No, I Will Not Fix Your #@$!! Computer → Broadcom 5752 NIC stuck in 10/100

System is a Precision 690 with the onboard Broadcom 5752 NIC. The OS is Server 2012.

I've noticed that the onboard NIC refuses to sync to my switch at anything more than 100/Full. It loaded a built in MS driver but I've also loaded the Dell driver that's built from a Bradcom 10.43.0.0 driver. Neither of these drivers lets me sync at gigabit speeds (yes, the switch is gigabit). I've even checked in the driver for the speed/duplex settings and gigabit isn't an option.

I downloaded the newest Broadcom driver for the 5752 (version is 15.something) and with that driver, the NIC throws a code 10 and refuses to start. Any ideas whats up with this thing?

Not sure this will be what you might be looking for but it looks like Broadcom is taking a page from MS and their updates.

»www.broadcom.com/support ··· ktop.php

There is a separate driver for 2012 x32 and x64, later ver. number (16.x.x) Hope it helps.

Yeah that's what I did was download the Server 2012 specific file. It indicates version is 15.6.0.10. Interestingly for x64 anyway, the zip file is the same for all of them (assume the same file for x86 but I didn't look). The date on the driver is actually January something not the June 4th the site indicates. Maybe the zip file contains an outdated driver and is messed up or something. Is there any way to find a list of previous or beta drivers for them? Never mind, found the archive page, I'll give a previous version a try.

Try another cable?

So the switch is capable of gigabit, but how is the interface configured? If it's Auto try explicitly setting it to 1000/full.Then that tells me it's a driver problem; the driver simply doesn't support gigabit.

Yeah I've tried different cables and different ports on the switch. Switch is set for Auto but I have not tried forcing it.

I just tried the archived driver for Server 2012, its version 15.4.0.17 and still throws a code 10 on the NIC upon reboot. I decided to check the speed/duplex options anyway and it's possible the driver hasn't fully loaded so it won't all be there but it still tells me the only options are 10/100, no 1000/full.

I agree it sounds like a driver problem but why would both Microsoft and Dell/Broadcom limit a gigabit NIC to 10/100? I don't recall ever having this problem with any previous OS on this same hardware, and with the exception of ESXi briefly, it was always Windows.

Broadcom just sucks. Anytime we get a new server, we kill (don't use) the Broadcom onboards. We get Intel add in cards and use those.

Just for giggles, I tried forcing the switch to 1000/full. Guess what, it isn't an option there either. 10 full/half and 100 full/half or auto are my only options just like in the driver. I assumed it might be because the NIC has already told it those are all it can handle so I looked at both a port already sync'd at gigabit and an unused port and neither of them offer 1000/full as an option. Switch is a HP Procuve 1810g-24 and it's running 2.10 (latest) firmware and DOES sync at 1000/full on other ports just fine.

They're better than Realtek cards, which I think is the only other single port gigabit card I have around here (and it's currently in use somewhere). I have an Intel quad port Pro/1000 GT in it and those do sync at 1000/full. I just wanted to use the onboard Broadcom as management but I'd like it to be gigabit.

So I looked in Broadcom's FAQ and found this jewel that explains why there's no 1000/full option.

Q: Why is there no option to force the speed and duplex to 1000/full?
A: Per the IEEE standard, the only IEEE complaint way to operate at 1000/full is through auto-negotiation. If the network switch being used is IEEE complaint, the link speed will auto-negotiate to 1000/full when auto is selected.

OK so I scanned through the INF with Notepad++ and found several sections setting the available speed/duplex options and in some of them 1000/full is an option. Thing is the 5752 is an OEM chip and is "hidden" under an OEM device listing instead of listed as BCM5752 like you'd expect it to be so without figuring out which OEM device the 5752 is listed as, I can't determine which speed/duplex parameters it's using (and possibly change it).

So what this tells me is that even if I ping Broadcom on this issue, they are going to tell me that since it's set for auto and syncing at only 100/full that's all my NIC/switch/cable is good for? Sorry but I don't buy that answer when the switch port and cable work just fine at 1000/full on other devices and the NIC has previously worked just fine at 1000/full on other switches (and this one too I believe).

I say it has previously sync'd to 1000/full on this switch but I'm not sure. I didn't even pay attention to it until I went to do a file transfer and it was only going at 100 meg speeds. I might be able to get them to help me on the code 10 issue but if it's like the older drivers supplied by Dell/MS, it may not matter.

I downloaded the legacy driver for 2003 x64, version 14.8.0.5. I had to manually load it in as even rolled back to the Dell 10.x driver, it still tried to tell me the driver was already up to date. Here's the thing though, the advanced tab now has a speed/duplex selection for "1 gb full auto" but there's duplicate speed/duplex options (as well as duplicate flow control, no others are duplicated). I left one at "auto" and the other to "1 gb full auto" and it refused to link to the switch at all. Setting it all back to "auto" syncs it at 100/full only still.

OK the main thing I wanted to check was if the port you were using was hard-coded to be 100/Full or something.

We had a case where we upgraded a server where the old NIC could only support 10/100, but the new server was gigabit. Server was set to Auto, as you say, but the switch port was hard-coded for 100/Full that they forgot about. What we wound up with was a half duplex link that was getting all kinds of errors and collisions. Once the switchport was set to Auto, we had a 1000/Full link.

My thoughts...
1. Check the inside of the connector on the system for out of place pins.
2. Try a live CD to see if it negotiates at 1 gigabit.
3. Try a cheap-o gigabit switch.
4. If under warranty, make a claim.

I've has the same issues on a personal NAS.
First got it, would not link at 1 gigabit. Found out the drop had messed a pin bent out of place (If I recall, it was position 7,8 shorting).
Month later, would not link at 1 gigabit again. After a few rounds of trying various things, warranty claim.
Currently, running at 1 gigabit just fine.

Haven't had a chance to really get into this prior to now but I've made some progress in my testing.

I've checked for any anomalies inside the connector, there are none.
I've tried a live CD (UBCD booting Parted Magic) and other OSes (currently have pfsense 2.1 on there), the OS doesn't seem to make a difference as it's all still syncing at 100/full.
I do have an unmanaged Linksys switch that's primarily 10/100 but has a couple gigabit ports on it. I haven't yet tried that switch but I have my doubts about that making a difference (I do plan to give it a try though).
It's long out of warranty and it has an Intel quad port gigabit card in it too so it's not like I'm stuck with just this port.

At this point, unless the unmanaged switch somehow makes a difference, I think it's something to do with the actual hardware itself. Come to think of it, back in June I had a nearby lightning strike that caused a lot of other failures including ports in other computers connected to the same switch as this computer (the switch was toast as well). The other computer's NICs are full on dead though, not stuck in 10/100 but maybe this got partially damaged?

That is what I think happened to me with a NAS I did an RMA on. It had also partly taken out an ASUS N66U. Only one port worked.

Also took out a wall wort, 5 port gigabit switch, and power supply for something else. Few other things went bezerk like a ups was stuck in alarm mode until it was power cycled.

This surge must have come in through the cable plant as primarily only cable related or connected equipment was damaged.

Cisco cable modem etherner port dead.
PSU for Motorola cable box dead (box OK but swapped just in case).
My TV (cable box on this TV had the failed PSU)
Living room TV (cable box OK)
Den TV (cable box OK) - this one turns on but no inputs work.
HP 1810g-24 switch
A couple of switch ports on another switch that were connected to the 1810.
A couple of NICs on other computers that were connected to the 1810.

Yeah it's crazy. I completely forgot about the power surge/lightning as being a possibility. Nothing else was damaged, just stuff connected to the cable system or my network. I think it came in through the cable system and hitched a ride on my network before dissipating.

As you found out, GigE relies on autonegotiation and generally can't be "forced" into it per the IEEE specs.

My thoughts are as follows :

- confirm cabling is 5e or better, and can connect two GigE devices _AT_ GigE speeds.
- may have to confirm cabling with a high(er) end fluke meter and ensure no cable length/NEXT/FEXT issues and is meeting GigE specs.
- as suggested, try with another GigE switch and see if there is a change in behavior.
- RMA the NIC if possible... or get another PCIe GigE NIC to replace it.

My 00000010bits

Regards

Cables are 5e and link up at gigabit speeds on other devices (such as the quad port NIC in the computer in question, and to another computer). It's the embedded NIC so no chance of RMA unless I did the entire system and that's been out of warranty for a while.

I think the best solution at this time is to just assume the surge/lightning messed up that NIC in some way and not use it except for something that can handle 10/100 limits.