Home   News   Concept   AMIGA-Compatible   Hardware   Forum   Questions+Answers   Pictures   Contact & Team

Welcome to the Natami / Amiga Forum

This forum is for AMIGA fans interested in the new NATAMI platform.
Please read the forum usage manual.



All TopicsNewsQAFeaturesTalkTEAMLogin to post    Create account
Do you have questions about the Natami?
Post it here and we will answer it!

Disk IO Performancepage  1 2 3 
Gunnar von Boehn
Germany
(Moderator)
Posts 5775
29 Apr 2011 11:45


We are going to beef up the disk IO speed.
What Amiga tool/benchmark would you recommend to measure disk speed?



Sergio Gabbiani
Italy

Posts 18
29 Apr 2011 12:19


If i remember well, sysInfo has some sort of speed test for devices...

I don't know the accuracy...

Matt Hey
USA

Posts 770
29 Apr 2011 13:58


Gunnar von Boehn wrote:

  What Amiga tool/benchmark would you recommend to measure disk speed?

SysSpeed...

EXTERNAL LINK 

Marcel Verdaasdonk
Netherlands

Posts 4043
29 Apr 2011 18:43


Gunnar bet you won't get it upto 133MB/s without using a disk packer. :p

Matt Hey
USA

Posts 770
29 Apr 2011 19:09


Marcel Verdaasdonk wrote:

Gunnar bet you won't get it upto 133MB/s without using a disk packer. :p

I get 20MB/s sustained (using SysSpeed) on my CSMK3 with 10k hard drive, PFS and using DMA and it's flying on an Amiga. Most hard drive accesses are instantaneous. The CPU becomes the bottleneck and the hard drive speed isn't an issue at a certain point. I expect 133 MB/s is way past that point on the Amiga :D.


Marcel Verdaasdonk
Netherlands

Posts 4043
29 Apr 2011 19:49


Matt don't inform Gunnar i was trying to Enrage him so we would have a faster then Thomas's PCI interface can handle HD interface seems moot but if the end result is worth it it's worth it. ;)

Darren Eveland
Canada
(Natami Team)
Posts 50
29 Apr 2011 20:08


You should use scsispeed for the most accurate result as it talks directly to the drive instead of going through the filesystem.

Christian Kummerow
Germany

Posts 316
30 Apr 2011 08:17


Darren Eveland wrote:

You should use scsispeed for the most accurate result as it talks directly to the drive instead of going through the filesystem.

But the end-user use the filesystem usual.

Matt Hey wrote:

  I get 20MB/s sustained (using SysSpeed) on my CSMK3 with 10k hard drive, PFS and using DMA and it's flying on an Amiga. Most hard drive accesses are instantaneous.

How old is your Harddrive, you should exchange it.
MK3 should have CSPPC SCSI Performance and it can reach much more
than 20MB/s. Thats a SCSI-3 controller.

Matt Hey
USA

Posts 770
30 Apr 2011 13:14


Christian Kummerow wrote:

  How old is your Harddrive, you should exchange it.
  MK3 should have CSPPC SCSI Performance and it can reach much more
  than 20MB/s. Thats a SCSI-3 controller.

You are right. I rechecked the speed and it's 30MB/s sustained. Here is typical SysSpeed results...

Create 1262 Op/s
Open 3056 Op/s
DirScan 16516 Op/s
Delete 1774 Op/s
Seek/Read 7770 Op/s

CreateFile 29.71 MB/s
WriteFile 32.10 MB/s
ReadFile 30.36 MB/s
RawRead 30.61 MB/s

The hard drive is a newer Maxtor Atlas 15k 16GB Ultra SCSI drive I bought new for $10 U.S. The smaller capacity Ultra SCSI hard drives can be found cheap. I hope I got all the information right this time :/.


P. Love
USA

Posts 47
04 May 2011 02:43


"beef up the disk IO speed"

I was kinda hoping for UltraDMA Mode 7.  :):)

CompactFlash Revision 6.0... UltraDMA Mode 7 (167MB/s)

EXTERNAL LINK 


SID Hervé
France

Posts 668
04 May 2011 20:10


Gunnar von Boehn wrote:

We are going to beef up the disk IO speed.
What Amiga tool/benchmark would you recommend to measure disk speed?

Hello

Long time ago, I tried scsibench and diskspeed, one of the two gave its source code. Due to the ambiguity of the Amiga SCSI, I still do not know if they took care of IDE and I did not have them tried on the Buddha card.

Anyway, maybe this will be of any help...

PS: Yes I know I'm late again...

Marcel Verdaasdonk
Netherlands

Posts 4043
05 May 2011 08:44


Phil, SID the little big Endian problem arise when reading or writing more then a Word to the HDD.
This is because the data bus directly connects to the IDE when using a GAYLE.
Having a 32 bit bus itself doesn't fix this problem because the interface width of the IDE being 16 bits.
This is one of the biggest reasons i was and at some points still am against the HDD controller on the PCI bus.

Boy am i wandering off topic to wrap this up, I do not get how this will not effect maximum performance.

Jakob Eriksson
Sweden
(Moderator)
Posts 1097
05 May 2011 12:23


Marcel, I don't get what you mean.  Will endianness issues make performance lower?

Megol .

Posts 695
05 May 2011 18:01


Jakob Eriksson wrote:

Marcel, I don't get what you mean.  Will endianness issues make performance lower?

No. The only "problem" is that the commands are little endian.

SID Hervé
France

Posts 668
05 May 2011 20:35


Hello
   
  I've never been interested in the IDE standard. I had and I still have a preference for the SCSI standard which I think is a better solution for the Amiga (it is just a user point of view), especially with a Blizzard 2060. Regarding the financial aspect, one might think that the pig is fed with jam.
   
  About NatAmi and IDE standard, it is not stupid to think that the team will not shoot itself in the foot. The IDE standard has already proven itself to other platforms.
   
  The endian problem has already been debated in the forum. And as far as I can remember, it's a minor problem that is easily bypassed.

Marcel Verdaasdonk
Netherlands

Posts 4043
05 May 2011 21:58


Megol . wrote:

 
Jakob Eriksson wrote:

  Marcel, I don't get what you mean.  Will endianness issues make performance lower?
 

  No. The only "problem" is that the commands are little endian.
 

 
  That isn't even a problem.
  What i am getting at is that ATA is a 16 bit bus and Natami uses 32 bit buses. ;)
  Little endian extension or better said little endian order of output is something that messes with through put on the classic Amiga's.(that is why you can only do word access, and not long word access to harddisk.)
SID is both right and wrong about it being solved, since there is more then one solution to the problem.
One is only doing word sized access this would omit the problem.
The other is doing a a hi lo word order swap in software.(solution in driver for most)
Or doing the swap in hardware, doing it in hardware to me strikes me as being done in the controller and strikes me as misplacement of the solution when it comes to bus utilization.

Gunnar von Boehn
Germany
(Moderator)
Posts 5775
06 May 2011 06:42


Its funny to read these posts.

Why should there be any problem with IDE transfers?

There is no problem at all.
Look at the A4000 and A1200 both have the IDE controller just like the NATAMI, (Ok NATAMI has not one but two controllers and NATAMI can some features more).

But that IDE worked fine with A4000 and A1200 should prove to everyone that there never was a problem.
This should be clear even to people with no knowledge about IDE.

This talk about endianes is just funny.

Wojtek P
Poland

Posts 1597
06 May 2011 15:22


Matt Hey wrote:

Christian Kummerow wrote:

  How old is your Harddrive, you should exchange it.
  MK3 should have CSPPC SCSI Performance and it can reach much more
  than 20MB/s. Thats a SCSI-3 controller.
 

 
  You are right. I rechecked the speed and it's 30MB/s sustained. Here is typical SysSpeed results...
 
  Create 1262 Op/s
  Open 3056 Op/s
  DirScan 16516 Op/s
  Delete 1774 Op/s
  Seek/Read 7770 Op/s
 
  CreateFile 29.71 MB/s
  WriteFile 32.10 MB/s
  ReadFile 30.36 MB/s
  RawRead 30.61 MB/s
 
  The hard drive is a newer Maxtor Atlas 15k 16GB Ultra SCSI drive I bought new for $10 U.S. The smaller capacity Ultra SCSI hard drives can be found cheap. I hope I got all the information right this time :/.
 

Your data only confirms that AmigaOS filesystem handling is very efficient - as RawRead is 1% faster only than ReadFile.

Is it on FFS or other filesystem?


Wojtek P
Poland

Posts 1597
06 May 2011 15:24


Gunnar von Boehn wrote:

Its funny to read these posts.
 
  Why should there be any problem with IDE transfers?

  There is no problem at all. The only problem with A1200 is SLOW IDE controller and not 100% compatible (not all disks works).

This is fixed on Natami so... that's all.


Chris Dennett
United Kingdom

Posts 136
06 May 2011 15:31


Yes, you have a better filesystem than FFS now that that PSF3 is open source (see EXTERNAL LINK ) :)
   
    Redistribution and use in source and binary forms, with or without
    modification, are permitted provided that the following conditions are met:
    1. Redistributions of source code must retain the above copyright
      notice, this list of conditions and the following disclaimer.
    2. Redistributions in binary form must reproduce the above copyright
      notice, this list of conditions and the following disclaimer in the
      documentation and/or other materials provided with the distribution.
    3. All advertising materials mentioning features or use of this software
      must display the following acknowledgement:
      This product includes software developed by Michiel Pelt.
    4. Neither the name of the author nor the
      names of its contributors may be used to endorse or promote products
      derived from this software without specific prior written permission.
   
    THIS SOFTWARE IS PROVIDED BY MICHIEL PELT ''AS IS'' AND ANY
    EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
    WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
    DISCLAIMED. IN NO EVENT SHALL MICHIEL PELT BE LIABLE FOR ANY
    DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
    (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
    LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
    ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
    (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
    SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
   
 
  There is also SFS ( EXTERNAL LINK )

posts 53page  1 2 3