-
Drive Performance
We’ve been experiencing some playback issues and are trying to figure out where our problems might lie, but are a little suspicious of a drive array, but it’s hard to judge for sure.
I would be very grateful if some people could post Xbench drive benchmark results for different known-good drive configurations, so I can get an idea of how ours are stacking up.
Xbench is a very small freeware benchmark application (stand-alone, no install required, can run from a USB stick) – available here:
https://www.xbench.com/ -
12 Replies
-
Dylan,
Almost everyone uses either the throughput tests that come with with Kona or Blackmagic drivers. There’s no compelling reason to get everyone to shift paradigms. Why don’t you post your numbers using one of those and we can better help interpret the data using a benchmark most are all familiar with.
David Roth Weiss
Director/Editor
David Weiss Productions, Inc.
Los AngelesPOST-PRODUCTION WITHOUT THE USUAL INSANITY ™
A forum host of Creative COW’s Apple Final Cut Pro, Business & Marketing, and Indie Film & Documentary forums.
-
BTW, if you list details of your FCP hardware and software it certainly makes life easier for those of us who try to help.
For instance, what type of array do you have? And, what capture card?
David Roth Weiss
Director/Editor
David Weiss Productions, Inc.
Los AngelesPOST-PRODUCTION WITHOUT THE USUAL INSANITY ™
A forum host of Creative COW’s Apple Final Cut Pro, Business & Marketing, and Indie Film & Documentary forums.
-
Well I have used the Blackmagic one, but it’s reporting impossibly slow results, so something is a amiss – according to the Blackmagic disk test I can’t even play DV video of my external SATA drive, but I can clearly demonstrate that is incorrect.
So I’m trying to get a little more information with a slightly more detailed test.
-
In this case the drive in question is an Addonics external array (4x 750GB drives on a Hardware Multiplier in RAID 1+0 configuration).
My suspicion is that the overhead of the hardware multiplier (which has to emulate a single SATA device to the system, as well as manage a hardware RAID array) is possibly an issue.
We’re quite possibly going to replace with a PM-aware SATA controller and a regular Port Multiplier in the array (shifting the burden to the controller card and operating system).
-
The legendary Bob Zelin is not a fan of the PM controllers, and he has suggested in several missives that they do tend to fail. I’ve stayed away from PM because of the things he’s said.
The reason the BM test and the Kona test are reliable for us is that they take many factors into account that directly apply to what we’re doing, while typical benchmarks leave many factors out of the equation. For example, the Kona test will also do a simutaneous DMA test which involves moving data through the Kona card, thus give realistic disk read/write speeds approximate to actual video capture using the card.
If the bechmark you are running is significantly different than your BM speed test, I think you may have a bottleneck somewhere that the benchmark app is missing.
David Roth Weiss
Director/Editor
David Weiss Productions, Inc.
Los AngelesPOST-PRODUCTION WITHOUT THE USUAL INSANITY ™
A forum host of Creative COW’s Apple Final Cut Pro, Business & Marketing, and Indie Film & Documentary forums.
-
An alternative is an SAS controller with SATA fanout cable to multiple SATA drives (which should be more reliable).
Or plain old FW800 (although we it feels like a pretty limiting bus when better ones are available).
For our purposes Fibrechannel will be overkill…
The problem I have with the BM tool is that it’s reporting numbers that are absurd (like 1.9MB/s read speed) which is insane, as if that were the case it would be barely usable for anything. So clearly the drive doesn’t perform very well in that test for some reason (our SMB network drives rate higher in the BM test tool)… But as the result is clearly wrong (we can achieve much better rates, even with BM codecs via the card) I am looking elsewhere for ideas.
-
Dylan,
I’m making a clone right now, but I’ll try the Xbench in a little while. I’ll be running it on my 4-drive SATA with Raid-0. Meanwhile, post your Xbench results so I can see them before I run my own for reference.
David Roth Weiss
Director/Editor
David Weiss Productions, Inc.
Los AngelesPOST-PRODUCTION WITHOUT THE USUAL INSANITY ™
A forum host of Creative COW’s Apple Final Cut Pro, Business & Marketing, and Indie Film & Documentary forums.
-
Here’s what I get from our external Addonics Hardware PM drive:
Results 74.88
System Info
Xbench Version 1.3
System Version 10.4.11 (8S2167)
Physical RAM 5120 MB
Model MacPro2,1
Drive Type External Disk 0
Disk Test 74.88
Sequential 77.40
Uncached Write 41.00 25.17 MB/sec [4K blocks]
Uncached Write 86.85 49.14 MB/sec [256K blocks]
Uncached Read 89.87 26.30 MB/sec [4K blocks]
Uncached Read 215.30 108.21 MB/sec [256K blocks]
Random 72.52
Uncached Write 30.09 3.19 MB/sec [4K blocks]
Uncached Write 216.89 69.43 MB/sec [256K blocks]
Uncached Read 99.29 0.70 MB/sec [4K blocks]
Uncached Read 138.03 25.61 MB/sec [256K blocks] -
Dylan,
I ran xbatch and I think the results below are meaningless. Meanwhile, my raid-0 is averaging like 215MBb/s read and write according to the Kona speed test.
Results 74.07
System Info
Xbench Version 1.3
System Version 10.5.2 (9C7010)
Physical RAM 5632 MB
Model PowerMac7,3
Processor PowerPC G5x2 @ 2.30 GHz
L1 Cache 64K (instruction), 32K (data)
L2 Cache 512K @ 2.30 GHz
Bus Frequency 1 GHz
Video Card ATY,RV351
Drive Type Maxtor 6B250S0 Maxtor 6B250S0
CPU Test 124.69
GCD Loop 101.67 5.36 Mops/sec
Floating Point Basic 104.56 2.48 Gflop/sec
AltiVec Basic 267.21 10.65 Gflop/sec
vecLib FFT 118.89 3.92 Gflop/sec
Floating Point Library 117.03 20.38 Mops/sec
Thread Test 90.28
Computation 91.11 1.85 Mops/sec, 4 threads
Lock Contention 89.46 3.85 Mlocks/sec, 4 threads
Memory Test 86.04
System 79.47
Allocate 179.54 659.33 Kalloc/sec
Fill 87.59 4258.97 MB/sec
Copy 48.16 994.80 MB/sec
Stream 93.80
Copy 87.60 1809.33 MB/sec [G5]
Scale 89.95 1858.41 MB/sec [G5]
Add 98.89 2106.54 MB/sec [G5]
Triad 99.99 2139.04 MB/sec [G5]
Quartz Graphics Test 114.31
Line 98.89 6.58 Klines/sec [50% alpha]
Rectangle 106.06 31.66 Krects/sec [50% alpha]
Circle 103.36 8.43 Kcircles/sec [50% alpha]
Bezier 103.82 2.62 Kbeziers/sec [50% alpha]
Text 204.34 12.78 Kchars/sec
OpenGL Graphics Test 94.73
Spinning Squares 94.73 120.18 frames/sec
User Interface Test 39.83
Elements 39.83 182.81 refresh/sec
Disk Test 51.61
Sequential 78.73
Uncached Write 80.31 49.31 MB/sec [4K blocks]
Uncached Write 79.15 44.78 MB/sec [256K blocks]
Uncached Read 65.06 19.04 MB/sec [4K blocks]
Uncached Read 96.64 48.57 MB/sec [256K blocks]
Random 38.39
Uncached Write 15.29 1.62 MB/sec [4K blocks]
Uncached Write 66.95 21.43 MB/sec [256K blocks]
Uncached Read 78.42 0.56 MB/sec [4K blocks]
Uncached Read 90.08 16.72 MB/sec [256K blocks]David Roth Weiss
Director/Editor
David Weiss Productions, Inc.
Los AngelesPOST-PRODUCTION WITHOUT THE USUAL INSANITY ™
A forum host of Creative COW’s Apple Final Cut Pro, Business & Marketing, and Indie Film & Documentary forums.
-
I guess I will never know.. 🙁
I’m certain that the Blackmagic test isn’t accurate (although perhaps signposts some sort of larger issue). And the Xbench test tells me nothing.
All I have is slightly inconsistent performance and suspicions.
Why can’t everything just work? Ever.
Reply to this Discussion! Login or Sign Up