VHC-CO-IT wrote:

 

To answer media 747's question, no spinning hard drive is going to even be on the same planet as a good SSD for your main booting drive.  I have a demo system with a little 60GB Intel maplecrest 330 SSD and when it's loading windows 7, the little floaty glowing ball things don't even touch.  They're supposed to form into a flag and then pulse while the OS loads but it finishes so fast, it doesn't even get that far.  It also loads the desktop, all applications, the antivirus, etc in 3 seconds.

 

My very high performance Seagate 7200 1TB drive in my home computer will do about 4 pulses on the flag and take at least 30 seconds to fully boot everything.  There's just no contest.  The only problem is, SSDs above 120-128GB are quite expensive!  Also, the majority of SSDs brands have quality control problems, inferior flash chips, bad wear leveling, etc.  It changes weekly but just look online to see which are the best.

SSD boot drives are expensive on a per GB basis, but as a boot drive it is quite cheap.  Since a boot drive doesn't need to be above 250GB (120GB is sufficient, really), there is no reason NOT to get an SSD.  I splurged on a quality 250GB SSD, and it was about the price as a 7200rpm 2TB drive so "expensive" it was not.

On the other hand, don't expect it to make a huge difference to your life.  Yeah, it boots faster but with modern operating systems, there is not a huge reason to reboot all that often.  When I boot in the morning, I really don't care whether it takes 40 seconds or 20 seconds.  If you really care about that 20 second difference, just consider that you will spend a whole lot more time researching the "right" SSD drive, troubleshooting it, downloading firmware, tweaking and posting here about it if it doesn't work as you expected.

Right now, I am running Windows 8 from a VHD on the 7200rpm drive.  (That means I am basically not using the SSD at all except occasionally when I need to boot into Windows 7 for something.)  And you know what?  I don't miss the speed of the SSD for boot-up and starting programs.  When Windows 8 is released, I will install it on the SSD, but it's not such a big deal.

As far as quality changing weekly, I suppose that is the case if you are scraping the bottom of the barrel.  If you go for quality name brands like Intel, Samsung and Crucial, their quality does not change by the day.

My advice would be not to fret too much about benchmarks and "performance" when shopping for an SSD boot drive and instead go for quality and reliability.  TRIM or no TRIM, the performance will deteriorate and the drive will be upgraded sooner or later, and cheaper, more reliable SSDa are just around the corner, so don't spend too much and don't get anything bigger than you need for "future proofing".

All the files were on the SSD; project, source and output files.  Any idea of how long you would have to wait before it is ready to be used?

Bill Gehrke wrote:

 

All the files were on the SSD; project, source and output files.  Any idea of how long you would have to wait before it is ready to be used?

Ask OCZ.  Their people definitely need to learn how to program and test their products instead of using end users as the beta testers but I use their drives because they're cheap, reliable, and can take over double the total amount of writes per chip before failing so they last a lot longer.  If I had unlimited money, I'd go with an intel 520 series with TRIM on.  They're THE fastest in certain tests, use top of the line chips, and their firmware is perfect.

TRIM or no TRIM, the performance will deteriorate and the drive will be upgraded sooner or later

That is completely incorrect.  That's just not how flash chips work.  With TRIM or firmware garbage collection, they don't suffer from fragmantation delay or any type of performance drop at all, unlike a spinning drive.  That's the reason people love SSDs.  They're fast and stay fast forever.

In the last 10 years, average everyday hard drives approximately doubled in speed.  Right now, a Patriot Pyro or Kingston HyperX are 5x faster than my extremely fast 1TB Seagate drive.  It's hard to say they're not worth it, considering that.

As far as quality changing weekly, I suppose that is the case if you are scraping the bottom of the barrel.  If you go for quality name brands like Intel, Samsung and Crucial, their quality does not change by the day.

Exactly the opposite is true.  They're still selling Patriot Torqx drives (ugh!).  Bad, old ones stick around forever.  It's the top of the line drives that change every week.  Patriot was the winner then Kingston made a faster one. Then OCZ developed the Everest II controller. Then someone else developed low voltage writes that don't require an erase command. Then OCZ developed RAID arrays between flash chips within their drives. Then some company called Angel something or other invents the fastest SSD in the world.  Then OCZ made a Revodrive, which runs at 1,600MB/s read and 1,500MB/s write.  Now Sandforce 2.0 is almost released, which enables firmware garbage collection by default if TRIM isn't available.

Since, over the last year, SSDs were glitchy, experimental, and failed too quickly, you have to keep an eye out for the latest of the late technology to get one that actually works.  You can't just say "oh, Samsung is good" and ride that for 2 years because someone will invent something better next month.

Harm Millaard wrote:

 

Your test was simply a bad test. For all I know, you had the drive on a SATA II controller by accident or were using the source file for the test on a hard drive on SATA III port #1 while testing the SSD on SATA III port #2, causing bandwidth issues.

 

How would you know? Your assumptions say it all.

 

 

 

When someone posts a benchmark that has end results that don't make any sense, you have to assume something was wrong with it.

Here is a retest of the OCZ Vertex 4.  I did a secure earase and then waited one hour.  I copied the test files (25.8 MB) to the SSD and ran the benchmark.  The results of writing the 12.0 GB  AVI file is 43 seconds.  Maybe I need to wait 24 hours before I run the test.  By the way we have test results on PPBM5 from over 100 users using a SSD for the project files on a somewhat similar test that a vast majority do not show this abnormal result.

TRIM takes an extremely long time to execute, so when deleting large files, expect not to use the SSD for some time afterwards or suffer horrible write performance degredation. This is because TRIM can not be queued as a command, unlike every other command executed in a computer. Compile that with the fact TRIM acts on blocks where more pages are marked for deletion first and there you have your answer. The Benchmarks with the SSD's were done under ideal conditions (whether on Tom's or PPBM5), hence why they show decent scores. Regardless of TRIM you will also see write performance degrade the fuller the SSD becomes, also for the same reason that TRIM commands like to wait for more pages in a block to be marked for deletion. This is to help reduce wear amplification.

SSD's are a great piece of hardware, they just haven't completely snuffed out HDD's for every application. In a Video Editing box, SSD's should really only be used for Read Heavy operations where data doesn't change much. Anytime you will be editing, moving, or deleting large files, a RAID 0 or RAID 3 on traditional HDD's is a much superior solution.

Also here is the change list from 1.4 to 1.5 firmware on the Vertex 4:

• Improved sequential file transfer performance for 128GB, 256GB and 512GB models
• Optimized idle garbage collection algorithms to extend the benefits of performance
  mode by enabling the feature across a greater percentage of the drive
• Improved HBA / RAID card compatibility
• Further improved compatibility with desktop and mobile ATA security features
• Corrected a corner case issue where the ‘Remaining Life’ SMART attribute could be reported incorrectly

I wouldn't exactly say write performance will drastically change because of these improvements, given the inherent shortcomings in SSDs I stated above.

Hello everyone, I have been watching this very interesting thread and have toiled over drive configurations for quite a bit of time over the last few systems I have bought and I can tell you that while it is important, don't spend too much time on it as there are more important things to focus on that will have a much greater impact on your machine and where you want to spend your $$.  I also find it interesting how much time I spend trying to tweak my system versus creating content, which is where my time should really be spent...but I digress.  I just thought I would share the items I have found to be no brainer decisions and ones that you would think would make a difference but don't.  The key however, is evaluating what kind of work you will be doing, and what is likely to have the most impact.  Listed below are things I have found over the years that have really made a difference and those that are a waste of time.

1. SSD vs Spinning disk for boot drive:  For your boot drive, this is a no brainer, get an SSD.  For pretty much everything else I have seen it have zero effect on PPro or After Effects, go with spinning disks (more on this below).

2. Make sure you have a CUDA enabled card, this will enhance your system performance with PPro more than anything else I have found (not as important with AE).

3. IF you are doing a lot of color correction, make sure you use the CUDA enabled effects...10X faster and if you have CS6 they have more enabled.

4. If you want a very fast way to do color correcting, take a look at Adobe Speedgrade, it is simply awesome and is fantasticlly fast regardless of the hard drive configuration, but runs all on your GPU.  Also, it comes iwth built in affects for making your video have the cinematic look and those plug ins can be very expensive, this one is free with production premium and is awesome.  Only kicker is it is not fully integrated yet and you should use it as the last part of your editing process (i.e. finish your transitions, effects, etc. and then do your color grading last).

5. Really watch what plug-ins you have.  As an example, most everyone has noise in their video and wants to remove it.  If you get denoiser from RedGiant vs. Neat, the render times differ by a factor of 10x because NEAT takes advantage of multiple threads and the GPU.  In other words get smart about what effects you use, which to use in After Effects vs. Premiere pro, get your workflow down, etc.  These will pay off 100x more dividends than the time you spend thinking about disks.

6. Fast RPM disks vs. 7200 RPM.  In my experience, this has had zero real world impact on AE or PPRo, I ahve 15K SAS drives and 7200 SATA drives in raid 0 and not much difference so save your money.  The reason is that the high RPM's are great for applications (like a database) that have non-sequential reads/writes and that doesn't factor in too much.

7. Definitely RAID your drives, I personally like RAID 0 with frequent backups approach as it tends to be the least costly, but I know I will get a lot of flack for saying it, bottom line, break up yoru drives as Harm has suggested in many of his posts, but don't make it a science project...there are much better ways to spend your time to get your system/workflow working faster.

8. Think through your workflow and determine what your bottlenecks are likely to be.  As an example, if you do a lot of color correction in PPRO, make sure you have CUDA and look at  ADobe SPeedgrade.  Do a lot of effects in After Effects?  Make sure that is the best place to do it, there are a lot of the same effects now in PPRo.

9. Use performance monitor on yoru system to see where your bottlenecks really are...are you really using all that memory you bought, does the system report you have a bottleneck with yoru disks, etc.  The sytem can tell you a lot about how it is running, so you can make more educated reasoning on where to spend your money.

I am out of time, and will continue to add to this list as I think of things and would appreciate people adding to the list of known good practices...just don't beat my recommendations up too badly:)

It should not affect performance, with the following exceptions, if the drive is spinning disk (i.e. not an SSD), you want to make sure you leave sufficient free space so it can be defragged and drives lose performance when they are really full.  Other than that, you should be fine.

All right, so no SSD for my other drives. Thank you for your response skeeze! Would this be the ideal way to set up my drives according to the "guidelines for disk usage" on the first page.

C: 256GB Msata SSD

D: 750GB 7200 RPM

E: 320GB 7200 RPM

F: 500GB 7200 RPM

Hi Harm,

I have tried to find the best performance 240 GB SSD for OS etc. but I'm a bid unsure, which manufactor / model to chose. What would you recommend, Harm?

I have been looking at the "red" Corsair Force Series GS 240GB.

Do you have any suggestions?

Thanks

/Ulf

Thanks a lot, Harm! :)

Ulf

I have the 180GB Corsair Force GT which has synchronous NAND flash memory like the OCZ Vertex3. I love it and it's super fast and have no issues.

Here's the problem:

That's the wrong Corsair SSD. The Force series SSDs are all based on Sandforce controllers. And Sandforce's track record has been spotty at best.

Actually, I stated "spotty" in a relative sense: When the SSDs based on the very same Sandforce SF-2200 series controller first came out, they introduced a slew of random BSOD issues in Windows, including BSODs that occurred even when the PC is doing nothing! (This occurred even in "good" brands of SSDs based on that particular controller.)

And did you know that the transfer rate claims on all of the Sandforce-based SSDs are based on highly compressible data? With non-compressible data, the true transfer speed of all Sandforce-based SSDs plummet severely - in some cases, to less than 100 MB/s! What's worse, the maximum read speed of most Sandforce-based SSDs with non-compressible data are also only half as fast as their claimed speed. Read well-written professional reviews of those SSDs and you see what I mean. (And don't rely on Newegg or any other user reviews on the account that their experience with this product category is highly variable, ranging from none to significant. And most of the positive reviews on Newegg of products that were judged "crappy" by professionals have either a horrible or outdated/obsolete/extinct reference product or absolutely no reference at all whatsoever to compare the said products to. In other words, those "positive" user reviews of products that have been judged "crappy" all have practically nothing to compare them to.)

And what you stated about the Crucial M4 SSD is true for the 64GB size. The 128GB version is listed as having a write speed of 170 MB/s (although it is because the 128GB SSD might have been two 64GB SSDs linked internally via a RAID 0 bridge).

Harm Millaard wrote:

 

Randall and Scott,

 

Thank you for your accurate and factual observations. They are a blessing after reading some ill founded fairy tales.

Harm,

I purchased a Crucial M4 SSD (the one with appro 240GB).  I will be installing Windows 8 and CS6 on it soon.  I have a 3-disk setup, per your chart at the beginning of this.  The other 2 disks are Caviar black 2TB and Seagate 3TB 7200rpm.

Given all we've learned, where should the pagefile go??

Hilarious…

How do I know ssd wont help performance that easy I have benchmarks to prove it how about you? That’s what I thought.. oh please show me your benchmarks where performance is improved…

Scott

ADK

a few benchmarks  note the ssd as os and caching was actually a few seconds slower than a standard sata.. also note the overclocked performance you know that thig you think should not be done.. lol 90% of posters on here have OCed systems..

Premiere Pro CS5 Testing

980X at 3.33GHz

12GB Redline at 1600 CL 6

Intel SAS 600 Controller

4 WD 1Tb Sata 64 Meg Cache 600 Drives in 2 Raid 0 arrays

Video material - AVCHD 1080P 24 Frame Each Cut to 30 minutes of material

Export Codec - H264 HDTV 1080P 24 Preset Default

4 Effects per Layer - Fast Color Corrector, Brightness & Contrast, Video Limiter, Sharpen

Each Layer Scaled to 50% for 4 frame PinP view.

285 GTX

3 Layers - 43:39

4 Layers - 52:46 plus 48 minutes effects render on 4th Layer

Quadro CX

3 Layers - 42:50

4 Layers - 47:34

980X at 4.0GHz

Quadro CX

3 Layers - 36:08

4 Layers - 45:22

980X at 4.0GHz

285 GTX

3 Layers - 36:13

4 Layers -  49:01 Plus 42 minutes effects render on 4th Layer

980X at 4.0GHz

285 GTX

4 Drive Raid 0

3 Layers - 35:10

4 Layers - 47:03

980X at 3.33GHz

285 GTX

4 Drive Raid 0

3 Layers - 41:47

4 Layers - 55:02

4 Layers Mercury Engine Software only - 2:10:00

980X at 3.33GHz

285 GTX

4 Drive Raid 0

1 WD 1TB Sata 64 Meg Cache 600 Drive for Media Cache and Temp Files

3 Layer - 41:38

4 Layer - 55:25

980X at 3.33GHz

285 GTX

4 Drive Raid 0

Intel SSD 160GB OS/Cache Drive

4 Layer - 55:27

same test

I7 2600K 4.7 GHz

16GB Blackline 1600 CL 9

570GTX

4 WD 1Tb Sata 64 Meg Cache 600 Drives in 2 Raid 0 arrays

3 Layer - 30:46

4 Layer - 33:36

I7 2600K 4.7 GHz Z68

16GB Blackline 1600 CL 9

570GTX

4 WD 1Tb Sata 64 Meg Cache 600 Drives Raid 0 and 80GB SSD Cache Drive

3 Layer - 31:16

4 Layer - 35:08

I7 2600K 4.7 GHz Z68

16GB Blackline 1600 CL 9

570GTX

2 WD 1Tb Sata 64 Meg Cache 600 Drives Raid 0 and 80GB SSD Cache Drive

Quick Sync

3 Layer - 13:36

4 Layer - 17:20

I7 2600K 4.7 GHz Z68

16GB Blackline 1600 CL 9

570GTX

2 WD 1Tb Sata 64 Meg Cache 600 Drives Raid 1 and 80GB SSD Cache Drive

Quick Sync

3 Layer - 13:51

4 Layer - 17:50

Hmmm...let's see...lol.  I think maybe the fact that you ran a video rendering test when we're actually talking about how quickly hard drives can load video clips into Premiere is part of the problem, lol.  May I perhaps suggest a better test?  Import a 10GB video file into Premiere from a spinning drive and see how long it takes then do it from an SSD and see how long it takes.  The SSD will be 4x faster.

The more specific reason why your test didn't make any sense for an SSD as a cache drive is that the filter you were running resulted in an output stream of data slower than the write speed of a spinning drive.  In other words, run a simpler, easier filter that can be processed at several hundred MB/s output stream speed and you'll see an immense different with an SSD as a cache drive.

That test is about as average you can get to a normal workflow and why its used.

i can play with synthetic tests all day long they are pointless and of no use.

import? the import is ONLY as fast as the drive being imported from

[This post has been edited to remove unacceptable personal comments]

Message was edited by: Jerry Klaimon