The pfSense Store

Author Topic: Watchguard XTM 5 Series  (Read 125290 times)

0 Members and 1 Guest are viewing this topic.

Offline 747Builder

  • Full Member
  • ***
  • Posts: 116
  • Karma: +1/-0
    • View Profile
Re: Watchguard XTM 5 Series
« Reply #555 on: June 30, 2017, 02:18:22 am »
Some Xeons will run in it either the socket 775 models or the socket 771 with the pin-mod. Not sure they're any faster than the C2Q though. I've not personally tried but there are reports in this thread.

Steve

the 771 Pin moddded Xeons are cheaper than the equivalent Core 2 Quad and the Xeons have a bigger L2 and L3 caches which theoretically makes them faster. but just like the C2Q's you have to watch the TDP values on them too.

Offline DeLorean

  • Full Member
  • ***
  • Posts: 128
  • Karma: +18/-0
    • View Profile
Re: Watchguard XTM 5 Series
« Reply #556 on: June 30, 2017, 02:20:15 am »
Delorean,

pin 3 isnt connected between the SPI header and the chip on mine but the *SHIELDED* cable i used to connect the adapter on my willems programmer works great and can read/write the flash chip just fine from the SPI header.

Thanks for the update.
In the meanwhile i have finally tried the BIOS update with flashrom trough serial console,
and succesfully updated 5 of these red boxes  :)
But now i'm sure that your programmer works, and this is my lifeline if anything goes wrong.
I have now flashed all XTM 5 boxes here in the house, and i'm safe for the moment.
Thanks for taking the time to investigate this. ;-)

ps : can you give a link of the shielded cable you used ?

Grtz
DeLorean

Offline 747Builder

  • Full Member
  • ***
  • Posts: 116
  • Karma: +1/-0
    • View Profile
Re: Watchguard XTM 5 Series
« Reply #557 on: June 30, 2017, 03:20:25 am »
Delorean,

pin 3 isnt connected between the SPI header and the chip on mine but the *SHIELDED* cable i used to connect the adapter on my willems programmer works great and can read/write the flash chip just fine from the SPI header.

Thanks for the update.
In the meanwhile i have finally tried the BIOS update with flashrom trough serial console,
and succesfully updated 5 of these red boxes  :)
But now i'm sure that your programmer works, and this is my lifeline if anything goes wrong.
I have now flashed all XTM 5 boxes here in the house, and i'm safe for the moment.
Thanks for taking the time to investigate this. ;-)

ps : can you give a link of the shielded cable you used ?

Grtz
DeLorean

i just searched for a shielded internal usb 2.0 cable on ebay. mines 18" long and grabbed the first one i seen dirt cheap. make sure you get the 9 pin or 10 pin cable, theres some 5 pin cables as well that wont work..*I* did have to dig out the block in one of the pin to plug into the adapter board side. just have to make sure you get both sides plugged in correctly so pin 1 is pin 1 on both sides.


my next step is updating the microcode in the Bios

Offline dlucas46

  • Newbie
  • *
  • Posts: 14
  • Karma: +4/-0
    • View Profile
Re: Watchguard XTM 5 Series
« Reply #558 on: July 14, 2017, 06:09:45 pm »
Those of you that are running Xeon processors have you had any problems with the temperature sensors?

I have just done my first 771 to 775 mod on one of these watchguards and want to confirm if I am seeing an error or usual behaviour since the microcode in the BIOS might be wrong.

I installed a L5420 and the dashboard widget is reporting over 70 Celsius but the BIOS is reporting about 40 Celsius.

I am using the Intel Core temp sensor option in settings and not None / ACPI.

Anybody else seen this behaviour before?  Which temperature is correct?

Should I redo the thermal paste?  I did not use Arctic silver but cheap gold stuff, as I had run out of the good stuff.  Since the two are reporting different temps I would think that the paste might be OK.

Thanks

Offline stephenw10

  • Administrator
  • Hero Member
  • *****
  • Posts: 11884
  • Karma: +461/-15
    • View Profile
Re: Watchguard XTM 5 Series
« Reply #559 on: July 17, 2017, 06:54:59 am »
It could very likely be incorrect.

The coretemp driver reads a delta value from the CPU sensor and removes that from a known maximum value for the CPU. But since you've modified the CPU to pretend to be something else it's probably using the wrong values there. No way to fix that in software that I know of. Check your Tjmax value against the documented value for that CPU:

Code: [Select]
[2.4.0-BETA][admin@xtm5.stevew.lan]/root: sysctl dev.cpu.0.coretemp
dev.cpu.0.coretemp.throttle_log: 0
dev.cpu.0.coretemp.tjmax: 85.0C
dev.cpu.0.coretemp.resolution: 1
dev.cpu.0.coretemp.delta: 53

Steve


Offline dlucas46

  • Newbie
  • *
  • Posts: 14
  • Karma: +4/-0
    • View Profile
Re: Watchguard XTM 5 Series
« Reply #560 on: July 17, 2017, 11:52:57 am »
Hi Steve,

I guessed Coretemp was the issue. 
The Tjmax for a L5420 is 70c but coretemp sets the processor to 105c.

As far as I can see this is because the code contains an else if clause for the core2duo but the Xeon has the same family model Id (0x17) and the same stepping (6) so the code executes and sets the wrong value.
I have modified the coretemp.c file on a 10.3 BSD and rebuilt it to set a tjmax of 70 instead.

Have just loaded it up and now have dashboard and widget reporting 40c instead of 75c.

I will do some testing and then post it here.



Offline DeLorean

  • Full Member
  • ***
  • Posts: 128
  • Karma: +18/-0
    • View Profile
Re: Watchguard XTM 5 Series
« Reply #561 on: July 17, 2017, 12:05:14 pm »
It could very likely be incorrect.

The coretemp driver reads a delta value from the CPU sensor and removes that from a known maximum value for the CPU. But since you've modified the CPU to pretend to be something else it's probably using the wrong values there. No way to fix that in software that I know of. Check your Tjmax value against the documented value for that CPU:

Code: [Select]
[2.4.0-BETA][admin@xtm5.stevew.lan]/root: sysctl dev.cpu.0.coretemp
dev.cpu.0.coretemp.throttle_log: 0
dev.cpu.0.coretemp.tjmax: 85.0C
dev.cpu.0.coretemp.resolution: 1
dev.cpu.0.coretemp.delta: 53

Steve

On YouTube there is a vid about LGA775 to LGA771 mod ,
and this vid explains the use of a tool for modifying / adding cpu instructionsets to the BIOS rom file,
can this be usefull for a XTM5 or not ?

This is the link of the vid :
https://www.youtube.com/watch?v=HUqBW-kq1_0


Grtz
DeLorean

Offline stephenw10

  • Administrator
  • Hero Member
  • *****
  • Posts: 11884
  • Karma: +461/-15
    • View Profile
Re: Watchguard XTM 5 Series
« Reply #562 on: July 17, 2017, 12:55:54 pm »
Mmm, that would probably work. Not sure I have the tools I used anymore. Definitely don't have the WinXP system I did all that on.  ;)

Be ready to recover from bricking though if you try it!

Steve

Offline 747Builder

  • Full Member
  • ***
  • Posts: 116
  • Karma: +1/-0
    • View Profile
Re: Watchguard XTM 5 Series
« Reply #563 on: July 17, 2017, 04:23:42 pm »
It could very likely be incorrect.

The coretemp driver reads a delta value from the CPU sensor and removes that from a known maximum value for the CPU. But since you've modified the CPU to pretend to be something else it's probably using the wrong values there. No way to fix that in software that I know of. Check your Tjmax value against the documented value for that CPU:

Code: [Select]
[2.4.0-BETA][admin@xtm5.stevew.lan]/root: sysctl dev.cpu.0.coretemp
dev.cpu.0.coretemp.throttle_log: 0
dev.cpu.0.coretemp.tjmax: 85.0C
dev.cpu.0.coretemp.resolution: 1
dev.cpu.0.coretemp.delta: 53

Steve

On YouTube there is a vid about LGA775 to LGA771 mod ,
and this vid explains the use of a tool for modifying / adding cpu instructionsets to the BIOS rom file,
can this be usefull for a XTM5 or not ?

This is the link of the vid :
https://www.youtube.com/watch?v=HUqBW-kq1_0


Grtz
DeLorean

Delorean,
what you linked to was updating the Microcode that creates workarounds for known hardware bugs in the CPU. i really doubt updating the microcode will fix the coretemp utility to display the correct value. i havent looked at coretemp code but i suspect since it doesnt dig deep enough to figure out the correct CPU and the true fix needs to come from the coretemp code and not the BIOS.

we had a number of servers deployed in the past (which are now retired and sitting in the pile to be sent to the recycler) with the L5420 CPU and as long as you have a decent heatsink and good paste on them , they *DONT* get hot nor hit >70c even under load.


Offline wubs

  • Newbie
  • *
  • Posts: 7
  • Karma: +0/-0
    • View Profile
Re: Watchguard XTM 5 Series
« Reply #564 on: July 19, 2017, 11:03:53 am »
I hope this isn't too off topic, but I've got an XTM 525 and I can only get WGXepc64 to control the case fan. Has anyone in this thread had any luck with controlling the 2 CPU and/or PSU fans? My only thought is to swap one of the CPU fans to run on the case fan connector and remove the case fan entirely, but then I'm still stuck with one fan at 100% all the time for the CPU cooler and also the one for the PSU. I also don't know if there is a way to specify specific fans in WGXepc64, but the default at least seems to be the case fan. I really just want to quiet this box down so it can hang out in my office without sounding like an airplane is taking off constantly.

Offline stephenw10

  • Administrator
  • Hero Member
  • *****
  • Posts: 11884
  • Karma: +461/-15
    • View Profile
Re: Watchguard XTM 5 Series
« Reply #565 on: July 19, 2017, 12:29:03 pm »
For some reason the CPU fans are controller by a separate chip that is connected only via SMBus. Despite the LPC controller having multiple FAN outputs.

WGXepc only has access to the LPC controller so cannot set the CPU fan speed, in the default fan setup at least.

https://forum.pfsense.org/index.php?topic=43574.msg717350#msg717350

Steve

Offline wubs

  • Newbie
  • *
  • Posts: 7
  • Karma: +0/-0
    • View Profile
Re: Watchguard XTM 5 Series
« Reply #566 on: July 19, 2017, 01:00:25 pm »
For some reason the CPU fans are controller by a separate chip that is connected only via SMBus. Despite the LPC controller having multiple FAN outputs.

WGXepc only has access to the LPC controller so cannot set the CPU fan speed, in the default fan setup at least.

https://forum.pfsense.org/index.php?topic=43574.msg717350#msg717350

Steve

I wonder if it would be possible to get around this with a hardware fan controller, perhaps controlled by USB or something, in front of these fans that would be supported by pfSense. I know most fans wouldn't support a simple potentiometer in front of them since there is feedback on fan speed provided to the system. I'd just worry about overvoltage in that scenario.

I guess that's just a ramble. I am going to see if I can control it with something like what's available here: https://www.newegg.com/Controller-Panels/SubCategory/ID-11

Edit:  I'm thinking of ordering this to bypass the system entirely. https://www.newegg.com/Product/Product.aspx?item=9SIACJF5S63669
« Last Edit: July 19, 2017, 01:05:11 pm by wubs »

Offline stephenw10

  • Administrator
  • Hero Member
  • *****
  • Posts: 11884
  • Karma: +461/-15
    • View Profile
Re: Watchguard XTM 5 Series
« Reply #567 on: July 19, 2017, 01:04:28 pm »
Pretty sure they are 3 pin fans on the XTM5 (mine is buried right now!) so no danger of overvolting etc. The third pin only provides speed output. If you disconnect it and the controller is set to adjust fan speed it might ramp up tomax but that's 12V which is fine.

I'm sure you could use a hardware controller. Personally I would just get some quieter fans if you need it.

Steve

Offline wubs

  • Newbie
  • *
  • Posts: 7
  • Karma: +0/-0
    • View Profile
Re: Watchguard XTM 5 Series
« Reply #568 on: July 19, 2017, 01:07:55 pm »

I'm sure you could use a hardware controller. Personally I would just get some quieter fans if you need it.

Steve

Do you have a recommendation for quieter 1U fans? I looked at Noctua, but they're crazy expensive.

Offline dlucas46

  • Newbie
  • *
  • Posts: 14
  • Karma: +4/-0
    • View Profile
Re: Watchguard XTM 5 Series
« Reply #569 on: July 19, 2017, 01:09:19 pm »
For some reason the CPU fans are controller by a separate chip that is connected only via SMBus. Despite the LPC controller having multiple FAN outputs.

WGXepc only has access to the LPC controller so cannot set the CPU fan speed, in the default fan setup at least.

https://forum.pfsense.org/index.php?topic=43574.msg717350#msg717350

Steve

They are indeed 3 pin fans. 

Personally I think the XTM 5 is quiet compared to the 750e. 

The fans run at about 6000 rpm on my unit, you could put a resistor on the 12v line and reduce the voltage.  Or you could simply buy quieter fans.