What do SSD drives do for your workflow? Any speedups to report?

[Asher Kelman]

We all know that mage files have to be read and rewritten numerous times and the speed of processing can be held back unless one has fast scratch disks. With the advent of solid state SSD drives, how have folk managed to speed up workflow?

What were the drives used and where are they physically located?

Asher

 

[Asher Kelman]

Seasoning SSD drives!

Here's a source from January 2009: They used seasoned SSD drves in a 32 BIT Windows operating system

Seasoning the drives (SSDs only)
Note: We preconditioned the drives so that our tests would deliver accurate sustained performance values. Without preconditioning, tests could deliver highly variable performance.
1. Plug in an SSD that you have securely erased or freshly performed a low-level format on. 2. Initialize the disk, and format it as NTFS. 3. With Iometer, run a one-second 128K sequential read test to the entire logical block addressing (LBA)
drive space. This enables all LBAs to have some content so the SSD does not have an artificially high reserve space. Note: We used Iometer 2008-06-22-rc1, available from http://sourceforge.net/projects/iometer.
4. Delete the IOBW.tst file from the SSD drive. 5. With Iometer, run a 5,700-second 128K sequential read test (request size aligned on 4K sector
boundaries) on 100 percent of the drive. This preconditions the drive.​

Question: Does this need for seasoning apply to Mac systems too? Can one buy an SSD drive "Pre-seasoned"?


Results Jan 09

"We created a test that included four tasks in Adobe Photoshop CS4: opening the Photoshop application, opening an image, rotating the image, and saving the modified image as a PSD file.

KEY FINDINGS
In our single-tasking tests, four SSDs
completed four typical Photoshop
tasks as much as 34 percent faster
TEST REPORT
than did four HDDs. (See Figure 1.)
In our multi-tasking tests, four SSDs
FEBRUARY 2006
completed four typical tasks up to 44 percent faster than did four HDDs when we ran the tasks in the foreground while unzipping a file in the background. (See Figure 1.) Compared to the single-tasking tests, the multi-tasking tests took the SSDs approximately 0.6 seconds longer, a penalty of only 7 percent, while the HDDs took approximately 3.6 seconds longer, a penalty of 27 percent. "Source​

 

[Cem_Usakligil]

We all know that mage files have to be read and rewritten numerous times and the speed of processing can be held back unless one has fast scratch disks. With the advent of solid state SSD drives, how have folk managed to speed up workflow?

What were the drives used and where are they physically located?

Asher

Hi Asher,

Due to a shortage of time just a short answer right now. There are many advantages to SSD drives but also certain disadvantages, the most obvious being the high prices. But more importantly, most SSD drives have slower write speeds than read and they also have limited write-cycle lifetimes. Also, deleting files leaves blocks of data unused on the disk, which can be only reclaimed by TRIM functionality at a later time. In short, SSD drives are ideal when the content is written once and read many, such as installing the OS and the applications on it. Using an SSD drive can speed up the startup times of a PC considerably and starting an application is also faster. The only MS OS which is fully equipped to work with SSD and to support the TRIM command natively is Windows 7. I am not sure about the OS X.

But beware: SSD drives are not suitable to be used as a fast scratch disk (due to the disadvantages I have mentioned)! One still has to use high performance (7200 rpm-10000rpm) hard disks, prefably configured as raid 0 or 10, as the scratch drive if one wants to achieve the best performance in image processing applications.

I haven't made the move to SSD yet as I can easily wait 30 seconds longer for my PC to start up. Will wait till I also upgrade the processor to i7 later this year.

Cheers,

 

[John Angulat]

I agree with Cem.
There are many drawbacks, cost being the greatest.
Figure 10x that of a regular drive
A 256GB (the largest currently available to the general population) prices out between $500 and $700!!!
I've a lot better things to save for!
Here's a good review of the drives: http://www.pcworld.com/article/170554/solidstate_drives_go_mainstream.html

 

[Asher Kelman]

SSD's seem to win as Scratch Drives on a Mac Pro 2.66 HZ Nehalem 12c GB RAM

Jan 2010 Result comparing OWC Mercury Extreme SSD cards with a Toshiba Hard drive array.

Photoshop CS with diglloydHuge benchmark, 56GB scratch file
OWC = Other World Computing Mercury Extreme SSD
E4P = Sonnet Tempo SATA Pro
Hitachi 7K2000 hard drive review​

Source, note the author discloses that MacSales.com is an advertiser on the review site.


So have you any reports to share or practical experience with SSD's?


Asher

 

[Cem_Usakligil]

Jan 2010 Result comparing OWC Mercury Extreme SSD cards with a Toshiba Hard drive array.

Photoshop CS with diglloydHuge benchmark, 56GB scratch file
OWC = Other World Computing Mercury Extreme SSD
E4P = Sonnet Tempo SATA Pro
Hitachi 7K2000 hard drive review​

Source, note the author discloses that MacSales.com is an advertiser on the review site.


So have you any reports to share or practical experience with SSD's?


Asher

Hi Asher,

With all due respect, you cannot just copy a graph showing the results of a benchmark test without disclosing the test details, equipment used, test methodology, etc. I am fine if you add the link to this test, but adding this graph by itself is misleading IMO. Besides, anyone who claims that the current technology SSD drives can be used as a scratch disk is simply providing unjust advice. Yes, for a performance test it can be setup as such. But if you use it in a production machine, I am sure that the performance will dwindle below that of a hard disk within a few months.


Cheers,

 

[Asher Kelman]

Hi Asher,

With all due respect, you cannot just copy a graph showing the results of a benchmark test without disclosing the test details, equipment used, test methodology, etc. I am fine if you add the link to this test, but adding this graph by itself is misleading IMO. Besides, anyone who claims that the current technology SSD drives can be used as a scratch disk is simply providing unjust advice. Yes, for a performance test it can be setup as such. But if you use it in a production machine, I am sure that the performance will dwindle below that of a hard disk within a few months.


Cheers,

Cem,

Did you miss the link, perhaps? If so, read the text and tell me if your concerns are covered.

Asher

 

[Asher Kelman]

Also, Cem, Cannot one rejuvenate and SSD drive every so often?

Here's a discussion going on for the past year on using the TRIM function to erase data in occupied cells and also the garbage cleaning function too. It does mention that the writes could get eventually slowed by 20-20% but that's it and that a good deal of the benefit in speed is in the reads.

Asher

 

[Cem_Usakligil]

Cem,

Did you miss the link, perhaps? If so, read the text and tell me if your concerns are covered.

Asher

Hi Asher,

Firstly, I apologize if my previous post has come over as harsh, it was not intended as such. I did not miss the link, it was there already. What I was trying to say was the fact that the graph cannot be interpreted correctly without reading the test conditions first. So why post only the graph here? It provides data but no information when presented like this. It would be better IMO if you'd just provide the link to the test and also a summary of their findings if necessary. Or if you present the graph, you have to also explain what is going on in that graph so that the presented data turns into information.


Cheers,

 

[Cem_Usakligil]

Also, Cem, Cannot one rejuvenate and SSD drive every so often?

Here's a discussion going on for the past year on using the TRIM function to erase data in occupied cells and also the garbage cleaning function too. It does mention that the writes could get eventually slowed by 20-20% but that's it and that a good deal of the benefit in speed is in the reads.

Asher

Hi Asher,

Let me try to reiterate what I wrote. I know that the main benefit is in the speed of reads. This makes the SSD an excellent choice for installing the OS and application on it.But when also write speed is important, such as when continuously writing and deleting data as in a scratch disk, then the benefits of SSD are for the short term. On the longer term, one has to rejuvenate the SSD every once in a while but also taking the limited write-cycle lifetimes into account, I do not think that the current generation of SSDs should be deployed as scratch disks. Perhaps the next generation or two will overcome these negative effects, we shall see.

Cheers,

 

[Asher Kelman]

Cem,

I just read the graph as one would read a piece of text. It has enough information there to understand at a glance given the brief introduction and to get a sense of the matter. For more, once click to the link and there's more. To me it's pretty clear that the SSD used in the circumstances of the test was very fast. If it degrades to 70% of that, so what, in a year it will be superceded and we'll recycle it anyway!

So far, Barefeats.com has a strongly supportive article here for using OWC's SSD drives which have a series of measures to try to overcome the issues of slowing down over repeated use. I'll look for other reviews and see if there's even more support or not.

Asher

 

[Cem_Usakligil]

...I just read the graph as one would read a piece of text. It has enough information there to understand at a glance given the brief introduction and to get a sense of the matter..

Sorry to still disagree, I will stop arguing after this post I promise . IMO "enough information" can easily become "misinformation" if one interprets is wrongly. By looking at this graph, and without reading the test details, I do not have any information about the following aspects of the test results:
- What is a diglloydHuge benchmark
- What raid configurations were deployed for the SSD array and for the HD array
- To what kind of raid controllers they were attached
- Where did one install the OS on, SSD or on HD (for each of the test configurations indicated)
- Which Hitachi hard disks and the capacity, just 7200 rpm is not enough info
- What is the capacity of the the SSD drives
- What other tasks were running in the background
- Whether the scratch disks have been "polluted" normally by using them in a CS4 production environment or were pristine

To me it's pretty clear that the SSD used in the circumstances of the test was very fast. If it degrades to 70% of that, so what, in a year it will be superceded and we'll recycle it anyway!

My point precisely; the circumstances of the test! That is what makes the results meaningful. If one would test the drives for a very subjective performance indicator which is not very important, and one would then present the results saying that the SSD array performs better, would it prove that SSD drives are definitely better? I don't think so.

In the meantime I have read about the diglloydHuge benchmark and it seems that it is a test script for CS4 which manipulates a 40,000 pixel wide image and creates a 56GB scratch file as a result. This is the most extreme situation whereby an SSD would win hands down from a HD, no discussions about it. But the time difference when working with our usual 5000 pixel wide images on a day-to-day basis will be less. Does it then justify investing 2000-3000 $ in the SSD array? I don't think so. I know I can invest that money much more wisely for real improvements to my workflow and the bottom line. And finally, I would not consider writing off such a huge investment after a year or so because the SSD drives have gone belly up.

Again: I agree with the added value of deploying SSD drives for the OS and the apps. The high investment is the only reason I haven't done that just yet. But I would never deploy SSD drives as a scratch disk unless the next gen SSD technology comes with improved lifecycles. That's all I wanted to say from the beginning.


Cheers,

 

[Asher Kelman]

Cem,

I'm really happy that you take the evaluation seriously. I read it with the same things in mind and if I was about to buy, would then check item by item. The first thing to ask if real user experience with the sort of files we use. Say a stitched pano with 10-40 images and with a machine that has say 12 GB RAM and so is not RAM crippled and see what difference it makes after 3 months worth of cycles with different files.

Barefeats.com quotes macsales .com, ie the OWC Mercury Enterprise level SSD drives as having

• Intelligent Block Management & Wear Leveling automatically distributes data evenly across the entire SSD.
  
  
• Intelligent Read Disturb Management spreads the active read/write across the flash components eliminating data corruption caused by constant use.
  
  
• Intelligent Recycling gradually re-writes data across the SSD over time to ensure data never gets corrupted.
  
  
• Redundant Array of Independent Silicon Elements protects the data on your drive similar to having a RAID setup.(A 200 GB OWC SSD drive is really a 250 GB drive and the extra space is for the redundancy)
  
  
• ECC Protection aides in data retention and extends drive life.

​

Is this just sales promoting mumbo-jumbo or "has OWC succeeded in extending the endurance and overall reliability the SSD without sacrificing performance?"

So far we have test results but then a jump over the question of longevity issue which was the problem with last years drives. One does not need huge SSD drives. 64 GB would be fine and 3 of them seems to max out the speed advantage in some tests by Barefeats.com in the unique conditions they chose to use.

So the extra information we need is the experience of a photoshop user not just a gamer and preferably not a site where income from advertising is important. In spite of these reservations, SSD might be the easiest way to speed up my Macbook Pro 17" as the RAM can't be increased above 2GB and I can't take advantage of a fast graphics card relieving the Intel processor as there are no swop outs for that important speed limiting component. I would appear that having CS4 on an SSD drive or on a superfast SD card would also give some speed increase. However, I think that the best investment of time and money is to make any of these tweaks on a Mac Pro Tower with an 8 core intel processor so every parameter can be optimized.

Asher

 

[Cem_Usakligil]

Cem,

I'm really happy that you take the evaluation seriously. I read it with the same things in mind and if I was about to buy, would then check item by item. The first thing to ask if real user experience with the sort of files we use. Say a stitched pano with 10-40 images and with a machine that has say 12 GB RAM and so is not RAM crippled and see what difference it makes after 3 months worth of cycles with different files.

So far we have test results but then a jump over the question of longevity issue which was the problem with last years drives. One does not need huge SSD drives. 64 GB would be fine and 3 of them seems to max out the speed advantage in some tests by Barefeats.com in the unique conditions they chose to use.

So the extra information we need is the experience of a photoshop user not just a gamer and preferably not a site where income from advertising is important. In spite of these reservations, SSD might be the easiest way to speed up my Macbook Pro 17" as the RAM can't be increased above 2GB and I can't take advantage of a fast graphics card relieving the Intel processor as there are no swop outs for that important speed limiting component. I would appear that having CS4 on an SSD drive or on a superfast SD card would also give some speed increase. However, I think that the best investment of time and money is to make any of these tweaks on a Mac Pro Tower with an 8 core intel processor so every parameter can be optimized.

Asher

Excellent, with this conclusion I fully agree!

Cheers,

 

[Asher Kelman]

Boost!