Introduction

In our last two articles we took 2 (test) and the Corsair force (test) two SSDs on the basis of the SandForce controller under the microscope with the OCZ vertex, to represent, inter alia, the performance with two different capacities. Both SSDs use the high IOPS firmware, which is designed to provide a higher performance for random writes with small files.

The current article should now serve extrememory introduce two models of the German manufacturer of SSD and to parse, which used the “normal” firmware to demonstrate any differences from the high IOPS firmware with the XLR8 plus a SandForce SSD.

The XLR8-M series is based on the current Indilinx controller and is equipped with Flash chips from Micron. The manufacturer 275 MB / s read and 175 MB / s in writing shall specify the maximum transfer rates. The warranty is two years.

The XLR8 plus sets extrememory SandForce controller SF1222 and installed again Micron Flash chips. The maximum transfer rates are as specified in other SandForce SSDs 285 MB / s read and 275 MB / s write. The warranty is two years as the M-series. The test sheet provided to us was a black metal. The exterior according to the manufacturer in silver will be held at the sales models. In addition to the SATA connector the XLR8 has plus a mini-USB connection.

Test system & methodology

Test system:

• Processor
  • Intel Core 2 Quad Q9550 @ 4.0 GHz
• Motherboard
  • Gigabyte EP45 ud4p, S775, P45 + ICH10R
• Memory
  • OCZ XTC Platinum DDR2-1000 (4 x 2 GB, 5-5-5-15)
• Graphics card
  • Sparkle 8800GTS 320 MB (G80)
• Power supply
  • BeQuiet dark power Pro 530 Watts
• SSDs
  • extrememory XLR8 (M series) and XLR8 plus
• Operating system
  • Microsoft Windows 7 Professional (64-bit)
• Benchmarks
  • ATTO disk benchmark v2. 34
  • AS SSD benchmark 1.4.3645.3568

Test methods:
For the comparability of results, it is of the utmost importance that the tested real scenarios in all subjects take place as equal as possible. To simulate a power loss may occur, we conducted twice the entire code in different States. First, the SSD with the program HDDErase was transferred to its original condition. Then, a previously created image was loaded by using the Windows 7 backup function. Thus, a full operating system with installed programs and virus protection, firewall and some Windows Sidebar gadgets on the SSD has been copied, so that no or very few Flash cell with invalid data arise. Then, we have measured various tests such as the boot time, the launch of applications, the copy, etc. manually with the stopwatch. A restart was performed each between similar tests such as, for example starting an application as files when you run the action in the memory are loaded. A new test without rebooting would measure the read / write performance of memory instead of the SSD part.

After the entire test course was once through, we have loaded the image again, filled the entire SSD several times with real files, disabled TRIM and the last 10 GB removed, to allow running benchmarks and real tests. Since we already have demonstrated in the past, that 7 there no performance loss when current SSDs using the TRIM command with Windows, the TRIM functionality of the operating system before the second test is disabled. Current SSDs should be with the help of GC without TRIM able to prevent a real loss of performance because otherwise many applications such as RAID arrays would be pointless.

In addition to the real scenarios from our last test, we again use the ATTO benchmark to check the manufacturer’s specifications to the sequential transfer rates. In addition, AS is also used SSD benchmark to examine various aspects of performance.

Benchmarks

ATTO

The reference tool of the most SSD manufacturers to measure the maximum possible transfer rates currently is the older ATTO benchmark. This displays the absolute maximum that is sequentially accessible with a hard drive or SSD. These values should not be confused with real transfer rates.

Note: by clicking on a SSD label in the legend of the corresponding graph is bold and it the individual values are represented as “Mouse RollOver”, which significantly improves the readability of graphs.

Sequential reading of different file sizes with the ATTO benchmark we could plus no relevant differences between the XLR8 and find the other tested SandForce SSDs. Transfer rates are the Intel Postville here always above those of the Indilinx competition or the SSDNow V +, depending on the file size a little under, or about the results. The XLR8 plus something is faster than the vertex 2 or the Corsair force only for sequential read files with a size of 4 KB. The indilinxbasierte XLR8 M is as expected at the level of the Corsair Nova, remains something behind the competition but with a file size of 16 KB.

The overview of the sequential write rates shows a similar picture. The results differ very little from the other SandForce SSDs. XLR8 plus to read 4 KB large files was somewhat better than the competition, it is located a little behind in this benchmark. The XLR8 M has a slightly different effect as the Corsair Nova, which probably is back to lead on the different Flash chips.

At this point should be noted that ATTO highly compressible data used to test the sequential transfer rates, which is why the SandForce SSDs in this benchmark achieve outstanding results, which show only the absolute maximum. One finds the minimum sequential transfer rates among other things with the AS SSD benchmark, because there not or barely-compressible data. The real sequential transfer rates are consequently always between these two benchmarks.

AS SSD benchmark

AS SSD benchmark is the relatively new work of a German programmer. It considered various aspects of performance of SSDs as the Crystal Diskmark and hard drives, the average of several passages but specifies instead of the maximum achieved values, making the results more correspond to reality.

There are no surprises at the sequential transfer rates with the AS SSD benchmark, the XLR8 is located right on M with the Corsair Nova and the sandforcebasierte plus model is just as fast as the vertex 2.

≪ blockquote > from our last article:

The initially higher sequential write speed must be regarded as a temporary bonus. The drop to the lower level is always occur after our current knowledge, no matter if TRIM is active or disabled and regardless of which driver is used. Depending on the use of the SSD, the fade time may delay but very long. If the lower level is reached, the write performance only with a secure erase can be restored. This behavior is probably related to the compression and encryption of data, as well as the absence of an external cache. Since the exact functioning of the DuraClass technology under NDA, we can offer currently unfortunately no better or more detailed explanation.

Random read 4 KB large file XLR8 are M and XLR8 plus again immediately with the competition. When writing the latter even with six percent before the vertex 2 could place themselves.

The “4K_64Thrd” indicate how well the SSD can handle random access with a size of 4 KB, when confronted with a queue of 64 commands or requests by the CPU. This is where “native command queuing’ (NCQ) plays a central role, as it the access via the store actually randomly distributed by software are optimized to provide maximum throughput. This is achieved through a review and reorganization of the chain of command. Reached while reading the XLR8 plus again a tie with the vertex 2. When writing, it remains sixteen percent behind the vertex 2, which can be explained by the different firmware in this test.

Erkläuterung to this test:
At this point, it should be noted that a queue of 64 commands in an SSD in a consumer PC is an unrealistically high value. The boot process of Windows represents for example a very high burden on a hard disk or SSD. But even with our Q9550 which is clocked at 4 GHz, the queue for a Ultradrive GX was almost always between 0 and 10. To get to a queue length of 64 commands or requests, one would need at least a massively overclocked Corei7 if not even multiple CPUs, as current SSDs just to quickly process the requests by the CPU. This is also the reason why a SSD more as a notebook with a dual-core low clocked speeds up a desktop PC with a quad core. The SSD is so fast that she faster processes the command queue to the part when she by the CPU extended is.

When the read access is XLR8 plus something surprisingly still behind the Corsair force with 60 GB, faster than the Indilinx drives but still eleven to thirteen percent. When writing the difference to the vertex 2 is again in the context of the measurement uncertainty. The XLR8 M achieved the same results as the Corsair Nova as expected both reading and writing.

The copy test of AS SSD benchmarks consists of three groups. “ISO” means that a large file on the SSD is copied. “Game” tests copying mixed file sizes and “Program” shows the Kopierleistung with many small files. The ISO test is the domain to 100 or 128 GB SSDs, as these offer the highest sequential transfer rates as expected. The XLR8 stay here M and the XLR8 plus just behind the vertex 2 back. In the second copy test are both models ahead of the competition. The XLR8 plus achieves amazingly even a lead of 21 percent compared to the vertex 2. In the third test, the gap with six percent fails again low. The XLR8 of M is even slightly behind the Nova.

It should again be noted that the speed of the SandForce SSDs depends on the compressibility of the data. So greater the files compress can which is the faster SSD.

Real tests

Boot times

During our test of the system start time, we have set as in the past on a completely fresh installation. To make the scenario as realistic as possible, the complete installation was used in all tests 64 bit with all drivers and various programs 7 Professional Windows. At startup, Pro loaded the program AntiVir, the Outpost Firewall automatically from Agnitum and a few widgets for weather, time and calendar. The POST was measured the time from the end until the full loading of the widgets and startup programs. Boot time was measured three times each. The arithmetic mean of these tests are shown in our diagram, to minimize fault-tolerance, which is created by manual measuring with the stopwatch.

On boot times the usual shows. The differences between the fastest and slowest SSD are minimal with five percent. The difference to the competing models from Corsair and OCZ is two percent in the XLR8 plus one percent in the XLR8 M maximum. Whether it is despite multiple measurements to normal deviations or whether the Micron chips at boot time are minimally slower, is not to assess.

Application startup

After analyzing the boot process we dedicate the starting times of applications ourselves now. Here we have set CS 4 and iTunes in particular on the programmes of Photoshop, since these require relatively long time to start. Photoshop has been opened directly with a seven megabyte big picture. As all launch consist mostly of read accesses affected not by the power loss, similar results is regardless of the application to start expected. In addition to the two individual program launches three applications via script were launched at the same time.

The application launches reveals a mixed picture. The launch of Photoshop is the great strength of the SandForce SSDs. With 3.7 seconds they are faster than the SSDs from Intel and Indilinx three times as fast as a conventional hard drive, and five to nineteen percent. Performance differences do not exist while the same. The start of the program iTunes is also very fast. Here, the XLR8 is M however just as fast as the plus model, and thus before the Corsair Nova. The XLR8 plus remains with up to 0.2 seconds behind the other SandForce SSDs. In the simultaneous opening of three applications, the SandForce SSDs are then again equally fast. The XLR8 M is minimal in this test behind the Corsair Nova.

Copy tests

In the following section, we focus the analysis of the copy given them. The maximum sequential write rates of small SSDs with 60 and 64 GB are lower than at 120 or 128 GB, which is reflected above all in this test. Also the comparatively low write performance, the Intel Postville when copying should manifest itself, if just the high sequential read rate partially could compensate this effect. At this point should be again noted that the complete second run, which checks the shared state, has been with disabled TRIM carried out. Thus, we can see in particular in the copy testing how well the “garbage collection”algorithm of controller prevents a loss of performance or minimized.

When you copy an ISO image of the SSD on the SSD with a size of about three gigabytes, especially the XLR8 makes a very good figure M. They achieved not peak of a SSDNow V +, is but at least three to four percent before the Corsair Nova and 17 per cent before the SandForce SSDs with capacity in used condition. As we have highlighted in recent articles, the minimum sequential write rate of the SandForce SSDs drops as soon as all the cells have been described only once. While it is explicitly not a poorly performing GC algorithm. This unique drop is a necessity in the context of the DuraClass technology to our knowledge. The XLR8 plus is as expected as fast as the vertex 2. Also, it lies at just after the one-off reduction in sequential write speed with large files with Intel’s Postville.

The second copy test indicate how well the candidate can handle copying many small files. For the umpteenth time plus are here no differences between the XLR8 and lockable, which are beyond the measurement tolerance of the vertex 2 or the XLR8 M and the Nova. The SandForce SSDs with 100 or 120 GB are the fastest in this test, and 12 to 22 percent ahead of Indilinx and Intel.

Program installation

After we 2009 have shown in our last SSD test, that the installation from a CD or DVD with a SSD is always equally quickly expires, we look at currently only the installation time of Office 2007, which is installed by the SSD to the SSD.

In this test, the XLR8 is as expected, plus front. The advantage to the hotel’s own M series is 10 to 14 percent. Compared to Intel’s Postville, the SandForce SSD is not quite as dominant, but still takes 5 to 13 percent less time for the same operation.

AntiVir

When tested with AntiVir, we conduct a complete review of the system drive C: after all previous tests. The indilinxbasierte XLR8 achieved again similar values such as the Corsair Nova. The plus model is V + current Toshiba controller with minimal earlier force Corsair or OCZ vertex 2, and just behind Kingston SSDNow.

Conclusion

The test of XLR8 M and the XLR8 rise of extrememory has us once again highlighted the need the Indilinx SSDs very well when sequential writing of large files, but not with the sandforcebasierten models compete just on application startup or installations can.

Also, we could see no difference in performance between the XLR8 plus with “normal” firmware and the vertex 2 with high IOPS firmware in our testing on average. The performance each SandForce SSD seems regardless of the manufacturer and the Flash chips used to be equally good.

< div >

So once again, the price seems a central purchasing criterion to be. Currently, you get the XLR8 M series from 162 euros, the plus model 155 euros. Thus the M series at 64 and 248 GB offers a minimally better price per GB while, for it is inferior in terms of performance the plus series in almost all areas. The XLR8 plus offers a comparatively cheap price in addition to an outstanding performance with only a few listings in our price. With a price of currently 249 euros, especially the variant we tested with a capacity of 120 GB is even cheaper than Intel’s Postville very attractive and the price per GB.

Due to the excellent performance of the SandForce controller and the excellent value for money the extrememory XLR8 has to plus 120 GB deserves our recommendation.

Overview:All SSD tests that have so far appeared on computer base
Price:View prices of SSDs