LEX's Track count Drive test - Glyph vs Rocstor

[LEX]

Part 1

I took it upon myself to do a little test, based on Glyph's "track count" test:
http://www.glyphtech.com/support/trackcount.html

Now I am going to compare their GT 050Q 160GB - eSATA port test verses a Rocstor 850 750Gig drive.

Any one can read the Glyph page on their specs per different drives and test.

So first there are some differences, so I will spell them out first:

1)
The "Density Test" is a collection of Glyph's proprietary Pro Tools test sessions that determine how many audio tracks a Glyph drive can reliably handle. It can run at 48k, 96k, and 192k, and has four levels of edit intensity.

I am only running a 48k test, as I don't ever do 96k and 192k (at this time)

Glyph also never mentioned whether the sessions were 16bit or 24bit, so I am going to assume their sessions were 24bit. 16 bit will yield HIGHER track counts.

2) Here are Glyph's "Density Levels"
Density Level 1 one edit every two seconds across every track in the test session
Density Level 2 one edit every second across every track in the test session.
Density Level 3 two edits every second across every track in the test session
Density Level 4 three edits every second across every track in the test session

I did not use all 4 density LEVELs.
Here are LEX's "Density Levels"

Density Level 0 No Edits - Just tracks
Density Level 2 one edit every second across every track in the test session.
Density Level 4 three edits every second across every track in the test session.
Density Level LEX - Six edits every second across every track in the test session.

3) There is another difference in their tests and mine. There are 2 types of Playback Methods:

a) Play and run
b) Loop

Play and run, the drive play head is just moving across the drive platter.
Looping requires the playhead to "jump" back to the starting point and continue playing.

With the 2 different types of "play", this torture test is an ADDON to Glyph's test.

4) I am going to assume based on Glyph's "Density Levels" and each density is 1 minute in length equaling a 4 minute audio file each.

I took a 5.1 mix and created 84 individual track files (that is 84 separate audio files, not 6 audio files duplicated across the tracks) at 6 minutes in length total.

48k, 24bit files.

5) Drive differences:

Glyph is using a 160 gig Seagate Barracuda 7200.9 SATAII, 8meg buffer, 7200 RPM, connected via eSATA PCI card. Full format, empty.
G5 Dual 2.0, OSX 10.4.9, ProTools 7.1, HD Accel 2

Rocstor 850 - Seagate Barracuda 750gig SATAII, 32meg buffer, 7200 RPM, Connected via FW800 to 400 cable to a FW 400 hub. Drive formatted is 698. Available room, 220 gigs.
Session size was 4.4 gigs.
Drive is about 75 percent full.
G4 Dual 1.25, PSX 10.4.9, ProTools 7.3.1, HD3 with 1 Accel card.

So basically we are going to test a SATAII drive running at SATAII speed verses a SATAII drive running at FW400 speed.

I should also mention Glyph's statement here:

Hard drives are fastest when they are nearly empty, so expect decreased performance when it starts to fill with data.

It should also be noted that smaller drive perform better as there are less platters to read.

End Part 1

LEX

[LEX]

Part 2:

(It should be noted that ProTools will sometimes play through the tracks with no error, but played again will through and error. So, In LEX's tests, each "Density Level" was played through multiple times to ensure no random error came up. If an error came on the 2nd playback, 1 track was reduced (from 84 tracks) until the session could play 5 times without error)

Glyph's eSATA results based on their web page:

GT 050Q 160GB - eSATA port - 48k Tracks
Density Level 1 96
Density Level 2 59
Density Level 3 44
Density Level 4 29
Level 1 on last 20% of drive 74
Level 2 on last 20% of drive 44

I am going to assume the last 20 percent means there is either data filling the other 80 percent, or the drive has been partitioned in to 80/20 segments.

I am also going to assume this is a "play" until the audio end, non looping.

LEX's Rocstor 850 results, via FW400 - 48k tracks - 24bit.

Straight Play, non looping -
Density Level 0 - 79
Density Level 2 - 77
Density Level 4 - 77
Density Level LEX - 74

Looping -
Density Level 0 - 57
Density Level 2 - 56
Density Level 4 - 55
Density Level LEX - 54

Now since the drive can read from the beginning of the file or the middle of the file, I edited it in this Density level way to make sure that it wasn't because the edits were getting smaller.

Here is the Density edits From the top of the audio file:

Density 4 - 1 minute
Density 2 - 1 minute
Density 0 - 1 minute
Density 4 - 30 seconds
Density 0 - 30 seconds
Density LEX - 30 seconds
Density 0 - 1 minute 30 seconds

Looping from any point resulted in 54 tracks playing back.
15 second loop, 30 second loop, 1 minute loop, 6 minute loop (entire selection) all yielded 54 track playback.

Part 2:

LEX

[LEX]

So everyone knows, I do use Rocstor.

But this isn't a What I use is the best Post. Rather, since Glyph posts "Torture Tests" online, I would do the same torture test, and then some to see what the results provided based on the DATA Glyph posted.

Had the results been poor for Rocstor, I still would have posted the results.
Honesty, I expected MY numbers to be lower.

I did these tests with my assistant, as to have a "witness" and knowledge of how far we CAN push the drives.

When I have a free and fresh formatted Rocstor 850, I'll run the test again to see if the numbers are higher.

I want to note again, this test drive was 75 percent full. What should also be noted is the files taking up 75 percent of the drive.
They are DV NTSC video files taking up 20 to 30 gig chunks of the drive at a time.

A side note on disk usage:

During Straight Play, non looping the drive activity was 80% to 90%. Density 0 to Density LEX.

During Looped Play, the drive activity was 50% to 65%, Density 0 to Density LEX, with Drive peaks hitting 90 percent on loop return.

So, whatever drive company you chose at least you will have a basis to go on and compare.
Like I said, when I have a fresh drive, I'll update this test.

Eventually, I'll try a 96k and 192k session as well, so eSATA connection may yield a better performace than a SATAII drive via FW400. I would expect that.

It is quite shocking though that even in the other Glyph tests, RAID 0,1, JBOD, FW800 and eSATA a larger drive, 75% full via FW400 BEAT ALL their drives with Density LEX, which is double the edits per second.

LEX

[TAFKAT]

Hey Lex,

Thanks for the time and energy on that.., some interesting numbers there.

I am quite surprised that the eSATA drive did so badly to be honest, but we need to look at the actual drives as well.

The drive in the Glyph is an old 8MB Cache drive, while the Rocstor had the newer 32MB Cache drive, so that could explain some of the variable...

Heres an idea if you can be bothered of course.. :wink:

Swap the HD's in the enclosures and see how the numbers pan out..

[LEX]

Hey Lex,

Thanks for the time and energy on that.., some interesting numbers there.

I am quite surprised that the eSATA drive did so badly to be honest, but we need to look at the actual drives as well.

The drive in the Glyph is an old 8MB Cache drive, while the Rocstor had the newer 32MB Cache drive, so that could explain some of the variable...

Heres an idea if you can be bothered of course.. :wink:

Swap the HD's in the enclosures and see how the numbers pan out..

I don't have a Glyph. I was working off their numbers and their test data on their website. Link in 1st post.

Yes, the 8meg buffer vs 32 megs. But again, we are talking a 75 percent full drive vs a fresh drive.

Look at the RAID numbers as well.

Again, there maybe some differences in higher Sample Rates.

This is why I took the test further. Again everything based off their information on their website.

LEX

[TAFKAT]

Oh O.K,

Seeing I am not that fluent of PT, can you forward a 48 track test session you created, I only have PTLE 8..

I'll run some tests up here when I get a chance on a basic eSata external HD I have lying around with a WD 500 GB / 16MB Cache HD, and see what I come up with.

Do you have a Rocstor with eSata handy ?

[LEX]

Oh O.K,

Being a lazy prick, can you forward a 48 track test session you created, I only have PTLE 8..

I'll run some tests up here when I get a chance on a basic eSata external HD I have lying around with a WD 500 GB / 16MB Cache HD, and see what I come up with.

Do you have a Rocstor with eSata handy ?

It is 4.4 gigs, the session.

It has to be individual audio files per track.

I have a eSATA drive, just no eSATA port on my systems. Remember, my test drive was a SATA2 drive, but connected via FW800 port with a FW800 to FW400 cable, into a FW400 hub into the native G4 FW ports.

Like I said, I based EVERYTHING on glyph's online torture test/performance reports and then added kickers to push it further.

So based on their data, I followed everything to the TEE and added more.
Unless there is a variable they are not including, as well as a test session for users to confirm.

Some of the difference's that "could" be possible could be PT hardware buffer size (though they never stated it), playback engine, CPU ect.
Though I would be willing to guess with PT, it wouldn't make a difference.

I do find the "lack" of information on Glyph's site interesting.
No indication 16bit/24bit, no file length or size, no other settings.

I wonder if they have left that out because if people test, users might find other drives performing equally or better but at cheaper prices.

I doubt anyone has tested Glyph's "torture" test other than me.

LEX

[Sam]

I would be interested to see what would happen if each of the edits was a unique file - as in all the edits had been bounced to new small files.....my experience is that when I do a lot of offline processing on small edits creating 'new versions' I end up with thousands of small <5mb files all referenced into the session and that is when things begin to struggle for me....maybe I can find an easy way to create such a session as a test bed....

[LEX]

I would be interested to see what would happen if each of the edits was a unique file - as in all the edits had been bounced to new small files.....my experience is that when I do a lot of offline processing on small edits creating 'new versions' I end up with thousands of small <5mb files all referenced into the session and that is when things begin to struggle for me....maybe I can find an easy way to create such a session as a test bed....

I thought of that as well, but followed the Glyph testing standard.

In all, there were 13600 different regions.

What would probably bring the drives to their knees would be this.

Chop, consolidate the audio files, then reorder them.
As consolidating them in order will create files in order will be read that way.
But once they are reordered, the drive will have to jump more randomly reducing track size.

Might have to do that test too, but that will take some time to create.

LEX

[TAFKAT]

It is 4.4 gigs, the session.

It has to be individual audio files per track.

Replace the files with the audio and sines from my DAWbench DSP session, the sines compress down to close to zero.. :D

You'll need to add an extra 3 sine waves to get the track count to 48.

That will be 48 Stereo tracks btw.., only 16bit/44.1 , but still a way to get the test session into some other hands..

[LEX]

Replace the files with the audio and sines from my DAWbench DSP session, the sines compress down to close to zero.. :D

You'll need to add an extra 3 sine waves to get the track count to 48.

That will be 48 Stereo tracks btw.., only 16bit/44.1 , but still a way to get the test session into some other hands..

How long are the audio files? I don't remember. The session test is 48k 24 bit.

Length matters.

LEX