Why does my FW800 HDD render at exactly the same speed as my Thunderbolt SSD RAID?

Siderz · Mar 31, 2013

In January I bought a LaCie 256GB SSD Thunderbolt RAID to replace my 1TB FW800 G-Drive, since I thought the G-Drive was a bottleneck, after all those posts telling you to buy RAIDs.

The fan in the LaCie is so damn loud that it's distracting, but I thought that I might as well work with it seeing as it's 'faster' than the G-Drive. The first thing I did was test it with Blackmagic speed test, the SSD RAID reads at about 400MBps, the G-Drive about 80MBps; it's definitely faster.

Just a couple of weeks ago I thought "Why not test it to see if it is actually any faster?", I was getting so tired of the fan and every time I turned my Mac on, the SSD would start as well...meaning I need to turn it off.

I tested it by rendering a sequence in Premiere Pro CS5.5*, turned off one drive, turned on the other, and rendered the same sequence, and surprisingly they rendered at exactly the same time! I then tried exporting, and, yep, same time again.

Then I realised that the videos I was using were already compressed to be uploaded to YouTube...and by then I had already put the SSD to one side. So this morning I finally plugged it back in and tested it again with clips from my Canon 650D which compresses clips into a .MOV at something like 50Mbps. I also put multiple layers on top which I set to different blending modes, just to give it more work. And still, they render/export at exactly the same times...well the HDD was 1 second slower, that's probably just the lag caused from finding the files. The SSD loads projects a heck of a lot faster, that's for sure.

So...why does a regular FW800 HDD have the same render performance as a Thunderbolt SSD RAID? Since everyone says that people ought to buy RAIDs and stuff.

I have the 2012 21.5" iMac with:

3.1GHz i7
16GB RAM
GeForce 650M 512MB
1TB Fusion Drive

I can't imagine there's a bottleneck in that.

Anyone know any other tests I can perform? Example projects to render? Considering my tests used my regular workflow, I don't see much point in trying other tests anyway.

*Rendered the sequence within Premiere Pro, exported using Media Encoder.

LethalWolfe · Mar 31, 2013

Things like rendering, exporting, transcoding, etc., are CPU intensive (and/or GPU intensive if you are using software that leverages the GPU). If you switched on the Activity Monitor in OSX and did the same tests I bet you'd see a lot of CPU usage and relatively little drive activity. In an extreme case, such as Uncompressed 1080p24 stored on a HDD connected via USB 1.1, the storage would impact rendering speeds.

If you did a test that was more disk I/O intensive, such as pulling multiple video streams, then the ThB RAID should allow you to pull more streams than the FW800 drive.

HobeSoundDarryl · Mar 31, 2013

It would appear that there is no bottleneck in the storage system, meaning maybe you have enough RAM for the CPU to work as fast as it can without ever having to wait on the SSD or Firewire setup. Conceptually, the SSD should yield a faster workflow but your setup may be that Firewire can keep up with the max demands of your workflow. Just a guess but I read your comments like there is no weaker link in your storage systems- either are "good enough" for what you are doing.

Siderz · Mar 31, 2013

LethalWolfe said:
Things like rendering, exporting, transcoding, etc., are CPU intensive (and/or GPU intensive if you are using software that leverages the GPU). If you switched on the Activity Monitor in OSX and did the same tests I bet you'd see a lot of CPU usage and relatively little drive activity. In an extreme case, such as Uncompressed 1080p24 stored on a HDD connected via USB 1.1, the storage would impact rendering speeds.

If you did a test that was more disk I/O intensive, such as pulling multiple video streams, then the ThB RAID should allow you to pull more streams than the FW800 drive.

Wait wait! Slow down!

How am I supposed to check drive activity on specific drives? I can see the graph, but not a way of selecting separate drives.

And how does the storage impact render speeds?

And the sequence I was rendering had 5 tracks composited on top of each other. However, they were all very dark scenes with very little colour in them, so I'm assuming this is much less intensive?

Also I've already looked at the CPU usage during rendering. Goes all the way up. Is there a way to check GPU usage or does it not work like that? I'd be interested to see the GPU as well.

----------

HobeSoundDarryl said:
It would appear that there is no bottleneck in the storage system, meaning maybe you have enough RAM for the CPU to work as fast as it can without ever having to wait on the SSD or Firewire setup. Conceptually, the SSD should yield a faster workflow but your setup may be that Firewire can keep up with the max demands of your workflow. Just a guess but I read your comments like there is no weaker link in your storage systems- either are "good enough" for what you are doing.

Interesting.

And the SSD is really good at loading projects quickly, all those thumbnails, and also makes conforming a lot quicker. Mustn't forget to mention that moving files across to it at lightning speed is excellent.

Well I'll sell it then if the FW800 one is adequate. I can't be doing with the loudness, nor a useless drive. I'm gonna spend the money on a monitor and maybe another [but used] FW800 G-Drive.

LethalWolfe · Mar 31, 2013

Siderz said:
How am I supposed to check drive activity on specific drives? I can see the graph, but not a way of selecting separate drives.

Only have one drive hooked up at a time? There might be a third party utility that can track individual drive I/O.

And how does the storage impact render speeds?

Storage would only impact render speeds if the CPU (or GPU) could process the footage faster than it could can be read/written from/to the drive. In your case the drives (both drives) are not the bottle neck in the rendering process hence the render performance being the same regardless of which drive you used.

And the sequence I was rendering had 5 tracks composited on top of each other. However, they were all very dark scenes with very little colour in them, so I'm assuming this is much less intensive?

It probably doesn't make a difference.

Is there a way to check GPU usage or does it not work like that? I'd be interested to see the GPU as well.

Not sure though I wouldn't be surprised if there was a third party utility for it.

HobeSoundDarryl · Apr 1, 2013

Siderz said:
And the SSD is really good at loading projects quickly, all those thumbnails, and also makes conforming a lot quicker. Mustn't forget to mention that moving files across to it at lightning speed is excellent.

Well I'll sell it then if the FW800 one is adequate. I can't be doing with the loudness, nor a useless drive. I'm gonna spend the money on a monitor and maybe another [but used] FW800 G-Drive.

Keep in mind, I'm just making a best guess based on what you've shared. A lot of people swear by using SSD raids for video editing. I suspect the opportunity with SSD raids is more in where the reading & writing function is used very heavily rather than how you seem to be using it. Where there is a lot of (especially) read activity, the SSD is going to show it's advantages. For example, I store big clip art collections on the SSD. Browsing all those individual files is faster than browsing them stored on a hard drive. There is an obvious speed gain in that scenario.

In my own case, I picked up one of those 4 bay OWC RAIDs for video editing on an iMac via Firewire. It has always seemed plenty fast for me. I've tried some smaller edits by putting the file on an SSD but did not "feel" a difference (other than initial load time). For rendering from FCP X, I could never "feel" a difference. I suspect for basic video editing needs, CPU + RAM + Graphics Card are more important (for speed objectives).

Now, if you were working for a Studio and needed rapid access to lot of individual clips, all those loads would certainly be faster from an SSD raid rather than some big bank of RAID hard drives.

BadLegBanks · Apr 1, 2013

Siderz said:
How am I supposed to check drive activity on specific drives? I can see the graph, but not a way of selecting separate drives.

...

Is there a way to check GPU usage or does it not work like that? I'd be interested to see the GPU as well.

If media creation is something you plan to do regularly, I'd recommend getting something like iStat Menus or atMonitor.

mjsanders5uk · Apr 2, 2013

Have you tired running the Blackmagic or Aja disk speed tests?

I think both programms are free downloads.

Siderz · Apr 2, 2013

mjsanders5uk said:
Have you tired running the Blackmagic or Aja disk speed tests?

I think both programms are free downloads.

Yes:

Siderz said:
The first thing I did was test it with Blackmagic speed test, the SSD RAID reads at about 400MBps, the G-Drive about 80MBps; it's definitely faster.

xStep · Apr 7, 2013

Siderz said:
And the sequence I was rendering had 5 tracks composited on top of each other. However, they were all very dark scenes with very little colour in them, so I'm assuming this is much less intensive?

Compositing two frames (or scenes) that are identical would go through the same processing as two frames that are unlike one another. Testing each pixel for equality is a waste of CPU cycles. It would be faster to simply perform pixel renders on a few identical pixels rather than spend time testing millions of pixels that are not identical. So, no, your render is not less intensive.

Siderz said:
Also I've already looked at the CPU usage during rendering. Goes all the way up. Is there a way to check GPU usage or does it not work like that? I'd be interested to see the GPU as well.

Looks like you have found your bottleneck. You are maxing out the CPUs. Doing this is invalidating any disk I/O tests.

Siderz · Apr 7, 2013

xStep said:
Looks like you have found your bottleneck. You are maxing out the CPUs. Doing this is invalidating any disk I/O tests.

I didn't actually mean it literally, attached is an image of how high it actually goes.

I doubt an i7-3770S could possibly be a bottleneck.

And besides, that's what I want right? For it to use up the whole CPU while rendering?

xStep · Apr 7, 2013

Siderz said:
I didn't actually mean it literally, attached is an image of how high it actually goes.

I doubt an i7-3770k could possibly be a bottleneck.

And besides, that's what I want right? For it to use up the whole CPU while rendering?

Oh yea, you can saturate the i7! See attachment. What I'm doing there is having Xcode analyze a large project.

Sure, you want the processors doing all of the work that they can. It's a balancing game. If your drives are too slow to deliver the items to be rendered or slow on writing that render back, then that would be a bottle neck to fix, if possible. If you are short on memory then potentially you'll be using the drives as a virtual memory system which wastes a great deal of time. As pointed out by someone else, you analyze your system via tools like Activity Monitor.

When you have 16.8MB of free memory, then closing Safari which is using 13.67GB of your 16GB can help the system out greatly. Yep, that has happened to me.

HobeSoundDarryl · Apr 7, 2013

Again, everything seems to keep supporting the idea that the storage system- either Firewire or Thunderbolt SSD- are both delivering the files to the processors fast enough. Thus, no meaningful advantage of Thunderbolt over Firewire in this case. In short, it appears that OPs firewire setup is good enough for his purposes and hardware "as is."

Obviously, Thunderbolt as a spec has the bandwidth to be much faster than Firewire but the rest of the system needs to capitalize on that speed. OP seems to be putting a pretty full load on his i7 and the FW is feeding it enough data that TB advantages don't make any difference (the processor/program really can't work with much more).

If he networked up a bunch of Macs and handed off rendering to a "farm" of CPUs, I bet he would find the point at which the TB does make a difference. At some point the CPU farm would be able to receive more video data than the FW setup could feed it. It seems it would be there where the difference would show itself.

MIDI_EVIL · May 23, 2013

BadLegBanks said:
If media creation is something you plan to do regularly, I'd recommend getting something like iStat Menus or atMonitor.

iStat menus is great for checking drive activity of all drives attached. It also shows detailed CPU and GPU usage.

Siderz · May 23, 2013

MIDI_EVIL said:
iStat menus is great for checking drive activity of all drives attached. It also shows detailed CPU and GPU usage.

I have iStat Pro, I don't like the iStat Menus form factor, and I probably wouldn't pay $16 (~£10) for it either.

Search

Search

Why does my FW800 HDD render at exactly the same speed as my Thunderbolt SSD RAID?

Siderz

macrumors 6502a

LethalWolfe

macrumors G3

HobeSoundDarryl

macrumors G5

Siderz

macrumors 6502a

LethalWolfe

macrumors G3

HobeSoundDarryl

macrumors G5

BadLegBanks

macrumors newbie

mjsanders5uk

macrumors regular

Siderz

macrumors 6502a

xStep

macrumors 68020

Siderz

macrumors 6502a

Attachments

xStep

macrumors 68020

Attachments

HobeSoundDarryl

macrumors G5

MIDI_EVIL

macrumors 65816

Siderz

macrumors 6502a

Our Staff