Intel NIC working on only 100BaseT, while onboard Broadcom NIC works at 1000BaseT

Status
Not open for further replies.

rakesh

Dabbler
Joined
Jan 26, 2014
Messages
16
I recently bought the Intel EXPI9402PT NIC in hopes of getting a slight improvement over my network speed when I have 2 clients hitting it simultaneously. With my previous onboard NIC, I was getting a top speed of 100 MB/s, but very rarely do I get that speed, most of the time I will get somewhere between 10-35 MB/s and even worse when 2 clients are hitting it at once. I was hoping that the new NIC would take the guesswork out of the speed and have it be consistently at 100 MB/s or so.

But, the new NIC seems to have degraded the speed somehow. The card has a PCI Express 4.0 connector and my motherboard didn't have a connector for that, so I just hooked it up to a x16 lane since PCI is backwards compatible. After I installed it, I installed a fresh copy of FreeNAS and mounted and shared one volume. I tested the speed and I get an average of 11 MB/s.

I haven't even tried to hit it with both clients yet since it already sucks so bad with one client. I also tried hooking up to the onboard NIC again and I an average speed of 100 MB/s. Am I missing something really obvious here?

The output of ifconfig with the onboard NIC is:

Code:
bge0: flags=8943<UP,BROADCAST,RUNNING,PROMISC,SIMPLEX,MULTICAST> metric 0 mtu 1500
        options=c0099<RXCSUM,VLAN_MTU,VLAN_HWTAGGING,VLAN_HWCSUM,VLAN_HWTSO,LINKSTATE>
        ether 00:26:2d:20:71:90
        inet 192.168.1.149 netmask 0xffffff00 broadcast 192.168.1.255
        nd6 options=9<PERFORMNUD,IFDISABLED>
        media: Ethernet autoselect (1000baseT <full-duplex>)
        status: active



The output of ifconfig with the Intel NIC is:

Code:
em1: flags=8943<UP,BROADCAST,RUNNING,PROMISC,SIMPLEX,MULTICAST> metric 0 mtu 1500
        options=40098<VLAN_MTU,VLAN_HWTAGGING,VLAN_HWCSUM,VLAN_HWTSO>
        ether 00:15:17:8e:e8:39
        inet 192.168.1.126 netmask 0xffffff00 broadcast 192.168.1.255
        nd6 options=9<PERFORMNUD,IFDISABLED>
        media: Ethernet autoselect (100baseTX <full-duplex>)
        status: active


As you can see, with the onboard NIC, it autoselects the right speed at 1000baseT, but with the Intel NIC, it only autoselects at 100. Is there anyway that I can force the Intel NIC to function at 1000MB/s instead of 100MB/s. I would think that is a problem because of my ethernet cable or something, but how come the onboard NIC works just fine?
 

cyberjock

Inactive Account
Joined
Mar 25, 2012
Messages
19,526
Can you post your full hardware specs?

Can you post a debug file with the Intel nic installed (system -> Advanced -> save debug)?

Any chance you got a "fake" Intel NIC? There's so many floating around now that this sounds like you may have been had. Did you buy it on ebay? Can you get 1Gb in another machine with the Intel NIC or in the same machine with a different OS?

I wouldn't recommend forcing 1Gb if its only negotiating 100Mb ;) At least... not yet.
 

rakesh

Dabbler
Joined
Jan 26, 2014
Messages
16
Can you post your full hardware specs?

Can you post a debug file with the Intel nic installed (system -> Advanced -> save debug)?

Any chance you got a "fake" Intel NIC? There's so many floating around now that this sounds like you may have been had. Did you buy it on ebay? Can you get 1Gb in another machine with the Intel NIC or in the same machine with a different OS?

I wouldn't recommend forcing 1Gb if its only negotiating 100Mb ;) At least... not yet.

Huh..I did but it from eBay. :( I guess I'm screwed...maybe. I'll try putting the card into another machine as soon as I get home. Also, is there any physical or visual indication that the card is a fake?

Anyways, my specs are very barebones, a repurposed home PC:
  • AMD Athlon II X2 240
  • 8GB RAM
Also attached is my debug log.
 

Attachments

  • debug-rakeshnas-20151024154344.tgz
    80.8 KB · Views: 211

cyberjock

Inactive Account
Joined
Mar 25, 2012
Messages
19,526
There are some physical signs that it is fake, but you'd have to Google them to see if anyone has posted any good "tell-tale" signs for your specific card.

Few things to think about:

1. 9.3 may give different results. You could try a fresh install of 9.3 to some spare USB stick just to "see if it matters".
2. You should avoid having really long tables (100m+), and ideally you want cat 5, 5e, or 6 cabling for 1Gb. For problems like this I'd try to use a 6 foot or shorter cable that is cat-6 just to make sure its not cabling. Companies have different expectations for signal and noise, and Intel may be more conservative and giving you 100Mb because of a marginal cable while Broadcom is less conservative and giving 1Gb. I've also seen marginal cables negotiate 1Gb on one brand NIC and 100Mb on another, but as soon as you start heavily taxing the NIC that had a 1Gb link you'll see lots of packet retransmits and such, so you aren't truly getting 1Gb of available throughput anyway.
3. Your debug was a bit weird. It had lots of errors of various kinds. I can only guess this is a working system and you removed the zpool drives so the zpool wouldn't mount when you made the debug file. This may or may not mask problems as the logs are stored on a temporary RAMdisk so the debug may not show the problem you are trying to identify.
 

gpsguy

Active Member
Joined
Jan 22, 2012
Messages
4,472
While your onboard NIC isn't an Intel, a Broadcom NIC isn't in the same league as a Realtek. I don't upgrading to an Intel is going to solve your problem(s).

At the end of the day, you have a 5 year old pc with a mismash of hard disks. Note: there is a BIOS update available for the machine. You have version A01, A06 is available.

As cyberjock said, we can't see what kind of zpool you have/had. I see you have the following hard disks: 3TB Seagate, 2TB Seagate, 320GB Seagate, and a 500GB WD Blue. How were you using them? A stripe (dangerous), two sets of mirrors (3TB & 2TB) + (320GB & 500GB), one volume per drive, or ???

You mention hitting speeds of 100 MB/s, but the average is much slower. Tell us what protocol you are using and the conditions for the testing. Are you using CIFS? When you get 100 MB/s are you copying a big file and when it's slow are you trying to copy 5000 little files?

Are you using any plugins or jails?

What are the spec's of your connected hardware? For example, my "old" pc was a little older than your "server" and had a spinning hard disk. My "new" pc (about a year old) has a 3.5GHz CPU with a fast SSD. I saw an increase in CIFS performance after migrating to new client hardware.

With my previous onboard NIC, I was getting a top speed of 100 MB/s, but very rarely do I get that speed, most of the time I will get somewhere between 10-35 MB/s and even worse when 2 clients are hitting it at once.
 
Status
Not open for further replies.
Top