Nice to hear the problems are not that severe. But those charts are in logarithmic scale; it does appear to lose about half its performance. Other SSD vendors, firmware or controllers generally have quite static performance. They only degrade in performance with no overprovisioning and little spare space. I prefer 'honest' specifications. Like Intel which gave specs that were below that of a fresh drive. You always had some headroom. I like this kind of honest specifications.
But even Intel is joining the zero-write specification frenzy with its Sandforce-powered SSDs. It's a big shame Intel didn't develop their X25-M controller even further, when Intel had a head start and a much more sophisticated controller than most.
As you may know, the Intel controller used internal 256K SRAM buffering, while other SSDs including the Crucial M4 and Samsung 830 uses DRAM write-back. For Intel, the DRAM was only used to cache the mapping tables, not to act as write-back buffer for write requests from the host. This made Intel have far superior latency and IOps, while being also more reliable. It's just a shame that Intel didn't update their controller to join the SATA/600 generation of SSD controllers. They could have easily beaten all competition with reliable and fast SSDs. Awhile ago it appeared that Intel did invest R&D in updating their controller to SATA/600 and focussing on consistent performacne instead of general high performance. But I fear their move has come too late; today, Intel is little more than yet another Sandforce vendor. They used to be so much more in the past.
But even Intel is joining the zero-write specification frenzy with its Sandforce-powered SSDs. It's a big shame Intel didn't develop their X25-M controller even further, when Intel had a head start and a much more sophisticated controller than most.
As you may know, the Intel controller used internal 256K SRAM buffering, while other SSDs including the Crucial M4 and Samsung 830 uses DRAM write-back. For Intel, the DRAM was only used to cache the mapping tables, not to act as write-back buffer for write requests from the host. This made Intel have far superior latency and IOps, while being also more reliable. It's just a shame that Intel didn't update their controller to join the SATA/600 generation of SSD controllers. They could have easily beaten all competition with reliable and fast SSDs. Awhile ago it appeared that Intel did invest R&D in updating their controller to SATA/600 and focussing on consistent performacne instead of general high performance. But I fear their move has come too late; today, Intel is little more than yet another Sandforce vendor. They used to be so much more in the past.
