To SSD or Not to SSD.

In my opinion, I equate the performance feeling (and benchmarkable speeds) to the difference seen when going from a single core to a upper-end quad core processor.So ya, big difference.


Be careful what SSD you buy. SSDs that use Indilinx's Barefoot controller (old now) have high failure rates. Sandforce based drives have the highest failure rates of modern SSDs (the successor to Indilinx Barefoot), although not as bad as Barefoot. Crucial C300 seem to have the highest long term reliability in overall performance without manual upkeep as well as failure rates. Intel makes good drives all around; last long, good speed.

Crucial C300 drives have pretty slow 4k random read speeds, but write speed is good. Also with C300, the larger drives scale write speeds higher and higher. So 64GB drive has slowest speed, 512GB has highest speed.

The newest Intel drives are unproven, and ALL new gen Intel drives since the first have had a major problem for the first month or two. So if you go Intel, get a G2 drive.



Mainstream SSD drives are not made to last more than 2-3 years!!! Cant stress that enough. If you want a drive to last a long time, you either need a good enterprise class drive (which you cant afford, trust me) or a large drive, or a drive with a really good controller. Ill look into which ones specifically have the longest lifetime, at least estimated. Many estimate at 1,000,000 hours MTBF, but they ALWAYS die sooner than that. you cant even truly measure a SSD lifetime in hours, thats a mechanical drive thing. SSDs have a certain amount of write endurance and that is how they are measured. They just aren't made to last a long time. cheap NAND chips dont have high write amounts (7000-10000 writes per cell), low capacities mean cells get used much faster (a 64GB drive only has 32GB of cells), write amplification means you are using a lot more cells than you think ( a few KB file uses a lot more than a few KB worth of cells on the drive), benchmarks wear drives out fast (writes lots of files all over the drive), and manual upkeep for max performance uses a lot of write cycles (writes data from one cell to another cell that has an open space, this takes a write to the partial cell and an erase to the original cell). All adds up to drives only lasting a few years.

Worth it? Yes, definitely. Can you afford to buy a new drive in 3 years if it breaks? I dont know, thats up to you.

 

Dont bother, Raptors are just a tiny bit faster than a good 1TB drive now days.

mmm.... more than that. Id say 300-400% faster. and if you dont run a defrag program once a week it could be even higher performance difference. If you havent run a defrag program in a few months, performance could easily be 500-600% difference.

You dont just look at a drive that has about double the max transfer speed, that is not where drives performance comes from. The only time that matters is when copying a movie file or soemthing from one place to another, not a super common occurrence. True performance is seen in seek times to find the files you need and how many inputs/outputs can be done per second. I/O is directly tied to random 4k benchmark numbers. a normal mechanical drive can do a few hundred I/Os per second, modern SSDs can do in the tens of thousands per second. Seek times on SSD are 0.05-0.2 miliseconds real average. HDD (mechanical drives) advertised average seek times are around 7 ms, real average is often around 10ms per file.


Do me a favor, go to one of your games folders, any game, and right click and select "properties". How many files are in that game alone? Now lets just pretend for ease of arguments sake that in order to play that game, all files had to be loaded. Really, not ALL files are loaded, but quite a lot of them are. So anyway, multiply the number of files, time 10 (for 10 milisecond average per file). That is how many milisecond it would take a normal mechanical hard drive to load that game. Now multiply the number of files, time 0.2 (highest average time for a SSD). Thats how many milisecond it would take a SSD to do the same thing.



Example for Bulletstorm:
3,769 files.

3,769 x 10 = 37,690ms (HDD time)
3,769 x 0.2 = 753.8ms (SSD time)
difference: 50x speed (or 5000%!!! LOLZ)
Thats just theoretical, but it proves my point. Not that ^^^ big a difference in reality but you get the general idea of how it works.

 

Newegg shell shocker deal on an Intel drive, this should last you 3-8 years Id guess. $177 after rebate:

http://www.newegg.com/Product/Produ...ntel-_-20167035&AID=10440897&PID=1727683&SID=



It has 60GB worth of "usable" cells to write data to, so better than a cheapo drive. The cells last 10,000-15,000 writes each as well. I dont remember specifically how many extra cells it has to swap around for wear leveling of the cells and backup as cells die, I think it is a lot, like an extra 30-40GB or something.

The early version of the Sandforce based drives (not this Intel drive) had less than Intel and Sandforce had 20GB extra cells "behind the scenes", this was later changed to only 10GB extra. People who dont know anything love this as they get more space "for free", but the cost is huge to drive lifetime as there is half as much space to keep the cells wearing evenly and as extra when cells start dieing. So I think the Intel drives have an extra 30-40GB of space as backup to extend life.

 

Do you have benchmarks showing this? All evidence I have seen shows them not really being too much faster.
Also, isnt the F4 faster than F3?

http://www.bit-tech.net/hardware/storage/2009/10/06/samsung-spinpoint-f3-1tb-review/9


The Velociraptor does win in random seek time in benchmark tests, however in actual tests where the drives are loading something/copying something/writing files, that advantage isnt seen anywhere in real performance for some reason. and Im not talking about total throughput for large files, but somewhat small files where the VR *should* win. If we look at the actual measured random read speeds between the VR and F3 drives, the VR only wins by a measly .35 megabytes per second.
Total throughput a modern 1TB hard drive should be capable of higher MB/s, although seek times do matter a LOT which the Raptor wins at. But in actual use it doesnt seem to have a big advantage for a Velociraptor over a modern drive. In that review, the Samsung F3 won the boot time test when it should have actually lost to the Velociraptor by a few seconds. And in Crysis load time the F3 only loses to the VR by 1 second. Those dont seem like a Raptor is significantly faster to me :/


But please, post some benchmarks of the Raptor vs other drives that show it as much faster and I will retract my previous statement about the drive.

 

Thats what I had Valars do too. Good advice.