SSDs make a world of difference

[That_Fox]

I've been using the MacBook Air since last year as my computer to edit when I'm traveling (upgraded to the stock HDD Rev C in June). And it's always been pretty slow, but bearable, as I was on a bit of a budget. But recently I've just been craving more speed as I know I'll be getting the 7D in a month or two. I expected a decent difference when reading to and writing to the disk, but boy was I surprised when it sped up every part of my computer!

Here (http://db.xbench.com/merge.xhtml?doc1=404096&unsetCookie=true) are the xbench results for my 1.86GHz Rev C MacBook Air with the default 120GB HDD. A measly 47, with only 1 result over 100, the disk test was 31.

Here (http://db.xbench.com/merge.xhtml?doc1=404349&unsetCookie=true) are the xbench results for my MacBook Air with the 128GB Runcore SSD. This time it scored 115, with only 4 results under 100, the disk test was 233.

The SSD drive increased every result significantly! This is phenomenal. Now it rivals my MacBook Pro (using a standard 500GB HDD) for speed and it actually feels speedy when editing. It's a large price (I paid $449 for my 128GB SSD, but am getting $40 back via Bing chashback) upfront price, but the results are well worth it IMO.

[JackLiu]

Great to know the performance enhancement. SSDs are rather pricey today. When prices drop I plan to get one or two for image storage.

[basroil]

It actually should not be able to batch photos as fast as a MBP even with only a 7200rpm hdd due to the vastly different core speeds. However, adding an ssd to a slow computer will make it usable in many cases

[That_Fox]

It actually should not be able to batch photos as fast as a MBP even with only a 7200rpm hdd due to the vastly different core speeds. However, adding an ssd to a slow computer will make it usable in many cases

You are correct, the CPU test was 170 on the MacBook Pro while it was 130 on the MacBook Air. So the MacBook Pro is about ~25% faster in regards to the processing capabilities. Although the MacBook Pro scored 50 on the disk test while the MacBook Air scored 233, that might close some of the difference.

Although surprisingly enough the MacBook Air scored 115 overall and the MacBook Pro scored 122. Which makes it only 6% slower than my MacBook Pro. My MacBook Pro has a 2.5GHz Core 2 Duo, 4GB RAM and a 500GB 5400RPM HDD. That's not to say that I'd prefer to edit on the MacBook Air, which, if given the choice I'd still edit on the MacBook Pro. But it definitely closes the gap quite a bit in regards to speed.

[hollis_f]

Great to know the performance enhancement. SSDs are rather pricey today. When prices drop I plan to get one or two for image storage.

Image storage doesn't really need the speed of an SSD. Image processing is where the benefits really show. So I store my processed images on an ordinary HDD. But my OS, Programs, Lightroom Catalog and the images I'm working on are on my SSD. An 80GB driove should be enough for all of that - and I reckon the price is low enough now to be worth it.

[Route246]

Be careful. You could lose everything. It may be fast but it is still very new technology with bugs.

http://www.techreport.com/discussions.x/17913

[That_Fox]

Be careful. You could lose everything. It may be fast but it is still very new technology with bugs.

http://www.techreport.com/discussions.x/17913

No need to worry about that, I always have three backups of all my data. Knock on wood, I've never lost any data.

[FZ1]

Be careful. You could lose everything. It may be fast but it is still very new technology with bugs.

http://www.techreport.com/discussions.x/17913
That is for Intel and Intel based drives (Kingston uses them and I think one other OEM) and specifically related to the latest firmware upgrade they released. In general, HDD's have a higher catastrophic failure rate due than SSD's to the moving parts. CF cards, thumb drives, etc all employ the same basic technology as the SSD's. Since Win 7 has been released with the TRIM function available, now is the time to start looking to buy SSD's if you have an interest.

[Route246]

Apples and oranges. HDD have a higher media failure rate, perhaps. This is a bug, not a failure. SSD is a very immature product. The media may be more reliable but the firmware is going to be bug-prone and the only thing that fixes that is time.

That is for Intel and Intel based drives (Kingston uses them and I think one other OEM) and specifically related to the latest firmware upgrade they released. In general, HDD's have a higher catastrophic failure rate due than SSD's to the moving parts. CF cards, thumb drives, etc all employ the same basic technology as the SSD's. Since Win 7 has been released with the TRIM function available, now is the time to start looking to buy SSD's if you have an interest.

[FZ1]

Apples and oranges. HDD have a higher media failure rate, perhaps. This is a bug, not a failure. SSD is a very immature product. The media may be more reliable but the firmware is going to be bug-prone and the only thing that fixes that is time.
Again, Intel based drives only and even though HDD's <> SSD's, it's relevant to look at the failure rate regardless of whether it is hardware or software related - the end result is the same. It's no more inaccurate to compare those than it is to lump all SSD's together and say they are buggy when they are comprised of different components. :cool:

[Route246]

I guess you are referencing MTBF. If not, then what are you referencing?

Failure rates measured using MTBF are a totally different thing when referencing failure rates due to a bug. If you can say that SSD are as bug-free as HDD then you must have information I'm not privy to. SSD are very early in their product life cycle. HDD are very much into their product life cycle and there is substantial learning curve advancement that's taken place in the past 30+ years.

Again, Intel based drives only and even though HDD's <> SSD's, it's relevant to look at the failure rate regardless of whether it is hardware or software related - the end result is the same. It's no more inaccurate to compare those than it is to lump all SSD's together and say they are buggy when they are comprised of different components. :cool:

[hollis_f]

Be careful. You could lose everything. It may be fast but it is still very new technology with bugs.

http://www.techreport.com/discussions.x/17913

That's why I never do a firmware update until a few hundred others have tried it out. The previous firmware for my motherboard caused big problems - to those that installed it straight away. I was unaffected because I waited. I'm not bleating about how terrible those m-boards are though. Same with my Intel SSD. No chance of mine being bricked as I didn't even try the firmware update.

[FZ1]

I guess you are referencing MTBF. If not, then what are you referencing?Failure rates measured using MTBF are a totally different thing when referencing failure rates due to a bug. Why? There's no qualification for the type of failure. It is what it is. Hard drives have controllers which are run by software (firmware) just the same as SSD's. Seagate just had a HUGE issue with their HDD drives "disappearing" and in some cases, loss of data. From what? Firmware issue.
If you can say that SSD are as bug-free as HDD then you must have information I'm not privy to. SSD are very early in their product life cycle. HDD are very much into their product life cycle and there is substantial learning curve advancement that's taken place in the past 30+ years. No, they aren't as bug free but things should be settling down very quickly now that Win 7 has been released, TRIM and garbage collection (NAND laundering) has been sorted out, and wear-leveling algorithms have been ironed out. I think OCZ has their Vertex, Summit and Agility lines pretty solid right now, once Intel irons out this issue (I hear they are very close) they should be good to go too. A lot of the others (Patriot, Corsair, Kingston & Super Talent) are sitting on the sidelines letting a few companies walk the minefield first - so to speak.

I have 8 SSD's sitting here on which I've ton a ton of benchmarks, tests, firmware upgrades, downgrades and sanity wipes and I have yet to lose data on any of them. Several are first generation models with no firmware updates possible and they have slowed some. But, it appears there are some things people can do to fix that. If you have one of the SSD's that do not and will not support TRIM or garbage collection - here is what appears to be a nice solution - LINK (http://www.ocztechnologyforum.com/forum/showthread.php?t=64753). I was able to restore performance to an OCZ Apex and a GSKill Titan drives.

BTW - I understand your viewpoint and I don't wish to come across as argumentative here, just trying to pass on what I know :)

[Route246]

I know and you won't ever see me buying Seagate ever again and I've bought thousands of their drives in the past for my job. The problem I have with Seagate is that they are still being cagey about this. They should cop to the problem and give the antidote to the users. It's still not available. You cannot get an update from Seagate for this. I have three drives sitting at home now that I can't use because of this. This is a major screw up on their part, especially the public relations part.
Why? There's no qualification for the type of failure. It is what it is. Hard drives have controllers which are run by software (firmware) just the same as SSD's. Seagate just had a HUGE issue with their HDD drives "disappearing" and in some cases, loss of data. From what? Firmware issue.

I don't think wear leveling algorithms are going to be settled for a long time. There is an inherent problem with wear on these devices that no amount of trickery and distribution techniques is going to overcome when you have paging involved. A paging device is going to be very noisy, it's the nature of all modern operating systems to do page trimming. As I understand it, it's not the reads that wear out the devices, it's the deletes and reclaims that do. So, if you have a device that has a lot of R/O data on it that consumes a fair amount of the available blocks, I'm not sure how the remaining free blocks are going to be leveled in such a manner that they can be assured a long life. Of course, the problem gets worse and worse as the device reaches higher and higher R/O occupancy. It comes down to use models. These devices are wicked-fast and it makes sense to use them as sandboxes, especially for PP and stuff like that. Any of these schemes also become degraded in effectiveness as the device becomes more and more full, resulting in fewer and fewer free blocks to play with. I guess you can do some sort of smart defrag operation but if the device is not fragmented there will be nothing to defrag.

No, they aren't as bug free but things should be settling down very quickly now that Win 7 has been released, TRIM and garbage collection (NAND laundering) has been sorted out, and wear-leveling algorithms have been ironed out. I think OCZ has their Vertex, Summit and Agility lines pretty solid right now, once Intel irons out this issue (I hear they are very close) they should be good to go too. A lot of the others (Patriot, Corsair, Kingston & Super Talent) are sitting on the sidelines letting a few companies walk the minefield first - so to speak.

My feeling is to wait and see about these devices. Aside from being extremely expensive, there is too much FUD out there now to make an informed decision. It's smelling like plasma TV which is pretty much dead technology. You will still find people who will defend "blacker blacks" until the end of time but I think the LED backlit technology that's becoming ubiquitous has pretty much silenced those people. The jury is still out for SSD to be used as a mainstream device. Many here may be too young to remember how bubble memory was going to revolutionize technology at one time. http://www.dvorak.org/blog/whatever-happened-to-bubble-memory/
I'm not sure if SSD is Bubble Memory 2.0 or not.


I have 8 SSD's sitting here on which I've ton a ton of benchmarks, tests, firmware upgrades, downgrades and sanity wipes and I have yet to lose data on any of them. Several are first generation models with no firmware updates possible and they have slowed some. But, it appears there are some things people can do to fix that. If you have one of the SSD's that do not and will not support TRIM or garbage collection - here is what appears to be a nice solution - LINK (http://www.ocztechnologyforum.com/forum/showthread.php?t=64753). I was able to restore performance to an OCZ Apex and a GSKill Titan drives.

BTW - I understand your viewpoint and I don't wish to come across as argumentative here, just trying to pass on what I know :)

[FZ1]

I know and you won't ever see me buying Seagate ever again and I've bought thousands of their drives in the past for my job. The problem I have with Seagate is that they are still being cagey about this. They should cop to the problem and give the antidote to the users. It's still not available. You cannot get an update from Seagate for this. I have three drives sitting at home now that I can't use because of this. This is a major screw up on their part, especially the public relations part. They got you too huh? I'm right there with you. I'm using them in a RAID 1 config so I'm hedging my bets. Seagate is officially off my "to buy" list.

I don't think wear leveling algorithms are going to be settled for a long time. There is an inherent problem with wear on these devices that no amount of trickery and distribution techniques is going to overcome when you have paging involved. A paging device is going to be very noisy, it's the nature of all modern operating systems to do page trimming. As I understand it, it's not the reads that wear out the devices, it's the deletes and reclaims that do. So, if you have a device that has a lot of R/O data on it that consumes a fair amount of the available blocks, I'm not sure how the remaining free blocks are going to be leveled in such a manner that they can be assured a long life. Of course, the problem gets worse and worse as the device reaches higher and higher R/O occupancy. It comes down to use models. These devices are wicked-fast and it makes sense to use them as sandboxes, especially for PP and stuff like that. Any of these schemes also become degraded in effectiveness as the device becomes more and more full, resulting in fewer and fewer free blocks to play with. I guess you can do some sort of smart defrag operation but if the device is not fragmented there will be nothing to defrag. It's almost a moot point really. If you are doing heavy duty writing, an SLC based drive will be the best bet. MLC drives will likely outlast the ~10 year NAND cell charge and most of us will be upgrading by then anyway.

My feeling is to wait and see about these devices. Aside from being extremely expensive, there is too much FUD out there now to make an informed decision. Yeah, that's one reason I try to chime in when I can. Bad info seems to propogate faster than good info. A lot of the problems people are having are self-inflicted or due to bad info they have been given. I don't blame you (or anyone) for waiting. If I didn't get them free (well, I paid for one), I would only have one of them. It looks like NAND prices are rather steady right now and no new drives have come out in the last few weeks to influence prices. There are a handful of new drives that will be coming out very soon, they should be good to go straight out of the box without the need for flashing firmware, etc. I'm still waiting on what their performance numbers look like but we are pretty much already starting to hit the ceiling for SATA II so if you are going to wait, wait for the SATA III drives and motherboards :)

[Route246]

They actually didn't get me. I never had one of those failures. It's the threat of getting one (I think I have affected drives but can't be sure). But, it's really the way they handled the whole mess that did it in for me.
They got you too huh? I'm right there with you. I'm using them in a RAID 1 config so I'm hedging my bets. Seagate is officially off my "to buy" list.
It's almost a moot point really. If you are doing heavy duty writing, an SLC based drive will be the best bet. MLC drives will likely outlast the ~10 year NAND cell charge and most of us will be upgrading by then anyway.

I settled on mirrored velociraptors. The performance is pretty good, but at least an order of magnitude less than SLC and a whole lot cheaper.

Yeah, that's one reason I try to chime in when I can. Bad info seems to propogate faster than good info. A lot of the problems people are having are self-inflicted or due to bad info they have been given. I don't blame you (or anyone) for waiting. If I didn't get them free (well, I paid for one), I would only have one of them. It looks like NAND prices are rather steady right now and no new drives have come out in the last few weeks to influence prices. There are a handful of new drives that will be coming out very soon, they should be good to go straight out of the box without the need for flashing firmware, etc. I'm still waiting on what their performance numbers look like but we are pretty much already starting to hit the ceiling for SATA II so if you are going to wait, wait for the SATA III drives and motherboards :)

[FZ1]

For those wanting to dip into the SSD realm without spending a mint, these new Kingston 40GB SSD's (same as the Intel) are supposed to be $85 this Wednesday on Newegg as their "Shellshocker" deal. If interested, better grab one early...I hear their supply is limited.
http://www.newegg.com/Product/Product.aspx?Item=N82E16820139081

One caveat...there will be a firmware upgrade upcoming to support TRIM. Here is a good review...
http://www.legitreviews.com/article/1111/1/

[hollis_f]

I guess you can do some sort of smart defrag operation but if the device is not fragmented there will be nothing to defrag.

I thought that the whole point of wear-levelling was to fragment the drive. You don't want files continuously written to, and deleted from, the beginning of the drive. You want stuff written to all portions of the drive to spread the wear. This will result in severe fragmentation.

With HDDs fragmentation is a bad thing because you waste a load of time moving the head to the right place and waiting for the data to come around. But that's no problem with SSDs, because it's as fast to read from one block as it is from any other. That's why the huge difference between HDDs and SSDs is access times.

I'm pretty sure that we've all been using wear-levelling for some time anyway. Isn't it used in CF cards for the same reason?

[FZ1]

I thought that the whole point of wear-levelling was to fragment the drive. You don't want files continuously written to, and deleted from, the beginning of the drive. You want stuff written to all portions of the drive to spread the wear. This will result in severe fragmentation.
Tru dat. However, there is another type of fragmentation on SSD's. Blocks are made up of cells and you can have a whole lot of partial blocks filled because once a block is written to, it can't be written to again until the entire block is erased. This is where NAND laundering/garbage collection comes in. It consolidates data into full blocks which frees up more blocks for writes without the need for an erase cycle. It works in tandem with TRIM.

[Route246]

Defrag in this context would be to scramble the eggs to get static data moved around. It's not actually to align blocks into contiguous chunks.

I thought that the whole point of wear-levelling was to fragment the drive. You don't want files continuously written to, and deleted from, the beginning of the drive. You want stuff written to all portions of the drive to spread the wear. This will result in severe fragmentation.

With HDDs fragmentation is a bad thing because you waste a load of time moving the head to the right place and waiting for the data to come around. But that's no problem with SSDs, because it's as fast to read from one block as it is from any other. That's why the huge difference between HDDs and SSDs is access times.

I'm pretty sure that we've all been using wear-levelling for some time anyway. Isn't it used in CF cards for the same reason?

[basroil]

Tru dat. However, there is another type of fragmentation on SSD's. Blocks are made up of cells and you can have a whole lot of partial blocks filled because once a block is written to, it can't be written to again until the entire block is erased. This is where NAND laundering/garbage collection comes in. It consolidates data into full blocks which frees up more blocks for writes without the need for an erase cycle. It works in tandem with TRIM.

Yup, exactly how it works on your old graphing calculator too! though no trim there)

[FZ1]

Those Kingston drives sold out on Newegg by 8am EST

[FZ1]

Here is a good technical article on SSD's...
http://www.lostcircuits.com/mambo//index.php?option=com_content&task=view&id=69&Itemid=1