Lightning-fast Vertex 3 Pro tested

The first benchmarks of an early test sample of OCZ’s Vertex 3 Pro have surfaced. The SSD has a new generation Sandforce controller and easily outperforms other SSDs in almost all benchmarks.

The website Anandtech received an early test sample from OCZ. The ssd consisted only of a pcb without a casing around it and is equipped with an early firmware version. Nevertheless, the SSD successfully passed the countless benchmarks. The Vertex 3 Pro is equipped with the SF-2500 controller, which will mainly be used for server SSDs. The SF-2200 controller is intended for mainstream SSDs. The test sample is also equipped with mlc 32nm nandflash memory from Toshiba, but the mainstream Vertex 3 SSDs will use 25nm memory from IMFT.

In the usual IOmeter test, the SSD is significantly faster with random writes of 4KB, but only with random reads of 4KB does the SF-2500 controller show its potential. In sequential reads and writes, the Vertex 3 Pro is approaching speeds of 400MB/s with maximums of over 500MB/s, made possible by the real-time compression the controller performs. Files that are not or less compressed are still written or read at over 200MB/s. The Vertex 3 Pro is also faster than the competition in various benchmarks that mimic daily use, although the differences here seem slightly smaller.

Not only does the SF2500 controller deliver much higher performance than any other controller on the market, it also solves a problem that plagued the SF1200 controller. If an SSD has been used a lot and all memory has been written to, performance will drop as soon as uncompressible data is written to the SSD. The SF1200 controller wants to load the memory as little as possible and unused blocks are not automatically emptied as long as it is not absolutely necessary. That is good for the life of the memory. Incompressible data, however, must be written to an empty block. So if all blocks are full, they must be emptied first.

The SF1200 controller can clear blocks at up to 80MB/s, bringing the maximum write speed of incompressible data to 80MB/s or less, while enabling write speeds of over 130MB/s. The SF2500 controller seems to have solved this problem. Unfortunately it is not clear how; maybe it can clear blocks faster than 80MB/s or the trim function is more aggressive and clears a certain number of blocks early as soon as incompressible data is written to the ssd.

The Vertex 3 Pro is very successful, but the question remains how reliable the new SSDs with a Sandforce 2200 or 2500 controller are. In any case, the first firmware versions of new controllers often have teething problems, wrinkles that are ironed out over time. The new Sandforce controller also has standard support for the SATA 600 interface and the various tests show that the use of a SATA 600 port is strongly recommended, because the performance on a SATA 300 port is not much. higher than that of many older SSDs.

Anandtech also elaborated on the expected lifespan of the flash memory. Due to the new production process, which evolved from 50 to 34 and now to 25nm, the lifespan of flash memory is decreasing. According to Intel, 50nm memory could be written and rewritten 10,000 times, but that number of cycles decreased to 5,000 with 34nm memory, while 25nm memory could only withstand 3,000 cycles before being unusable. According to Anandtech, however, it would take more than 300 years for even 25nm memory to go through all its cycles if the SSD is used very intensively. However, the flash memory will die much sooner than after 300 years.