Amiga.org
Amiga computer related discussion => Amiga Hardware Issues and discussion => Topic started by: AmigaMance on June 05, 2009, 07:41:34 AM
-
Hello.
I would like to know how fast ADSL connection an Amiga 1200 can handle, through a PCMCIA ethernet card?
I can imagine that the speed of the CPU is a primary factor, but are there other HW/SW factors that impose an upper speed limit?
Has anyone encounter any other kind of problems with fast ADSL, *besides* CPU saturation?
-
Sa fast as your PCMCIA card is.
So if you have 10MBit ethernet card or 11MBit wireless card that is the limit.
-
Depends on your CPU.
Absolutely no more than 10/11Mbits as they is the theoretical limit of the ethernet cards.
In practice more realistically like 6Mbits.
But you'd be mad to want to download large volumes of data using the Amiga, use your main PC and then transfer the data across.
-
Answer is simple.
PCMCIA Bandwidth is about 2.5Mb/Sec @ best & as alexh already says, your CPU will have an impact on things too!
Kin
-
10 Mbps ethernet = 1.16 MB/s
PC Card 16-bit 255ns Byte mode = 3.92 MB/s
PC Card 16-bit 100ns Word mode = 20 MB/s
Amiga 1200, 14 MHz = 71 ns cycle.
And PC Card 16-bit don't have any DMA capability.
The CPU will be the worst bottleneck.
-
IIRC Piru benchmarked cnet.device and got something about 800-900 kB/s (can not remember exact figures).
-
IIRC Piru benchmarked cnet.device and got something about 800-900 kB/s (can not remember exact figures).
Not a particularly useful figure without knowing which CPU @ what speed.
Assume it is 060@50MHz and this probably is the top figure you'll ever get from an A1200 and that puts the maximum speed at 7Mbit/s.
I feel the figure will be lower with EC020@14MHz
PC Card 16-bit 255ns Byte mode = 3.92 MB/s
PC Card 16-bit 100ns Word mode = 20 MB/s
Amiga 1200, 14 MHz = 71 ns cycle.
Call me dumb, but what is the relevance of these stats?
-
PC Card 16-bit 255ns Byte mode = 3.92 MB/s
PC Card 16-bit 100ns Word mode = 20 MB/s
Amiga 1200, 14 MHz = 71 ns cycle.
Call me dumb, but what is the relevant info here?
-
Assume it is 060@50MHz and this probably is the top figure you'll ever get from an A1200
Isn't motherboard-access on the A1200 slower with a 060 than with a 030 ? Or is this a rumor ?
-
Not a particularly useful figure without knowing which CPU @ what speed.
Assume it is 060@50MHz and this probably is the top figure you'll ever get from an A1200 and that puts the maximum speed at 7Mbit/s.
Here's an excerpt from the original readme:
My network consists of an A1200 603e+/240MHz/060/50MHz connected via
coax cable to a P120 laptop. Software on the Amiga is AmigaOS 3.5 +
Genesis 1.0.6 and on the laptop Debian 2.2. The amiga has Apollo RE450CT
pcmcia card and the laptop has 3com EtherLink III pcmcia card. FTP
download from laptop have reached 786Kbytes/Sec. On overclocked 060
systems speeds of >819Kbytes/Sec have been reported.
-
And it is, of course, more complicated than that. Data rates use powers of ten, so 10 Mbps = 10,000,000 bps = 1,250,000 bytes per second. Throughput in TCP/IP is a ratio of receive window size to latency, bounded by bandwidth.
For example, a default AmiTCP configuration with a receive window of 4,096 bytes has a maximum transatlantic (~45ms latency) throughput of 728,177 bps = ~89 kilobytes per second. To approach the theoretical maximum of 10 Mbps, you need a receive window size of 56,940 bytes (always a multiple of the TCP MSS of 1,460 bytes).
Transamerican latency is ~85ms, so you'd need a receive window size of 128,480 bytes (64,240 bytes scaled once) to reach 10 Mbps. I don't know off the top of my head if AmiTCP supports window scaling.
None of these numbers takes into account system speed and application and stack processing time.
-
Thank you all for your answers!
I have one more relevant question. Has any of you benchmarked cnet and 3c589.device at the same configuration to see which one consumes less CPU? I suppose that there should be some difference. I have a cnet compatible card only and i can't do it myself.
-
Both should be idle unless a packet is sent or received or some other monitored interrupt is triggered. Received packets should only be sent to the protocol stack if there's a pending I/O request; otherwise, the ISR will drop/ignore the packet.
-
lets change the CPU to 240Mhz running a PPC OS. how fast will it go now?
-
With a 10 Mbps interface? Exactly the same speed. With a 100 Mbps interface? The upper bound is higher, assuming your peripheral bus provides enough bandwidth, but the same basic rules re: actual throughput still apply. I don't think any classic Amiga system, regardless of processor, can process data quickly enough for a 1 Gbps interface. A G-Rex on a Blizzard PPC might reach half that if the system was doing nothing but pushing raw Ethernet frames. Anything faster than that is obviously out of the question unless you build a custom interface with a very, very, very large buffer.
-
lets change the CPU to 240Mhz running a PPC OS. how fast will it go now?
If ur using iBrowse 2.4, forget it. - An 060 @ 50Mhz under OS3.9 just totally trashes a 240Mhz PPC under OS4.0 Classic in desktop speed alone. :D
PMSL!!
Kin