3Com 64-bit 66MHz PCI-X gigabit server NIC for $135 shipped @Dell Home

[docinthebox]

Dell Home has the 3Com 10/100/1000 PCI-X Server NIC (3C996-T) for:

$178
- $17.80 (10% off coupon)
- $25 ($25 off $150 coupon)
---------------------------------------
= $135.20 shipped (free shipping, no minimum)

This gigabit NIC can run on a 32-/64-bit 33/66/133 MHz bus, which will be perfect for my next system when the AMD-760MPX motherboards hit the market. I've been thinking how I can make use of the new 768 south bridge which has a 66-MHz/64-bit/32-bit PCI bus.

Of course, the next question is how can I make use of the gigabit bandwidth.

 

[craigg]

I have a few of these new in box and will sell one for $100, if anyone wants it email me at xxxxxxxxxxx Sorry...all sold

 

[jjjpflynn]

Docinthebox,
Better check into this story at The Inquirer before you assume you'll be able to use the NIC. At least if you haven't gotten the MB yet, there'll be time for them to figure out the issue and come up with a solution. Sorry to bear bad news.

http://213.219.40.69/24120102.htm

 

[grimreap]

On the back of the box it says PCI/PCI-X and some other nonsense that kind of implied to me that the card is backwards compatible and would work in a PCI slot, I'm positive there are some drawbacks if thats the case, maybe can only do 100mbit...

But I didn't want to open the box and try it out so again im not positive on this....

 

[docinthebox]

Thanks for the tip, guys. It seems like a problem with the Tyan motherboard though, not with the AMD-760MP(X) chipset. I hope this problem won't exist in Asus' A7M266-D when it comes out in January.

According to the article, it could be a BIOS problem. I think it could be a driver issue also. The reseller in the article said "But the problem is more serious than this and is not a driver problem . 3Com's 3c996 causes the brand new dual Athlon boards (Tyan Thunder K7 as well as Tiger) to become unstable, they freeze sometimes. Once I saw this happen before the OS - Linux in this case - had even started, so it is emphatically not a driver problem," which puzzled me because a driver problem could also cause a system to freeze, say for example if your driver thread deadlocks with a kernel thread. Why is it "emphatically not a driver problem" only because it freezes? IMHO, this doesn't make sense. Maybe the reseller is a friend of 3Com's and want to blame Tyan/AMD?

Edit: My bad... I was too blind to see the phrase "...Once I saw this happen **before the OS had even started**..." So I agree with the reseller's comment now - it cannot be a driver problem. The symptom is consistent with a BIOS problem.

 

[seniorguy]

Can somebody elaborate the performance advantage of using this than a 32 bit NIC card?

 

[docinthebox]

If you have two PC's hooked up with a gigabit LAN, the data path from the CPU on one PC to the CPU on the other PC looks like this:

CPU -> FSB -> Northbridge -> PCI bus -> NIC -> cable -> NIC -> PCI bus -> Northbridge -> FSB -> CPU

If you're using DDR SDRAM, the FSB has a data rate of 2.1 GB/sec (gigabyte, not gigabit). Assuming you're using a cable appropriate for a gigabit LAN (at least CAT 5e), you can see that the PCI bus on both sides can easily become the bottleneck.

The data rate of a conventional 32-bit 33-MHz PCI bus = 32 bit * 33 MHz = 1.1 Gb/sec (gigabit). When you take into consideration that this data rate of the PCI bus is shared between all your SCSI devices, your other LAN cards (if any), and the Southbridge (through which the USB and EIDE buses merge into the PCI bus), you can see that it's hard to achieve a 1.1Gb/sec data rate from CPU to CPU if you only have a conventional PCI bus. A 64-bit 66-MHz PCI bus will quadruple this data rate to 4.2 Gb/sec and now you have a lot of leeway to share the PCI bus with other devices as well.