Hi all,

I have two PCs. One running Win2k Pro, the other WinXP Pro. Both with 10/100M network cards (from the same manufacturer) with a direct cable (CAT5) connection.

The cards are configurated in 100 Full mode speed.

When I transfer one file of 300 Mbytes between the PCs, the transfer (estimated and real) time is around 3-5 minutes.

My questions are:

1-) If the connection speed is 100M, the transfer time should not be 3-5 SECONDS instead minutes?

2-) What 100M stands for? 100MBytes/s or 100Mbits/s? In the case of 100MBytes/s the transfer time shoul be around 3 seconds. If the correct answer is 100Mbits/s, dividing by 8 (to get the speed in MBytes/s) give 12.5MBytes/s. Even with this "slow" speed, the transfer time shoul be around 24 seconds! So, something is wrong.

Also, I measured the connection speed with the software "ICB 2000 v4.1", the average speed was around 25MBytes/s. So, the time should be 12 seconds.

Any ideas, sugestions, hints?

Best regards,

Hatuey

Take a look at post #6611 in this forum.

1.Its a fallacy. You'll never get 100% utilization of the bandwidth.

2.The 100M is megabits. A rule of thumb is to count 12 bits for every byte of data sent (network overhead).

Disk access speed plays a big part in copying files over a network. Also DMA, memory, fragmentation ratio and other running apps and services.

Your tested connection speed sounds about right. Using 12 bits per byte, your transfer rate is running 12 to 20 Mb per sec. Take a look at the Network Tab in XP's Task Manager for the utilization. Probably doesn't hit 30%.

Forgot to mention that you might want to try testing using NetBuie. I might try it another day (soon) and will post back.

I downloaded ICB and try that same file transfer, NT4 to XP (post 6611). ICB topped out at 16Mb/s, but like I'd said, my systems aren't the greatest.

So I tried the transfer using NetBeui. ICB wouldn't recognize that I was using that protocol, so there are no numbers from it.

Timing the 85,038,592 byte transfer resulted in this (this was done NT4 to NT4 once again):

This PC to other PC @ 100Mhz - 23s, equaling 29.4 Mb transfer rate.

Other PC to this PC @ 100Mhz - 27s, equaling 25.2 Mb transfer rate.

Other PC from this PC @ 100Mhz - 15s, equaling 45.3 MMb transfer rate.

If you read that other post (6611), you'd see that I shaved off 5s in using NetBeui to push the data over. While only 4s were shaved off pulling the data back.

I haven't pinned down why there is that 12s difference (beyond the machine differences) in regards to the direction of the transfer. I still think its because of the disk read/write transfer speeds of the IDE controllers used.

This PC uses more CPU power (27%) for about a 8MB R/W transfer rate. The other PC uses about 1-3% CPU for a 16MB read rate vs a 2MB write rate. When I'm at the other PC (slow write rate) pulling the data from this PC, I'm writing to the supposedly slow write rate disk. So you see my dilemma.

Thanks a lot Michael.

I really was surprised, because I used to think that the connection between PCs using network cards should be faster than using other kind of connection (in the PC-PC USB connection the speed is around 10MBytes/s and via LPT is around 4MBytes/s).

Hatuey

Ya, but try running a 100m USB or parallel cable. Or having more than two systems connected together. Those options will also run into the disk transfer speed problem. Haven't tested them, so I have no comparison.

Remember, a network is for a lot of machines to communicate together. I once worked on a 10Mhz network that was normally 80% utilized by 30 PCs (and it was slower than hell, too many collisions).