10Mbps network only going @ 1Mbps?

[cococola1]

help please? I've got 2 computers connected thru a 10Mbps Netgear EN104tp ethernet hub. Each computer is using a Netgear FA310TX card. I tested out the file transfer times for 200mb, 500mb and 1000mb and they all averaged out to be 1Mbps transfer from 1 computer to another. I called up tech support and they blamed it on the IRQ's being shared. I can see the Netgear card is using IRQ 9, which is shared with Vortex AU8810 PCI audio (sound card), IRQ holder for PCI steering, and Vortex AU8810 Multifunction PCI platfrom. I'm also using a USR modem, video card and a PCI USB card that are only using other IRQs. I think this isnt too much, and would be a basic setup for many people. So does this mean other people are putting up with the 1Mbps when it's supposed to go 10Mbps? How can I take full advantage of 10Mbps w/o removing any cards?

both computers running win98 SR2.

 

[Xanathar]

Are you measuring in Bytes or Bits? And How?

 

[cococola1]

ah crap.. it's worse than i thought. It's really 0.25 Megabytes per second. I just did it again and it took 20 minutes to transfer 300 megabytes of files. I tried to force them to transfer 10megs/sec by changing the settings, but it's still the same.

 

[Quickfingerz]

If your connection was going at full 10baseT (10mbits aka 1.25 mbytes/sec) a 300 mbyte file would take you around 5 minutes. This means you are at around 1/4 1.25 mbytes/sec or roughly 300kbytes/sec which isn't too bad.

Just remember... 10baseT = 1.25 mbytes/sec and 100baseT = 12.5 mbytes.sec

remember this is the theoretical max. In most cases you will be traveling at around 800 kbytes/s and 6 mbytes/s respectively depending on your hardware.

 

[jeebus21]

how many ports does your hub have? You should know that a hub just takes the bandwith of its maximum speed (10/100/whatever) and divides that across the number of ports. If you have a 10-port hub, then you should see roughly 1Mbps on each connection. You would have to purchase a switch to get better performance (but it sometimes deteriorates performance amongst only a few computers...). I would guess your problem is the hub, not the computers.

 

[esung]

jeebus21: no, that's not how it works. You are figuring the theoritical AVERAGE bandwidth PER connection , NOT per how many physical port! say I have 4 computer sharing a 10Mbps network, so the average bandwidth is 2.5Mbps per computer/connection.. even if I have a 16 port hub. Since there's only 4 connection to the hub. But if only 1 computer is transfering data from the other computer (1 connection only), even there are 4 computers, that computer can enjoy the full 10Mbps bandwidth since no one else is talking.

 

[FirmPete]

Don't forget that when you have a speed of 10 mega bit the protocol you are using takes also takes it's amount of bits to operate.

TCP/IP is much "heavier" on bits to operate than for instance Netbuieieyuu ^%#$%Netbuie whatever.

Spellcheck: NetBEUI.

Try to get Netmedic, it's freeware and lets you see how fast things are going

 

[jeebus21]

Nevermind, I was under the impression that he had all the ports filled with something...well, computers. I see now that he only has 2.

 

[cococola1]

yeah. i've connected 2 computers to this 4 port hub. but i could have sworn i got these two computers to transfer files at 10 megabytes per second before! was that just a glitch?

moral of the story?
"10baseT = 1.25 mbytes/sec and 100baseT = 12.5 mbytes.sec"

thanks!

 

[Ender78]

Ethernet is actually very bad and not very efficient. Although you have a 10 Megabit hub, its unlikely that you are able to utilize more that 60-70% of it (its just the way it is with Ethernet, at anything higher than 50% chances are no data is flowing well at all). You also may want to watch the colisions light on the hub, if it is flashing very often, you have a problem. Many cheap hubs are the cause of poor network performance. The other thing to look at is distance, how far away is the second machine, where is the cable going through ( if for instance it is going through the walls of your house it may be getting a great deal of interferance from wiring or other media)

 

[cococola1]

ethernet is bad? I see it is a cost effective means of transfering files and sharing a analog internet connection between two home networked computers. What do you recommend that would be better? shorter cat5 cables and a 100 megabit hub?

 

[Ender78]

CocoCola1

I guess the word BAD was a little strong. What I was trying to say is that Ethernet is not as efficient as many people believe it to be. Its still by far the most widely used technology for networks.

As for the distance, how far are you going ? With cheap and old cabling (and if this cabling is picking up any interferance) you may get a lot of colisions.

A switched network is best as it pretty much eliminates collisions. Looking at CAT6 may also be worth it (it will let you use Gigabit Ethernet down the road).

 

[cococola1]

the computers are not separated far from each other at all. actually, they're right next to each other. however, they are each using 15' cat5 ethernet cables, this way i can extend them in case i need to. i'll take a look into a switched network. thanks so much for your advice... all of you and ender78 too.

 

[TheViper6]

Check your cables, bad cables kill bandwidth, watch full duplex too, on a hub its suicide

 

[SuperKen!]

jeebus21, even if he did have all the ports filled up on the hub, your equation on bandwidth is not correct.

 

[Ender78]

Buy a little more expensive hub at a store with a refund policy. I have an ACER hub here that works REALLY well, I had a buddy that bought some cheapie $30 hub and it works like doodie, Make sure you get a powered hub(one with its own power supply).

 

[cococola1]

ender78, how does it perform "really well"? Does does it transfer more bytes per second? the Netgear EN104tp came as a kit w/ two network cards and its own power adapter which i thought was not too shabby for ~$40.

-coke!

 

[Dark]

Ender78:

<< A switched network is best as it pretty much eliminates collisions >>

It's not totally true. In fact Fullduplex eliminates collisions not switches. If for example one of your nic is set to halfduplex and the swith to autosense...well you'll see collisions even if you use an all switched network. I learnt my lesson thx to xanathar, spidey07 and damaged I don't forget CTR

 

[Cfour]

What I would do is setup and FTp across yur network and transfer files that way, it will tell your right there how fast you are transferring.

-Tony

 

[spirites]

hmm...Dark, I believe you are quite mistaken
switches stop collisions since they store mac addresses and only let one packet through at a time to the appropriate nic. Hubs on the other hand just keep bouncing all signals/packets to all machines hoping one will respond and packets collide all the time. Full duplex does nothing but double the available bandwith with a dedicated transmit and a dedicated receive.

 

[Damaged]

spirites,

Not if it's set to half-duplex.

You're right if it were switched full-duplex though.

 

[Dark]

Spririt: I'm not mistaken Hubs use CSMA/CD to access the media while when using switches with full duplex, CSMA/CD and loopback detection are disabled. If u use switches without using fullduplex, then CSMA/CD is not disabled and collisions may occur. Since halfduplex means using the same pair to receive and transmit, what would happen if the switche tries to send infos to the nic at the same time the nic sends info to the switch? -> collision.
Still swithes even in halfduplex will reduce the collision domain...so two nics &quot;directly plugged&quot; in a switches won't collide with each other...
Cheers.

 

[RockyMaivia]

I was thinking about getting a Linksys 5 port workgroup hub. Anybody know about the quality of those?