D945GCLF2 Board, Atom330



  • Just built a new box with the Atom330, D945GCLF2 Board, and a Apex MI-100 Case. Installed 1.2.3 RC and so far so good, I thought i was going to have problems with the Realtek onboard NIC. It was detected and everything seems to be working so far in testing. Now I just have to see how it runs under a lot more load.



  • I'm using the same board and the onboard NIC has given me zero problems. The onboard goes to my gigabit switch and the Intel PRO/100+ Dual Port NIC in the PCI slot goes to 2 DSL modems for load balancing. Very solid setup with zero problems.



  • Hmmm… Good to hear I wont have problems with this setup. :)



  • Using the same board in a production environment with 30+ clients without issues as well.

    Setup:
    D945GCLF2
    2GB Kingston DDR2 ValueRam
    Intel Pro/100 adapter (WAN)
    Onboard Realtek (LAN)
    Some old Maxtor 15GB PATA drive (on UDMA100)

    It has actually been working since PFsense 1.2.2 but with a major quirk:
    In 1.2.2, PFsense has issues mounting SATA drives (gets stuck trying to mount the partition) whether I set the host controller to Native/ Legacy mode.
    Using a PATA drive works 1/2 the time but only if I set the controller to Legacy mode.  The other half of the time, it gets stuck over and over again; takes about 15 reboots to get it to boot up into PFsense once.

    I gave up and went over to using IPcop 1.22.  With advanced QoS, it did the job but performance was significantly lacking compared to PFsense.  Latencies were quite a bit higher (20-30ms increase on 1/2 the packets).  However, it didn't lock-up like with the Linksys 350N running DD-WRT I tried previously.

    I've just re-deployed PFsense 1.2.3 RC2 on the same machine and now everything works like a charm.  SATA/ PATA drives both work regardless of the controller mode.  The machine consistently boots into PFsense.
    Note that these are the exact same drives I used for 1.2.2 so the failure to mount the drives isn't a hardware fault.



  • Sounds like something a bios update might possibly resolve



  • I had grief with this board.  Seemed pfsense was unresponsive after a few hours.  After much playing around with it (tested on windows install), I came to the conclusion that it was the onboard realtek network that was actually the problem.  I installed an intel network card and the problem went away in windows, and pfsense.  Now it's running and have not touched it in months.

    My conclusion.. 
    Pfsense good, realtek bad….

    Probably should have RMA'd the board...



  • @databeestje:

    Sounds like something a bios update might possibly resolve

    BIOS has already been updated.  I think it's more to do with the FreeBSD version (or the drivers included for that matter) since the version of 1.2.2 I used was still based on 6.x, IIRC.  In any case at all, 1.2.3RC2 July 30th snapshot works just fine now.



  • @tester_02:

    I had grief with this board.  Seemed pfsense was unresponsive after a few hours.  After much playing around with it (tested on windows install), I came to the conclusion that it was the onboard realtek network that was actually the problem.  I installed an intel network card and the problem went away in windows, and pfsense.  Now it's running and have not touched it in months.
     
      My conclusion.. 
    Pfsense good, realtek bad….

    Probably should have RMA'd the board...

    Looks like I spoke too soon.  Yeah, I just encountered this problem (existing connections work fine but the box rejects new connections).  I went to console and restarted the web configurator and that seems to solve the problem.  Will be looking into buying an Intel Pro MT dual-gbe adapter soon.  Current using the onboard for LAN and an Intel Pro/100 for WAN.



  • @dreamslacker:

    Looks like I spoke too soon.  Yeah, I just encountered this problem (existing connections work fine but the box rejects new connections).

    That seems to be different, sounds like state exhaustion, check your RRD graph to see if you're near/at the limit (10,000 default). You can increase it under System -> Advanced.



  • @cmb:

    That seems to be different, sounds like state exhaustion, check your RRD graph to see if you're near/at the limit (10,000 default). You can increase it under System -> Advanced.

    I have my state table limit set at 256,000 since day 1.
    From experience with PFsense at home (I thoroughly abused the box before deploying it in a production environment), I understand that this phenomenon occurs under some circumstances:

    • State Table Limit is reached. (Tried torrenting to the tune of 120,000+ connections from 2 computers)

    • Connection bandwidth is saturated and causes backlogging of packets.

    • Hardware just died; experienced HDD crash twice last year.

    The thing is, when the problem occurred, I had only 1 client online and bandwidth usage was approximately only about 3% of what my static line delivers.  I've never encountered more than 13,000 states even with 36 clients online at the same time.
    The QoS does it job and I have both U/L and D/L capped at 90% of the rated line speed along with the usual bandwidth hogs further limited to 85% of that 90% so I know it's not because of excessive bandwidth usage.

    Going to the actual PFsense box and using the console to restart the Web-Configurator did the job for me though.  Either that or it was a freak coincidence that the box went out of lock-out just as I restarted the Web-Con.



  • Hmmm… So there is problems with this board then still? I didn't experience it in testing but...
    I updated the bios, and am running 1.2.3, so I guess I will cross my fingers and see. This is going to be a pain if the realtek starts doing as you say.



  • @cwilkinson:

    Hmmm… So there is problems with this board then still? I didn't experience it in testing but... Damn I just deployed one, I guess I will see...

    It doesn't seem to occur very often if at all.  I've had the box online for 2½ weeks in production (with 1.2.3) and only encountered this problem yesterday.  IMO, it could be a problem with the onboard Realtek NIC.  I'm not exactly fond of Realtek chipsets but I had to make do with it.
    I already had hell trying to convince my partner in charge of accounts that a dedicated box was needed over the WRT-350N he had lying around.
    Needless to say, he only budged after the WRT-350N hung within 5 minutes of deployment and we were bleeding 3 times the cost of the Pfsense box everyday for 3 days from the downtime.  Now, I'm just gonna order a Pro/1000 MT dual port adapter off ebay and charge it directly to accounts.  =P



  • lol ya well the reason I went with Pfsense for this client is because all the consumer dlink and linksys products were locking up and he had to keep rebooting them, don't want this to happen again with this system…
    Wonder if I could put a Trendnet USB NIC in to replace the onboard if this occurs... Don't know if pfsense would pick it up though.



  • @cwilkinson:

    lol ya well the reason I went with Pfsense for this client is because all the consumer dlink and linksys products were locking up and he had to keep rebooting them, don't want this to happen again with this system…
    Wonder if I could put a Trendnet USB NIC in to replace the onboard if this occurs... Don't know if pfsense would pick it up though.

    As long as the state tables and QoS is set-up properly, you shouldn't have any issues with the system locking up (hardware issues aside, of course).
    Might want to pick-up an Intel dual-port PCI(X) adapter off ebay though.  They're a lot cheaper than from most online retail stores like Newegg.  I have a lower regard for USB ethernet adapters than even Realtek NICs.

    Anyway, when I had the box on PFsense 1.2.x previously, it was stable for more than 80 days.  A power outage killed the uptime and of course, as I mentioned before, it took too many reboots to get the hdd to mount then.



  • All I can say is that the problem I had was not pfsense related, as it appeared also in windows.  It would last 1+ days then the the box would stop responding (but if I went to the physical box all appeared fine except no network).  I at first tried updating the bios (thought that made the most sense), but it still caused me problems.  I was alsmost going to RMA and trash it, but I put another nic (intel) and all the problems went away.  It was not worth the both to RMA at that point, and it just simply works.
      Based on they have sold many of these boards, and most people are working ok, I blame mine as an isolated hardware defect.
      Good board otherwise.  cheap and normally works….



  • Would be nice if Realtek just quit making NIC cards period!!  But given the low price point over Intel they rather go cheap with it.  So I always keep spare Intel NICs around and put em in routers.  I haven't had any problems with 4 port 100Mb D-Link server card which I also got off of ebay on the cheap.  It works so well that I picked up another card as a spare, also off of ebay.



  • @Darkk:

    Would be nice if Realtek just quit making NIC cards period!!  But given the low price point over Intel they rather go cheap with it.  So I always keep spare Intel NICs around and put em in routers.  I haven't had any problems with 4 port 100Mb D-Link server card which I also got off of ebay on the cheap.   It works so well that I picked up another card as a spare, also off of ebay.

    I do know some guys working for motherboard manufacturers and it isn't just price point.  Realtek has managed to make their chipsets very easy to integrate on motherboards.  Comparatively, one of them at least, has had to send development motherboards (and) trace designs to Marvell's German engineering to get goofball issues resolved.  Implementing Broadcom wasn't even possible and Intel chipsets were about 4 to 5 times the price.



  • If it was not for Realtek and Ralink making cheap wired and wireless for the masses the internet as we know it wouldn't exist.

    Or atleast it would make adoption mildly prohibative.

    One thing Realtek and Ralink have well under control is working with partners, if that's making a network card, drivers or a mainboard.

    They also talk to opensource developers in general without jumping through nda hoopla.



  • @databeestje:

    If it was not for Realtek and Ralink making cheap wired and wireless for the masses the internet as we know it wouldn't exist.

    Or atleast it would make adoption mildly prohibative.

    One thing Realtek and Ralink have well under control is working with partners, if that's making a network card, drivers or a mainboard.

    They also talk to opensource developers in general without jumping through nda hoopla.

    If what you say is true then why so many issues with the Nix / FreeBSD flavors?  Heck, even some Window users experience issues with them.

    I don't think hardware is the issue, it's the bad drivers or poorly implmented into an OS.  Let's hope RealTek keeps trying to improve things.



  • @Darkk:

    I don't think hardware is the issue, it's the bad drivers or poorly implmented into an OS.   Let's hope RealTek keeps trying to improve things.

    Hardware bugs that vary from one implementation to another are the problem. There are countless work arounds that keep getting added to the Realtek drivers in FreeBSD (and no doubt other OSes as well, I don't follow others as closely though) as people keep finding hardware that's buggy in new and different ways from other implementations of the exact same chipset.



  • @cmb:

    Hardware bugs that vary from one implementation to another are the problem. There are countless work arounds that keep getting added to the Realtek drivers in FreeBSD (and no doubt other OSes as well, I don't follow others as closely though) as people keep finding hardware that's buggy in new and different ways from other implementations of the exact same chipset.

    Aye aye.  Some of the issues are due to implementation of "features" in new OSes that don't play well with the firmware on the NIC's too.
    eg.  The problem my friend faced with the Yukon was that 1/2 the chips don't work well in Vista without a new driver patch because the chip wouldn't respond properly to the PCI-e Link Power Management.
    Oddly, the same driver paired with an external card on the same chipset works fine.  It's just the integrated chipset that won't play well.  There are no issues with WinXP or other OS flavours either.  The only solution was to manually disable the Link power state management in Vista Device Manager or update to a "newer" driver provided by the manufacturer (which does exactly just that).



  • Well, I've had the box up and running in production for about 3 days so far and no problems, yet.



  • System has been running for 5 days without a hiccup now.
    Uptime is 9 days though (since the last hiccup resolved without a reboot).

    It could have been a backend issue on my ISP side but I never got a notification from them.  Time to cause them a little more grief I guess.
    Had already made them replace the CPE (Cisco 1710 to Cisco 877) just because I didn't like the fact that they gave me a end-of-line and end-of-shelflife CPE.  Just told them that the 1710 was flaky and churned out some "evidence" of the hop between my router and the CPE taking as long as 100ms.
    The engineer who came down had a nice time though.  I let him play World of Warcraft for 2 hours (I run a CyberCafe) and told the n.o.c. that he was running tests for me during that time.

    Only thing I couldn't figure out is why they actually deployed the connection the way they did it.  All the other ISP's provide an IPoA connection and the CPE is a Bridge/ Gateway combo.  Thus, allowing me to deploy my own bridge cum gateway if I don't like their equipment.  Or for that matter, actually modifying their CPE for my uses; did it before by flashing a ST516 and converting it into bridged mode (I told them and they said it was ok as long as I return the modem in good working condition).

    This ISP actually has the CPE as a BGP router so I can technically access their "internal" network.  They conveniently gave me the cisco console cable too.  Or more so, the engineer was too lazy to lug the box of accessories back.  He didn't have an issue when I told him I could re-configure the 877 so long as I didn't lock the n.o.c. out from the 877.



  • I just broke 90 days uptime with pfsense running on this board. Sadly, I'm going to have to take the router down for maintenance in the next few days as I'm having a dedicated 100 amp circuit run to my office. Why? Because I'm running running 14 computers on 1 20amp circuit which isn't really smart. ;) Hehe.


Log in to reply