Cheap and fast SSD for running OS

[Fortunex]

Hi,

I plan on purchasing a small SSD just for my OS, I'm thinking 32GB would be enough for a dual-boot XP/W7, is that enough storage?

What model would you guys suggest?

 

[r_manic]

Windows 7 needs 16GB for install (and I think goes down to 10 after setup). That leaves you little, so I suggest you go for a 64GB SSD?

 

[sub mesa]

There is no such thing as "cheap and fast". Either you go for a cheap SSD which is very slow but at least you have the other benefits of SSD storage (low read latency, low power consumption, high reliability) - OR - you go for a fast SSD with an intelligent controller and additional DRAM memory chip acting as buffer/cache.

Fast but still affordable SSDs are OCZ Vertex and Intel X25-M 80GB.

 

[daship]

OCZ vertex all the way but try for 2 in raid 0

 

[Onus]

I thought the Windows 7 RC 64bit version needed 20GB for installation...

 

[r_manic]

Minimum is 16GB, but I'd rather work with 20GB: How To: Install Windows 7 on an XP or Vista Machine

OP, take a look at Tom Hardware's newest storage benchmarks: I think the fastest SSD is the Intel X25-M if I'm not mistaken.

 

[sub mesa]

All you did wasted tested its sequential read capabilities. Even cheap SSDs without advanced controllers with its own DRAM buffercache will display good read capabilities, though do have less parallel flash channels so other SSDs can go even faster like 280MB/s (SATA 3Gbps) or 1GB/s (PCIe x8). Still, due to their read latency being extremely low compared with mechanical disks they would display good boot times and general good responsiveness when mostly reading.

For writing, however, its easy to distinguish a good controller from a bad one, and this is where the price aspect comes from. A good SSD would have a low write latency even for small and random writes, whereas a bad SSD the write latency for small and random writes is going skyroof. This is because these controllers have no capability of 'hiding' the difficult process of reprogramming flash memory chips, which first requires an read of a whole cell block (640KiB or sometimes larger than 1MiB), then an erase-cycle on the whole cell block, then a program (write) cycle on the whole cell block. You see, if you want to write 223 bytes the whole cell block will have to be updated.

All this causes frequent 'hick-ups' when applications are writing, causing the UI to become non-responsive. The mouse still moves but the program doesn't respond to any input, or part of the screen is garbed. After some time the program finally finished what it was doing and it becomes responsive again. These frequent hickups are frequently associated and documented on the web, specifically those who employ a JMicron JMF-602 controller.

In order to remedy the situation, you need an advanced SSD controller with multiple channels that has a write-back buffer using a DRAM memory chip (just like you can find on any modern HDD, next to its controller). Intel X25-M has been reported to work very well in random write situations, although a firmware flash is necessary to make this performance permanent.

As you can see, even large and knowledgeable companies like Intel struggle with how to get their controller just right. And the SSDs i give you are just a prelude of what is to come: PCIe flash disks hooked up to your second PCIe II x16, x8 or x4 slot, with sequential transfers higher than 1GB/s both read and write, and at least 100.000 random read/write IOps. Such disks would primarily be aimed at server oriented systems, but will quickly find its way into enthousiast and high-end workstations. This is an exciting time for storage-enthousiasts, and it is clear that flash memory storage will shatter the storage bottleneck imposed still today by mechnical disks with their serial operation, high latency and unreliable operation.