-
Updated 2023-07-12: Hello, Guest! Welcome back, and be sure to check out this follow-up post about our outage a week or so ago.
External LocalTalk / serial port clocks
- Thread starter cheesestraws
- Start date
cheesestraws
Well-known member
"how tp melt your TashTalk"
Phipli
Well-known member
Mk.558
Well-known member
Arise ye from the depths of the grave...
In the depths of the Apple Service Source 2.5, lies this page:
From this we can deduce whether this GPi pin is a factor in the speed of the fabled (and often unreliable) Farallon EtherWave Mac/PB or Printer Adapter devices when they push the throughtput higher. Today I wanted to find out whether that was the case. I used a SE/30 and a IIsi: both ran System 6.0.8 on a RAM disk and copied a 3MB (..ish, it's 3.09MB) file over a network to the RAM disk. Looking closely at the IIsi motherboard, it's difficult for me to say where the pins even go on the serial port, so if I wanted to "Add" the GPi pin manually with a wire to the serial chip, well...that's be an exercise.
Below I have a table of the results.
If the Farallon stingray was using the GPi pin for synchronous transfers, it would show in the data. Looking back at my notes, a Quadra 700 and a IIci both did it at about 65KB/s, but they were using System 7 or 7.1. However nobody can call a IIci slow, or a Quadra 700, so they're quite capable for this measurement.
Without the speed bonus, the same file would take about 02:20.xx to 02:25.xx to upload and download, which corresponds to the LocalTalk ceiling of 230400bps, or about 20KiB/s. I wonder how this information can help us clock the serial port faster for stuff like AirTalk and TashTalk. The speed bonus is very much noticeable. The record of 80KiB/sec that a PowerBook 1400cs pulled is probably peak (as if 230400bps = roughly 20KiB/s, then 921600 = roughly 80KiB/s), but it's also a fussy adapter. These and other developments will be included in the next update of the Guide...whenever that happens. Maybe next year. Would be cute to test it on a IIfx and a Quadra 950.
In the depths of the Apple Service Source 2.5, lies this page:
From this we can deduce whether this GPi pin is a factor in the speed of the fabled (and often unreliable) Farallon EtherWave Mac/PB or Printer Adapter devices when they push the throughtput higher. Today I wanted to find out whether that was the case. I used a SE/30 and a IIsi: both ran System 6.0.8 on a RAM disk and copied a 3MB (..ish, it's 3.09MB) file over a network to the RAM disk. Looking closely at the IIsi motherboard, it's difficult for me to say where the pins even go on the serial port, so if I wanted to "Add" the GPi pin manually with a wire to the serial chip, well...that's be an exercise.
Below I have a table of the results.
| Machine | Download time | Upload time | Download KiB/s | Upload KiB/s | Download Kbps | Upload Kbps |
| SE/30 | 0:47.32s | 0:51.36s | 64.9 | 59.8 | 519.35 | 478.50 |
| IIsi | 0:32.36s | 0:36.44s | 94.9 | 84.3 | 759.46 | 674.42 |
If the Farallon stingray was using the GPi pin for synchronous transfers, it would show in the data. Looking back at my notes, a Quadra 700 and a IIci both did it at about 65KB/s, but they were using System 7 or 7.1. However nobody can call a IIci slow, or a Quadra 700, so they're quite capable for this measurement.
Without the speed bonus, the same file would take about 02:20.xx to 02:25.xx to upload and download, which corresponds to the LocalTalk ceiling of 230400bps, or about 20KiB/s. I wonder how this information can help us clock the serial port faster for stuff like AirTalk and TashTalk. The speed bonus is very much noticeable. The record of 80KiB/sec that a PowerBook 1400cs pulled is probably peak (as if 230400bps = roughly 20KiB/s, then 921600 = roughly 80KiB/s), but it's also a fussy adapter. These and other developments will be included in the next update of the Guide...whenever that happens. Maybe next year. Would be cute to test it on a IIfx and a Quadra 950.
Last edited:
Mk.558
Well-known member
The IIfx wins again.
Here's the results with a IIfx. It's not the most ... compliant user of this adapter. Serial Switch must be set to Compatible, as Faster will just mean it can't do anything with the adapter unless Printer Compatibility Mode is checked and Network CDEV is set to LocalTalk Built-in.
Here's the Performance program doing a speed test directly. File size of 3242649 bytes was used for the tests on the SE/30 and IIsi.
According to Snooper 2.0, the IIfx pushes memory around at about 17.4MB/s. While I doubt that is a factor here, I'm sure it has to do with some other factors. Maybe we'll never know how much you can push a I/O Serial or DMA serial port with RS422 and AFP.
Here's the results with a IIfx. It's not the most ... compliant user of this adapter. Serial Switch must be set to Compatible, as Faster will just mean it can't do anything with the adapter unless Printer Compatibility Mode is checked and Network CDEV is set to LocalTalk Built-in.
| Machine | Download time | Upload time | Download KiB/s | Upload KiB/s | Download Kbps | Upload Kbps |
| IIfx, 3186293bytes | 0:30.33s | 0:33.04 | 101.3 | 93.0 | 810.29 | 743.83 |
| IIfx, 3242649bytes | 0:30.79s | 0:33.60s | 99.7 | 91.4 | 798.18 | 731.43 |
Here's the Performance program doing a speed test directly. File size of 3242649 bytes was used for the tests on the SE/30 and IIsi.
According to Snooper 2.0, the IIfx pushes memory around at about 17.4MB/s. While I doubt that is a factor here, I'm sure it has to do with some other factors. Maybe we'll never know how much you can push a I/O Serial or DMA serial port with RS422 and AFP.
Similar threads
