 2 x 8GB DDR4-3200 Memory (Running at DDR4-2666 speeds)
  • MSI Radeon RX 580 Gaming 8GB
  • Intel Optane SSD 800P 118GB (Boot Drive)
  • Intel Optane Memory M10 Series 64GB SSD (Caching Drive)
  • Western Digital 8TB Red Hard Drive

AMD

  • AMD Ryzen 7 2700X
  • Gigabyte X470 Aorus Gaming 7 (BIOS F4G)
  • 2 x 8GB DDR4-3200 Memory (Running at DDR4-2997 speeds)
  • MSI Radeon RX 580 Gaming 8GB
  • Intel Optane SSD 800P 118GB (Boot Drive)
  • Samsung 860 EVO 250GB SSD (Caching Drive)
  • Western Digital 8TB Red Hard Drive

Editor's Note: The first revision of this article incorrectly omitted the RAM caching option in StoreMI. the 2GB RAM cache tier was enabled for all of this testing, but we would expect very similar results without the RAM cache enabled. 

You’ll notice that we used an Optane SSD 800P for our boot drive in both configurations – this is simply acting as the OS host during our testing of both StoreMI and Optane Caching as secondary storage options. Only in a couple of cases did we need to move to having the StoreMI/Optane Cache solutions as the primary boot arrangement to get proper results.

PCMark 10

Developed by the global standards organization, UL, PCMark is an industry standard benchmark that aims to quantify overall system performance for a wide variety of users through a comprehensive series of test. Scenarios tested include Productivity, Digital Content Creation, and Essentials.

Test Setup

For this testing, PCMark 10 was installed to the secondary storage drive on both the AMD and Intel test platforms. Once installed, the “Standard” workload was run, and the resulting data gathered. For system configurations featuring drive caching solutions, an idle period of one-hour was observed, before rebooting the system to purge the contents of RAM and running the test again. This one-hour period allows the caching solution to properly perform its duties.

View Full Size

Comparing the two offerings, we see while Intel’s solution provides a 3.8% performance increase with caching enabled but StoreMI provides an over 7% performance increase when enabled on the Ryzen 7 2700X and a Samsung 860 EVO 250GB SSD. Do note, however, that the Intel system scored higher overall, with the AMD system having more to gain as it was starting from a lower value.

SYSmark 2014 SE

SYSmark 2014 is a benchmark that aims to emulate real-world usage in several scenarios including Office Productivity, Data/Financial Analysis and Media Creation. Additionally, the Responsiveness test includes tasks like application and file loading, in order to quantify responsiveness in typical user scenarios.

Test Setup

SYSmark 2014 SE testing was performed with the Western Digital 8TB Hard Drive as the primary boot drive. Once a baseline score was recorded, the various caching solutions were enabled.

View Full Size

Both the Intel Optane Memory caching solution and StoreMI saw performance increases when utilized in conjunction with SYSmark. With the 64GB Optane M10 as a caching device, the Intel platform saw a 30% increase in overall score.

With the 250GB Samsung 860 EVO as a caching drive, the AMD system saw a 26% increase in system performance compared to just the hard drive.

Adobe Lightroom CC

Used by creative professionals and hobbyists alike, Adobe Lightroom is a popular application for cataloging and editing RAW photos. Users import RAW photos from their camera, which are then converted to the generic Adobe RAW format and copied to their Lightroom Catalog. As users import photos, the Lightroom catalog continues to grow, requiring a high-capacity storage device.

Test Setup

Since users only import their photos to Lightroom one time, this is not a test that can be repeated multiple times to take advantage of caching solutions. However, it does provide insight into write-back caching provided by each unique caching solution. In this scenario, the Lightroom catalog is being stored on a secondary hard drive, and 700 Sony ARW RAW photos totaling 5.8GB are imported. Performance is measured by the total time elapsed by the importing process.

View Full Size

While we see StoreMI provide a modest performance increase over the HDD only solution, the 6.5% speed increase pales in comparison to the almost 20% improvement provided by Intel’s Optane Memory Caching solution.

Content Management

Another pain point for digital content creators is the process of archiving projects for future use and posterity. While the majority of video editing takes place on fast storage devices like SSDs, these projects are often archived onto slower, mechanical storage devices.

Test Setup

In this scenario, a project folder totaling 50GB is transferred from the fast, but low capacity Intel 800P 118GB SSD to the 8TB Western Digital Hard Drive. This test is aimed at providing further insight into how caching solutions can accelerate large file transfers.

View Full Size

Both solutions provide a significant speedup for the file copy to the secondary hard drive with the accelerated systems in place. The 2700X with StoreMI sees a 90% improvement in speed while the 8700K with Optane Caching sees more than a 2x speedup.

Ashes of the Singularity: Escalation and Grand Theft Auto V

With increasing file sizes, PC games are often relegated to lower mechanical storage solutions instead of faster, but lower capacity flash storage solutions. Ashes of the Singularity: Escalation and Grand Theft Auto V represent expansive gaming titles that load in large data sets when you enter the world.

Test Setup

With a Steam game library configured on the secondary drive, Ashes of the Singularity: Escalation was installed, and then the initial main menu load time was recorded. For Grand Theft Auto V, the scenario being timed was loading an end-game save file in the “Story Mode.”

For caching setups, 5 minutes were afforded between runs, and then a system reboot was performed to purge the system RAM contents.

View Full Size

View Full Size

Both Ashes of the Singularity and Grand Theft Auto V exhibit the same behavior. The first time the games are launched with StoreMI enabled, the load time increases when compared to loading just from the hard drive, but only slightly.

However, once the game has been loaded and is in the cache, loading times speed up to match SSD-level performance. The Intel 8700K system runs a bit faster on Ashes but is in a virtual tie with the 2700X platform on GTAV.

Effective Cache Size Testing

To evaluate what effect larger cache drive sizes had on normal user workflow, SYSmark 2014 SE was utilized. Emulating consumer workloads, SYSmark 2014 SE helps us gain insight into how varied, but repetitive tasks are handled by each caching technology, especially in regards to how cache size effects overall performance.

Test Setup

In this configuration, Windows 10 and SYSmark 2014 SE were installed to an 8TB Western Digital Red Hard Drive, and caching was enabled on both platforms, the 64GB Optane M10 Module for the Intel platform, and a 250GB Samsung 860 EVO for the AMD StoreMI platform. SYSmark was then run for 5 iterations, with the preconditioning feature turned off, to better emulate a standard users’ workflow.

View Full Size

Here we can see that while the Optane Memory caching solution has consistency issues in how much it is speeding up the SYSmark Responsiveness test, the StoreMI solution with its much higher capacity 250GB cache shows consistent results across all 5 test iterations. Since you cannot enable Optane Caching with a larger than 64GB drive, this remains a competitive advantage to AMD’s StoreMI technology.

Allyn did find the inconsistent Optane caching behavior we observed to be 'odd' and not what he typically sees when performing repetitive tasks in his prior testing. We are not sure of the cause of this inconsistency, but it is possible that our specific configuration did not play nicely with repeated SYSmark runs.

Closing Thoughts

First, do not consider this story to be the deep dive technical break down of Optane Memory Caching versus StoreMI that someone like Allyn might do for us in the near future. Instead, this story is really meant to be a sanity check on the state of Ryzen’s ability to match one of the unique features that has made Intel systems stand out for a while: storage caching. Based on our testing, it appears that is indeed the case: StoreMI is a suitable substitute for Optane Memory Caching and can even provide advantages in some areas.

In general, all else being equal, the Optane storage solution does provide faster-accelerated state performance than StoreMI does with a SATA SSD. That kind of makes sense: Allyn has proven over and over that Optane is the fastest storage solution on the market in terms of raw performance. When properly configured and operating in a cached state, Optane over NVMe SHOULD be faster than TLC over SATA. The differences are not that dramatic and consumers would be hard pressed to the see the difference in the real world.

View Full Size

However, the differences for BOTH platforms comparing HDD-only and HDD+cache performance are, as expected, staggering. I don’t think it’s a surprise to anyone that hard drives are slow and generally suck for enthusiasts that are used to SSDs in any form or function. Using a cached or tiered storage system offers the best of both worlds – capacity and speed.

StoreMI does need a bit of work still. The user interface is just plain bad and it's complicated to undo: you won’t be able to simply rollback a StoreMI enabled system. (As with everything in life, make sure you are taking precautions with ALL STORAGE and backing up to external network systems or Dropbox, etc.) Optane Memory lets you easily enable and disable the cache with a simple reboot. The inherent differences of caching and tiered storage systems are partially to blame for this headache for AMD, but in reality, I see no reason that Enmotus couldn’t solve this problem.

It has room to improve, but AMD StoreMI definitely closes a feature gap that the Ryzen platform had compared to Intel mainstream systems. 

Review Terms and Disclosure
All Information as of the Date of Publication
How product was obtained: The product(s) was on loan from AMD and Intel for the purpose of this review.
What happens to the product after review: The product remains the property of AMD and Intel but is on extended loan for future testing and product comparisons.
Company involvement: AMD and Intel had no control over the content of the review and was not consulted prior to publication.
PC Perspective Compensation: Neither PC Perspective nor any of its staff were paid or compensated in any way by AMD or Intel for this review.
Advertising Disclosure: AMD has purchased advertising at PC Perspective during the past twelve months.
Affiliate links: This article contains affiliate links to online retailers. PC Perspective may receive compensation for purchases through those links.
Consulting Disclosure: AMD and Intel are current clients of Shrout Research.

August 28, 2018 | 04:36 PM - Posted by pdjblum

great stuff

first time i have seen the two compared

August 29, 2018 | 11:02 AM - Posted by Power (not verified)

The test is a bit pointless. 64GB Optane is twice the price of 256GB Samsung Evo SSD.

August 28, 2018 | 07:28 PM - Posted by ipkh

Did you try storemi with the optane drive?

August 30, 2018 | 10:10 AM - Posted by Ryan Shrout

We did, but didn't see a lot of differences. If anything, Optane seemed a bit slower on StoreMI than the SATA SSD we used.

August 28, 2018 | 07:31 PM - Posted by Gary Fox (not verified)

Gee, I wonder what a DRAM cache would do.

August 30, 2018 | 10:10 AM - Posted by Ryan Shrout

Actually it WAS enabled, we updated the testing setup above.

August 28, 2018 | 09:59 PM - Posted by Zoea

Could you compare StoreMI to Samsung's RAPID mode from 5.2.1 Magician maybe please?

August 29, 2018 | 12:09 AM - Posted by Allyn Malventano

RAPID is a DRAM cache that goes way faster than StoreMI (but is a smaller cache and is also volatile). Actually, StoreMI's DRAM caching function doesn't really do much in my experience.

August 29, 2018 | 05:04 AM - Posted by Zoea

Thanks Allyn

I guess that explains a lot ~ lol

If you don't mind, would you recommend a 970 pro on this basis as an OS drive and am I likely to notice different from 850 evo 500Gb I currently am using?

August 29, 2018 | 03:42 PM - Posted by Mostly Sane (not verified)

How far we've come.... I remember using a MS-DOS-based RAM disk to cut out the near-constant temp file thrashing on my _4_GB HDD, back in the Netscape/AOL days. Probably doubled the life of that old platter, lol!

August 29, 2018 | 06:16 PM - Posted by ReRe (not verified)

Why would you compare performance of a SATA interface SSD to an NVMe device? And how are you not doing any IO testing? This is a sad excuse of a comparison and not even worth publishing unless the intention is to be completely irrelevant.

August 30, 2018 | 10:11 AM - Posted by Ryan Shrout

We tested this way because this is how AMD is presenting the technology to consumer. The "NVMe" device you refer to is the Optane drive and we HAVE to use that to get Optane Memory caching to work. 

September 4, 2018 | 09:07 AM - Posted by fla56 (not verified)

er well either use the SSD as the boot drive and Optane as the cache in both cases OR use Optane / SSD cache and HDD boot

but this test looks like Apples to fish

September 15, 2018 | 03:58 AM - Posted by Anonymous_Man (not verified)

This is Apples to rotten fish in every aspect.

Also, Optane will not do much good in StoreMI, because StoreMI is NOT a cache solution. It will speed it up significantly only for a few moment before it is erased and demoted back to HDD. But it should be enough if all you want to do is the benchmark. If you use something like 9xx NVMe drive, the content will stay there, permanently, unless 9xx drive is full. Which in case of 32GB Optane, will be full in a a very short time.

August 30, 2018 | 02:28 PM - Posted by Syl (not verified)

I used Storemi for a few months and came across some issues, enough issues that I had to nuke the entire Storemi drive from orbit because Documentation for Storemi/FuzeDrive is essentially nonexistant.

Out of curiosity I decided to go for Primocache afterwards, which has both been faster, more efficient, and more flexible for my situation. I think it definitely is worth looking at.

August 30, 2018 | 04:07 PM - Posted by elites2012

seen linus tech review. not something i want to mess with. took too long to get it to run correctly.

August 31, 2018 | 05:45 PM - Posted by TheRiceKing

That is false. Linus reviewed Intel Optane but never "StoreMI" or "Enmotus FuzeDrive"

I am a gamer and a photographer and it is so easy and beneficial to use that I ditched the AMD branded StoreMI and went and purchased the FuzedDrive Plus. This technology is very real and extremely good.

FYI- The Plus version is even better.

FuzeDrive Plus
Up to 1TB Fast Tier SSD1
NVMe, Optane, m.2, SATA SSD
4GB FuzeRAM
Windows 10
$59.99 <-- Cheap

I am using it with a 1TB m.2 and a 6TB HDD. I don't know how I can ever go back to anything else until 6TB SSD are affordable enough. We are years away from that. I would love to see a review comparison of AMD StoreMI vs FuzeDrive Plus. [Yes, I know is made by the same company with AMD branding but the software specs are different]

September 4, 2018 | 11:11 PM - Posted by Alberrt89 (not verified)

Its still going to be AMD free vs FuzeDrive $60........
But I would love to see an all three comparison, as well as any other kind of memory optimizer.

September 3, 2018 | 04:18 AM - Posted by Prodeous@Work (not verified)

I'd expect better results with the 860 Pro, or an M.2 drive. Any plans to update testing with other SSD's on AMD's solution?

September 4, 2018 | 09:10 AM - Posted by fla56 (not verified)

i really fail to see the point of this test and suggest you should explain:

-who on Earth would use Optane as a boot drive then cache it with an SSD?

-where is testing of typical usage case of Optane cache + SSD/HDD

September 4, 2018 | 11:16 PM - Posted by Alberrt89 (not verified)

Did the AMD hating idiots at PCper notice it was 2018 ? And you still havn't updated to a DISQUS comment service ? Instead of giving thumbs up or down I'm suddenly a robot under the current system.

September 15, 2018 | 03:50 AM - Posted by Anonymous_Man (not verified)

OK.. this is not correct, and very wrong in every aspect.

First, you are comparing SATA SSD to Optane, while StoreMI able to use NVMe such as 9 series. My result is higher in 9 series than this.

Second, AMD is NOT caching solution. There is no need for second run, unless you use Fast_Last strategy. StoreMI is a tiered solution, where data is permanently written. Not a cache by all means. That 2GB cache is a bonus.

Third, if you use Fast_First, then initial speed will utilize 960/970 performance, as shown in my own testing. There is NO HDD ONLY term in StoreMI, especially in Fast_First strategy. There is only SSD only, or tiered. Since the first run will be executed from SSD. HDD will only be written when SSD part is full in Fast_First strategy. Only when you use Fast_Last strategy, then your result would make sense.

Next time you write an article, please learn on WHAT YOU WILL WRITE before you blabbering something you don't even know.

September 15, 2018 | 03:53 AM - Posted by Anonymous_Man (not verified)

Another thing to consider, Virtual HDD means that StoreMI will install HDD driver in Windows, and Virtual SSD means that StoreMI will install SSD driver in Windows for its virtual mapped drive. Virtual HDD is slower since HDD won't support TRIM or any command to clean SSD side of things, and won't respond to PCMark or any software that send SSD specific commands. Virtual HDD means, that StoreMI will behave like HDD.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

CAPTCHA
This question is for testing whether you are a human visitor and to prevent automated spam submissions.