OCZ will soon publish firmware 1.4 for its Vertex 4 to improve performance. OCZ promises that? Improvements on sequential speeds:
-Vertex 4128 Go: 550/420 MB / s read / write (against 535/200)
-Vertex 4256 Go: 550/465 MB / s read / write (against 535/380)
-Vertex 4512 Go: 550/475 MB / s read / write (against 535/475)
If there was a slight improvement in reading, it is mostly the side of the writing on models 256 and 128 GB gains are most francs. On the latter, the performance is more than doubled!
If this rate is confirmed in practice it will be far from the 120/128 GB SSD the most efficient in this area since we did not measure more than 331 MB / s in our last test. Note, however, that such a write speed is not really useful in conventional use.
First tests
OCZ has provided a preview a pre-release version of the firmware, the 1.4RC3 for 256 and 512 GB versions. We have updated a Vertex 4 256 GB successfully launched and much of our test protocol . Here are the results:
Synthetic IOMeter tests in reading show a noticeable improvement in performance with sequential read blocks 2MB If we stay below the Vertex 3 and Crucial M4 with a single command, with two commands the Vertex 4 is greater than the M4 and shoulder to shoulder with the Vertex 3.
There is also a rebalancing performance during sequential reads on small blocks of 4 KB, firmware increasing performance with one or two orders at the expense of performance with more simultaneous orders. The Vertex 3 and especially the Crucial M4, however, remain much higher in this area.
In random play with blocks of 4KB Vertex 4 firmware already dominated and it changes nothing.
We also performed some tests on ATTO to show the behavior of the new firmware sequential read when varying the block size and the number of simultaneous orders:
With two commands, the new firmware offers significant gains to a size of 256 KB blocks. The Crucial M4 remains significantly faster on small blocks. If we go to four commands the Crucial M4 digs an even bigger gap with small blocks and the new firmware changes nothing.
The writing was already an area in which the Vertex 4 was well placed and the Vertex 4 further improves its performance in the field with sequential blocks of 2 MB Crucial M4 and Vertex 3 can not fight.
What about the difference in more practice tests?
In reading files the new firmware will fill part of the gap facing the Crucial M4 and Vertex 3 which remain higher and even back with small files. By writing against its advantage is amplified.
In the timed application measures the gains are significant, except for the rescan sources in Visual Studio which is slower. Overall we approach sometimes even exceeds the M4.
Conclusion
The arrival of the 1.4 firmware for the Vertex 4 is good news because it improves the performance of sequential read on small blocks, an area where it was lagging behind, while bringing significant gains in terms of writing sequential which was already flying high.
The sequential read performance behind the small blocks, however, still not completely resolved and if application performance are improving the Vertex 4 should on average be slightly behind the best SSD. Remains that once again it should be noted that the differences between good SSD is minimal and not really noticeable in use. Similarly, an important flow in writing will not be really useful in practice and eventually other factors such as price, warranty or reliability which will further guide your choice.
Side tariff Vertex 4 is well placed but positioned less aggressive than others, which is offset by its 5-year warranty. Reliability was by cons never been the strong point of OCZ in the past and it is hoped that the Vertex 4 will confirm the improving trend noted in the manufacturer on in recent months after the disastrous returns recorded on some models last year.
The 120/128 GB SSD Comparative will of course be updated shortly after the launch of the final version of firmware 1.4, scheduled for later this month.
Update: Since the comparative SSD 120/128 GB has been updated. We dug this occasion the subject and it appears that the write performance gain is not without consideration.
0 commentaires:
Enregistrer un commentaire