exstand/yano, freewing, pccommcard transferspeed ?

[seal1]

hello,

how fast are your pccomms, freewing´s or extand´s interfaces out there ?

lastweekend I build a yano/extand-parallel-port-2-replay-interface.

http://forums.segaxtreme.net/index.php?showtopic=15686

using his vxd under w98se. http://www1.odn.ne.jp/psx-alternative/laboratory_01.html

his vxd-driver routes from virtual pccomm-isa-port to LPT, so any soft using pccomm-card should work with yano/extand.

how fast should this be ?

(from my amiga-pc-laplink-experiences I was able to do 120 kb/s over LPT-port.)

using charles mcdonald´s sat.exe:

my yano-interface is only 35 kb/s fast vs. original PCCOMMS-ISA-LINK-CARD 100 kb/s.

(I stopped per hand the transfer of 500kb file, all my kb/s-values rounded.)

(tested also with a psx and caetla´s psexe.exe, which has a native extand-mode, so I can test pccomm-isa 100 kb/s vs. native extand 50 kb/s vs. virtual-pccom-isa-port 35 kb/s.)

other problem - power voltage for interface:

with original datel AR transfer of big files is unstable at 3v, 4.5v, 6v or 7.5v

with EMS 4in1 it works stable from 3v-6volt (7.5v not tested)

(gamebuster on psx wants atleast 6volt to work)

I used HCT coz my local dealer had no HC chips.

the AR has for the 74__245 a HC the EMS a ALS, so this is maybe the problem, if LOW-output from the HC (inside the AR) is over 0.8volt it is in undefined range for the HCT (used on my yano-interface).

regards,

seal1

 

[antime]

The speeds you're getting sound reasonable. The Commlink transfers one byte at a time, the Freewing and clones split that up into two nybble transfers, and using the virtualization driver you're taking an exception for each read/write which adds a substantial overhead.

 

[seal1]

Originally posted by antime@Tue, 2005-05-24 @ 05:23 AM

The speeds you're getting sound reasonable. The Commlink transfers one byte at a time, the Freewing and clones split that up into two nybble transfers, and using the virtualization driver you're taking an exception for each read/write which adds a substantial overhead.

[post=134440]Quoted post[/post]​

I have no freewing ;-)

yano/exstands != freewing,

yano transfers full byte

freewing two nibbles, so freewing should half of speed of yano or freewing 4x slower than pccommslink

ISA-bus can do 1-2 MB/s? vs. old LPT only 150 kb/s -> pccommslink has better basis.

but LPT-timing depends most times on ISA-bus and some modern modes of LPT can do 500 kb/s and more.

but old EPP-mode not ? for use of yano´s interface port must be EPP not EPP+ECP, not ECP, not NORMAL.

I coded a native-yano-mode and benchmark-support for sat.exe and got of course better speeds around:

66 kb/s vs. yano´s vxd 36 kb/s vs. 114 kb/s pccommlink

and no speeddifferences between w98se and winxp even many sources say LPT under winxp ist slower !

66kb/s is enough for me now, but without the "send byte, read byte back for check"-behavior over LPT should 120-130 kb/s be possible.

 

[antime]

Originally posted by seal1@Mon, 2005-06-06 @ 07:54 PM

yano transfers full byte

freewing two nibbles, so freewing should half of speed of yano or freewing 4x slower than pccommslink

Looking at the VxD sources you have to do exactly the same number of port accesses for FreeWing and ExStand, so I would guess their performance are very close.

66kb/s is enough for me now, but without the "send byte, read byte back for check"-behavior over LPT should 120-130 kb/s be possible.

Probably. It might also be possible to improve speeds by using a high-performance PCI parallel port card, to get rid of the ISA bottleneck.

 

[seal1]

Originally posted by antime@Wed, 2005-06-08 @ 07:07 AM

Looking at the VxD sources you have to do exactly the same number of port accesses for FreeWing and ExStand, so I would guess their performance are very close.

oh yes, both routines do 5 port-accesses, never counted ... thought freewing takes 2 read-accesses for 1 byte so it must take longer

Probably. It might also be possible to improve speeds by using a high-performance PCI parallel port card, to get rid of the ISA bottleneck.

[post=135074]Quoted post[/post]​

via ISA you can do atleast 1-2 MB/s.

bottleneck is the old-styles LPT-port in EPP mode.

u got experience with highspeed-pci-lpt-port-cards ?

 

[antime]

Your current biggest limiting factor is that you have to do all signalling manually. The speed advantage from EPP/ECP comes from the hardware's ability to handle that for you, but the ExStand isn't an EPP device. From what I can tell, the EPP requirement comes from needing a standard programming interface which was lacking with earlier byte-wide parallel port implementations. To reach ISA bus limits I would assume you would have to do nothing but continuous 16-bit DMA transfers, but that's not really relevant here.

I've never used any parallel port cards, but I do remember seeing adverts for some that could support a large number of ports at high speeds. Maybe those speeds are only achievable with ECP hardware, using on-board DMA buffers.