New Alix board for 2013

[jahonix]

@stephenw10:

@senser:

Looks like it fits into the old enclosures (the alix.2 ones).

This now seems unlikely, at least without some modification, since they are relying on the enclosure to cool the CPU.

Rumor has it that the newer version cases (with antenna holes) will fit the new boards.
There's gonna be a heat spreader in place which is said to be "a bit tricky to position correctly". Black cases will give better cooling results as compared to other colors.
FWIW

[stephenw10]

@jahonix:

Black cases will give better cooling results as compared to other colors.

Some time ago I was trying to convince a colleague of mine of this fact. Despite being a very experienced engineer he didn't believe me, I actually had to run a test to prove it.
Do they do it in absolute black?  ;)

Steve

[jahonix]

@stephenw10:

Do they do it in absolute black?  ;)

Absolutely    ;D

[dotdash]

How much more black could this be…

[stephenw10]

;D

[edmund]

@dotdash:

How much more black could this be…

The trick is to use the new black LEDs …

[marama]

Should I expect the VPN accelerators (Soekris vpn1411) to work in the new device? I would love to know what 3DES throughput to expect from this new device, with and without 3DES accel. If current Alix has some 32Mbit/s 3DES with vpn1441 accel, you think some 50-60 Mbit/s would be within reach?

[stephenw10]

That specific card won't be usable as the new Alix only has mini-PCIe slots so it won't fit.
I would expect the software throughput tot be greater than that card is capable of anyway.

Steve

[marama]

@stephenw10:

That specific card won't be usable as the new Alix only has mini-PCIe slots so it won't fit.
I would expect the software throughput tot be greater than that card is capable of anyway.

Oh, OK then.
I am in a bit of dilemma here… we will be upgrading from current ~15Mbs to 30 or more. Current 3DES throughput of Alix without the accel. is enough to fill the current bandwidth, but I have a constant 100% CPU usage. Once we upgrade to >30, Alix with accel would also be enough, but I'd have 100% CPU usage again. Is having constant ~100% usage something bad in the long run? The box does DHCP serving, DNS forwarding, NATing, firewalling... would those services get degraded because of near 100% CPU usage? If so, I will probably want to get some mini-ITX box, but I had good expirience with Alix devices so I'm hoping the new board would enough to cover the needs of the office using ~30 Mbit/s. Did any of you guys here get a demo board, maybe you can feed us with some numbers and benchmarks?
Once the new box comes out, do you expect to see some compatible VPN accelerators or you'd say we'll have to stick with software only encryption?

[watercooled]

The Jaguar-based processor planned for the release version supports the AES-NI instruction set, so provided it's supported in software, it would allow for accelerated throughput when using the AES cipher.

Either way, the new CPU cores are significantly faster than the current Geode CPU, so as stephenw10 said, it's likely the new ALIX will be faster in software than the current one is with accelerators.

[Applied]

@watercooled:

The Jaguar-based processor planned for the release version

The release version will be based on the the older Ontario core (T40E) without AES-NI support, not on Jaguar.

When they say, Jaguar's on their roadmap, I'd take this as "further down the road". This design should be a much smaller step up than from Alix to the APU - but it's no simple drop-in replacement either. I think AES-NI support in software won't be a big issue - but I wouldn't hold my breath for an "AES-NI-capable Alix" hardware release earlier than 2015.

The pending 1st generation APU should be a magnitude faster than the Alix though. I should alsol be faster on crypto, even without hardware support.

[stephenw10]

It would be very interesting to see some acceleration using the graphics cores that would presumably otherwise be unused. I see people talking about that possibility.

Steve

[watercooled]

I seem to remember reading (or possibly misreading) that the release version would use the Kyoto APU.

Of course, it becomes a one-chip solution with that so the board would need to be largely re-designed from the two-chip Brazos platform.

[Applied]

@watercooled:

I seem to remember reading (or possibly misreading) that the release version would use the Kyoto APU.

That would be the Opteron on the roadmap.

I was referring to Pascal Dorniers post on the PC Engines support forum:

"Production boards will change from T40N (9W TDP) to T40E (6W TDP)"

http://www.pcengines.info/forums/?page=post&id=6DBDDAB1-20E9-48CE-99A8-371F1C91D239&fid=DCB0643F-CE4D-4CAA-A3BA-72135A57B61D&cachecommand=bypass&pageindex=8

[watercooled]

Yeah, I'm just using the platform/market/core names interchangeably sorry. Worded another way, I thought the Jaguar part would be used on the release version, with the Bobcat platform only released in limited quantities.

Now I think of it though, that doesn't make a ton of sense.

Like I say I've probably just misread it somewhere. Either way, no biggie. :)

[nesense]

Hello gents, I got the APU board prototype yesterday with the recommended package:

apu1b board
case1d2blku enclosure (black gives best cooling)
ac12veur2 AC adapter (or ac12vus)
msata16a m-SATA SSD (optional)
wle200nx miniPCI wifi + 2 pigsma + 2 antsmadb (optional)

First of all I tried flashing the embedded NanoBSD image to a SanDisk 4GB SDHC card (2) and it failed to mount the filesystem after boot, I also tried the Memstick image on the sdcard and it failed too, I'm not sure why but from the documentation provided with the board it says that the sdcard reader is connected through USB on board.

Next thing I tried was flashing the embedded nanobsd image on a usb flash drive and that also failed, I ended up booting it successfully with flashing the memstick-serial image on the usb flash drive, and installing the OS on the m-SATA, since installing on SDCard also failed using this method (got incorrect block/geometry I think)

the provided WiFi kept giving me kernel panic, I tried mixed G+N mode and it crashed with auto channel selection, when I set it to channel 11 it didn't crash the kernel but the wifi card failed to start and the interface kept showing as DOWN, also 802.11g failed and crashed the kernel:

Fatal trap 12: page fault while in kernel mode
cpuid = 1; apic id = 01
fault virtual address  = 0xe
fault code              = supervisor read data, page not present
instruction pointer    = 0x20:0xffffffff802e830f
stack pointer          = 0x28:0xffffff802f059750
frame pointer          = 0x28:0xffffff8000299000
code segment            = base 0x0, limit 0xfffff, type 0x1b
                        = DPL 0, pres 1, long 1, def32 0, gran 1
processor eflags        = interrupt enabled, resume, IOPL = 0
current process        = 74391 (ifconfig)
trap number            = 12
panic: page fault
cpuid = 1
panic: bufwrite: buffer is not busy???
cpuid = 1

ath0: unable to reset hardware; hal status 3
aath0: ath_chan_set: unable to reset channel 11 (2462 MHz, flags 0x480), hal status 3
th0: ath_reset: unable to reset hardware; hal status 3

Fatal trap 12: page fault while in kernel mode
cpuid = 0; apic id = 00
fault virtual a

Fatal trap 12: page fault while in kernel mode
cpuid = 1; apic id = 01
fault virtual address  = 0xe
fault code              = supervisor read data, page not present
instruction pointer    = 0x20:0xffffffff802eaedd
stack pointer          = 0x28:0xffffff80000209f0
frame pointer          = 0x28:0xffffff8000299000
code segment            = base 0x0, limit 0xfffff, type 0x1b
                        = DPL 0, pres 1, long 1, def32 0, gran 1
processor eflags        = interrupt enabled, resume, IOPL = 0
current process        = 11 (swi4: clock)
trap number            = 12
panic: page fault
cpuid = 1
ddress  = 0xe
fault code              = supervisor read data, page not present
instruction pointer    = 0x20:0xffffffff802e830f
stack pointer          = 0x28:0xffffff802f02c750
frame pointer          = 0x28:0xffffff8000299000
code segment            = base 0x0, limit 0xfffff, type 0x1b
                        = DPL 0, pres 1, long 1, def32 0, gran 1
processor eflags        = interrupt enabled, resume, IOPL = 0
current process        = 45851 (ifconfig)
trap number            = 12

Packages managing the LEDs aren't working.

[athurdent]

@nesense:

Hello gents, I got the APU board prototype yesterday with the recommended package:

When you find the time it would be great if you could do some performance tests, like NAT / Routing performance with or without traffic shaping.

[nesense]

When you find the time it would be great if you could do some performance tests, like NAT / Routing performance with or without traffic shaping.

using iperf to test throughput, default values, traffic shaper DISABLED:

TCP window size:  129 KByte (default)
–----------------------------------------------------------
[  4] local 192.168.10.10 port 59068 connected with 192.168.10.1 port 5001
[ ID] Interval      Transfer    Bandwidth
[  4]  0.0-10.0 sec  388 MBytes  326 Mbits/sec

Traffic shaper ENABLED using HFSC on 2 LAN and 1 WAN (voip + all p2p protocols + all network games + some other applications):

–----------------------------------------------------------
Client connecting to 192.168.10.1, TCP port 5001
TCP window size:  129 KByte (default)

[  4] local 192.168.10.10 port 59150 connected with 192.168.10.1 port 5001
[ ID] Interval      Transfer    Bandwidth
[  4]  0.0-10.0 sec  336 MBytes  282 Mbits/sec

will test using netio next.

[stephenw10]

How are you testing that? A throughput test is what's needed, iperf running on two separate machine not on the pfSense box.
326Mbps seems disappointingly slow.  :-\

Steve

[nesense]

@stephenw10:

How are you testing that? A throughput test is what's needed, iperf running on two separate machine not on the pfSense box.
326Mbps seems disappointingly slow.  :-\

Steve

i'm running iperf server on pfsense and client directly connected to it through 1gbit ethernet macbook port