Caching Large Hard Drives with Solid State Drives

There are three prominent technologies and related hardware that have emerged in the past year, with the noble goal of speeding up computer program loading and execution times. All three involve the use of a solid state drive to cache programs and data that reside on a larger hard drive. The motivation is to simultaneously take advantage of the lower per unit storage cost of hard drives and the higher speeds of SSDs:

premium-cpm

 

  1. In this method, a large hard drive and a small, separate, fast SSD are connected into a Raid 0 configuration on a motherboard that hosts the caching capability. The hard drive and SSD are physically connected to the motherboard by SATA cables, or with a separate version of the SSD on a printed circuit board, connected by mSATA (mini-SATA) means.This technology is exemplified by Intel’s Smart Response Technology, which is currently available on motherboards with the Z68 chipset. One significant distinction of this technology is the need to create a new Windows installation on the RAID-connected HDD and SSD. Pictured below is the Intel 311 Series Enterprise SSD that’s recommended by Intel for use with their Smart Response Technology caching solution, with a series 313 replacement announced.

 

Intel-311-SSD3

Intel 311 Series 20GB 2.5″ SATA II SLC Enterprise SSD

 

  1. This method features a hybrid hardware package that combines a hard drive and small caching SSD on one board. An example is OCZ Technology’s RevoDrive, which includes a 1 TB 5400 RPM hard drive with 100 GB SSD for caching. The board-mounted package is connected to a motherboard PCI-e x4, x8, or x16 slot. When new, the hard drive and SSD are recognized by the BIOS and Windows separately, but after execution of the included Dataplex™ software, they are “married” and are recognized as the bootable C: drive. Before Dataplex is applied it’s possible to clone an existing hard drive onto the RevoDrive’s hard drive, so that existing Windows and other programs do not need to be reinstalled.

 

RevoDrive

OCZ Technology’s RevoDrive combines a 100 GB SSD and 5400 RPM 1TB drive on a PCI-e X4 expansion card.

 

  1. The third method connects a separate large hard drive and a small, fast SSD by SATA connectors, and it then combines both drives, through furnished software. An example is provided by the OCZ Technology Synapse Caching SSD with included Dataplex software, which is available in two memory sizes: 64 MB and 128 MB.In this example, the caching SSD is merged to an existing bootable hard drive, which can be operated on any modern motherboard that accommodates the two drives. After Dataplex is installed, Windows “sees” the merged drives as the C: drive. As needed, the two drives can be unmerged by uninstalling the Dataplex software. This is the technology I tested for this article.

 

Synapse

OCZ Synapse 2.5” SATA III caching SSD

 

According to OCZ, the Synapse SSDs are optimized for caching applications, to provide users with SSD-level performance across the entire capacity of the Dataplex-linked HDD. The OCZ Synapse SSD series includes both 64 GB and 128 GB models, though only about half of each SSD is available for caching, as indicated in this note from OCZ:

 

“In the case of the Synapse SSD, 50% NAND flash “over-provisioning” is used to accommodate performance and software features.”

 

Because it’s not intuitive that only about half the physical capacity of the caching SSD is used for cache storage, it’s appropriate to learn a bit more about this subject. Technically, “over-provisioning” of an SSD relates to the percentage of physical capacity that is allocated to the SSD controller (and not given to the user) in order to decrease “Write Amplification,” which in turn, is explained by a excerpt from Wikipedia.org:

 

“Write amplification (WA) is an undesirable phenomenon associated with flash memory and solid-state drives (SSDs) where the actual amount of physical information written is a multiple of the logical amount intended to be written.”

 

According to OCZ, the Dataplex caching software, which is user-downloaded, registered, and installed after Synapse SSD purchase, is provided to “dynamically manage the use of both the SSD and HDD for superior overall storage performance, and this combination creates an environment where the most frequently used “hot” data stays on the ultra-fast SSD, while the “cold” data remains on the larger capacity HDD (provided by the user). Advanced caching algorithms learn user behavior and adapt storage policies to ensure optimal performance for each individual user, maximizing productivity for the most demanded programs and applications.”

 

Synapse series caching SSDs are designed to integrate seamlessly into any modern PC platform, to ensure optimal compatibility along with performance gains. The Synapse series SSDs are provided with a 3-year warranty. At the time I received the sample, the 64 GB and 128 GB products were selling for about $145 and $245 US, respectively, though prices had decreased by the time I completed testing and writing, to about $98 and $210 US, less redeemable rebates of $20 US for each.

 

Here are nominal specifications for the two Synapse drives.

 

OCZ Synapse Cache SSD Specs

  • Available in 64GB and 128GB capacities – models (SYN-25SAT3-64G
    and SYN-25SAT3-128G)
  • SATA III (6Gb/s)
  • MLC NAND Flash
  • 64GB max performance – max read: up to 550 MB/s, max write: up to 490 MB/s, random write 4k: 75,000 IOPS
  • 128GB max performance:– max read: up to 550 MB/s, max write: up to 510 MB/s, random write 4k: 80,000 IOPS
  • Included Dataplex caching software, with license key
  • ECC recovery, data encryption
  • 2.5″ form factor
  • Power consumption: 2.7W active, 1.5W idle
  • Included 3.5″ desktop adapter bracket
  • Compatible with Windows 7 32-bit and 64-bit
  • MTBF: 2 million hours
  • 3-year warranty

 

Note from OCZ: “Rated speeds may vary slightly depending on the benchmark used, drivers, windows version, bios version and file size. Rated performance is achieved by using the Asus P8P67 Pro motherboard in conjunction with a 1TB HDD, using ATTO, IOMeter, and PC Mark Vantage.”

 

The Dataplex license key is only valid on one machine and with one combination of hard drive and SSD. Dataplex uses various components to identify a PC (memory, OS, CPU Id, BIOS, Ethernet card), and if two or more components change, it is considered a different machine. If you wish to change only one component, Dataplex will automatically revalidate the license as long as there is an internet connection when the PC is rebooted. Users must uninstall Dataplex to release the license prior to changing two or more components in their systems. Licenses cannot be released after the system is no longer valid, and users should contact OCZ technical support to reset the license if this occurs. During the period of testing, my registration of the Synapse SSD led to my receiving a message that announced an update to the Dataplex software. After downloading the updated version, I needed uninstall, download, and reinstall Dataplex, with required re-entry of the registration key for both downloading and installation.

 

SynapseBack

Back side of Synapse SSD, showing label for Dataplex License Key (blanked out) and other information, before mounting to furnished 2.5” to 3.5” adapter

 

The sample OCZ caching SSD I tested was the larger of the two models, with 128 GB of memory. As mentioned above, because of design and operational considerations, only about half the Synapse SSD memory, or 65 GB, was shown as available for caching. Still, that’s a lot of memory, considering that if the SSD cached every file in FSX, with Acceleration Park installed, only about 18 GB memory would be used.

 

Test Results

 

I conducted loading times and determined benchmarks results of the SATA III hard drive, without the Synapse caching SSD connected, and then I connected the SSD to a SATA III terminal on the Intel DP67BG motherboard. Next, I downloaded and installed the required Dataplex program from OCZ’s website, which merged the caching SSD with the existing bootable C: drive. At this point, the Synapse SSD was no longer visible to Windows 7.

 

To measure the effectiveness and make it more relevant to flight simulation enthusiasts, I conducted several timings and benchmarks, including:

  • Windows boot time, from pressing computer On button, until main desktop screen appeared
  • Time to load FSXMark07 test flight, from clicking a taskbar shortcut until cockpit and scenery appeared
  • FSX framerate performance, as measured with FSXMark07 procedure and FRAPS for counting framerates, as described in readme file for FSXMark07.zip, available from www.Avsim.com.
  • Time to load X-Plane 10.04 beta 3, from clicking desktop shortcut to X-Plane.exe, until cockpit and scenery appeared
  • PCMark Vantage: HDD Test Suite
  • PCMark Vantage: Gaming Test

 

After installing the Synapse caching SSD and the Dataplex software, I conducted program loading times and benchmarks, repeating each until it appeared the test results had stabilized and programs and related data were fully cached. This indicates that the booting and program data originally on the hard drive, were subsequently read from the SSD, not from the hard drive. The question is whether the results are positive and whether investment in the caching SSD is worthwhile.

 

Here are results, with Bold blue print indicating the best result of each test:

 

Caching Performance Measure

Without

SSD

Run 1

Run 2

Run 3

  1. Windows boot time, seconds

74

58

31

28

  1. Time to start FSXMark07 flight, sec.

59

55

19

18

  1. FSXMark07 Performance, FPS

46

47

47

46

  1. Time to load X-Plane 10 Demo, sec.

92

78

54

53

  1. PC Mark Vantage HDD Test Suite

4947

8738

27191

26633

  1. PC Mark Vantage Gaming Test

11941

15901

21384

21011

 

It’s interesting to compare the results posted on the OCZ website for the 128 GB Synapse SSD, and results from my testing, as summarized on line 6. In both cases, testing shows that after Dataplex is used to merge the SSD with the HDD, the result from PCMark Vantage HDD test suite increases slightly after the first reboot, and by the second reboot (Run 2), caching is fully accomplished. The slight decrease in my third run result compared to my second run is likely a simple test variation. On the other hand, my results for my second and third runs being in the high twenty thousands, compare to OCZ’s results being over forty thousand for runs 2 and 3, is not easily explained, though it was suggested to me that the difference might be attributable to my testing with an i5 series processor, compared to OCZ testing with an i7 series processor, and it’s unknown what other influences might have been at work.

 

OCZ128Test4

Results of OCZ’s Synapse caching tests

In further pursuit of testing, I coincidentally received components to build a new test system with an Intel i7-3820 processor and X79 LGA2011 series motherboard. Sure enough, there is a measureable difference in the caching test results, as shown in the combined graph below. In performing these tests, both processors were set to the same core frequency, and I used the same power supply unit, memory, hard drive, and Synapse caching SSD.

 

DH128-CombinedTest5

Results of my tests with i5-2500K and i7-3820 processors

 

Here are the data for the PC Mark Vantage HDD tests with two different processors and motherboards:

 

*Caching Performance Test Results, with PCMark Vantage HDD Test Suite

Without

SSD

Run 1

Run 2

Run 3

Intel i5-2500K processor and P67 motherboard

4947

8738

27191

26633

Intel i7-3820 processor and X79 motherboard

4749

6991

33132

33579

*All other components in these tests were the same, including the 1TB SATA III hard drive, 2 x 8 GB 1600 MHz DDR3 memory, and 650W power supply unit.

 

FSX Results. The results for the FSXMark07 framerate benchmarking runs stand out as not benefitting from SSD caching. This is similar to other tests I’ve conducted on a computer system in which both Windows 7 and FSX are run directly from SSD’s, with no framerate benefit compared to running from a hard drive. They’re also consistent with other tests I’ve run in the past three years, in which I’ve shown that FSX framerate performance is almost entirely related to CPU core speed, with only slight variations from hardware changes, given as least four real cores. On the other hand, the PCMark Vantage Gaming score did improve two-fold, so there is progress for general gaming.

 

DH128GamingTest

Caching results for PCMark Gaming Test

Summary

 

SSDs are becoming increasing popular as their prices decrease, and the hybrid solution of purchasing a relatively small capacity, fast SSD to cache programs that reside on large, low cost hard drive is currently popular. For sure, tests are positive for Windows and tested program startup times, and selected performance benchmarks are also positive. However, if you’re looking for enhanced framerate performance from FSX (and I’m highly confident this would be true for FS2004 also), it’s not to be found in current hybrid SSD/HDD caching solutions, nor in previous testing, have I found it to be true for operating programs on SSDs alone. It might be faster for other gaming applications, though for some, the faster boots and software starts, subject to then-current cache contents, may alone be attractive. Also attractive are the steadily decreasing prices of SSDs, including the Synapse product tested for this review.

 

By Doug Horton