Re: The SSD myth unraveled
 

Can you please explain this statement more? I'm actually specing out a new PC and have tentatively chosen a VelociRaptor for the boot drive.

Newegg.com - Western Digital VelociRaptor WD1500HLHX 150GB 10000 RPM 32MB Cache SATA 6.0Gb/s 3.5" Internal Hard Drive -Bare Drive

Are you saying that a 72000 RPM drive is actually faster? I've never heard of such a thing before.

Re: The SSD myth unraveled
 

The reasons why I speced out a VelociRaptor are really quite simple: 1) I've had only one WD drive act flaky on me, and I was able to return it w/ zero hassle. 2) The OCZ SSD reviews that I've read are not very encouraging.

Newegg.com - Computer Hardware, Hard Drives, SSD, Internal SSD, OCZ Technology, 120GB

Re: The SSD myth unraveled
 

Peter,

The 150 GB Velociraptor is one generation old and achieves a transfer rate of around 105 - 110 MB/s, but for instance the Samsung Spinpoint F4 320 GB achieves almost 140 MB/s. Of course the newest generation Velociraptors are almost equal in performance, but then the Samsung is only € 30.

Re: The SSD myth unraveled
 

Harm,

Here's a link of all the VelociRaptors currently available from newegg:

Newegg.com - Computer Hardware, Hard Drives, Internal Hard Drives, Western Digital, 10000 RPM

I have to think that the newest drives are those that are SATA 6.0Gig and have 32MB of cache (including the 150GB ones). Would that be correct?

And you're saying that the Spinpoint is just as fast as new VelociRaptor, right?


P.S. Looking at some reviews on the web, it seems that the drives are very close when it comes to continuous reads or writes. But random access is considerably faster w/ the VelociRaptor.

Re: The SSD myth unraveled
 

Peter,

You are correct that the SATA3 disks are the new generation and the one Steve and I talked about were the old generation.

The Samsung is not as fast in random access because of the lower rotational speed and thus the higher access times but for sequential work it is about equally fast. Surprising for the price.

Before Steve's remarks or my own get misinterpreted, let me restate that an SSD for the OS makes perfect sense if all the other components of your system are nicely balanced and meet your requirements.

But..., yeah there is always a but, only as the last optimizing step. The seconds won during the loading of a program are quickly and effectively offset if the rest of the system is not up to its task. I take one example, Steve Kalle's own system ranks only # 121 if we were to include his data into the new PPBM5 benchmark at 265 seconds (his results can still be found on the Old benchmark page), despite his SSD and despite his dual Xeons. I think that my own system, which is a very simple 920 and without an SSD at rank # 11 with 157 seconds does a better job for far less money.

If I were to make investments into a new computer system, it would start with CPU (i7-3960X), mobo (X79), raid controller (Areca ARC-2080i-24) and raid cache (4 GB), conventional disks for the arrays (20 x Hitachi 7K3000 2 TB), memory (4 x 8GB or 8 x 4 GB), video card (GTX 600 series), hot-swappable disk trays, and last a couple of Crucial M4 or Mushkin Enhanced Chronos SSD's in raid10. That would be my last step.

Re: The SSD myth unraveled
 

Thanks for the list of components! BTW, what's your take on memory C.A.S.? I've looked at C.A.S. 9, 8 and 7, and am just not sure if it makes sense to get the faster memory.

C.A.S. 9: Newegg.com - G.SKILL Sniper Gaming Series 12GB (3 x 4GB) 240-Pin DDR3 SDRAM DDR3 1600 (PC3 12800) Desktop Memory Model F3-12800CL9T-12GBSR

C.A.S. 8: Newegg.com - Mushkin Enhanced Blackline 12GB (3 x 4GB) 240-Pin DDR3 SDRAM DDR3 1600 (PC3 12800) Desktop Memory Model 999010

C.A.S. 7: Newegg.com - G.SKILL Ripjaws Series 12GB (3 x 4GB) 240-Pin DDR3 SDRAM DDR3 1600 (PC3 12800) Desktop Memory Model F3-12800CL7T-12GBRM

Right now I'm leaning towards the Mushkin C.A.S. 8's.

Re: The SSD myth unraveled
 

The differences in performance will be marginal at best but likely not even noticeable.

Re: The SSD myth unraveled
 

One thing that people sometimes forget when comparing spinning disk HDDs is that capacity can be traded for speed. For example, if you expect to need 100 GB of space, then the VelociRaptor will be 2/3 full, and will need to use the slower inner portion. But if you get a 2 TB drive, it will only be 1/20th full and will stay 100% on the outer fast portion. That difference may be enough to make the 2TB drive faster. (Tomshardware and other benchmark sites should be able to help with the determination.) Plus, when you decide you no longer need the speed, you can fill the 2TB drive up to the brim.

Re: The SSD myth unraveled
 

Hey Harm,

I didn't read your 1st response to me so I am starting from here.

I think you will agree that for an OS, the most important feature is random access speed. This is why I rank in order from fastest to slowest: SSD (Intel), other SSDs, 15k SAS, 10k V-Raptor, 7200rpm Sata. The write speed is not important, which is why the 80GB Intel X25 at only 70MB/s beats all mechanical disks when used as an OS drive. Furthermore, the performance margin between a SSD and any disk is so large that not even 4 10k drives in Raid can compete - I know this because I once had 4 Raptors in Raid 0 for the OS and then moved to an Intel X25 G1 (1st generation). The performance booting up and loading programs was night and day. Ever since then, I have fallen in love with 'good' SSDs. At that time a couple years ago, all non-Intel SSDs had the write problem you talked about, but the Intel X25 G1 had a firmware update that managed the 'garbage' left from deleted files.

Your argument about my system being slow even though I have a SSD is misguided. The speed of encoding to MPEG2, H264 or uncompressed has absolutely NOTHING to do with the OS drive. The PPBM results cannot and do not illustrate how fast a system is while someone is working and performing everyday tasks. Also, it ONLY shows how fast a PC can encode to a few formats with Premiere Pro....it doesn't show ANYTHING else. Everyone uses multiple programs so PPBM only provides a small part of the performance picture. One program which shows very different results is After Effects. AE likes more cores and more ram whereas PPro likes a faster clock speed and isn't optimized very well for more cores.

Most people I know click the render button and leave for the day or leave to do something else. It seems as though you think that everyone performs only a few tasks (encoding) and uses only Premiere Pro. I don't know about you, but less than 10% of my day is spent on a computer rendering or encoding with PPro, AE, C4D and Nuke.

Say what you want about my HP Z800 but its the only system capable of encoding through AME while designing in Cinema 4D and not getting bogged down every time I hit the "Render View" button (in C4D), which happens every few minutes. For the work I do, the extra money spent on 12 cores is worth every penny. The time saved and reliability of the SSD as an OS drive is worth every penny.

But saying that you only save a few seconds during boot up is being disingenuous. If using Vista or XP, the time saved is even larger. Boot up ends once your computer is useable and you can open windows and programs. With a disk, when your desktop appears, Windows is still loading for many seconds. With a SSD, when your desktop appears, Windows only takes a few seconds to finish loading and becomes useable almost instantly.

Re: The SSD myth unraveled
 

You guys are missing the fact that having Premiere open within the last hour, few hours or day (I don't know how long Windows 7 holds data in its pre-fetch) allows it to open much faster versus closing it at the end of the day and then coming in the next day and opening it. Same thing happens with almost all other programs. The 19 seconds to open Premiere is after opening Premiere a day after it was last opened. When I open Premiere, close it and open it again right away, it only takes about 7 seconds to open on my V-Raptor. However, with my Intel SSD, it only takes 7 seconds the first time and less than 5 seconds after that. (I have several plug-ins for Premiere including several from Red Giant and even more plug-ins with AE. Plug-ins seem to add the most noticeable amount of time to load)

One last comment about all of Harm's arguments and this 'debate': using the 'value' argument as the entire basis for buying a computer or other equipment is not the ONLY thing to consider. This value argument is the same as 'good enough'. Many people here think that a Matrox Mini is 'good enough' for color accuracy. To get the next level of color accuracy, it costs at least $3000 more than a Mini plus HDTV. (either a Quadro plus EIzo or Aja plus 17" broadcast monitor) Based on your arguments, people are fine with the 'good enough' level of performance from the Mini and mechanical disks. Just because we experience the law of diminishing returns with computer parts doesn't mean we should all just settle for the best 'value'.

I find it funny that you recommend a high-end Areca raid controller for everyone with 20 drives but a SSD is overkill. You and I both have large raid arrays, but do we need so many drives for speed? Nope. We need it for reliability. What is far more reliable than disks....SSDs (for full disclosure, I am not comparing Maxtor drives and Patriot SSDs or other low quality drives. For SSDs, I greatly prefer Intel which have proven to be far more reliable than the best 15k SAS drives).

On a related but completely different topic, if anyone plays a lot of games with their PlayStation 3, you will be blown away at how much faster a SSD is as the OS drive. With my favorite game of all time, Gran Turismo 5, load times are cut by 50% or more.

Re: The SSD myth unraveled
 

The one thing I wouldn't argue is SSD being more reliable than HDD. The only advantage I've noticed is HDD usually give warning signs before failing where as SSD are known to just die. SSD reliability varies with manufacture where as HDD are pretty much all the same. Either way you should have a back up or if your using it as an OS, worst case scenario your reinstall the OS on a new drive.

Re: The SSD myth unraveled
 

The warning sign I got from my HDD was bunch of corrupted data as the drive tried to be "SMART" behind my (and the OS's) back!! It will actively try to recover bad sectors without telling you (at microcode level) and then fail silently. What genius though of that?! Only when I used a GNU/Linux boot disk to interrogate the SMART logs of the HDD at a low level did I manage to build the real picture of what happened to my data.

Of course, SSD use exact same concept for bad sector recovery. I'll be happier when everything is RAID-10 in affordable consumer gear :-)

Re: The SSD myth unraveled
 

I tried testing again with the prefetch cleared. This time it took more than 30 seconds.

Nonetheless, I can now see how Harm got those fast launch results: I believe that his workflow requires less than 24 hours (or whatever the length that Windows 7 retains prefetch data) between Premiere Pro sessions. That's simply not long enough to obtain a valid comparison between a hard drive and an SSD.

I also did a retest on this i3-2100 system after Premiere Pro had not run on that system for over 24 hours. This time, with a new prefetch file replacing an expired one, it took about 21 seconds to launch Premiere Pro CS5.51.

Re: The SSD myth unraveled
 

How long does it take to open a large project (more than an hour edited) in Premiere if:

If an SSD is the program/OS drive and HDDs are the media and write drives?
If an SSD is also the temp media drive along with OS/Program?

Re: The SSD myth unraveled
 

The problem with such a question is that if there is not a common large project, all answers are meaningless, because a comparable base figure is lacking.

I do know that opening a 6 hour 15 minute project with 526 individual clips (75.7 GB), two sequences, once in AVCHD 1080/25i and once in AVCHD 1080/50P (duplicates actually, apart from the framerate and interlaced versus progressive) loads in 3 seconds, displaying the timeline but with media pending, 1 second later that message is gone, then the rest of the time is spent on loading the clip heads and is ready for editing in 54 seconds when all the heads are displayed in that 06:15:00:00 timeline. Makes no difference which timeline is open when loading.

Is that fast or slow? Don't know because I have no idea how much time that same project would take on an all SSD system.

BTW, this is on an old i7-920 system with conventional HDD's only.

@Randall,

I get these startup times after turning on my machine in the morning and booting Windows. If the prefetch cache causes this, then all the more reason to diminish the impact of SSD's, because it is quite normal to start PR once a day and if the prefetch cache makes the loading of PR happen in 7 seconds on a system with conventional disks, then that is great. Personally I'm not looking at ways to slow down my system just to make a point.