M225 Crucial SSD 128GB

[Zachary Sarah]

After the disastrous performance of its first SSD architected around the controller Phison who has always wondered what he was doing since he was not really designed to find themselves in a DSS, Crucial turned course and made appeal to Korean Indilinx Barefoot and its controller. This is one of the most popular controllers found now in many OCZ, Corsair, G. Skill, Super Talent, Patriot, etc. Crucial not innovating with its M225 series but had been talking about these templates to launch them with very low prices. Some even wondered if they did not include flash memory cheaper.

In reality, the M225 are just clones of other SSD-based controller Indilinx Barefoot, neither more nor less. Where it marks its difference is sometimes at the price or look at the warranty, set at 5 years. Inside one finds then the controller Indilinx Barefoot with 64 MB of external cache as a memory chip Elpida. It was also entitled to chip MLC NAND Flash Samsung original, the same as in the majority of SSD-based controller Barefoot.

The M225 is available in 64, 128 and 256 GB they are references CT64M225, CT128M225 and CT256M225. They all have an aluminum hull, making them very lightweight: 69 grams for our 128GB version in terms of performance, this version Crucial Announces 128 GB to 250 MB / s read and 190MB / s write. Data somewhat optimistic in writing as we shall see.

The support of TRIM is provided by the 1819 firmware available on the crucial website and for those not using Windows 7, it remains the utility named Wiper.exe TRIM manual. The performance is however not affected by this firmware that beyond supporting the TRIM fixes several bugs, adds support for 34-nanometer NAND flash and controller support Indilinx ECO version supporting this fine engraving. A 1916 version was recently posted in the forum Crucial, January 13, 2010 to be precise, and brings support Garbage Collection. This feature is useful if your OS does not support the TRIM or if you want to use two SSDs in RAID. However, we prefer the manual TRIM compared to Garbage Collection, which generates many entries are not necessarily useful, kicking the life span of NAND Flash chips.

Curve Flows

Analyze the flow of the DSS through IOMeter reading and writing, sequential and random.

Sequential Playback

Curve a bit special which tends to slow start but really took off from a file size of 32K This is similar to other SSD-based Barefoot Indilinx as OCZ Vertex with a final speed exceeding 240 MB / s. In comparison with other SSD market is well placed but unable to compete with the excellent Intel SSD regarding the management of small files. He cons by a par with the Samsung but the latter offers a more regular curve beyond a file size of 16 K. Note that this "shape" of the flow curve has been reproduced with other SSD-based Barefoot on Indilinx, it therefore seems to be a "normal" behavior of this controller.

Sequential Write

In writing, we find that break even when processing files larger than 16 KB but also of 8 KB, with a decrease of performance before taking off to a level that exceeds 130 MB / s, far in the final 190 MB / s announced by Crucial. But other SSD Indilinx we have bestowed the same benefits. Nevertheless this remains a good level in all but a little distant from the tenors as the Intel X25-E 32GB and a notch below the Samsung PB22-J that are not yet the most swift of the range of Korean manufacturer. For cons, the rates are the level of Intel and Samsung from the beginning, with small files, then, except for files from 0.5 and 1 K.

Shuffle

As often random, there is a downturn in performance with the files of 256 KB and 512 KB, but here we see the same phenomenon also with files of 128 K. Finally, we achieved good rates, similar to Other DSS Indilinx. If we dwell on the file size under 128 KB, we obtain the same level pretty much as the Intel X25-E 32 GB, which is a very good performance.

Random Writing

Here is "THE" oddity of this SSD Indilinx, random entries in the form of a flat curve. The problem is that the rates are worse with the medium sized files and large files. Conversely, they are excellent with small files and only the Intel X25-E 32GB VelociRaptor and do better. It is a much better level than the Samsung PB22-J and clearly above all SSD architected around the JMicron controller JMF602.

Synthesis flow

This graph confirms that it has excellent flow in reading but that the gap between the rates in writing is larger than that announced by Crucial. You'll also notice the rather low rates in writing random, this should not obscure the fact that they are against very good with small files...

H2benchw: flow "before / after"

History verify the effects of "wear" or use the DSS, following its curves flow in sequential h2benchw making his test on the entire disc capacity, unformatted. The test "before" is made after a secure erase and test "after" shall be made once all the testing protocol is completed, which includes the very (too) requiring random testing of writing. We also conducted tests after running the utility provided by Indilinx Wiper and to operate manually TRIM supposed bring the performance to their original levels.

Flows into Reading


As shown, the level of performance is "reached" by the execution of our test protocol. The only operation responsible for this degradation is running 100% random writes on the SSD. Indeed, the random reads and writes and sequential reads this not so devastating effect on performance. We run the utility Wiper to overcome this problem. The operation is done under Windows and takes a little over a minute. After the execution, performance has found an excellent level, a notch below but performance obtained after presenting the Secure Erase SSD in new condition. There is also a gap in performance between 40 and 50% of disk capacity after implementation of the first run of h2benchw, behavior is also typical controller Indilinx. The second run solves this problem allowing the disk to reorganize and ultimately found rates very close to state new.

Drinking write

In writing, the degradation is more important as we move from 215 Mb / s at an average of 25 Mb / s. After running the utility Wiper for TRIM, the first run under h2benchw shows a marked improvement performance once again despite a fall in the middle of the test, is also found on the G. Skill Falcon based on the same controller Indilinx. The second run will prevent the hole but gave a general level a little lower. A third run gave results exactly the same. But generally, the utility is doing its office and avoid having to do a secure erase to recover performance as is the case with other SSD. Note that in 7, TRIM has the same effect. Note that flows are more important here than in IOMeter or during practical tests, is found on all SSD Indilinx.

Our opinion

The Crucial M225 128GB does not surprise since it uses the controller Indilinx Barefoot we now know very well. It covers many small files and offers excellent flow reading. The maximum write rates are the same barrel as the OCZ Vertex and its clones. As with all DSS origin Indilinx were some "oddities" such as breaks in the curves of flows to the 8-16 KB in reading and writing sequential. More strange, the random flow curve which is not bad with small files, it is even quite good, but it never took off toward high throughput with large files as is the case with other SSD we’ve tested.

Then there is the usual strong performance degradation when the player "spent" under an operating system does not support the TRIM or the proper firmware was not installed. While our testing protocol is demanding and extreme in some respects but it has not caused problems to others as the Samsung SSD PB22-D example.

Fortunately, the TRIM or Wiper Utility to manually call the TRIM function is there to get the performance but the ideal would have been a DSS does not require such a utility or at least a DSS whose performance does not deteriorate to this point compared to the competition.

We also welcome the general treatment of the Comptroller Indilinx with small files, a notch better than the Samsung PB22-J, but worse than the Intel X25-E. To conclude, we can say that it is a very good product that will please many users. In terms of prices at time of publication, in January, therefore, the 128 GB version is trading from 350 EUR, 5 year warranty included, advantage. This is expensive for a capacity of 128 GB, gigabytes of 2.73 euros while the Intel X25-M 160GB Postville is in turn displayed from 398 euros or 2.49 euros from Giga. The latter is better read, better deals with smaller files but fishery flows in writing that leveled to 100 Mb / s. wrenching choice but in one case as in the other, you have to do a good SSD. ..