I was thinking about using the SSD, there are some good things; no moving parts (except for the MEMS oscillator and the fan), lower latency between NAND flash and the SoC (think hard drive arm seek time - 100x faster than the hard drive's seek times - seek time here is within 5 - 20 nanoseconds).
However, there are few caveat emptors that kinda kept me away from it (I have couple NAND flash-based products here, 16GB SDHC memory cards and a few jump drives - they don't see that much works compared to what the SSDs must endure throughout its life - Linux usually take easy on the SSD, as swap spaces are rarely used) and I know how the flash works.
Here's a short summary: a floating-gate transistor that's left uncharged (or discharged during erase) is summarily read "0", to write a "1" the buffer network in the bus controller applies 5 to 20 Volts to gate and drain of the floating-gate transistor, depositing the charges on the middle gate (not connected to anything at all), and the controller check its state, making sure it's "written" as "1", it apply the charge on the gate using voltage as low as 10% of the original programming requirements, energizing the transistor substrate below the floating gate. If it's discharged or left untouched, applying the voltage to the gate doesn't do anything at all, hence "0".
Now as high voltage is involved, there's always a risk of wearing the transistor down as you hammer the charge into it hard, in order to set the floating gate as "1". The life expectancy of the flash chip (the Silicon die) is pretty long compared to the hard drives (in order of few decades to 100 years), however the rewriting cycle is shorter than a regular hard drive; 250,000 to 10 million cycles before it fails completely.
Now, I want to hear some of your opinions and expert advices before I decide to go with SSD (even though I would prefer to wait for the magnetic RAM based SSDs as they have no rewrite weardowns, it's all magnetic like a hard drive, a primary advantage I want as I deal with "constantly moving data" like video editing). Your rundowns of the pros and cons would be appreciated. (better to have more than one data set so I can park the percentage before I consider that it's worth the risks.)
However, there are few caveat emptors that kinda kept me away from it (I have couple NAND flash-based products here, 16GB SDHC memory cards and a few jump drives - they don't see that much works compared to what the SSDs must endure throughout its life - Linux usually take easy on the SSD, as swap spaces are rarely used) and I know how the flash works.
Here's a short summary: a floating-gate transistor that's left uncharged (or discharged during erase) is summarily read "0", to write a "1" the buffer network in the bus controller applies 5 to 20 Volts to gate and drain of the floating-gate transistor, depositing the charges on the middle gate (not connected to anything at all), and the controller check its state, making sure it's "written" as "1", it apply the charge on the gate using voltage as low as 10% of the original programming requirements, energizing the transistor substrate below the floating gate. If it's discharged or left untouched, applying the voltage to the gate doesn't do anything at all, hence "0".
Now as high voltage is involved, there's always a risk of wearing the transistor down as you hammer the charge into it hard, in order to set the floating gate as "1". The life expectancy of the flash chip (the Silicon die) is pretty long compared to the hard drives (in order of few decades to 100 years), however the rewriting cycle is shorter than a regular hard drive; 250,000 to 10 million cycles before it fails completely.
Now, I want to hear some of your opinions and expert advices before I decide to go with SSD (even though I would prefer to wait for the magnetic RAM based SSDs as they have no rewrite weardowns, it's all magnetic like a hard drive, a primary advantage I want as I deal with "constantly moving data" like video editing). Your rundowns of the pros and cons would be appreciated. (better to have more than one data set so I can park the percentage before I consider that it's worth the risks.)