NIC article

Good article, though lacking in number of cards compared. Might have been interesting to see most common Home cards - Linksys, D-Link, NetGear, etc. Also, what about integrated, like on my Asus P2D-LS or CUR-DLS or my Compaq ProLiant servers (DL360, DL380, ML370)? How about more info on the common routers/hubs/switches?

11 answers Last reply
More about article
  1. We are in the process of getting more NIC cards for testing. Once we have more cards, I will be doing a follow up article that will look at cards from most of the companies that you suggest. Since this was the first article in the networking area, we wanted to get an idea if this was something that folks really wanted to know about. I agree that more cards need to be looked at, and if the responce is positive, I am sure that they will give me the green light to do it.
  2. I agree with BCinSC. I would've liked to see some Netgear cards in there especially. I bought a Netgear FA310tx partially because that's what Tom's uses in most of their test systems, and partially because a couple of my friends have it.

    <font color=blue>Quarter pounder inside</font color=blue>
    <font color=red>Change the Sig of the Week!!!</font color=red>
  3. definately a great idea for a new addition to the site!! its strange how so many sites completely ignore networking and related hardware when doing reviews. i agree that budget cards that are typically used by home users (like us) should be added to the benchmarks. a common question in this forum is 'what is the difference between business class NIC's and consumer class NIc's.......good job.

    ignore everything i say
  4. Please try to spend more time in researching when making affirmations.
    The 3CSOHO100-TX card IS supported by the Linux driver, for both 2.2 and 2.4 kernels; it is a recent addition, but you should be able to find it in all distributions released this year. Even more, the Linux driver (3c59x) uses it in exactly the same way as the fully-featured 3c905C-TX. (Of course, WOL and other facilities which depend on the hardware features will not work).

    Using 3C905C-TX cards and Linux, one can transfer over 90 MBit/s with low CPU usage. Around 90 MBit/s is obtainable with the Intel card; cards based on designs similar to DEC-21040 (driven by the tulip driver) also can transfer over 90 MBit/s. This is not to start a Linux vs. Windows war, but just to show that 100 Mbit/s cards deserve their name (which is not clear from your 60+ Mbit/s figures).
  5. I'm glad that Tom's has added this to the site... I feel it's an overlooked area. Most people just say 'Bah, a NIC is a NIC,' but we don't really know if this is the case unless someone takes the time to do some objective testing. I'm glad you guys have stepped up to the plate.

    Something you might want to consider for future tests is to look at the performance of the cards when the FSB has been overclocked. I've heard that some brands more than others have a tendency to lock up when the FSB is pumped too high. This would be good to know, so that the overclockers who read this site have some idea of what NIC is best for their rig. I hope you'll consider adding this as a test to the next NIC review.
  6. Great article,

    but with one flaw (atleast for me). I couldnt find the tcp-responsetime. I play a LOT of online games, mainly fps and ping matters. In addition to your lost response page could you add a realworld reference like ping in q3 or cs on your lanserver. Thanks again for an execellent article and please consider my requests.

    Can I borrow your candy?
  7. Post deleted by error
  8. How do I delete my second post?

    Can I borrow your candy?
  9. Click on the edit button, and then there's a button below the little typing window to delete the post.

    <font color=blue>Quarter pounder inside</font color=blue>
    <font color=red>Change the Sig of the Week!!!</font color=red>
  10. Good, I'm glad this review was done, but it raises questions for me regarding this 3CR990-TX-97 from 3COM. I should say off the bat that the only reason I desire network performance is for multiplayer games. I could care less how long it takes for a file to float through the ether. Also price is not an object (within reason), and since these cards all fall within my price range I would like to consider all of them. Also any CPU offloading I see as a huge benefit for keeping my older machines useful.

    It looks like from a description of some of the features that the card is designed to offload the system from network related computations. I'm guessing this was done to free up these resources for other uses. The author goes on to praise this card as the greatest thing in the world for WIN2K. It all made sense to me, and I was listning.

    My problem is the test results. I can't see how this card could possibly have gained a reputation as the "best", especially based on the posted results. It is clearly inferior in most of the results, granted the CPU utilization under WIN2K doing an FTP trnsfer is the lowest. I guess my quesstion is: is this one measly result the reason that this card is so highly praised as being the best?

    Unfortunately I am unable to translate the posted results into something meaningful for what I am doing. Does this card do anything for quake III performance while playing multiplayer games across a LAN? Or is this something that only a super IT guy with with an electron microscope can get any benefit from?

  11. I am glad that Tom is giving attention to this area. I have done my own testing with some of these cards. I do disagree with some areas of the reveiw though because the Intel Pro 100 is much more reliable than the other cards. On Linux 2.2 if you disconnected the cable to the network for even a moment with any other card (AMD,3Com 3c905 cards) it would not regain the connection even if you restarted networking and Samba. You would have to restart the machine. (Which of course is a driver problem but nobody else but Intel bothered to work this detail out even had a simular problem with Netware5.0). Second the test fails to demonstrate the what happen when multiple users hit one machine. As in a server or wrkstation being hit by multiple reuests. This matters in a real world Enviroment. Try it you will be surprised how much the Intel Pro 100 shines over the other cards in this area. Third the fact is that spending a mere $100-$200 more for a higher end card no matter which brand could save you thousands in hardware for your server. Which makes a high end nic a cheap way to deal with server bottle necks. Fourth you need to look into load balancing features of the cards. Most Techs just throw in a 3Com and forget about it. It is a bad choice to go cheap on your server in this area.

    I have a comparison chart with win98 winNT win2k NW5.0 and Linux 2.2 comparing the operating systems and their hardware on a test network set wy classmates and me set up at our tech school.

    Email me with the subject "Nic article" and I will send you the comparison as an attachment in reply.
Ask a new question

Read More