Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

chefwong

macrumors 6502
Original poster
Jan 17, 2008
466
30
I was planning to do the whole optibay coversion on my new 13" MBP - but have decided to just wait till TRIM support is supported.

For those who have jumped the gun....are you guys just wiping free space every once in awhile...
 
I was planning to do the whole optibay coversion on my new 13" MBP - but have decided to just wait till TRIM support is supported.

For those who have jumped the gun....are you guys just wiping free space every once in awhile...

I've had SSD on my MacBook Air (first a rev C, now a rev D), and I don't worry about the write performance degradation. I've seen benchmarks on performance degradation but I haven't seen enough of a problem in my personal use to bother with it.

I created a RAID-0 array in my Mac Pro, and it gets 80% of the benefit of SSD without the cost or hassle. I purchased a bracket to mount an SSD into my Mac Pro but I haven't pulled the trigger on an SSD yet for the Mac Pro.
 
Last edited:
I was planning to do the whole optibay coversion on my new 13" MBP - but have decided to just wait till TRIM support is supported.

For those who have jumped the gun....are you guys just wiping free space every once in awhile...

Wouldn't it be possible to use TRIM in bootcamp with Intel RST (if you go with the X25)?

Or does having OSX on the same partition prevent TRIM?
 
Several modern SSD controllers do internal garbage collection and don't need TRIM, such as the Sandforce controller.
 
Several modern SSD controllers do internal garbage collection and don't need TRIM, such as the Sandforce controller.

Bingo. Any drive from OWC (macsales.com) is safe. Had their 120 GB Mercury Elite in my Mac Pro for over six months now and it's just as fast today as it was when I first installed it.

Just waiting for my tax refund to come in to jettison the optical drive in my MacBook Pro and replace it with an SSD. Once you go SSD, you will never go back!
 
Bingo. Any drive from OWC (macsales.com) is safe. Had their 120 GB Mercury Elite in my Mac Pro for over six months now and it's just as fast today as it was when I first installed it.

Just waiting for my tax refund to come in to jettison the optical drive in my MacBook Pro and replace it with an SSD. Once you go SSD, you will never go back!

I have a OWC Mercury extreme Pro 120GB SSD. Had to send it back to company due to performance issues. After I got it back, installed it again, while no apparent performance issues, now it does not like to reboot. Looks like its going back again.

After this, no more SSD drives for me, I am going back...lol
 
While it would be nice to have TRIM support, it's certainly not necessary. Many of us have been getting along fine for 2 years without it and a LOT of SSD owners (including those on Win7), run their SSD's in RAID0 which doesn't support TRIM regardless.
 
I havent needed to wipe mine. The Sandforce based SSDs are the fastest right now and they have built in garbage collection. Its not quite as good as full TRIM support, but it still works really well.
 
I have a OWC Mercury extreme Pro 120GB SSD. Had to send it back to company due to performance issues. After I got it back, installed it again, while no apparent performance issues, now it does not like to reboot. Looks like its going back again.

After this, no more SSD drives for me, I am going back...lol

Instead of giving up all together you can use your ExpressCard slot or even your SD card slot to just put your OS, apps, and a few other things you need all the time on so that you still get the speed for opening and running apps and rebooting the comp/OS, but you can keep the massive space of the HDDs without having to worry about garbage bits until lion comes out or configuring third party TRIM solutions. Also if you use your optical bay a lot, as I do, then you don't have to give that up either. I use a 48GB Wintec SSD card and have barely had any issues at all with it (the only ones being permission/file path issues from moving apps over) and I kept my 500GB HDD to store my libraries and bootcamp partition etc.
 
No problem/slowdown at all with the OCZ Vertex 2 drive in my MBP optibay.

Enjoy your wait
JohnG
 
I was planning to do the whole optibay coversion on my new 13" MBP - but have decided to just wait till TRIM support is supported.

While this is the Mac Pro forum, I can tell you that the new MacBook Pros have 6Gbs SATA III on the main drive connection and 3Gbs SATA II on the optical drive bay. So try to get a SATA III SSD and replace the HD with the SSD and move the HD to the optical drive bay.

For those who have jumped the gun....are you guys just wiping free space every once in awhile...

The OWC Extreme Pro SSD, which I use for my Mac Pro boot drive, has a Sandforce controller that does very efficient garbage collection so TRIM is less of an issue. I've never felt the drive slow down in any way after over 6 months of heavy use.
 
I have 3 SSDs striped and I've never had any issues with decreased performance; I think the key is to leave 30% or so of free space.

With a Mac Pro, it shouldn't be a problem since you can put in more drives.
 
While this is the Mac Pro forum, I can tell you that the new MacBook Pros have 6Gbs SATA III on the main drive connection and 3Gbs SATA II on the optical drive bay. So try to get a SATA III SSD and replace the HD with the SSD and move the HD to the optical drive bay.

The 6Gbs SSD's (there are very few) only get you burst peformance. The Optical Bay doesn't offer sudden motion detection. If the laptop drive you have has it, it was probably disabled to not conflict with the MacBook Pro's built in sudden motion detection.

Either way, if you aren't careful putting the mechanical hard drive in the optical bay may dramatically shorten it's life for what is, right now at least, a minimal performance gain on the SSD - if you even get a 6Gbs capable SSD.
 
I think the key is to leave 30% or so of free space.

This has no bearing on SSD performance. All you will be doing is not utilizing some very expensive flash memory you paid top dollar for - not exactly what you want to do, I hope!

Remember, from the SSD drive's perspective the whole drive is "used". It has no visibility into what is "real" and what is "free". If the blocks are offered to the computer as usable space, from the drive's perspective it has valuable data on it.

That's why TRIM exists, and it's why the SandForce drives have extra space for garbage collection without having to rely on TRIM.

Both TRIM and extra capacity accomplish the same thing - allowing the controller enough known free space to consolidate partially written to flash memory cells to create totally empty cells that permit top speed writing. Again, this is irrespective of blocks, sectors and fragmentation as you think about it with traditional hard drives. Just as the SSD is ignorant as to what is going on above it, the operating system is ignorant as to where the data really resides on the SSD - even with TRIM. The SSD controller shuffles data around at a low level between flash cells *all the time*. Not just for consolidation of partially used blocks to ensure top write performance, but also for wear leveling (ensuring that all cells get an equal number of writes).

That's really all there is to it. If you have a totally empty flash cell to write to, you get maximum write performance. If you don't, write performance suffers. This is at a low, low block level. Below the file system, below partitioning - basically if the storage is offered to the computer for use, it's considered used.

So fill your SSD up! You paid for that storage!
 
My personal SSD experience:

Intel 80gb Gen 2:

Used this in the first gen 2008 Unibody Macbook. It was awesome for that computer. I parted ways with it to put money towards another purchase. I miss it to this day. That SSD complimented the Macbook perfectly. Never had any problems

OCZ 240gb Vertex 2 3.5"

Had this installed in a 27" iMac i7. It served as the primary disk. I do not like having an external disk running the entire time so I did not set my home folder or stored data on one. I used maybe 100gb total on this SSD. One day the computer froze. I rebooted the system, and it would not log back in again. The disk was shot. I tried to partition it and restore it, nothing. It stopped working. I panicked because its cost was close to $500 at the time. Luckily OCZ sent a new once I RMA'd it. I sold the new sealed one I eBay, as I had given up on the SSD.

They're great if you need that speed for the price performance. I would totally do another SSD, but personally I would like to see something in the ~200gb range that has increased reliability at a lower price.

I understand new technology comes with a hefty price tag. But I also understand that comparing this new technology, to modern HDDs, that the HDDs have an advantage in the case of time. They have been available a lot longer, and have had time to mature in their technology, and decrease in price.

For now, I am sticking to a Velociraptor 150gb boot drive and a 1tb Caviar black for my home folder. I have no problems and I am satisfied with the performance to price ratio.
 
There was at least one published report indicating that Intel has enjoyed much lower RMA numbers than many other manufacturers.... particularly OCZ. Your experience seems like another data point in support of this.

I've run 3x Intel SSD G1 drives for 2 years continuously now without a hiccup. (Knock on wood). I'm waiting for the new Intel G3 drives before I upgrade, which I hope have the same reliability as their predecessors. TRIM is irrelevant to me since I run my drives in RAID0 which doesn't support TRIM regardless of OSX support for it.
 
This has no bearing on SSD performance.
Though unused capacity has no bearing on throughputs, it does have an effect on the drive's ability to remap dead cells = usable service life in terms of writes (not all SSD's have additional capacity for wear leveling not available to the user, but even those that do, may not actually have enough). It's unfortunate, but this is a limitation of NAND based Flash memory (limited number of write cycles per cell from the NAND manufacturer; no limit typically published for reads).

Wear Leveling just shifts the writes across unused cells evenly at the time of the write execution. It's not actually increasing the write cycle limits for the Flash used in the SSD, as that's fixed due to physics at the gate level. Now as the disk fills, each available cell will be written to more often, thus reducing their operational lifespan that much faster than a disk with more capacity available to it.

So filling one up to the max usable capacity isn't a good idea for a drive that's written to on a regular enough basis.

Even though OS/applications use is by far more reads (random to be specific) than writes, you still need to make allowances for temp folders/files used by the OS (you could push it a bit here, but it's a good idea to be careful yet, as cost per GB isn't that great yet for SSD's). For a scratch disk, the entire disk should be dedicated to that, as it will increase the lifespan. Fortunately, small disks are suitable for things like scratch, and the cost is now low enough that most users would be able to consider them disposable when they die (~$100USD for 40GB).
 
This has no bearing on SSD performance. All you will be doing is not utilizing some very expensive flash memory you paid top dollar for - not exactly what you want to do, I hope!

Remember, from the SSD drive's perspective the whole drive is "used". It has no visibility into what is "real" and what is "free". If the blocks are offered to the computer as usable space, from the drive's perspective it has valuable data on it.

That's why TRIM exists, and it's why the SandForce drives have extra space for garbage collection without having to rely on TRIM.

Both TRIM and extra capacity accomplish the same thing - allowing the controller enough known free space to consolidate partially written to flash memory cells to create totally empty cells that permit top speed writing. Again, this is irrespective of blocks, sectors and fragmentation as you think about it with traditional hard drives. Just as the SSD is ignorant as to what is going on above it, the operating system is ignorant as to where the data really resides on the SSD - even with TRIM. The SSD controller shuffles data around at a low level between flash cells *all the time*. Not just for consolidation of partially used blocks to ensure top write performance, but also for wear leveling (ensuring that all cells get an equal number of writes).

That's really all there is to it. If you have a totally empty flash cell to write to, you get maximum write performance. If you don't, write performance suffers. This is at a low, low block level. Below the file system, below partitioning - basically if the storage is offered to the computer for use, it's considered used.

So fill your SSD up! You paid for that storage!

This is VERY bad advice and I hope no one makes the mistake of doing what this new poster suggests.

cheers
JohnG
 
So fill your SSD up! You paid for that storage!
This part is absolutely wrong due to a complete misunderstanding about ssd's and performance degradation! The problem with performance degradation is NOT cells that are filled but should be empty. No, performance degradation is when cells are filled. Period. In other words the more you put on your ssd the slower it will become. Since you don't want to throw away stuff you want to keep this will cause the dreaded performance degradation.

That's why people don't partition the entire drive and that's why you absolutely should never ever fill the drive completely!

The performance degradation that is caused by cells that aren't being used can be recovered by the use of TRIM and/or garbage collection. The extra amount of memory cells on the ssd are used for wear levelling and partly for GC (not sure about TRIM). GC and wear levelling work together to keep the drive as clean as possible by using the least amount of program/erase cycles. But like nanofrog explains it is more due to the wear levelling. They did take a good look at SCSI drives :)
 
This part is absolutely wrong due to a complete misunderstanding about ssd's and performance degradation! The problem with performance degradation is NOT cells that are filled but should be empty. No, performance degradation is when cells are filled. Period. In other words the more you put on your ssd the slower it will become. Since you don't want to throw away stuff you want to keep this will cause the dreaded performance degradation.

That's why people don't partition the entire drive and that's why you absolutely should never ever fill the drive completely!

The performance degradation that is caused by cells that aren't being used can be recovered by the use of TRIM and/or garbage collection. The extra amount of memory cells on the ssd are used for wear levelling and partly for GC (not sure about TRIM). GC and wear levelling work together to keep the drive as clean as possible by using the least amount of program/erase cycles. But like nanofrog explains it is more due to the wear levelling. They did take a good look at SCSI drives :)

But it's also important to be specific when you talk about performance degradation... only write performance is affected. Not read performance. And if your drive is largely full, then chances are you aren't writing much to it anyway (no space) - you're probably reading most of the time. In fact, something like 80% of all desktop storage I/O is reading anyway. So let's not blow this out of proportion.
 
But it's also important to be specific when you talk about performance degradation... only write performance is affected. Not read performance. And if your drive is largely full, then chances are you aren't writing much to it anyway (no space) - you're probably reading most of the time. In fact, something like 80% of all desktop storage I/O is reading anyway. So let's not blow this out of proportion.
You still want to write to an ssd so it still is somewhat important. I've seen several reviews (I think some from Anandtech and bit-tech) where read performance was also affected. Write performance was by far the most affected and very noticeable (which is why I was referring to the program/erase cycle). But again, performance degradation does indeed affect both read and write speeds. Due to things like good gc algorithms, TRIM, read speeds not being that much affected by performance degradation and most people not filling their drives (for example you need 10% free space if you want to update OS X), the performance degradation isn't a problem at all.
 
I've seen several reviews (I think some from Anandtech and bit-tech) where read performance was also affected.
I'm not sure which article you're reffering to, but there's a good one from Anand (2009, when the X25-M released) that helps to describe SSD's performance degradation issue, but only writes, where expected (not reads). If you can post links to what you're reffering to, it could be helpful.

I'm wondering if there's some mixed operations in the system's queue that may have surfaced during real world testing (which would be shown in their results), such as writing temp data during an application launch/operation when all operations are to the same storage pool, such as a single disk.

This is why the Intel X25-M and other SSDs get slower the more you use them, and it’s also why the write speeds drop the most while the read speeds stay about the same. When writing to an empty page the SSD can write very quickly, but when writing to a page that already has data in it there’s additional overhead that must be dealt with thus reducing the write speeds.
Source (bottom of pg. 8).
 
You still want to write to an ssd so it still is somewhat important. I've seen several reviews (I think some from Anandtech and bit-tech) where read performance was also affected. Write performance was by far the most affected and very noticeable (which is why I was referring to the program/erase cycle). But again, performance degradation does indeed affect both read and write speeds. Due to things like good gc algorithms, TRIM, read speeds not being that much affected by performance degradation and most people not filling their drives (for example you need 10% free space if you want to update OS X), the performance degradation isn't a problem at all.

Again, I think having a full SSD is likely to create more issues for users because it's actually full than due to the performance degradation. Personally, as my drive fills up, I'm less concerned about the speed with which I can write to it than the fact that I'm running out of space to write anything at all. :p

BTW, a quick search of Anand revealed this tutorial on write amplification and how it impacts performance...

This is why the Intel X25-M and other SSDs get slower the more you use them, and it’s also why the write speeds drop the most while the read speeds stay about the same. When writing to an empty page the SSD can write very quickly, but when writing to a page that already has data in it there’s additional overhead that must be dealt with thus reducing the write speeds.

If you can find some links that show the impact on read performance, please share. The only situation I can think of where read performance would be impacted is where the drive is under heavy load and excessive write amplification is impacting the drives ability to service read requests. But for the queue depths most desktop users are dealing with, this is very unlikely to be an issue or perceptible.

EDIT: Nano... great minds think alike! :D (or fools seldom differ :p)
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.