There have been many questions about how to set up your disks.
Where do I put my media cache files, where the page file, and what about my preview files?
All these allocations can be set in PR, so I made this overview to help you find some settings that may be beneficial. It is not a law to do it like this, it is a generic approach that would suit many users, but depending on source material, workflow and backup possibilities, it is not unthinkable you need to deviate from this approach in your individual case.
The reasoning behind this overview is that you want to distibute disk access across as many disks as possible and get the best performance.
Look for yourself:
I hope this helps to remove doubts you may have had about your setup or to find a setup that improves performance.
425
Replies
Where I live (Australia) Seagate Barracuda 14's are readily available, and very cheap.
They're 8 times less expensive per GB than say; Velociraptors.
I guess your comment : "a high failure rate, many are DOA or develop screeching noises" explains why they're so inexpensive here.
Disk drive prices in Australia are normally much higher than in other countries.
Decisions, decisions 
. I'm eagerly waiting to see what you end up choosing for this build >
http://ppbm6.com/Planning.html before I purchase components for mine.
Edit: Why so much of my message came out highlighted, I have no idea.
I hope it's still readable.
You did read the last page of that article explaining that those results are completely incorrect, right? Anyway...
Tomshardware has a little oversight there. If you instantly fill like 60GB of data onto a 128GB drive then a couple minutes later, attempt to use it, that's not "natural" use of the drive. The firmware would get extremely confused and didn't have time to update its index of which chips have which write counts, attempt to move the data around intelligently, do garbage collection, etc. It also switches into something called "storage mode" instead of "performance" mode which results in a very short time performance drop as it moves stuff around to adapt.
Also, that entire test said "out of the box" and didn't say anything about updating the firmware. Well, there's your problem. And this data is from June 25th, which is before the 1.5 firmware was even released so they weren't using it for the test. It even says, they used the 1.4 firmware. That's at least 2 versions ago. 1.4 even had drive detection problems in the BIOS. It was useless! Any benchmark using it is wrong.
Also, OCZ's own website claims they already fixed this sort of problem in the changelog for 1.5, likely by telling the wear leveling to ignore files over a certain size or to terminate wear leveling chip write count searching after a certain time period or by limiting the period of time for processing to change over into bulk storage mode.
Also, read speed never degrades on any drive ever.
Also, other manufacturers don't have this specific problem anyway as long as they do garbage collection (except Kingston). But as the final page states, this isn't a problem, it's a feature and it only affects the drives for a couple minutes after doing something stupid like filling it half full with data all at once.
If you read the final page..."Effectively what this means is that drives that are less than half full will enjoy further optimized performance and after crossing more than half full the garbage collection algorithm will re-optimize the drive for maximum efficiency based on a larger data footprint. During this transition there may be a small latency hit, but this is a onetime event, and overall performance quickly improves as the drive is now optimized for the larger amount of storage."
It's a one time deal. Like I said, you write a sequential 50-60GB data file, it gets a little confused about what you're using the drive for and re-optimizes itself the maintain the best speed BUT only temporarily until the firmware adapts and then it operates at full speed again. From the article: "From our observations on a partitioned drive, “storage mode” is encountered when sustained write activity exceeds 50% of the available free space." So if you slowly fill the drive up over a month, it won't do that.
And the biggest also is also, Premiere cache drives are usually around 100% empty lol. They're cache drives. They cache stuff, not store it long term 
P.S. why aren't they using cold fusion for the forum? lol.
Premiere cache drives are usually around 100% empty lol. They're cache drives. They cache stuff, not store it long term
Incorrect. When you use the program you will notice this is again a faulty statement. Had you ever used the program, you would know that the cache files are used from the moment you import footage into the project, till you remove the project from your disk.
I forgot, that's how it works by default but that's not how I use mine. When the final project is done and encoded and there's a low chance I'll edit it later, I either delete the cache files and they'll just have to regernate on the fly if I decide to reopen the project later or I move them to my long term storage drive within the project preferences, which is the way everyone should do it unless they have a reeeeeeeeally big SSD.
Large, bulky files are fine on SSDs as long as you have room but if they're not frequently used, throw them on a stable low powered drive or something instead. I use a 640GB green drive for backup images of all my software CDs and DVDs. It spins at a low speed, gives off almost no heat, and basically is extremely unlikely to fail because of that. It's around 130GB so that'd be a total waste of space on an SSD considering I pretty much never use those images unless the original disc gets destroyed
VHC-CO-IT wrote:
Also, that entire test said "out of the box" and didn't say anything about updating the firmware. Well, there's your problem. And this data is from June 25th, which is before the 1.5 firmware was even released so they weren't using it for the test. It even says, they used the 1.4 firmware. That's at least 2 versions ago. 1.4 even had drive detection problems in the BIOS. It was useless! Any benchmark using it is wrong.
Also, OCZ's own website claims they already fixed this sort of problem in the changelog for 1.5, likely by telling the wear leveling to ignore files over a certain size or to terminate wear leveling chip write count searching after a certain time period or by limiting the period of time for processing to change over into bulk storage mode.
Also, read speed never degrades on any drive ever.
Also, other manufacturers don't have this specific problem anyway as long as they do garbage collection (except Kingston). But as the final page states, this isn't a problem, it's a feature and it only affects the drives for a couple minutes after doing something stupid like filling it half full with data all at once.
If you read the final page..."Effectively what this means is that drives that are less than half full will enjoy further optimized performance and after crossing more than half full the garbage collection algorithm will re-optimize the drive for maximum efficiency based on a larger data footprint. During this transition there may be a small latency hit, but this is a onetime event, and overall performance quickly improves as the drive is now optimized for the larger amount of storage."
It's a one time deal. Like I said, you write a sequential 50-60GB data file, it gets a little confused about what you're using the drive for and re-optimizes itself the maintain the best speed BUT only temporarily until the firmware adapts and then it operates at full speed again. From the article: "From our observations on a partitioned drive, “storage mode” is encountered when sustained write activity exceeds 50% of the available free space." So if you slowly fill the drive up over a month, it won't do that.
And the biggest also is also, Premiere cache drives are usually around 100% empty lol. They're cache drives. They cache stuff, not store it long term
I cannot tell if these comments were meant for me as you were replying to Harm. I said nothing about "out of the box" on the SSD testing--I specifically stated that the drive I got was firmware 1.5 and therefore I did not have to update it. Also our PPBM5 test running in Premiere does write a 13 GB file. so their so called fix would never be usefull as a Premiere Project file location.
AdChoices