Editing and Hard Drives

Dach wrote on 3/10/2014, 5:08 PM
Okay, so several years ago I made the decision to put WD Green (slow drives) hard drives into my system build. I am not sure what my motivation was, but in any case its time to upgrade.

Can anyone speak to the cache that is on the hard drives. Through out the years they have been doubling and I notice that some of the new ones have 128 MB cache. When editing, in theory is this something that would be beneficial?

Any recommendations? For budget reason I am still looking to go with a spindle drive, not SSD.



videoITguy wrote on 3/10/2014, 5:39 PM
Never ever never Green drives in an editing machine - they are for green users of internet browse crowd.

AND never blue drives, and no, not red drives either - just plain black.
Chienworks wrote on 3/10/2014, 5:58 PM
Green drives work beautifully in my editing machine. They transfer data 5 times faster than the Raptors i used to use a few years ago. They also run much cooler than the 7200rpm+ drives.
paul_w wrote on 3/10/2014, 6:32 PM
Greens here too. No problem.

videoITguy wrote on 3/10/2014, 7:11 PM
The op subject is "editing" which could mean -he does not say-- pulling multiple streams in uncompressed .avi containers...or something else. I would never install a Green drive for this. You can use Blue for ordinary low-volume moderate speed editing and you can use RED as a sub for the black. But black is black and is the standard. It's what you are going to do with them that makes them different.
riredale wrote on 3/10/2014, 7:53 PM
I've never worked with uncompressed files, only DV, HDV, and MPEG2, so maybe that explains it. The green drives are plenty fast here.
Dach wrote on 3/10/2014, 8:21 PM
I see the appeal to the WD Blacks, but I am curious to what specific feature makes it the best. The 5 year warranty is unique, but 7200 RPM, 64 MB Cache and Sata6 connection can be found in other drives.

Allow me to propose this question. While editing will real time play back be approved on a 7200 RPM vs. 5400 RPM Drive.

While I'll appreciate the additional storage that comes with any drive, I'm just looking for the most out of the upgrade.
videoITguy wrote on 3/10/2014, 8:30 PM
Drive performance depends on what you are measuring and how you are measuring it.
Even 5400 rpm drives can deliver moderate speed streams without difficulty.
As for cache aboard, it does not make much difference until you get to special configurations, for example Raid assembly and even then the cache that is most important is the cache on the Raid backplane not the drive.
Dach wrote on 3/10/2014, 8:52 PM
I am coming to the conclusion that perhaps GB per dollar, not considering the warranty is coming into play.

Any thoughts on the Seagate, Hybrid drives?
john_dennis wrote on 3/10/2014, 10:14 PM

"Any thoughts on the Seagate, Hybrid drives?"

It's theoretically possible for a small project (<8 GB) that you could have most of the assets in cache if you worked on it long enough to have it displace everything else you might have passed through cache from the platters. In that case it would run like an SSD. If you use different assets (perhaps by moving from project to project for a little bit at the time) the performance would likely be more like a regular hard drive.

Personally, I have an SSD for the boot disk and spinning disks for video assets. I had an SSD for a project disk for a month or so and it didn't suck, but I just put it into another system that lost its SSD. I'm not suffering pangs of grief at losing the SSD for projects.

I use Hitachi Deskstar 7200 RPM disks with 64 MB of cache in both machines I use for video.

I have an Hitachi "Cool, Green" drive that I bought to back up my personal pictures and video but it's hard to justify since the first thing I do when I build a system is set a power profile to make sure the drives never spin down.

Virtually all my current drives are 2TB and I just bought a 2TB Seagate to hold video for playing on my Blu-ray player.

I recently bought 8 Western Digital Blacks at work to replace the drives in two security video recorders that run 24/7/365 and record 480 frames per second from 82 cameras between them. Though I don't have anything against Western Digital, the only reason I continue to use them in this application is because that's what the manufacturer shipped in the machines. One set of Western Digital Blacks ran from October, 2007 until the first failure last week. I policy replaced the other 4 drive set in December before the first failure. Shame on me for not doing both sets sooner.

I'm somewhat brand agnostic since I've owned disks from every manufacturer including lots of names that no longer exist.

GeeBax wrote on 3/11/2014, 1:06 AM
The new computer I am building at present has two 120GB Sandisk Extreme SSDs, one for the system drive and the other to backup the system.

It then has a Sandisk Extreme 480GB drive for editing and working space.

I also have two WD Reds of 4TB each as archive and storage drives.

No project I have ever worked on has exceeded 480GB, so the working SSD is going to be fine. The 8TB of archive space is fine for offloading projects once they are finished, and I have another 8TB of NAS hanging off the system as well.
Steve Mann wrote on 3/11/2014, 3:24 PM
The disc cache is where write data is cached. Data on hard discs are stored in clusters, one at a time, and clusters are usually not sequential. So, when a cluster is written and the next cluster is approaching the head, if the data are in the cache, the data gets written sooner. Otherwise, if the data is not in the cache or there is no cache, then the data data won't get written until the processor provides it.

Cache gives the drives very impressive burst data rates. The ones that the marketing department likes to highlight. And burst data rates means little to video editing.

The only data rate that makes any difference to video editing is the sustained data rate. That's how fast the disc can read or write data without the cache. Also, the sustained data rate is a difficult spec to uncover.
videoITguy wrote on 3/11/2014, 3:34 PM
Because harddrive manufacturing is an ever-changing source and model number change in distribution ..and not to mention assembly batching as well it is hard to find a place where you can evaluate that drive a is better than b.
Occassionally a few test sites do pop-up from time to time.
When I buy a new drive ( i've bought hundreds over 15 years) I use the old character based BlackMagic Design test for NLE suitability to evaluate the incoming drive performance.

YOU will be surprised at how large a variance actual delivered drives can have in perfomance. For example in one model year a test of a Maxtor against WD Black showed that operation thru the Sata3GBs interface the WD Black was nearly twice better. Yet the Maxtor retailed for more money, was a good reliable drive, and in every measure but NLE performance would have been a good bet.
DrLumen wrote on 3/12/2014, 6:40 PM
Like Steve mentioned, higher sustained is better for, well, pretty much everything - not just video editing.

For this reason faster drive speeds are better overall. WD Greens are cheap but slow. WD Black drives is my preferred drive. I had rather pay a bit more for overall speed than go cheap and get impatient or frustrated by lags.

intel i-4790k / Asus Z97 Pro / 32GB Crucial RAM / Nvidia GTX 560Ti / 500GB Samsung SSD / 256 GB Samsung SSD / 2-WDC 4TB Black HDD's / 2-WDC 1TB HDD's / 2-HP 23" Monitors / Various MIDI gear, controllers and audio interfaces

riredale wrote on 3/12/2014, 7:03 PM
But I think an argument can be made that after a certain point an increase in transfer speed is moot. Certainly for the stuff I do it is--all my renders are processor-bound, not transfer-rate bound.

And the flip side of a slower-spinning drive, the random-access delay, rarely becomes a factor, either, since renders usually involve big chunks of data, not itty-bitty stuff from here and there.

Or at least these are my assumptions. Someone probably has data showing green versus black on render performance.
fldave wrote on 3/12/2014, 7:27 PM
I'll never go back to Green drives. Black for sure. They perform pretty comparably, but I've always had the green go bad after 1 year, or whatever warranty period there is, and the black just keep going and going, well over 5 years in my experience. Interrupting a big project to have to replace drives and (hopefully) recover your data, it's just not worth the time nor risk to me.

A few years ago, someone reported the 5200 small drives performed better than the 7200. The smaller size/speed ratio actually spun linearly faster than than the 7200. That may have been some company just wanting to sell more 5200rpm drives, so take it with a grain of salt.

My 7200rpm Caviar Black allow me to view uncompressed or HuffYUV 1080i at full speed (no or few effects) on my Intel i5 quad 16GB RAM machine.

john_dennis wrote on 3/12/2014, 7:54 PM

"...an argument can be made that after a certain point an increase in transfer speed is moot."

I hate to be the one to cloud an otherwise delightful anecdotal thread with something as obnoxious as real data, but the point where drive performance becomes moot is around a couple megabytes per second for a single stream of compressed source being rendered to a compressed target.

Notice in this real-time total disk activity measurement with all 4 cores near 100%, the total transfer rate for all disk activity was ~ 1 megabyte per second with peaks in the sample time as high as 10 megabytes per second. This was on a Xeon server with 10,000 RPM SAS drives. The source video was 35 megabit per second AVC video, 48 kHz Stereo PCM audio in a MOV wrapper being rendered using the 25 mbps Mainconcept Blu-ray template.

Now, I'll grant that those who use uncompressed source files and/or render to uncompressed intermediate files will have a higher disk throughput requirement during renders, but I'll let them post their own screenshots.

Someone discuss timeline responsiveness.

videoITguy wrote on 3/12/2014, 8:33 PM
hmmm, seems to me those doubting thomases here ought to just read my first post in this thread and conclude " ...."
DrLumen wrote on 3/13/2014, 3:33 PM
While rendering is a time/cpu consuming process, personally, I don't consider rendering being the sole process of editing. Naturally if you have compressed clips, color correction, masks, effects, etc, more time will be spent rendering each frame rather than file operations.

However, you should see a marked improvement with the import of media or transfer of files across drives or on the lan with a 7200rpm drive.

About 15+ years ago I replaced a 7200 rpm drive with a 5400. Same exact system but going from a Seagate 7200 rpm drive to Seagate 5400 rpm drive. There was a marked increase of sluggishness. There was also much more dependence on page files/virtual ram then as well. No benchmarks; just personal experience.

intel i-4790k / Asus Z97 Pro / 32GB Crucial RAM / Nvidia GTX 560Ti / 500GB Samsung SSD / 256 GB Samsung SSD / 2-WDC 4TB Black HDD's / 2-WDC 1TB HDD's / 2-HP 23" Monitors / Various MIDI gear, controllers and audio interfaces

VideoFreq wrote on 3/18/2014, 1:12 AM
The slower the drive, the more memory you will need. I would borrow the money and get an SDD. You WILL NOT BELIEVE the difference and the issues it will solve. Loading, editing, backing up, copying - ALL four to five times faster. Vegas loads in two seconds. A two hour project loads in ten seconds. Rendering is flawless and crashless and in 32 bit Floating Point. Having the correct video card helps too.
john_dennis wrote on 3/18/2014, 1:42 AM

My curiousity got the best of me so I decided to test the effects of hard disks on renders when the source files are uncompressed.
I rendered three one minute sections of video with the following characteristics (typical):

Complete name : C:\Users\John\Desktop\Uncompressed Screenshots\Section 3.avi
Format : AVI
Format/Info : Audio Video Interleave
Format profile : OpenDML
File size : 5.86 GiB
Duration : 1mn 3s
Overall bit rate : 797 Mbps
TCOD : 1232898333
TCDO : 1863945416

ID : 0
Format : YUV
Codec ID : UYVY
Codec ID/Info : Uncompressed 16bpp. YUV 4:2:2 (Y sample at every pixel, U and V sampled at every second pixel horizontally on each line). A macropixel contains 2 pixels in 1 u_int32.
Duration : 1mn 3s
Bit rate : 795 Mbps
Width : 1 920 pixels
Height : 1 080 pixels
Display aspect ratio : 16:9
Frame rate : 23.976 fps
Color space : YUV
Chroma subsampling : 4:2:2
Compression mode : Lossless
Bits/(Pixel*Frame) : 16.000
Stream size : 5.84 GiB (100%)

ID : 1
Format : PCM
Format settings, Endianness : Little
Format settings, Sign : Signed
Codec ID : 1
Duration : 1mn 3s
Bit rate mode : Constant
Bit rate : 1 536 Kbps
Channel(s) : 2 channels
Sampling rate : 48.0 KHz
Bit depth : 16 bits
Stream size : 11.6 MiB (0%)
Alignment : Aligned on interleaves
Interleave, duration : 249 ms (5.98 video frames)
Interleave, preload duration : 250 ms

I created a project with no artistic value with the 1920x1080-24p screen split into three sections and part of each video showing at all times. This makes Vegas Pro 12 use the data from all three source files to produce the output file.

The first example shows the three uncompressed streams on three separate physical 7200 RPM spinning disks. The render target is an SSD and the media type is uncompressed avi typical of the Mediainfo above. Using 28% of CPU the system wrote 93.4 megabytes per second. The render promises to complete in ~67 seconds.

I then moved the three uncompressed source files to one of the 7200 RPM spinning disks and rendered to uncompressed video on the SSD.

The CPU load dropped to ~8% and the individual disk throughput dropped to ~23 megabytes per second. The total render time rose to 277 seconds. The single spindle was a bottleneck when all the uncompressed video had to come from one disk.

Thinking the SSD with no seek time would be faster, I moved all three uncompressed files to the SSD and rendered uncompressed to a separate 7200 RPM spinning disk.

The SSD was almost able to move as much data (~84.8 megabytes per second) as when the data was on three separate 7200 RPM spinning disks (~93.4 megabytes per second). The render promised to complete in 73 seconds.

Finally, I rendered to the 25 mbps Blu-ray template with all three uncompressed streams being read from one 7200 RPM spinning disk.

With the CPUs at ~57%, the data from the source disk at ~23 megabytes per second, and the avc file being written to a 7200 RPM spinning

disk at < 1megabyte per second the ~ one minute render promised to finish in 273 seconds.


Multiple uncompressed sources should be on separate spindles.

An SSD will work up to about three simultaneous streams, or until you run out of space, or until you run out of money.

For compressed output, disk speed has less effect on throughput even with uncompressed source files.

For compressed output with compressed input files, disk speed for modern disks has little to no effect on throughput.

VidMus wrote on 3/18/2014, 1:49 AM
"Vegas loads in two seconds."

I must have a problem because Vegas does not load in 2 seconds on my system. It is definitely faster but not as much as it should be.

I did a clone using 'Casper' from my hard drive to the SSD. Maybe that is why it is not quite up to par?

I am very reluctant to do a clean install of Windows because in order to use my current install disk I would have to call Microsoft and give them the numbers from my computer so they can give me their numbers to get it activated.

The problem is, my hearing is way too poor to do this anymore. I wish it were possible to use the internet with the numbers for people with bad speech hearing like me. Microsoft does not meet the needs of those with hearing disabilities in this regard. At least as far as I know. My speech hearing has been bad long enough now that it is getting difficult for me to say the numbers as well.

Still, they will have to drag me away from my cameras and computer system to get me to stop doing videos!

Sorry about the OT but I wanted to express a point.

VidMus wrote on 3/18/2014, 2:04 AM

Unless I missed something in your post going from an SSD to SSD would give more accurate results.

I have no spinners on my system at all. I have a 256 gig boot and a 500 gig project drive. Both SSD's.

Going from an SSD to a 7200 rpm drive, the slower drive will be a bottleneck. Even though the seek time of an SSD is virtually zero the transfer of bits from and then to the same SSD will be limited by the interface.

Copying a file from one SSD to another, Windows will show a certain rate. Going from one folder to another on the same SSD, the rate will be slower. And the first part of the test is with both SSD's being the same speed and also being the same speed of read and write which mine are. And yes, I did this test.

DrLumen wrote on 3/18/2014, 2:11 PM
Good info John.

intel i-4790k / Asus Z97 Pro / 32GB Crucial RAM / Nvidia GTX 560Ti / 500GB Samsung SSD / 256 GB Samsung SSD / 2-WDC 4TB Black HDD's / 2-WDC 1TB HDD's / 2-HP 23" Monitors / Various MIDI gear, controllers and audio interfaces

john_dennis wrote on 3/19/2014, 3:23 PM

Unless I missed something in your post going from an SSD to SSD would give more accurate results."

You apparently missed the rationale of the test. The exercise was not to test raw disk speed, but overall system throughput given different input / output data rates, physical data sources and Vegas Pro 12 overhead. Running SSD to SSD might have made it faster but not more accurate. The measurements speak for themselves.

Going from an SSD to a 7200 rpm drive, the slower drive will be a bottleneck.

I don’t currently have two SSDs in my machine though every machine has one. Last night I configured a RAM DISK and wrote to that instead of the 7200 RPM disk using the SSD for the three uncompressed source files. The result was somewhat disconcerting. The write to the RAM DISK went much faster but the reads from the SSD didn’t improve. I’m not ready to post or make any conclusions on that test.

Perhaps you would like to post some SSD to SSD measurements?