General Network Speed Question....

[essenz]

I have a variety of vintage macs (IIci, IIfx, LC III, Powerbook 3400, Quadra 605, Quadra 700, PowerMac 8600). They all have 10Base-T NICs in them (Quadra 700 and PoweMac 8600 are built-in).

I've noticed that KBps (Bytes not bits) varies wildly between them, with the less powerful machines being the slowest.

For example, the Quadra 605 on my local network gets a MAX of about 42KBps transfer throughput, thats 336kbps on a 10Mbps LAN link - which seems really low. In this case, I'm downloading a file from a more modern iMac on the local network running a web server, the download is being done in Netscape 4.X on the Quadra 605. Is that a reasonable trasnfer throughput? Is the bottleneck the CPU horsepower?

On my PowerMac 8600, I get a much better transfer rate, like 220KBps throughput (1.7Mbps) - better but still far below the max link capacity.

Its more of curiosity, though it is annoying how long it takes to get stuff onto that Quadra 605 when 42KBps is my max speed.

 

[ArmorAlley]

Out of curiosity, is your NuBus bus overloaded? Do you have a graphics card running a 1024x768 at millions of colours and an LVD SCSI card with a fast disk flooding the bus? If so, your question may be partly answered.

How are you measuring your throughput? Using a connection to the Internet with Netscape 4? If so, have you considered building up a simple matrix of your machines? Say, how many seconds does it take to transfer xMB of mixed files from Mac A to Mac B. I, personally, use 95MB it just needs to be the same files in every test. Then try every permutation with the same set of mixed files and then you can compare the results.

Also bear in mind the state of fragmentation of your hard drives, the size of the system cache may play a role and you should also look at Speed Doubler from Connectix. You will need to make a physical floppy from an image of Speed Doubler if you wish to install it.

A final point is disk to disk transfer within the same machine, and epecially a RAM disk. To get the maximum possible transfer speed from a machine, you need to create a RAM disk and time the transfer of the aforementioned files from the disk to the RAM disk and vice versa. Once you have that, then measure the time taken for the files from one disk to another within the same machine, should you have two disks in one machine. This will give you an idea of the speed differential between machine and network speeds.

Someone here will surely know of a very useful set of utilities that measures and analyses network speed and I hope that they will share them.

 

[cruff]

One thing that could affect the transfer rate is if the Ethernet interface uses DMA for data transfers or CPU polled I/O. Not having DMA available will lower the throughput drastically.

 

[essenz]

So some followup details, its not internet traffic, its local network. The web server is a 2016 iMac. All my vintage Macs have SCSI2SD cards in them, not sure if the SCSI2SD is limiting file IO which is then slowing me down to 42KBps. The PowerMac 8600 is not SCSI2SD, its using a PC SATA card attached to a more modern SSD drive.

Regarding Nubus... the IIci and IIfx have Nubus ethernet cards, PM 8600 and Q700 have built-in ethernet, and the LC III and Q605 have PDS style ethernet cards, video on those in onboard.

I'll confirm the make/model of the ethernet cards, maybe from that I can confirm DMA details. 

For transfer rates I'm just using the guage within Netscape while doing the download, It peaks at 42KBps on the Q605.

Also, the machines are connected to a modern 10/100/1000 switch, maybe having them on an older 10/100 Hub would be better.