Maximum Polyphony Homebuilt DAW Piano Thread

joojoo1234

Distinguished
Jun 28, 2009
84
0
18,640
Building a DAW Please don't complain about the organization of this thread as it is casually done.

I would not do a raid 0 for anything but piano samples as its better to put the samples on different disks.

On a quest for maximum polyphony on a DAW for piano samples this is what I came up with and I decided to do a somewhat casual thread on my results if someone happens to google and find this..

When creating the Raid 0 and choosing the cluster size on...

WINDOWS XP PRO ON WD SINGLE 320GB BLUE

DYNAMIC DRIVES SOFTWARE RAID 0 IN XP PRO ON 4 320GB WD BLUE DRIVES
8GB RAID PARTITIONS TOTAL (2000mb PER DRIVE)
RUNNING TASCAM GVI not gigastudio but a vst relative.... will buy Kontakt in the future no doubt.

Cluster size I tested and my polyphony and piano response.

BAD TO GOOD
16K BAD
1024K BAD AS 16K
32K BETTER BUT BAD
4026K QUITE DECENT BUT NOT AS GOOD AS 512K
512K GREAT
64K VERY GREAT (the best)

SUBJECTIVE BUT I BET YOU CAN TRUST IT... WITH FASTER DRIVES PERHAPS 512K WOULD BE A BETTER CHOICE

I am getting about 250-300 usable polyphony... starts to get haywire at 250 but often goes to 300 occassionally can get to 325 or so.

Processor use is low. Samples had no gigapulse.
Sample was Sampletekk Post PMI Emperor using the 24/12 choice

WINDOWS XP PRO ON 320GB BLUE
4X BLUE 320GB IN RAID 0 (they are almost totally quite)

I now think that a better way then striping to get maximum polyphony would be to divide the piano samples by the number of drives. I know in GVI you can combine different pianos so there is probably no reason why you could not design three versions (instruments) of say the sampletekk emperor piano with the required missing octaves and combine these. Assuming you could combine three instruments. You obviously would not have to combine the instruments but just direct the midi toward each gvi instance. I may do this if I can figure out how to do it. I am not an expert in using samplers. I think now that if I take four 320gb drives and divide the piano up over these four that I will get better polyphony. I would guess this based on the fact that I had these on an F1 Samsung 1TB and got 150 polyphony with indexing of the drive on...(I forgot to turn it off) and would have gotten more if the computer was more properly setup for music daw work.

TURN OFF INDEXING OF DRIVES!!!!
AND GO THROUGH THE XP STEPS TO MAKE XP FLY AS A DAW AS THEY ARE CRUCIAL


SPECS.....................................................
AMD X4 940 cpu at normal clock speed
corsair 800 4gb ram

Model
Brand CORSAIR
Model TWIN2X4096-6400C5
Type 240-Pin DDR2 SDRAM
Tech Spec
Capacity 4GB (2 x 2GB)
Speed DDR2 800 (PC2 6400)
Cas Latency 5
Timing 5-5-5-18
Voltage 1.8V
Multi-channel Kit Dual Channel Kit
Heat Spreader Yes
Manufacturer Warranty
Parts Lifetime limited
Labor Lifetime limited

motherboard is
GIGABYTE GA-MA790GP-UD4H AM2+/AM2 AMD 790GX HDMI ATX AMD Motherboard

Model
Brand GIGABYTE
Model GA-MA790GP-UD4H
Supported CPU
CPU Socket Type AM2+/AM2
CPU Type Phenom FX / Phenom / Athlon 64 FX / Athlon 64 X2
FSB 2600MHz Hyper Transport (5200 MT/s)
Chipsets
North Bridge AMD 790GX
South Bridge AMD SB750
Memory
Number of Memory Slots 4×240pin
Memory Standard DDR2 1200 / 1066 / 800
(Whether 1200 MHz memory speed is supported depends on the CPU being used)
Maximum Memory Supported 16GB
Channel Supported Dual Channel
Expansion Slots
PCI Express 2.0 x16 1 x PCI Express x16 slot, running at x16 (PCIEX16_1)
1 x PCI Express x16 slot, running at x8 (PCIEX8_1)
PCI Express x1 3
PCI Slots 2
Storage Devices
PATA 1 x ATA100 2 Dev. Max
SATA 3Gb/s 6
SATA RAID 0/1/5/10/JBOD
Onboard Video
Onboard Video Chipset ATI Radeon HD3300 graphics built-in 128MB DDR3 1333MHz SidePort Memory
Onboard Audio
Audio Chipset Realtek ALC889A
Audio Channels 8 Channels
Onboard LAN
LAN Chipset Realtek 8111C
Max LAN Speed 10/100/1000Mbps
Rear Panel Ports
PS/2 2
Video Ports D-Sub + DVI
HDMI 1 x HDMI
USB 4 x USB 2.0
IEEE 1394 1 x IEEE 1394a
S/PDIF Out 1 x Optical
Audio Ports 6 Ports
Onboard USB
Onboard USB 8 x USB 2.0
Onboard 1394
Onboard 1394 2x 1394a
Physical Spec
Form Factor ATX
Dimensions 12.0" x 9.6"
Power Pin 24 Pin

Onboard video is being used.


500 watt power supply
Model
Brand OCZ
Model OCZ550FTY
Series Fatal1ty
Spec
Type ATX12V / EPS12V
Maximum Power 550W
Fans 120mm fan
PFC Active
Main Connector 20+4Pin
+12V Rails 2
PCI-Express Connector 1 x 6-Pin, 1 x 6+2-Pin
SATA Power Connector 6
SLI Ready
CrossFire Ready
Modular Yes
Efficiency 82% Typical
Over Voltage Protection Yes
Input Voltage 100 - 240 V
Input Frequency Range 50/60 Hz
Input Current 10A
Output +3.3V@25A,+5V@25A,+12V1@25A,+12V2@25A,-12V@0.3A, +5VSB@2.5A
MTBF >100,000 Hours
Features
Connectors 1 x Main connector (20+4Pin)
1 x 12V(P4)
1 x 12V(8Pin)
4 x Peripheral
6 x SATA
2 x Floppy
2 x PCI-E
Features OCZ PowerWhisper Technology
 

joojoo1234

Distinguished
Jun 28, 2009
84
0
18,640


I have thought of this... I just wanted more gigabytes or I would have gone ssd. Maybe I should have in truth.

315 mb/s seq sustained read on sandrasoft benchmarking for the 4 drive raid 0 512 cluster.

327 mb/s seq sustained read on sandrasoft benchmarking for the 4 drive raid 0 64 cluster. This shows that my subjective playing experience is right and that it was slightly better or faster.
 

KyleSTL

Distinguished
Aug 17, 2007
1,678
0
19,790

I do not understand this statement. Please elaborate.

Are you recording many tracks at very high quality levels? Why do you need such ample HDD bandwidth? Sorry for the questions if they're stupid, but I know nothing about computer-based audio recording.
 

joojoo1234

Distinguished
Jun 28, 2009
84
0
18,640



Should explain. It is well known that sample libraries (orchestral for sure) desire to have each section of the orchestra on its own disk for optimal performance rather than on a raid 0. This is very easily accomplished and configured. Piano samples are different in that the requirements of that one instrument can be like a huge orchestra polyphony wise. The software is not designed to take a flute or a grand piano and divide the samples over more than one disk. This is why a raid 0 partition for max polyphony makes sense. A better way though would be finding a way to divide the piano samples between the disks without using raid 0. This is a matter of creating multiple incomplete pianos and loading them and driving them... which is what I working on doing right now. But I am not sure its worth the work and if it is too much of a pain then I will not do it. I am using GVI. This is what I will use for this experiment.
 

joojoo1234

Distinguished
Jun 28, 2009
84
0
18,640



The reason I am doing this is I have specific needs out of the piano that normal pop musicians don't face.... alot of sustain.
 

joojoo1234

Distinguished
Jun 28, 2009
84
0
18,640
Well anyone who sees this thread who is interested in my raid results.... 3 dynamic disks are better than 4. With four my seek times went from hard drive and sandra spec of 8.9 or so to 16 ms seek time. With three drives my random access time stayed the same as a single drive or even improved a little bit. Plus my sandra file read for sequential and random is close to the same for sequential and even improved for the random... and by a good margin. 3 320gb blue caviar drives for 49.00 dollars in dynamic raid 0 with cluster 64 seems in my little study to be what I am looking for.... lets see if my polyphony is improved due to those seek times improving by such a wide margin.
 

joojoo1234

Distinguished
Jun 28, 2009
84
0
18,640
SiSoftware Sandra for 3 caviar blues 320gb raid 0 dynamic

Benchmark Results
Drive Index : 230.45MB/s
Results Interpretation : Higher index values are better.
Random Access Time : 7.7ms
Results Interpretation : Lower index values are better.

Detailed Benchmark Results
Buffered Read : 455.89MB/s
Sequential Read : 319MB/s
Random Read : 92.41MB/s
Buffered Write : 308.75MB/s
Sequential Write : 309.79MB/s
Random Write : 132.37MB/s
Random Access Time : 7.69ms

Performance Test Status
Result ID
Platform Compliance : x86
System Timer : 3.58MHz
Operating System Disk Cache Used : No
Use Overlapped I/O : Yes
I/O Queue Depth : 4 request(s)
Test File Size : 3.44GB
File Fragments : 3
Block Size : 1MB

Drive
Drive Type : Hard Disk
Total Size : 5.86GB
Free Space : 5.83GB, 99%
Cluster Size : 64kB

Performance Tips
Notice 5008 : To change benchmarks, click Options.
Notice 5004 : Synthetic benchmark. May not tally with 'real-life' performance.
Notice 5006 : Only compare the results with ones obtained using the same version!
Notice 5209 : Consider using the Removable Storage/Flash Benchmark for Flash devices.
Tip 5202 : Use cache on to measure Windows performance.
Tip 2 : Double-click tip or press Enter while a tip is selected for more information about the tip.





__________________________




SiSoftware Sandra 4 caviar blues 320gb dynamic

Benchmark Results
Drive Index : 219MB/s
Results Interpretation : Higher index values are better.
Random Access Time : 13.66ms
Results Interpretation : Lower index values are better.

Detailed Benchmark Results
Buffered Read : 378.52MB/s
Sequential Read : 330.14MB/s
Random Read : 60MB/s
Buffered Write : 135.13MB/s
Sequential Write : 257MB/s
Random Write : 132.22MB/s
Random Access Time : 13.66ms

Performance Test Status
Result ID
Platform Compliance : x86
System Timer : 3.58MHz
Operating System Disk Cache Used : No
Use Overlapped I/O : Yes
I/O Queue Depth : 4 request(s)
Test File Size : 3.44GB
File Fragments : 2
Block Size : 1MB

Drive
Drive Type : Hard Disk
Total Size : 7.81GB
Free Space : 7.77GB, 99%
Cluster Size : 64kB

Performance Tips
Notice 5008 : To change benchmarks, click Options.
Notice 5004 : Synthetic benchmark. May not tally with 'real-life' performance.
Notice 5006 : Only compare the results with ones obtained using the same version!
Notice 5209 : Consider using the Removable Storage/Flash Benchmark for Flash devices.
Tip 5202 : Use cache on to measure Windows performance.
Tip 2 : Double-click tip or press Enter while a tip is selected for more information about the tip.
 

KyleSTL

Distinguished
Aug 17, 2007
1,678
0
19,790
So what is the advantage then to putting each section of the orchestra on a different logical partition of the same RAID array? I could understand totally independent disks, but I don't understand the reasoning for logical data separation.
 

joojoo1234

Distinguished
Jun 28, 2009
84
0
18,640
I should be clear. If its a raid 0 then it is piano and in one partition. If not a raid - independent disks - which is what I would do for orchestra section samples. As an update I am getting 600 notes of polyphony using two raids at once - one for my piano samples and one for my new realization - a large swap file - 18GB. Normally you would never want a large swap. This is a unique situation of having hundreds of 24bit 44khz audio files filling up your remaining memory. This swap file when it is raid 0 is very effective at doing its duty of virtual ram. The swap is a raid 0 of two drives to be specific and then I am also having a second non-raided drive (main drive) as a swap. Obviously since it is not a dynamic disk I cannot include it in a raid. I have achieved my goal and this works beautifully. 3 drive raid 0 for delivering and a 2 raid 0 drive for swap - plus a second 3gb swap on C:/ for a total swap size of 21gb. I have 5 caviars and a Samsung f1 in this case so it gives me good flexibility. WORKS AWESOME. Again I want to remind anyone doing this that they need to go through the xp tweaking steps for audio setup that are on the internet. Turn off indexing of drives and focus on background services are two crucial steps. With that swap in place my polyphony and the throughput on my computer is like a 64bit computing environment in a way. My 32bit ram limit is busted. Obviously programs aren't in swap so that is not a perfect analogy but it gives me a hell of a lot of streaming data room. Obviously the swap file is placed on quiet disks plus C: which is relatively unbusy.
 

joojoo1234

Distinguished
Jun 28, 2009
84
0
18,640
Also the swap raid is done with 64gb cluster dynamic disks raid 0. Now my decision is how to backup. I don't know about mirroring at all... can you mirror a raid 0 on dynamic disks? Anyway.. The data is all replaceable on these raids. But for speed of recovery I am going to look into how to ghost this... in some way. You might want to do a scheduled defrag on the swap file... I am assuming.?