NF7-S SATA RAID-0, WD-SE: Anyone wanna compare disk benchmarks?

[Wayne Youngman]

Greetings Abit and Storage Newsgroup,

I have just had a bit of spare time to finally do some proper tests on my
SATA RAID-0 set-up. I was hoping that some of you out there would be able
to run some tests yourself so we can compare results. Doesn't matter what
disk system you have (ATA/66 or SCSI etc), I just want some real world
results. . .

I have three results posted using three different programs:

DiskSpeed32 (v3.0.0.4)
http://www.waynes.spamtrap.btinternet.co.uk/Disktest/DiskSpeed32.jpg

HD Tach v2.7
http://www.waynes.spamtrap.btinternet.co.uk/Disktest/HDTach_v2_7.jpg

SiSoftware Sandra MAX
http://www.waynes.spamtrap.btinternet.co.uk/Disktest/Sandra_Max3.jpg


I would be interested to see some other results, here is my old system:
http://www.waynes.spamtrap.btinternet.co.uk/Disktest/DiskSpeed32_ATA66.jpg
http://www.waynes.spamtrap.btinternet.co.uk/Disktest/HDTach_v2_7_ATA66.jpg

--
Wayne ][

Barton (AQXEA) XP2500+ @ 2.2GHz (10x220) - 1.775vCore
CoolerMaster Aero 7 Lite - 3,200rpm
ABIT NF7-S (v2.0 - BIOS#14)
512MB Dual TwiSTER PC3500 @ DDR440 1:1 (9,3,3,2.0 - 2.8v)
Sapphire Atlantis 9800 - 3.3ns Samsung (325/290 Default)
240GB (2x120GB) WD-SE SATA RAID-0 (16k Stripe)
Antec SX630II Mini-Tower Case Inc 300w PSU
2 x CoolerMaster 80mm Blue Neon Fans
WinXP-PRO-SP1
Cat 3.7 - DX9.0b

 

[Chris]

Ran a quick test in Sandra2004 and got the following:
Drive index 52981 kB/s (2x SATA 200Gb 7.2krpm 8Mb cache in Raid 0) - I
didn't bother shutting down any applications/background processes/etc.
Buffered/Sequential/Random read: 146/80/8 Mb/s
Buffered/Sequential/Random write: 135/79/17 Mb/s
Avg. access time 7 ms
For some reason, my test file size is 1023Mb, while yours is 511Mb?
Anyways, not sure how the drive index is calculated from the individual
numbers, but it almost has to weigh sequential reads very heavily since your
index is higher but the only individual number that's higher in your test is
the sequential read number.

I noticed in your HD Tach result that your CPU% is very high, ideally that
should be less than 5%. My first guess would be that you have a very small
stripe size - too small a stripe size really overtaxes your CPU.

 

[Wayne Youngman]

Ran a quick test in Sandra2004 and got the following:
Drive index 52981 kB/s (2x SATA 200Gb 7.2krpm 8Mb cache in Raid 0) - I
didn't bother shutting down any applications/background processes/etc.
Buffered/Sequential/Random read: 146/80/8 Mb/s
Buffered/Sequential/Random write: 135/79/17 Mb/s
Avg. access time 7 ms
For some reason, my test file size is 1023Mb, while yours is 511Mb?
Anyways, not sure how the drive index is calculated from the individual
numbers, but it almost has to weigh sequential reads very heavily since your
index is higher but the only individual number that's higher in your test is
the sequential read number.

I noticed in your HD Tach result that your CPU% is very high, ideally that
should be less than 5%. My first guess would be that you have a very small
stripe size - too small a stripe size really overtaxes your CPU.

Hi Chris,

thanks for the post, quite interesting. You got 146MB/s Buffered Read! I
figure you are using the INTEL ICH5 Native RAID controllers?
--
Wayne ][

Barton (AQXEA) XP2500+ @ 2.2GHz (10x220) - 1.775vCore
CoolerMaster Aero 7 Lite - 3,200rpm
ABIT NF7-S (v2.0 - BIOS#14)
512MB Dual TwiSTER PC3500 @ DDR440 1:1 (9,3,3,2.0 - 2.7v)
Sapphire Atlantis 9800 - 3.3ns Samsung (325/290 Default)
240GB (2x120GB) WD-SE SATA RAID-0 (16k Stripe)
Antec SX630II Mini-Tower Case Inc 300w PSU
2 x CoolerMaster 80mm Blue Neon Fans
WinXP-PRO-SP1
Cat 3.7 - DX9.0b

 

[Chip]

Greetings Abit and Storage Newsgroup,

I have just had a bit of spare time to finally do some proper tests on my
SATA RAID-0 set-up. I was hoping that some of you out there would be able
to run some tests yourself so we can compare results. Doesn't matter what
disk system you have (ATA/66 or SCSI etc), I just want some real world
results. . .

Do me a favour, will you? Download and run the ATTO disk benchmark,
"bench32.exe". Set the "Total Length" to 32 MB and then run it. This is
perhaps the most widely used disk benchmark these days.

If you haven't got it already, you can download it here (as part of ATTO's
SCSI toolset package):

http://www.attotech.com/software/files/eptscsi.exe

I get writes of around 107~108MB/s and reads around 103.

I just ran Sandra for you. Here are the results:

Overall score : 64939kb/s
Buffered Read : 106 MB/s
Sequential Read : 99 MB/s
Random Read : 8 MB/s
Buffered Write : 103 MB/s
Sequential Write : 96 MB/s
Random Write : 21 MB/s
Average Access Time : 7 ms (estimated)

This is with 2 x Maxtor 120MB sata's also in an nf7-s.

Cheers,

Chip.

 

[Chip]

test


Hi Chris,

thanks for the post, quite interesting. You got 146MB/s Buffered Read! I
figure you are using the INTEL ICH5 Native RAID controllers?
--

You see, this is exactly my point. Although the Intel controller allows
higher burst (i.e. buffered) performance (since it isn't limited by the PCI
bus' 133MB/s bandwidth), this makes very little difference to the real-world
peformance of the disk subsystem, since it isn't *in* burst mode for very
long or very often. What really matters are the sequential and random
numbers.

Its for exactly the same reason that the performance of ATA100 interface
disks (e.g. Western Digital) is the same as for ATA133 disk (Maxtor)
although the interface allows higher burst rates, it makes little or no
difference to the actuall performance of the drive.

Chip.

 

[Folkert Rienstra]

You see, this is exactly my point. Although the Intel controller allows
higher burst (i.e. buffered) performance (since it isn't limited by the PCI
bus' 133MB/s bandwidth),

this makes very little difference to the real-world peformance of the disk
subsystem,

In the case of SATA - which is point to point - but not with ATA's 2 drive
per channel.

since it isn't *in* burst mode for very long or very often.

It is in burst mode ALL of the time as that is the only way (most) buses
work, at 33/66/100/133/150 MB/s, in bursts, since obviously no device
(with those particular interfaces) can sustain those (particular) rates.

What really matters are the sequential and random numbers.

Actually the burstrates would be an indication of what might happen
when two drives are accessed simultaniously on the same channel.

Its for exactly the same reason that the performance of ATA100 interface
disks (e.g. Western Digital) is the same as for ATA133 disk (Maxtor)

Not with 2 drives on the same channel accessed simultaniously when both
drives have STRs of over 45MB/s.

although the interface allows higher burst rates, it makes little or no
difference to the actuall performance of the drive.

Drive indeed, as in 'single' drive.

 

[tHatDudeUK]

You see, this is exactly my point. Although the Intel controller allows
higher burst (i.e. buffered) performance (since it isn't limited by the PCI
bus' 133MB/s bandwidth), this makes very little difference to the real-world
peformance of the disk subsystem, since it isn't *in* burst mode for very
long or very often. What really matters are the sequential and random
numbers.

Unless of course you use hardware which also wants to use a bit of PCI
bandwidth.

 

[Chip]

Unless of course you use hardware which also wants to use a bit of PCI
bandwidth.

That a completely separate issue. What I meant above is "What really
matters (as far as hard disk performance goes) are the sequential and random
numbers."

How the disk is attached to the system, through what controller and on which
bus etc its a separate issue that will determine other system performance
aspects..... like whether the sound works or not.

Chip.

 

[Folkert Rienstra]

That a completely separate issue. What I meant above is "What really matters
(as far as hard disk performance goes) are the sequential and random numbers."

How the disk is attached to the system, through what controller and on which bus
etc its a separate issue that will determine other system performance aspects.....

like whether the sound works or not.

.... or the drives perform or not.

 

[Colon Terminus]

I, too, just ran the test.
Drive index 57238 (2x ATA100 120GB 7.2krpm RAID 0)
Ran from fresh boot with all normal processes running.
Buffered/Sequential/Random read: 94/91/9 Mb/s
Buffered/Sequential/Random write: 74/71/15 Mb/s
Avg. access time 7 ms

I'd have to agree with Chris that sequential reads weigh VERY heavily in the
index.

Ran a quick test in Sandra2004 and got the following:
Drive index 52981 kB/s (2x SATA 200Gb 7.2krpm 8Mb cache in Raid 0) - I
didn't bother shutting down any applications/background processes/etc.
Buffered/Sequential/Random read: 146/80/8 Mb/s
Buffered/Sequential/Random write: 135/79/17 Mb/s
Avg. access time 7 ms
For some reason, my test file size is 1023Mb, while yours is 511Mb?
Anyways, not sure how the drive index is calculated from the individual
numbers, but it almost has to weigh sequential reads very heavily since your
index is higher but the only individual number that's higher in your test is
the sequential read number.

I noticed in your HD Tach result that your CPU% is very high, ideally that
should be less than 5%. My first guess would be that you have a very small
stripe size - too small a stripe size really overtaxes your CPU.

Greetings Abit and Storage Newsgroup,

I have just had a bit of spare time to finally do some proper tests on my
SATA RAID-0 set-up. I was hoping that some of you out there would be able
to run some tests yourself so we can compare results. Doesn't matter what
disk system you have (ATA/66 or SCSI etc), I just want some real world
results. . .

I have three results posted using three different programs:

DiskSpeed32 (v3.0.0.4)
http://www.waynes.spamtrap.btinternet.co.uk/Disktest/DiskSpeed32.jpg

HD Tach v2.7
http://www.waynes.spamtrap.btinternet.co.uk/Disktest/HDTach_v2_7.jpg

SiSoftware Sandra MAX
http://www.waynes.spamtrap.btinternet.co.uk/Disktest/Sandra_Max3.jpg


I would be interested to see some other results, here is my old system:
http://www.waynes.spamtrap.btinternet.co.uk/Disktest/DiskSpeed32_ATA66.jpg
http://www.waynes.spamtrap.btinternet.co.uk/Disktest/HDTach_v2_7_ATA66.jpg

--
Wayne ][

Barton (AQXEA) XP2500+ @ 2.2GHz (10x220) - 1.775vCore
CoolerMaster Aero 7 Lite - 3,200rpm
ABIT NF7-S (v2.0 - BIOS#14)
512MB Dual TwiSTER PC3500 @ DDR440 1:1 (9,3,3,2.0 - 2.8v)
Sapphire Atlantis 9800 - 3.3ns Samsung (325/290 Default)
240GB (2x120GB) WD-SE SATA RAID-0 (16k Stripe)
Antec SX630II Mini-Tower Case Inc 300w PSU
2 x CoolerMaster 80mm Blue Neon Fans
WinXP-PRO-SP1
Cat 3.7 - DX9.0b

 

[Wayne Youngman]

Do me a favour, will you? Download and run the ATTO disk benchmark,
"bench32.exe". Set the "Total Length" to 32 MB and then run it. This is
perhaps the most widely used disk benchmark these days.

If you haven't got it already, you can download it here (as part of ATTO's
SCSI toolset package):

http://www.attotech.com/software/files/eptscsi.exe

Hi,

I did download that *ATTO ExpressPro-Tools* but I have no idea how to use
it's *Disk Benchmark*?. There is allot of parameters to configure. Any
tips?

 

[Wayne Youngman]

I, too, just ran the test.
Drive index 57238 (2x ATA100 120GB 7.2krpm RAID 0)
Ran from fresh boot with all normal processes running.
Buffered/Sequential/Random read: 94/91/9 Mb/s
Buffered/Sequential/Random write: 74/71/15 Mb/s
Avg. access time 7 ms

I'd have to agree with Chris that sequential reads weigh VERY heavily in the
index.

Hi,

thanks for the feedback. Your still getting better scores than me using
your ATA/100 RAID-0 than I am getting using SATA RAID-0. Did you try using
HDTach v2.7 yet?

http://www.waynes.spamtrap.btinternet.co.uk/Disktest/HDTach_v2_7.jpg

--
Wayne ][

Barton (AQXEA) XP2500+ @ 2.2GHz (10x220) - 1.775vCore
CoolerMaster Aero 7 Lite - 3,200rpm
ABIT NF7-S (v2.0 - BIOS#14)
512MB Dual TwiSTER PC3500 @ DDR440 1:1 (9,3,3,2.0 - 2.8v)
Sapphire Atlantis 9800 - 3.3ns Samsung (325/290 Default)
240GB (2x120GB) WD-SE SATA RAID-0 (16k Stripe)
Antec SX630II Mini-Tower Case Inc 300w PSU
2 x CoolerMaster 80mm Blue Neon Fans
WinXP-PRO-SP1
Cat 3.7 - DX9.0b

 

[Z Man]

Here are my Hdtach 2.70 stats:

Drive #1: Western Digital 200GB, 8MB cache, 7200RPM:

Random Access 13.8ms
Read Burst Speed: 67.1MB/s
Read Speed Average: 47.0MB/s
CPU Utilization: 8.7%

Drive #2: Seagate 18GB ST318452LW, 15000RPM

Random Access 5.9ms
Read Burst Speed: 81.9MB/s
Read Speed Average: 55.3MB/s
CPU Utilization: 5.7%

And, just for fun, Drive #3, Apple Ipod, 30GB, Firewire

Random Access 22.8ms
Read Burst Speed: 19.5MB/s
Read Speed Average: 12.3MB/s
CPU Utilization: 1.9%

I have two 15k drives in my office, one of which is the world's fastest
drive. I'll try to post those results later today or tomorrow.

I, too, just ran the test.
Drive index 57238 (2x ATA100 120GB 7.2krpm RAID 0)
Ran from fresh boot with all normal processes running.
Buffered/Sequential/Random read: 94/91/9 Mb/s
Buffered/Sequential/Random write: 74/71/15 Mb/s
Avg. access time 7 ms

I'd have to agree with Chris that sequential reads weigh VERY heavily in the
index.

Ran a quick test in Sandra2004 and got the following:
Drive index 52981 kB/s (2x SATA 200Gb 7.2krpm 8Mb cache in Raid 0) - I
didn't bother shutting down any applications/background processes/etc.
Buffered/Sequential/Random read: 146/80/8 Mb/s
Buffered/Sequential/Random write: 135/79/17 Mb/s
Avg. access time 7 ms
For some reason, my test file size is 1023Mb, while yours is 511Mb?
Anyways, not sure how the drive index is calculated from the individual
numbers, but it almost has to weigh sequential reads very heavily since your
index is higher but the only individual number that's higher in your

test

is
the sequential read number.

I noticed in your HD Tach result that your CPU% is very high, ideally that
should be less than 5%. My first guess would be that you have a very small
stripe size - too small a stripe size really overtaxes your CPU.

Greetings Abit and Storage Newsgroup,

I have just had a bit of spare time to finally do some proper tests on my
SATA RAID-0 set-up. I was hoping that some of you out there would be able
to run some tests yourself so we can compare results. Doesn't matter what
disk system you have (ATA/66 or SCSI etc), I just want some real world
results. . .

I have three results posted using three different programs:

DiskSpeed32 (v3.0.0.4)
http://www.waynes.spamtrap.btinternet.co.uk/Disktest/DiskSpeed32.jpg

HD Tach v2.7
http://www.waynes.spamtrap.btinternet.co.uk/Disktest/HDTach_v2_7.jpg

SiSoftware Sandra MAX
http://www.waynes.spamtrap.btinternet.co.uk/Disktest/Sandra_Max3.jpg


I would be interested to see some other results, here is my old system:
http://www.waynes.spamtrap.btinternet.co.uk/Disktest/DiskSpeed32_ATA66.jpg
http://www.waynes.spamtrap.btinternet.co.uk/Disktest/HDTach_v2_7_ATA66.jpg

--
Wayne ][

Barton (AQXEA) XP2500+ @ 2.2GHz (10x220) - 1.775vCore
CoolerMaster Aero 7 Lite - 3,200rpm
ABIT NF7-S (v2.0 - BIOS#14)
512MB Dual TwiSTER PC3500 @ DDR440 1:1 (9,3,3,2.0 - 2.8v)
Sapphire Atlantis 9800 - 3.3ns Samsung (325/290 Default)
240GB (2x120GB) WD-SE SATA RAID-0 (16k Stripe)
Antec SX630II Mini-Tower Case Inc 300w PSU
2 x CoolerMaster 80mm Blue Neon Fans
WinXP-PRO-SP1
Cat 3.7 - DX9.0b

 

[Chip]

Hi,

I did download that *ATTO ExpressPro-Tools* but I have no idea how to use
it's *Disk Benchmark*?. There is allot of parameters to configure. Any
tips?

Sure. Leave everything alone apart from "Total Length", which you typically
set to 32MB. Then hit the start button. Takes perhaps 1 minute to run.

Thanks,

Chip.

 

[Jon]

HD Tach version 2.70


Drive: PhysicalDrive1 74.3gb
Access time: 7.6ms
CPU utilization: 3.4%
64 zones to be tested (1123120kb zones).
Burst speed: 118.2MB/s
Average read speed: 65