FastATA - How fast can you go?
- Thread starter fitzsteve
- Start date
- Replies 49
- Views 8297
Being a relative noob to all this, is there a handy guide such as Andy's one for accelerators which details the various steps to speeding up your miggy?
I have a CF card fitted to one of kippers external adapters, so its easy to backup, but looking at all your charts, my speeds are awful. Given I have a 68060@80 I might be interested in what I can do to further enhance it.
I have a CF card fitted to one of kippers external adapters, so its easy to backup, but looking at all your charts, my speeds are awful. Given I have a 68060@80 I might be interested in what I can do to further enhance it.
Justin
Active member
Being a relative noob to all this, is there a handy guide such as Andy's one for accelerators which details the various steps to speeding up your miggy?
I have a CF card fitted to one of kippers external adapters, so its easy to backup, but looking at all your charts, my speeds are awful. Given I have a 68060@80 I might be interested in what I can do to further enhance it.
what speeds do you have? and which version of FastATA are you using?
No Fast ATA (yet) - as am relatively clueless hence my question regarding what the options are and how they all rate.
Hi Mate,
I have an IDEFix Express, this boosts my IDE to about 3.5mb/sec and my FastATA about 6.2mb/sec, incidentally I've decided to sell my FastATA & SSD as I'm a little short on cash!
https://www.amibay.com/showthread.php?p=534127#post534127
I have an IDEFix Express, this boosts my IDE to about 3.5mb/sec and my FastATA about 6.2mb/sec, incidentally I've decided to sell my FastATA & SSD as I'm a little short on cash!
https://www.amibay.com/showthread.php?p=534127#post534127
Hmm - none of these will play nicely with my external CF card adapter as far as I can see, and I haven't had to open the case for quite a while which is nice, so I am quite attached to that feature.
I suppose I could try and play around with an alternative way of having the cf card sticking out somewhere.
does the IDEFix interfere with an Indy MK2
I suppose I could try and play around with an alternative way of having the cf card sticking out somewhere.
does the IDEFix interfere with an Indy MK2
i have simillar speed with Cyberstom mk2 060 50mhz
A good point was raised in this thread about Transfer Speeds and CPU Load.
IDE, by its definition is a Programmed Input / Output (PIO) is very CPU intensive.
Using the Native IDE and GVP's Fast SCSI 2 (DMA) as an example
In the above test, we are using a stock 7.4MHz 68000, this achives 1.67MB persecond with just over 7% CPU load to achieve this.
CPU : 68000 @ 7.4MHz
RAM : 8MB of FAST RAM
DMA : YES
FFS : Standard FFS Format
KS? : Kickstart 3.1
OSx : Classic Workbench
This is an Apollo 030 Mk3 with 64MB of RAM running on an A1200, as you can see it manages about 1.64MB per second, only slightly slower than the GVP - however as you can see it has a CPU load of over 99% so the CPU is spending all of its time transfering data as opposed to processing it.
CPU : 68030 @ 40MHz
RAM : 64MB of FAST RAM
DMA : NO
SFS : Smart File Format
KS? : Kickstart 3.1
OSx : Classic Workbench
__________________________________
If we were to work out the processing load one could look at a simple method as this
68000@7.4MHz = 0.9 MIPS - achieves - 1.67MB Per Second Transfer @ 7% Load
Equates to [ 0.063 MIPS per 1.67 MB Per Second ]
68030@40MHz = 8.5 MIPS - achieves - 1.65MB Per Second Transfer @99% Load
Equates to [ 8.42 MIPS per 1.4 MB Per Second ]
to put this into contrast my 060@50Mhz (64MB RAM) provided the following
68060@50Mhz = 65.2 MIPS - Achieves - 1.87MB Per Second Transfer @ 45.6% Load
Equates to [ 29.34 MIPS per 1.8 MB Per Second ]
As you can see a non-managed / off loaded solution is quite detrimental to the workings of the CPU as it spends more time (processing power) in fetching data as opposed to processing it.
In the case of the 060 above we can see an extreme case where the CPU has work increasingly harder to maintain throughput of data. Arguably the result shows that I am using the processing power of an 040@40Mhz to achieve a piddly 1.8MB Per Second Transfer
On the flip of this
Cyberstorm Mk2 060@50Mhz + 128MB of RAM + DMA SCSI
68060@50Mhz = 65.2 MIPS - Achieves - 7.4MB Per Second Transfer @ 28.3% Load
Equates to [ 18.45 MIPS per 7.4 MB Per Second ]
Interestingly, while we have a much faster transfer Rate I am still using the processing power of an 040@25Mhz to achieve this result.
So in a nut shell, there;s more to transferring data, it is how it is transferred.
(sorry for the lack of pictures on the last to two - sadly I lost them =( )
IDE, by its definition is a Programmed Input / Output (PIO) is very CPU intensive.
Using the Native IDE and GVP's Fast SCSI 2 (DMA) as an example
In the above test, we are using a stock 7.4MHz 68000, this achives 1.67MB persecond with just over 7% CPU load to achieve this.
CPU : 68000 @ 7.4MHz
RAM : 8MB of FAST RAM
DMA : YES
FFS : Standard FFS Format
KS? : Kickstart 3.1
OSx : Classic Workbench
This is an Apollo 030 Mk3 with 64MB of RAM running on an A1200, as you can see it manages about 1.64MB per second, only slightly slower than the GVP - however as you can see it has a CPU load of over 99% so the CPU is spending all of its time transfering data as opposed to processing it.
CPU : 68030 @ 40MHz
RAM : 64MB of FAST RAM
DMA : NO
SFS : Smart File Format
KS? : Kickstart 3.1
OSx : Classic Workbench
__________________________________
If we were to work out the processing load one could look at a simple method as this
68000@7.4MHz = 0.9 MIPS - achieves - 1.67MB Per Second Transfer @ 7% Load
Equates to [ 0.063 MIPS per 1.67 MB Per Second ]
68030@40MHz = 8.5 MIPS - achieves - 1.65MB Per Second Transfer @99% Load
Equates to [ 8.42 MIPS per 1.4 MB Per Second ]
to put this into contrast my 060@50Mhz (64MB RAM) provided the following
68060@50Mhz = 65.2 MIPS - Achieves - 1.87MB Per Second Transfer @ 45.6% Load
Equates to [ 29.34 MIPS per 1.8 MB Per Second ]
As you can see a non-managed / off loaded solution is quite detrimental to the workings of the CPU as it spends more time (processing power) in fetching data as opposed to processing it.
In the case of the 060 above we can see an extreme case where the CPU has work increasingly harder to maintain throughput of data. Arguably the result shows that I am using the processing power of an 040@40Mhz to achieve a piddly 1.8MB Per Second Transfer
On the flip of this
Cyberstorm Mk2 060@50Mhz + 128MB of RAM + DMA SCSI
68060@50Mhz = 65.2 MIPS - Achieves - 7.4MB Per Second Transfer @ 28.3% Load
Equates to [ 18.45 MIPS per 7.4 MB Per Second ]
Interestingly, while we have a much faster transfer Rate I am still using the processing power of an 040@25Mhz to achieve this result.
So in a nut shell, there;s more to transferring data, it is how it is transferred.
(sorry for the lack of pictures on the last to two - sadly I lost them =( )
hooverphonique
Member
File systems shouldn't matter in this sort of test as you need to use software that just reads blocks/sectors directly using the device driver.. If you use software that uses a file to read/write through, you're getting the wrong benchmark..
software like sysinfo will be using the device driver directly..
software like sysinfo will be using the device driver directly..