Solid state drives offer some extremely fast data access and load times. The problem is that they offer much less overall storage space and come with some relatively high price tags when compared to hard drives. Enterprise class servers have been using solid state drives as a form of cache between the server and their hard drive arrays as a means to boost data access performance without the extremely high cost of a full SSD array. Now this same technology has entered the personal computer market with Intel's new Z68 chipset in the form of Smart Response Technology. This article looks at the technology, how to set it up and whether or not there are tangible benefits of using the it to help boost the computers overall performance.
Setup of Smart Response Technology
Using the Smart Response Technology with the new Intel Z68 based computers is extremely easy. All that is really needed is a hard drive, a solid state drive, the Intel driver and one setting in the systems BIOS. The most complicated step is to BIOS setting. Essentially, the BIOS setting for the hard drive controller needs to be set to the RAID setting rather than ACHI or Legacy IDE modes. Consult your motherboard documentations for how to access the BIOS to make the change.
Once the operating system has been installed on the hard drive and loaded with the Intel Rapid Storage Technology driver, it is time to setup the solid state drive. Format the solid state drive with then NTFS file system. Then launch the Rapid Storage Technology program. Go into the Accelerate Tab and select enable. It will then ask you how much of the SSD up to 64GB you want to use for the cache and what mode (discussed further below) to use. Once that is done, the cache is setup and should be running.
Enhanced vs. Maximized
During the setup process, the cache can be set to an Enhanced or Maximized mode. This will affect the performance of the cache via how it writes data to the drives. Enhanced mode uses a method called write-through. In this mode, when data is written to the drive, it is written to both the cache and the hard drive at the same time. This keeps the performance for writes to the slowest writing device which typically is the hard drive.
Maximized mode uses a system called write-back. In this case, when data is written to the system, it is written to the faster cache first and then back filled to the slower hard drive. This gives the fastest write performance possible but has one big problem. In the event of a power failure or crash, it is possible that data will be corrupted on the hard drive if it has not been fully written. As a result, this mode is not recommended for any form of critical data system.
Performance
In order to see how effective the new Smart Response Technology is, I setup a test system with the following hardware:
- Motherboard: ASRock Z68 Pro3
- Processor: Intel Core i5-2500k (default speeds)
- Memory: 8GB (2x4GB) G.SKILL Ripjaws DDR3 1600MHz
- Hard Drives: Two WD Caviar SE16 640GB SATA in RAID 0
- Solid State Drive: OCZ Agility 3 60GB SATA III
The big difference in my setup compared to what many will use is the RAID 0 setup. The Smart Reponse Technology can work with a single hard drive or a RAID array. RAID arrays are designed for improved performance. Most tests of the technology to date have been done with single drives so I wanted to see if it will give a performance boost to a system that is already using an existing technology to boost performance. To demonstrate this, below I've taken the CrystalMark benchmark data for just the RAID array:
- CrystalMark - Two WD Caviar SE16 640GB in RAID 0
- Sequential: 129.5 MB/s Read, 164.8 MB/s Write
- 512k: 29.32 MB/s Read, 64.84 MB/s Write
- 4k: .376 MB/s Read, 1.901 MB/s Write
- 4k QD32: 1.598 MB/s Read, 2.124 MB/s Write
Next, I ran the same benchmark across the OCZ Agility 3 60GB SSD to get its performance baseline:
- CrystalMark - OCZ Agility 3 60GB SSD
- Sequential: 171.2 MB/s Read, 75.25 MB/s Write
- 512k: 163.9 MB/s Read, 75.5 MB/s Write
- 4k: 24.34 MB/s Read, 57.5 MB/s Write
- 4k QD32: 48.39 MB/s Read, 72.88 MB/s Write
Finally, I enabled the caching with the Enhanced mode between the RAID 0 and the SSD and ran CrystalMark:
- CrystalMark - RAID 0 + SSD Caching
- Sequential: 158.6 MB/s Read, 74.18 MB/s Write
- 512k: 155.7 MB/s Read, 62.08 MB/s Write
- 4k: 22.99 MB/s Read, 1.981 MB/s Write
- 4k QD32: 78.54 MB/s Read, 2.286 MB/s Write
These results show that in terms of data writes, the system is slowed down to the slower of the two devices because of the write-through method. This greatly diminishes the sequentially written data as the RAID 0 was faster than the SSD. On the other hand, reading data from the system which is the primary purpose of the caching has been improved. It isn't as dramatic on the sequential data but it is a huge improvement when it comes to random data reads.
This method of testing is synthetic though. So to take it a step further, I timed a few different tasks on the system over multiple passes to see how the caching improved their performance. I decided to look at four different tasks to see how the cache affected the system. First, I did a cold boot to Windows 7 login screen minus the hardware POST time. Second, I launched the Unigine graphics benchmark from launch until the benchmark started. Third, I tested out loading a saved game from Fallout 3 from the load screen to being able to play. Finally, I tested opening 30 photos simultaneously in Photoshop Elements. Below are the results:
- Time To Run (Cold Boot / Unigine / Fallout 3 / Photoshop Elements)
- No SSD Cache: 28 sec / 40 sec / 13 sec / 19 sec
- SSD Cache - Pass 1: 23 sec / 35 sec / 13 sec / 19 sec
- SSD Cache - Pass 2: 18 sec / 24 sec / 8 sec / 19 sec
- SSD Cache - Pass 3: 16 sec / 24 sec / 7 sec / 18 sec
- SSD Cache - Pass 4: 15 sec / 24 sec / 7 sec / 18 sec
The most interesting result from this test was Photoshop seeing no benefit when loading multiple graphics into the program with the cache compared to the standard RAID setup. This shows that not all programs will see benefits from the cache. On the other hand, the Windows boot sequence saw and nearly 50% reduction in the amount of time it took to get into the system as did loading a save game from Fallout 3. The Unigine benchmark also saw a good 25% reduction in loading time from the caching. Thus, programs that have to load a lot of data from the drive will see benefits.
Conclusions
Intel's Smart Response Technology is a good step forward in terms of boosting performance but its benefits are mixed when you factor in the costs of a solid state drive. A 60GB SSD will cost roughly $100 to $150. For someone using a single drive that plays a lot of games or other programs that read a lot of data, the benefits are very tangible. But, if you are willing to spend around $250 to $300 on the solid state drive, you can make that you primary boot and program drive with a secondary data drive that will offer even higher levels of speeds in general. The result is a very niche product.
To be honest, the benefits seem most beneficial to those that may have an older SSD laying around than those putting together a new system. When the cost of a small SSD plus hard drive is roughly the same as buying two solid performing hard drives in a RAID 0 or spending a bit more for a larger SSD with additional performance and a data hard, using the cache system doesn't provide enough solid benefits. It may be much more worthwhile to use more affordable SSDs of 20 to 40GB capacities for under $100 as a stopgap measure for those that can't afford a larger capacity SSD as a primary system drive.
