Slow LAN - surely not normal?

McCarthy

Platinum Member
Oct 9, 1999
2,567
0
76
Got a real basic LAN here, two computers (sometimes 3) hooked up through a four port hub. DSL comes in and is shared via same hub. TCP/IP for internet access and NetBEUI for LAN on both machines. One's running W98, other is W2K. 10bt hub and one NIC is 10 only, other is 10/100.

I've had various combos over time (W98 on both, ISA/PCI/USB NICs) but have never seen more than 500KB/s transfer rates. At present I'm getting around a whooping 300KB/s at best. Level of internet activity makes for a lot of collisions, but overall transfer rate doesn't change, even with the DSL removed from the loop it stays that low. Powered hub, Belkin cables. Always works, but always slow.

Searching good ol Google for 'slow lan performance' hasn't turned up anything useful yet, figured maybe someone would know "oh, it's _____" so here I be, asking for your knowledge.

Thanks
--Mc
 

Mucman

Diamond Member
Oct 10, 1999
7,246
1
0
Try using FTP to test file transfer speed... I am not sure what protocol overhead there is with NetBEUI. Here's what I would try :

- Disconnect the DSL from the hub.
- Setup a simple FTP server on the best computer
- Download a 100MB file and check the speeds...

Try replacing cables if the speed isn't around 1MB/s
 

JackMDS

Elite Member
Super Moderator
Oct 25, 1999
29,553
430
126
Your LAN is basically 10Mb/sec. LAN.

Is it 500kb/sec. (b=bit), or 500kB/sec. (B=Byte).

If it is 500kB/sec. then it is within the normal range of a 10Mb/sec. LAN.
 

McCarthy

Platinum Member
Oct 9, 1999
2,567
0
76
500 kilobytes max, but that's when both were W98. Now the best I see is around or under 300. Thought 10 megabit should be right around 1 megabyte give or take because of overhead.

Ok, MB is 1048576 bits, x 10/8192 = 1280KB divide 300 that = 23% of theorhetical throughput. Going to edit to fix those kb to KB, always forgetting to do those right.

Still seems slow. If that is normal, why is it normal? Surely NetBEUI doesn't have over 50-70% overhead??
 

Mucman

Diamond Member
Oct 10, 1999
7,246
1
0
Just try using FTP and find out :) I have a large post somewhere in the networking forum where I was benchmarking my SAMBA box and found out all sorts of bottlenecks... althougth not getting 10Mbit/s seems odd.
 

spidey07

No Lifer
Aug 4, 2000
65,469
5
76
Also lots of collisions can slow your down to a crawl, however you shouldn't have many with just a single file transfer between machines. If you want blistering speed then get 100 Megabit cards and a 100 meg switch.
 

JackMDS

Elite Member
Super Moderator
Oct 25, 1999
29,553
430
126
When evaluating LAN speed with normal functional home computers. The ?speed? is a combination of the Network hardware (Hub/Switch, NIC, Cable, collision etc.), and also the computer power, the hard drive speed and status, the cache status, the software used, computer doing other computation, etc.

As a result the theoretical 1.25MB/sec. of a 10Mb/sec. network, can spread between 500kB/sec. to 900kB/sec.
 

zephyrprime

Diamond Member
Feb 18, 2001
7,512
2
81
When evaluating LAN speed with normal functional home computers. The ?speed? is a combination of the Network hardware (Hub/Switch, NIC, Cable, collision etc.), and also the computer power, the hard drive speed and status, the cache status, the software used, computer doing other computation, etc.
I've heard such rationalizations before but I've always thought they were bunk. Would anyone be satisfied if your GF4 gave half the frames per second as someone elses with the same computer setup?
McCarthy, I asked a question simliar to yours months ago but I was never able to get a satifactory answer to my problem. I ultimately solved most of it though. I'll share what I learned with you. Much of it may be irrelavent but I hope some of it helps.

I read long ago that windows 95 had some problems with networking and would simply never run to full speed on a 100mbps network. I realize that you only have a 10mbps network but the problem may be related. Also, the problem was apparently never solved in any of the version of windows 9x and was only addressed with windows NT.

You may notice that your transfer rates are a particular value when win 2k starts the transfer and is a very different value when w98 initiates the transfer. Many people have this problem and there is no solution to it! (Thanks microsoft!) I had this problem and I read on usenet that many other people have this problem too.

My own problem was like this, when transferring data from my win2k machine to win98, I would only get ~3MB/s on my 100 Mb network. I asked on this forum for some help but people gave me lame-o replies about system not being fast enough and my hard drive not being fast enough. Whatever. Last time I checked, hard drive speeds surpassed 3MB/s long, long ago. Anyway, I was hoping that migrating to Windows XP would solve the problem.

It didn't. The problem was still there. So I tried swapping my linksys network card with a Netgear network card. This worked to a degree. My speeds went up to 8MB/s. Now I thought, "I want to try super premium cards. I want gigabit ethernet!". So I tried out a generic national semiconductor based network card which I got yesterday. Keep in mind that my network switch was still only 100mbps so I wasn't expecting gigabit ethernet speeds. The gigabit nic helped even more! My network speeds were not up to 8MB/s. I expect that if I can replace my remaining 100mbps SMC nic with a better brand, I'd get even better speeds. I've read that the max theoretical speed out of 100mbps is 11.79MB/s not including the overhead of the netork protocol used by your operating system.

I've also gone online and read paper issues of pc magizine. Old issues of pc magazine showed that was a definite difference between brands of ethernet cards. Many simply did not run at max speeds. All these reviews are pretty old though because 100mbps has been around for so long. I couldn't find any 10mbps reviews. No doubt because 10mbps is so old. Any such reviews would be really unreliable anyway since all the cards reviewed would be ISA and the chipsets under scrutiny would have long since been discontinued.

I was able to find some gigabit ethernet reviews. gigabit comparison As you can see, there was significant deviance between nics and even between switches (check of the cisco switch, it stinks.).

So given all this, I believe that most brands of NICS simply suck but most people don't notice. The only way to get a good nic is to rely on benchmarks but unfortunately there are very few benchmarks produced and older standards like 100mbps and 10mbps have no modern benchmarks available at all! And even gigabit ethernet benchmarks are beginning to be aged because none of the chipsets in the article I link are currently in production. I would recommend you buy a Intel nic because Intel scored consistantly well in both gigabit ethernet tests and the old 100 mbps tests. Unfortunately, I sense that you don't have the most money to spend so this may not be an option for you. If you want, you can have my netgear or linksys card that I tossed if you pay for shipping. Or maybe you can try some cards that your friends may have.

And lastly, this is a little off topic but while I was reseaching networking problems on usenet I discovered that there are many people that have a problem with XP and trying to browse win 9x and 2k machines. Browsing and file transfers are mind bogglingly slow. Many solutions have been proposed including registry hacks and stuff but these only help a small percentage of people. Many people seem to experience this XP networking problem but there is no fix.
 

JackMDS

Elite Member
Super Moderator
Oct 25, 1999
29,553
430
126
Tom's hardware compared three NICs. (Aug. 2001)
I used Pricewatch.com to find the best price for these NICs.

1. 3Com 3CR990-TX-97 - $80
2. 3Com 3C905C-TX - $30
3. SMC 1211TX - $13

Quote from: http://www6.tomshardware.com/network/01q3/010820/nic-11.html#conclusion

"Well, there is some difference between Network cards. The difference was not that great. Some cards did have better performance than others, but over all most of the cards performed within acceptable ranges. "

If you read the rest, you can infer that if there is a difference it will impact high volume cooperate Network, not three computers sharing the Internet at home.


I've heard such rationalizations before but I've always thought they were bunk.
BTW - Rationalization is on the list of Freud's Defense Mechanisms. It considered much healthier Defense Mechanism then another Defense Mechanism on the list "Realty Denial".
 

McCarthy

Platinum Member
Oct 9, 1999
2,567
0
76
zephyrprime, I've wondered for awhile now if it'd pick up if I went to W2K on both, sounds like chances are it would. As for cards, I can't see spending money on new NICs in the hopes that they may do what they say better than others that didn't do what they said, if you get what I'm saying :)

Curious about firewire networking, at $15 a card (the VIA chipset ones at newegg, gotta see if they're ok first) I'd have more functionality out of them for hooking up other devices and have a theoretical 400Mb/s to play with, less than giga, but with some upside. Distance isn't any problem for me, the computers are five feet apart.

Maybe that's what I'll do. At present the internet functions are working fine, just near useless for working with large files for some video editing I've been wanting to do. Can't find my D-Link PCI card right now, but when I had the D-Link PCI and my Netgear PCI I had the best transfers, but again that's when I was W98/W98 so I can't swear that it was the cards making the difference then. Either way, if going to 100Mb is only going to get me 8Mb and I'm going to have to have two NICs per machine to do it I figure I might as well give the Firewire route a try, eh? Of course after doing a search here to see if there are any major problems!

--Mc
 

zephyrprime

Diamond Member
Feb 18, 2001
7,512
2
81
I tried that program that tom's hardware used to test network equipment, QCheck. It's a pretty good program. Anyway, for TCP throughput, I got 80Mbps using a 100KB data size and 83.3Mbps using a 1000KB data size. These scores are comparable to the speeds I was getting before using my own "click and drag a file" network test. It's interesting to note that the network card on my machine is a SMC1211 which was one of the cards tested by Tom but Tom's score for the card was only 50Mbps.

As far as firewire networking goes - it's an interesting idea and may be a cheap and effective solution to your problems.