[Server-devel] [XSCE] Re: are NUC5i3RYH & NUC6i3SYH really limited to 12 WiFi clients??

Adam Holt holt at laptop.org
Wed Aug 31 17:38:55 EDT 2016


On Wed, Aug 31, 2016 at 5:16 PM, T Gillett <tgillett at gmail.com> wrote:

> Unfortunately it appears that in recent NUC versions (Skylake), Intel have
> changed from using PCIe cards for the wifi module to fixed modules on the
> motherboard thus preventing any changes to the wifi hardware.
>
>    http://arstechnica.com/gadgets/2016/03/review-much-improved
> -iris-gpu-makes-the-skylake-nuc-a-major-upgrade/
>

Tragic, I too noticed that.

Interestingly, Gigabyte BRIX devices are using the Intel wifi and bluetooth
> modules as M.2 cards.
>
>     http://www.gigabyte.com.au/products/list.aspx?s=47&ck=104
>

 Excellent information!

Curt Thompson's also recommended we look into the $130+ ZOTAC CI323 he's
had great luck with, or more to the point higher-end ZOTAC units that
accept Legit Internal WiFi modules as you describe?

(Almost anything that's not artificially capped to 12 WiFi devices, as
Intel's seems to be...)

Clarif: many of our deployments now demand that WiFi be fully contained
*inside* the school server / Internet-in-a-Box unit, keeping a lid on
endless theft/configuration/breakage/loss/sustainability hassles.


On Mon, Aug 29, 2016 at 8:19 AM, T Gillett <tgillett at gmail.com> wrote:
>
>> One more thought - have you looked at the dedicated Access Points that
>> are made by Intel to see what they are using for the wifi chipset and
>> drivers?
>>
>>              http://www.intel.com/content/www/us/en/education/products/co
>> ntent-access-point.html
>>
>> This may give you a clue as to a suitable wifi card for the NUC.
>>
>> T
>>
>> On Mon, Aug 29, 2016 at 8:13 AM, T Gillett <tgillett at gmail.com> wrote:
>>
>>> Hi Adam
>>>
>>> "Any other angles of attack??"
>>>
>>> I thought of testing the NUCs with a copy of OpenWrt just to see if the
>>> wifi would behave differently.
>>> This has been found to be the case with some types of router hardware
>>> (eg TP-Link) where the (closed) OEM firmware will not support large numbers
>>> of wifi connections, but the hardware works perfectly well using OpenWrt
>>> and its open source Linux wifi drivers.
>>>
>>> Unfortunately it seems that while running OpenWrt on x86 is well enough
>>> supported, the Intel wifi chipsets used in the NUCs are not so well
>>> supported. (There may be a message in this. Intel probably never intended
>>> the devices to be used as serious wifi Access Points)
>>>
>>> An alternative approach may be to change the wifi in the NUC for
>>> something for which there are Linux drivers available that are known to
>>> work as APs.
>>>
>>> I see that Intel offer a range of wifi cards for the NUC and it may be
>>> that one of their offerings will work correctly as an AP if set up with a
>>> suitable Linux driver.
>>>
>>> The wifi cards appear to be half size PCI-e cards, so an alternative may
>>> be to use a card with a different chipset, from a different supplier.
>>>
>>> I realise that changing the wifi card is not an ideal solution if the
>>> NUC comes with a card already installed, but it seems that you are getting
>>> to the bottom of the barrel in terms of finding a solution.
>>>
>>> Can you buy the NUCs without a wifi card fitted?
>>>
>>> Of course adding a 'real' wifi router on the network port of the NUC
>>> would give you a good solution from a wifi perspective (particularly as you
>>> can use better antennas), albeit at the cost of having two pieces of
>>> hardware to install and support.
>>>
>>> Regards
>>> Terry
>>>
>>> On Mon, Aug 29, 2016 at 1:26 AM, Adam Holt <holt at laptop.org> wrote:
>>>
>>>> On Sat, Aug 27, 2016 at 12:55 AM, Anish Mangal <anishmg at umich.edu>
>>>> wrote:
>>>>
>>>>> FWIW, on an older 2nd gen NUC I also see max 10-12 connections on the
>>>>> WiFi. I dont know if I have tried the latest kernel on that, so your
>>>>> finding is atleast consistent with older gen NUCs.
>>>>>
>>>>
>>>> FYI Kernel 4.4.14 (Fedora 22's latest upon yum update) is no better:
>>>> 12 simultaneous WiFi connections is still the maximum we're able to
>>>> sustain from random OS's to NUC6i3SYH's internal WiFi (*)
>>>>
>>>> Sadly the latest WiFi driver from Intel also does not help: (for 6i3's
>>>> internal WiFi "IntelĀ® Dual Band Wireless-AC 8260")
>>>> http://www.intel.com/content/www/us/en/support/network-and-i
>>>> -o/wireless-networking/000005511.html
>>>> https://wireless.wiki.kernel.org/_media/en/users/drivers/iwl
>>>> wifi-8000-ucode-25.30.13.0.tgz
>>>>
>>>> (*) Curious Anomaly: Android WiFi connections made within the "first 12
>>>> WiFi connections" can hold on to their DHCP/WiFi connections, remaining
>>>> active and usable as 13th and 14th connections etc, as other laptops
>>>> connect, *until they disconnect from WiFi*.  But when 12 laptops (or
>>>> iOS) have connected to WiFi, and one of these Android WiFi connections
>>>> happens to drop, it will not be able to reconnect to WiFi -- until one of
>>>> the 12 laptops (or iOS) disconnects.  Baffling that DHCP issues more than
>>>> 12 IP simultaneous IP addresses in these exceptional cases, when Android
>>>> has snuck in connecting early on.
>>>>
>>>> Other OS's (Windows, Linux, Mac, iOS) however do not demonstrate
>>>> Android's unusual/sneaky/resilient behavior.  In any case, this Android
>>>> curiosity (however tantalizing, no idea how Android's apparently able to
>>>> circumvent the "max 12" limit) does not solve the larger/general problem of
>>>> supporting all OS's!
>>>>
>>>> Any other angles of attack??
>>>>
>>>> Beyond that number of users, one should anyway consider a router.
>>>>>
>>>>> Off the shelf routers like 701nd support around 15 users, but with
>>>>> openwrt they have been reported to support upwards of 30.
>>>>>
>>>>> On Sat, Aug 27, 2016 at 4:56 AM, James Cameron <quozl at laptop.org>
>>>>> wrote:
>>>>>
>>>>>> Doesn't sound right.
>>>>>>
>>>>>> What happened to stop the test?
>>>>>>
>>>>>> Isn't the CentOS 7.2 kernel used in your test way older than the RPi3
>>>>>> kernel?
>>>>>>
>>>>>> It would appear that CentOS 7.2 released with kernel 3.10 dated 30
>>>>>> June 2013, with minimal changes patched into it since.  Perhaps it
>>>>>> needs another fix.
>>>>>>
>>>>>> RPi3 kernel with Rasbian is 4.4 dated 1 November 2015.
>>>>>>
>>>>>> Bisect the problem broadly.  Try the latest kernel.
>>>>>>
>>>>>> There have been many wireless driver and wireless networking changes
>>>>>> between the two kernels.  There's a possibility it may be one of them
>>>>>> you have hit.
>>>>>>
>>>>>> --
>>>>>> James Cameron
>>>>>> http://quozl.netrek.org/
>>>>>> _______________________________________________
>>>>>> Server-devel mailing list
>>>>>> Server-devel at lists.laptop.org
>>>>>> http://lists.laptop.org/listinfo/server-devel
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Anish
>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> Server-devel mailing list
>>>>> Server-devel at lists.laptop.org
>>>>> http://lists.laptop.org/listinfo/server-devel
>>>>>
>>>>> --
>>>>> <http://lists.laptop.org/listinfo/server-devel>
>>>>> Unsung Heroes of OLPC, interviewed live @
>>>>> <http://lists.laptop.org/listinfo/server-devel>http://unleashkids.org
>>>>> !
>>>>>
>>>>
>>>
>>
>


-- 
Unsung Heroes of OLPC, interviewed live @ http://unleashkids.org !
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.laptop.org/pipermail/server-devel/attachments/20160831/dd4f1d1b/attachment.html>


More information about the Server-devel mailing list