Re: Quick APU2 review

[Chris Cappuccio]

Stuart Henderson [s...@spacehopper.org] wrote:
> >>> Heat: The APUs have an innovative design where the CPU heat sink
> >>> is coupled to the case.  Since this is typically assembled by the
> >>> customer, a lot of attention is drawn to it and people obsess over
> >>> the CPU temperature.  It's a nonissue.
> 
> I've got hold of an APU2C2 now. Agreed - but most of the temperature
> comments where about APU1 where this was not the case, especially with
> the version of OpenBSD that was around when the hardware first became
> available they really were uncomfortably warm. With the 2C2 I'm now
> happy to use it pretty much anywhere I'd use an ALIX.

The APU2 case is slightly cooler. 

My experience is that the heatsink pad failed on some units, it became almost
brittle.

The affected units would crash every few days...

I took apart some 20 initially assembled units with the heatsink pads, and
found some pads brittle, others were perfect.

I just replaced the heatsink pads on all of my APUs (probably 50 by now) with
heatsink paste and they all work perfectly, no crashing ever. Some serve DNS
and DHCP to networks with tens of thousands of devices, some act as redundant
routers, some just collect information via GPIO pins.

The units with low temps and heatsink pads got hotter with the paste, but the
units with bad heatsink pads got cooler with the paste.

It may seem like obsession, but I just want them to be stable :)

The APU1 is fine even in a high temp environment as long as the heatsink
material is working properly.

(PC Engines does not endorse replacing the heatsink pad with paste.)

Chris

Re: Quick APU2 review

[Stuart Henderson]

>>> Heat: The APUs have an innovative design where the CPU heat sink
>>> is coupled to the case.  Since this is typically assembled by the
>>> customer, a lot of attention is drawn to it and people obsess over
>>> the CPU temperature.  It's a nonissue.

I've got hold of an APU2C2 now. Agreed - but most of the temperature
comments where about APU1 where this was not the case, especially with
the version of OpenBSD that was around when the hardware first became
available they really were uncomfortably warm. With the 2C2 I'm now
happy to use it pretty much anywhere I'd use an ALIX.

Re: Quick APU2 review

[sven falempin]

On Wed, Apr 20, 2016 at 4:32 AM, Stuart Henderson 
wrote:

> On 2016-04-19, sven falempin  wrote:
> >> This : https://github.com/jasperla/openbsd-wip ??
>
> Yes
>
> > 5   ftp http://download.flashrom.org/releases/flashrom-0.9.9.tar.bz2
>
> No, don't do it this way. Use the port in openbsd-wip.
>
> Or if you aren't completely comfortable with this, just use the
> vendor's tool.
>
> http://pcengines.ch/howto.htm#bios
> http://pcengines.ch/howto.htm#TinyCoreLinux
>
>
> On 2016-04-19, Chris Cappuccio  wrote:
> >> Like : echo /bin/ksh > /etc/rc.securelevel ??
> >
> > echo sysctl kern.securelevel=-1 >/etc/rc.securelevel
>
> Yes
>
> >> No patch here : wont work ?
>
> The patch is in the flashrom directory in openbsd-wip.
>
> > Get the securelevel right first, then worry about the patch
>
> I would do it the other way round. Only run with securelevel=-1
> for the absolute minimum time needed, just when you're trying to
> actually update the bios. Don't go building ports (especially
> ones that aren't even committed) like that.
>
> Thank you all,

yes i used tiny core after.

I saw than on 5.8 and 5.9 -stable SDCARD is not working
i can boot but after there is a driver error (5.9 go a bit further)
it s the "AMD Hudson-2" who handle it apparently.

If the commit exists in -current can you point it out so i can backport it
in 5.9-stable ?

(i prefer beeing on stable, and 6 month is a bit long).

-- 
-
() ascii ribbon campaign - against html e-mail
/\

Re: Quick APU2 review

[Stuart Henderson]

On 2016-04-19, sven falempin  wrote:
>> This : https://github.com/jasperla/openbsd-wip ??

Yes

> 5   ftp http://download.flashrom.org/releases/flashrom-0.9.9.tar.bz2

No, don't do it this way. Use the port in openbsd-wip.

Or if you aren't completely comfortable with this, just use the
vendor's tool.

http://pcengines.ch/howto.htm#bios
http://pcengines.ch/howto.htm#TinyCoreLinux


On 2016-04-19, Chris Cappuccio  wrote:
>> Like : echo /bin/ksh > /etc/rc.securelevel ??
>
> echo sysctl kern.securelevel=-1 >/etc/rc.securelevel

Yes

>> No patch here : wont work ?

The patch is in the flashrom directory in openbsd-wip.

> Get the securelevel right first, then worry about the patch

I would do it the other way round. Only run with securelevel=-1
for the absolute minimum time needed, just when you're trying to
actually update the bios. Don't go building ports (especially
ones that aren't even committed) like that.

Re: Quick APU2 review

[Chris Cappuccio]

sven falempin [sven.falem...@gmail.com] wrote:
>
> This : https://github.com/jasperla/openbsd-wip ??
> Like : echo /bin/ksh > /etc/rc.securelevel ??
> 

echo sysctl kern.securelevel=-1 >/etc/rc.securelevel

> 
> No patch here : wont work ?
>

Get the securelevel right first, then worry about the patch

Re: Quick APU2 review

[sven falempin]

On Tue, Apr 19, 2016 at 1:06 PM, Stuart Henderson 
wrote:

> On 2016/04/19 12:33, sven falempin wrote:
> >
> >
> > On Mon, Apr 18, 2016 at 6:07 PM, Stuart Henderson 
> > wrote:
> >
> > On 2016-04-18, Christian Weisgerber  wrote:
> > > On 2016-04-18, Stuart Henderson  wrote:
> > >
> > >> From a different machine though. Compared to APU1 the APU2 has
> > 4x the L2
> > >> cache, RAM is clocked a quarter faster, twice the number of cpu
> > cores,
> > >> and a few more cpu features (e.g. AES-NI, RDRAND).
> > >
> > > For the record: The APU2 does not have RDRAND.
> >
> > Ah, I thought I saw a page suggesting that the cpu did, sorry for
> > misinformation.
> > O
> >
> >
> >
> > Is it really possible to flash the BIOS from openBSD ?
>
> A few people reported that flashrom in openbsd-wip works. You'll need to
> patch pciutils (see pciutils.diff in the openbsd-wip dir) and boot with
> securelevel temporarily set to -1 to use it.
>
>
This : https://github.com/jasperla/openbsd-wip ??
Like : echo /bin/ksh > /etc/rc.securelevel ??


No patch here : wont work ?
5   ftp http://download.flashrom.org/releases/flashrom-0.9.9.tar.bz2
6   bunzip2 ./flashrom-0.9.9.tar.bz2
19  grep -A 12 OpenBSD ./README
20  sudo pkg_add gmake
21  sudo pkg_add pciutils
23  sudo pkg_add libusb-compat
24  gmake
25  find ./ -name flashrom
$ file flashrom
flashrom: ELF 64-bit LSB shared object, x86-64, version 1
$ ldd ./flashrom
./flashrom:
StartEnd  Type Open Ref GrpRef Name
1d198350 1d19839b7000 exe  10   0  ./flashrom
1d1c3a10f000 1d1c3a524000 rlib 01   0
 /usr/lib/libz.so.5.0
1d1bcd0fe000 1d1bcd503000 rlib 01   0
 /usr/local/lib/libusb.so.10.0
1d1c15b35000 1d1c15f4 rlib 02   0
 /usr/local/lib/libusb-1.0.so.1.0
1d1c01646000 1d1c01b2 rlib 01   0
 /usr/lib/libc.so.80.1
1d1c0873d000 1d1c08b4f000 rlib 04   0
 /usr/lib/libpthread.so.19.0
1d1c6130 1d1c6130 rtld 01   0
 /usr/libexec/ld.so
-- 
-
() ascii ribbon campaign - against html e-mail
/\

Re: Quick APU2 review

[Stuart Henderson]

On 2016/04/19 12:33, sven falempin wrote:
> 
> 
> On Mon, Apr 18, 2016 at 6:07 PM, Stuart Henderson 
> wrote:
> 
> On 2016-04-18, Christian Weisgerber  wrote:
> > On 2016-04-18, Stuart Henderson  wrote:
> >
> >> From a different machine though. Compared to APU1 the APU2 has
> 4x the L2
> >> cache, RAM is clocked a quarter faster, twice the number of cpu
> cores,
> >> and a few more cpu features (e.g. AES-NI, RDRAND).
> >
> > For the record: The APU2 does not have RDRAND.
> 
> Ah, I thought I saw a page suggesting that the cpu did, sorry for
> misinformation.
> O
> 
> 
> 
> Is it really possible to flash the BIOS from openBSD ?

A few people reported that flashrom in openbsd-wip works. You'll need to
patch pciutils (see pciutils.diff in the openbsd-wip dir) and boot with
securelevel temporarily set to -1 to use it.

Re: Quick APU2 review

[sven falempin]

On Mon, Apr 18, 2016 at 6:07 PM, Stuart Henderson 
wrote:

> On 2016-04-18, Christian Weisgerber  wrote:
> > On 2016-04-18, Stuart Henderson  wrote:
> >
> >> From a different machine though. Compared to APU1 the APU2 has 4x the L2
> >> cache, RAM is clocked a quarter faster, twice the number of cpu cores,
> >> and a few more cpu features (e.g. AES-NI, RDRAND).
> >
> > For the record: The APU2 does not have RDRAND.
>
> Ah, I thought I saw a page suggesting that the cpu did, sorry for
> misinformation.
> O
>
>
Is it really possible to flash the BIOS from openBSD ?

-- 
-
() ascii ribbon campaign - against html e-mail
/\

Re: Quick APU2 review

[Stuart Henderson]

On 2016-04-18, Christian Weisgerber  wrote:
> On 2016-04-18, Stuart Henderson  wrote:
>
>> From a different machine though. Compared to APU1 the APU2 has 4x the L2
>> cache, RAM is clocked a quarter faster, twice the number of cpu cores,
>> and a few more cpu features (e.g. AES-NI, RDRAND).
>
> For the record: The APU2 does not have RDRAND.

Ah, I thought I saw a page suggesting that the cpu did, sorry for 
misinformation.
O

Re: Quick APU2 review

[Christian Weisgerber]

On 2016-04-18, Stuart Henderson  wrote:

> From a different machine though. Compared to APU1 the APU2 has 4x the L2
> cache, RAM is clocked a quarter faster, twice the number of cpu cores,
> and a few more cpu features (e.g. AES-NI, RDRAND).

For the record: The APU2 does not have RDRAND.

-- 
Christian "naddy" Weisgerber  na...@mips.inka.de

Re: Quick APU2 review

[Boudewijn Dijkstra]

Op Fri, 15 Apr 2016 18:12:41 +0200 schreef Christian Weisgerber  
:

A "make -j4 build" took exactly 120 minutes.


Using which physical disk type(s)?



--
Gemaakt met Opera's e-mailprogramma: http://www.opera.com/mail/

Re: Quick APU2 review

[sven falempin]

On Mon, Apr 18, 2016 at 10:11 AM, sven falempin 
wrote:

> I had problem with my USB3 key, but i do not truct the key that much
>
> Use (A)uto layout, (E)dit auto layout, or create (C)ustom layout? [a]
> /dev/rsd1a: 521.3MB in 1067648 sectors of 512 bytes
> 4 cylinder groups of 130.33MB, 8341 blocks, 16768 inodes each
> /dev/rsd1k: 4554.6MB in 9327776 sectors of 512 bytes
> 23 cylinder groups of 202.47MB, 12958 blocks, 25984 inodes each
> newfs: cg 0: bad magic number
> newfs: fsinit1 failed
> /dev/rsd1d: 826.1MB in 1691840 sectors of 512 bytes
> 5 cylinder groups of 202.47MB, 12958 blocks, 25984 inodes each
> /dev/rsd1f: 1341.3MB in 2747008 sectors of 512 bytes
> 7 cylinder groups of 202.47MB, 12958 blocks, 25984 inodes each
> newfs: reduced number of fragments per cylinder group from 99424 to 99032
> to enlarge last cylinder group
> /dev/rsd1g: 776.8MB in 1590848 sectors of 512 bytes
> 5 cylinder groups of 193.42MB, 12379 blocks, 24832 inodes each
> /dev/rsd1h: 2930.6MB in 6001920 sectors of 512 bytes
> 15 cylinder groups of 202.47MB, 12958 blocks, 25984 inodes each
> /dev/rsd1j: 1653.0MB in 3385440 sectors of 512 bytes
> 9 cylinder groups of 202.47MB, 12958 blocks, 25984 inodes each
> newfs: cg 0: bad magic number
> newfs: fsinit1 failed
> /dev/rsd1i: 1200.5MB in 2458656 sectors of 512 bytes
> 6 cylinder groups of 202.47MB, 12958 blocks, 25984 inodes each
> newfs: cg 0: bad magic number
> newfs: fsinit1 failed
> /dev/rsd1e: 1227.4MB in 2513760 sectors of 512 bytes
> 7 cylinder groups of 202.47MB, 12958 blocks, 25984 inodes each
> Available disks are: sd0.
> Which disk do you wish to initialize? (or 'done') [done]
> /dev/sd1a (8d4b156cca50297b.a) on /mnt type ffs (rw, asynchronous, local)
> mount_ffs: 8d4b156cca50297b.k on /mnt/home: Invalid argument
>
>
> Perks the preboot log:
> Mainboard PCEngines apu2 Enable.
> APIC: 00 missing read_resources
> APIC: 01 missing read_resources
> APIC: 02 missing read_resources
> APIC: 03 missing read_resources
> SeaBIOS (version rel-1.8.0-190-gc029eab-20151109_141843-wim-ws)
> BUILD: gcc:  binutils: (GNU Binutils) 2.23.2
> SeaBIOS (version rel-1.8.0-190-gc029eab-20151109_141843-wim-ws)
> BUILD: gcc:  binutils: (GNU Binutils) 2.23.2
> Found coreboot cbmem console @ 77fdf000
> Found mainboard PC Engines PCEngines apu2
> Relocating init from 0x000eb200 to 0x77f66110 (size 32352)
> Found CBFS header at 0xfc50
> multiboot: eax=0, ebx=0
> boot order:
> 1: /pci@i0cf8/usb@10/usb-*@1
> 2: /pci@i0cf8/usb@10/usb-*@2
> 3: /pci@i0cf8/usb@10/usb-*@3
> 4: /pci@i0cf8/usb@10/usb-*@4
> 5: /pci@i0cf8/*@11/drive@0/disk@0
> 6: /pci@i0cf8/*@11/drive@1/disk@0
> 7: /rom@genroms/pxe.rom
> 8: pxen0
> 9: scon1
> 10:
> Found 21 PCI devices (max PCI bus is 03)
> Copying SMBIOS entry point from 0x77fb7000 to 0x000f3040
> Copying ACPI RSDP from 0x77fb8000 to 0x000f3010
> Copying MPTABLE from 0x77fdc000/77fdc010 to 0x000f2e60
> Copying PIR from 0x77fdd000 to 0x000f2e30
> Using pmtimer, ioport 0x818
> Scan for VGA option rom
> Running option rom at c000:0003
>
> Google, Inc.
> Serial Graphics Adapter 08/22/15
> SGABIOS $Id: sgabios.S 8 2010-04-22 00:03:40Z nlaredo $
> (wiv@coreboot-Virtual-Ma
> chine)
> Sat Aug 22 09:25:30 UTC 2015
> Term: 80x24
> IO4 0
> Turning on vga text mode console
> [.]
> SeaBIOS (version rel-1.8.0-190-gc029eab-20151109_141843-wim-ws)
> XHCI init on dev 00:10.0: regs @ 0xfeb22000, 4 ports, 32 slots, 32 byte
> contexts
> XHCIextcap 0x1 @ feb22500
> XHCIprotocol USB  3.00, 2 ports (offset 1), def 0
> XHCIprotocol USB  2.00, 2 ports (offset 3), def 10
> XHCIextcap 0xa @ feb22540
> Found 2 serial ports
> ATA controller 1 at 4010/4020/0 (irq 0 dev 88)
> EHCI init on dev 00:13.0 (regs=0xfeb25420)
> ATA controller 2 at 4018/4024/0 (irq 0 dev 88)
> Searching bootorder for: /pci@i0cf8/*@14,7
> Searching bootorder for: /rom@img/memtest
> Searching bootorder for: /rom@img/setup
> XHCI port #3: 0x00200e03, powered, enabled, pls 0, speed 3 [High]
> Searching bootorder for: /pci@i0cf8/usb@10/storage@3/*@0/*@0,0
> Searching bootorder for: /pci@i0cf8/usb@10/usb-*@3
> USB MSC vendor='Kingston' product='DataTraveler G2' rev='1.00' type=0
> removable=
> 1
> USB MSC blksize=512 sectors=3919872
> Initialized USB HUB (0 ports used)
> All threads complete.
> Scan for option roms
> PCengines Press F10 key now for boot menu:
> Searching bootorder for: HALT
> drive 0x000f2dc0: PCHS=0/0/0 translation=lba LCHS=972/64/63 s=3919872
> Space available for UMB: c1000-ef000, f-f2dc0
> Returned 258048 bytes of ZoneHigh
> e820 map has 6 items:
>   0:  - 0009f800 = 1 RAM
>   1: 0009f800 - 000a = 2 RESERVED
>   2: 000f - 0010 = 2 RESERVED
>   3: 0010 - 77fad000 = 1 RAM
>   4: 77fad000 - 7800 = 2 RESERVED
>   5: f800 - 

Re: Quick APU2 review

[sven falempin]

And finally just after this test, CPU was 104degC according to sysctl, so
if you want to use it, get a box !


-- 
-
() ascii ribbon campaign - against html e-mail
/\

Re: Quick APU2 review

[Christian Weisgerber]

Boudewijn Dijkstra:

> > A "make -j4 build" took exactly 120 minutes.
> 
> Using which physical disk type(s)?

A no-name 16 GB mSATA SSD.
http://www.apu-board.de/produkte/datapower-msata.html

-- 
Christian "naddy" Weisgerber  na...@mips.inka.de

Re: Quick APU2 review

[sven falempin]

I had problem with my USB3 key, but i do not truct the key that much

Use (A)uto layout, (E)dit auto layout, or create (C)ustom layout? [a]
/dev/rsd1a: 521.3MB in 1067648 sectors of 512 bytes
4 cylinder groups of 130.33MB, 8341 blocks, 16768 inodes each
/dev/rsd1k: 4554.6MB in 9327776 sectors of 512 bytes
23 cylinder groups of 202.47MB, 12958 blocks, 25984 inodes each
newfs: cg 0: bad magic number
newfs: fsinit1 failed
/dev/rsd1d: 826.1MB in 1691840 sectors of 512 bytes
5 cylinder groups of 202.47MB, 12958 blocks, 25984 inodes each
/dev/rsd1f: 1341.3MB in 2747008 sectors of 512 bytes
7 cylinder groups of 202.47MB, 12958 blocks, 25984 inodes each
newfs: reduced number of fragments per cylinder group from 99424 to 99032
to enlarge last cylinder group
/dev/rsd1g: 776.8MB in 1590848 sectors of 512 bytes
5 cylinder groups of 193.42MB, 12379 blocks, 24832 inodes each
/dev/rsd1h: 2930.6MB in 6001920 sectors of 512 bytes
15 cylinder groups of 202.47MB, 12958 blocks, 25984 inodes each
/dev/rsd1j: 1653.0MB in 3385440 sectors of 512 bytes
9 cylinder groups of 202.47MB, 12958 blocks, 25984 inodes each
newfs: cg 0: bad magic number
newfs: fsinit1 failed
/dev/rsd1i: 1200.5MB in 2458656 sectors of 512 bytes
6 cylinder groups of 202.47MB, 12958 blocks, 25984 inodes each
newfs: cg 0: bad magic number
newfs: fsinit1 failed
/dev/rsd1e: 1227.4MB in 2513760 sectors of 512 bytes
7 cylinder groups of 202.47MB, 12958 blocks, 25984 inodes each
Available disks are: sd0.
Which disk do you wish to initialize? (or 'done') [done]
/dev/sd1a (8d4b156cca50297b.a) on /mnt type ffs (rw, asynchronous, local)
mount_ffs: 8d4b156cca50297b.k on /mnt/home: Invalid argument


Perks the preboot log:
Mainboard PCEngines apu2 Enable.
APIC: 00 missing read_resources
APIC: 01 missing read_resources
APIC: 02 missing read_resources
APIC: 03 missing read_resources
SeaBIOS (version rel-1.8.0-190-gc029eab-20151109_141843-wim-ws)
BUILD: gcc:  binutils: (GNU Binutils) 2.23.2
SeaBIOS (version rel-1.8.0-190-gc029eab-20151109_141843-wim-ws)
BUILD: gcc:  binutils: (GNU Binutils) 2.23.2
Found coreboot cbmem console @ 77fdf000
Found mainboard PC Engines PCEngines apu2
Relocating init from 0x000eb200 to 0x77f66110 (size 32352)
Found CBFS header at 0xfc50
multiboot: eax=0, ebx=0
boot order:
1: /pci@i0cf8/usb@10/usb-*@1
2: /pci@i0cf8/usb@10/usb-*@2
3: /pci@i0cf8/usb@10/usb-*@3
4: /pci@i0cf8/usb@10/usb-*@4
5: /pci@i0cf8/*@11/drive@0/disk@0
6: /pci@i0cf8/*@11/drive@1/disk@0
7: /rom@genroms/pxe.rom
8: pxen0
9: scon1
10:
Found 21 PCI devices (max PCI bus is 03)
Copying SMBIOS entry point from 0x77fb7000 to 0x000f3040
Copying ACPI RSDP from 0x77fb8000 to 0x000f3010
Copying MPTABLE from 0x77fdc000/77fdc010 to 0x000f2e60
Copying PIR from 0x77fdd000 to 0x000f2e30
Using pmtimer, ioport 0x818
Scan for VGA option rom
Running option rom at c000:0003

Google, Inc.
Serial Graphics Adapter 08/22/15
SGABIOS $Id: sgabios.S 8 2010-04-22 00:03:40Z nlaredo $
(wiv@coreboot-Virtual-Ma
  chine)
Sat Aug 22 09:25:30 UTC 2015
Term: 80x24
IO4 0
Turning on vga text mode console
[.]
SeaBIOS (version rel-1.8.0-190-gc029eab-20151109_141843-wim-ws)
XHCI init on dev 00:10.0: regs @ 0xfeb22000, 4 ports, 32 slots, 32 byte
contexts
XHCIextcap 0x1 @ feb22500
XHCIprotocol USB  3.00, 2 ports (offset 1), def 0
XHCIprotocol USB  2.00, 2 ports (offset 3), def 10
XHCIextcap 0xa @ feb22540
Found 2 serial ports
ATA controller 1 at 4010/4020/0 (irq 0 dev 88)
EHCI init on dev 00:13.0 (regs=0xfeb25420)
ATA controller 2 at 4018/4024/0 (irq 0 dev 88)
Searching bootorder for: /pci@i0cf8/*@14,7
Searching bootorder for: /rom@img/memtest
Searching bootorder for: /rom@img/setup
XHCI port #3: 0x00200e03, powered, enabled, pls 0, speed 3 [High]
Searching bootorder for: /pci@i0cf8/usb@10/storage@3/*@0/*@0,0
Searching bootorder for: /pci@i0cf8/usb@10/usb-*@3
USB MSC vendor='Kingston' product='DataTraveler G2' rev='1.00' type=0
removable=
1
USB MSC blksize=512 sectors=3919872
Initialized USB HUB (0 ports used)
All threads complete.
Scan for option roms
PCengines Press F10 key now for boot menu:
Searching bootorder for: HALT
drive 0x000f2dc0: PCHS=0/0/0 translation=lba LCHS=972/64/63 s=3919872
Space available for UMB: c1000-ef000, f-f2dc0
Returned 258048 bytes of ZoneHigh
e820 map has 6 items:
  0:  - 0009f800 = 1 RAM
  1: 0009f800 - 000a = 2 RESERVED
  2: 000f - 0010 = 2 RESERVED
  3: 0010 - 77fad000 = 1 RAM
  4: 77fad000 - 7800 = 2 RESERVED
  5: f800 - fc00 = 2 RESERVED
enter handle_19:
  NULL
Booting from Hard Disk...
Booting from :7c00
Using drive 0, partition 3.
Loading.
probing: pc0 com0 com1 mem[638K 1918M a20=on]
disk: hd0+
>> OpenBSD/amd64 BOOT 3.28
switching console to com0

-- 

Re: Quick APU2 review

[sven falempin]

http://s4.postimg.org/5ov9malvh/back.jpg
http://s1.postimg.org/qqiiqvfi7/front.jpg

wait for it.



On Mon, Apr 18, 2016 at 7:29 AM, Mihai Popescu  wrote:

> > But his point is still valid.
>
> Yes? APU1x is old and tested, so I can bet that benchmarks are readily
> available. It is over-popular already, full of examples and tests. The
> user was interested in APU2x wich is totally different.
>
> > He knew he had an inferior machine but it was still able to saturate a
> 150Mb/s line.
>
> Where is your logic? Inferior looking at what? How much is 150Mb/s?
>
> > These requests for 'Real World' numbers are almost always stupid,
>
> No! Those are the real ones, the rest are just numbers.
>
> because the people asking are almost certain to never need the max
> amount of bandwidth even a modest machine can supply.
>
> Wrong again. The market is full of applications demading power and
> power. Look in the browser world!
>
> > Really? Are you actually considering this box for use at the
> telecommunications provider you work for?
>
> You have no idea how many boxes like that are deployed in the
> production environments.
> My provider is offering 1Gb/sec in its own ISP network, at my door for
> aprox. 10 euro/month. That is 1ooo Mb/sec and it can be reach most of
> the time. So , do I really need a box to handle that? And this is home
> install.
>
>


-- 
-
() ascii ribbon campaign - against html e-mail
/\

Re: Quick APU2 review

[Mihai Popescu]

> But his point is still valid.

Yes? APU1x is old and tested, so I can bet that benchmarks are readily
available. It is over-popular already, full of examples and tests. The
user was interested in APU2x wich is totally different.

> He knew he had an inferior machine but it was still able to saturate a 
> 150Mb/s line.

Where is your logic? Inferior looking at what? How much is 150Mb/s?

> These requests for 'Real World' numbers are almost always stupid,

No! Those are the real ones, the rest are just numbers.

because the people asking are almost certain to never need the max
amount of bandwidth even a modest machine can supply.

Wrong again. The market is full of applications demading power and
power. Look in the browser world!

> Really? Are you actually considering this box for use at the 
> telecommunications provider you work for?

You have no idea how many boxes like that are deployed in the
production environments.
My provider is offering 1Gb/sec in its own ISP network, at my door for
aprox. 10 euro/month. That is 1ooo Mb/sec and it can be reach most of
the time. So , do I really need a box to handle that? And this is home
install.

Re: Quick APU2 review

[Raul Miller]

On Mon, Apr 18, 2016 at 12:02 AM, Eric Furman  wrote:
> These requests for 'Real World' numbers are almost always
> stupid, because the people asking are almost certain to never
> need the max amount of bandwidth even a modest machine
> can supply. Really? Are you actually considering this box
> for use at the telecommunications provider you work for?

Mm... I think you should have left out this paragraph.

(1) It's true: technically most people really do not "need" computers.
Some food, and a place to sleep - those are things people actually
need.

(2) That said, there are other circumstances where bandwidth or
throughput might matter to someone's opinion of how well they are
doing (for example when they have to deal with a number of local
machines -- perhaps because they are working with groups of other
people, or perhaps because they have a variety of machines at home --
and are moving large things between these machines).

(3) Machine performance can matter even when neither bandwidth nor
throughput is a bottleneck, if the machine is multipurpose.

(4) Being stupid should be thought of as a part of the learning
process - something to get over, certainly, but an aspect of something
everyone has to go through.

And, I imagine I am being stupid, here, myself - since I have nothing
to contribute about the actual performance of any APU2 machines. So:
if you want to tell me about how stupid I am, please do it by emailing
me directly, and do not burden this list with your reactions to my
stupidity.

Thanks,

-- 
Raul

Re: Quick APU2 review

[Eric Furman]

On Sun, Apr 17, 2016, at 08:26 PM, Stuart Henderson wrote:
> On 2016-04-15, Daniel Ouellet  wrote:
> >> That's nice.  I don't have a ferrari, I have a rather basic truck.
> >> 
> >> You are off topic.
> >
> > Sorry Theo,
> >
> > He asked for
> >
> > "real world through put?"
> >
> > I provided some to be helpful.
> 
> From a different machine though. Compared to APU1 the APU2 has 4x the L2
> cache, RAM is clocked a quarter faster, twice the number of cpu cores,
> and a few more cpu features (e.g. AES-NI, RDRAND). On the downside the
> bios is still a bit in flux (note it should be possible to flash from
> OpenBSD, but you need a patched version of pciutils).
> 
> It depends on the workload but in certain cases these changes will make
> a huge difference.
> 

But his point is still valid. He knew he had an inferior machine
but it was still able to saturate a 150Mb/s line. That means
unless the person asking for throughput data is using a 
DS4/NA then he probably will be Okay.
These requests for 'Real World' numbers are almost always
stupid, because the people asking are almost certain to never
need the max amount of bandwidth even a modest machine
can supply. Really? Are you actually considering this box
for use at the telecommunications provider you work for?

Re: Quick APU2 review

[Stuart Henderson]

On 2016-04-15, Daniel Ouellet  wrote:
>> That's nice.  I don't have a ferrari, I have a rather basic truck.
>> 
>> You are off topic.
>
> Sorry Theo,
>
> He asked for
>
> "real world through put?"
>
> I provided some to be helpful.

>From a different machine though. Compared to APU1 the APU2 has 4x the L2
cache, RAM is clocked a quarter faster, twice the number of cpu cores,
and a few more cpu features (e.g. AES-NI, RDRAND). On the downside the
bios is still a bit in flux (note it should be possible to flash from
OpenBSD, but you need a patched version of pciutils).

It depends on the workload but in certain cases these changes will make
a huge difference.

Re: Quick APU2 review

[Stuart Henderson]

On 2016-04-15, Heine Lysemose  wrote:
> Can you give some real world through put? How much can you push through it
> from a NAT’et device? And what is the device stats when doing so?

This depends on several things. IPsec or not (and how is it configured).
pppoe or routing or bridging. kern.pool_debug or not. PF rulesets. Packet sizes.

Re: Quick APU2 review

[Niels]

APU2 is quite a different product than APU1.
Different processor(s), different network interfaces (using different
drivers).


> On 15 Apr 2016, at 21:49, Daniel Ouellet  wrote:
>
>> That's nice.  I don't have a ferrari, I have a rather basic truck.
>>
>> You are off topic.
>
> Sorry Theo,
>
> He asked for
>
> "real world through put?"
>
> I provided some to be helpful.

Re: Quick APU2 review

[Theo de Raadt]

>> That's nice.  I don't have a ferrari, I have a rather basic truck.
>> 
>> You are off topic.
>
>Sorry Theo,
>
>He asked for
>
>"real world through put?"
>
>I provided some to be helpful.

You provided data from an ENTIRELY DIFFERENT MACHINE.

That is unhelpful.

Re: Quick APU2 review

[Daniel Ouellet]

> That's nice.  I don't have a ferrari, I have a rather basic truck.
> 
> You are off topic.

Sorry Theo,

He asked for

"real world through put?"

I provided some to be helpful.

Re: Quick APU2 review

[Theo de Raadt]

>From owner-misc+M157084=deraadt=cvs.openbsd@openbsd.org Fri Apr 15 
>13:39:59 2016
>Delivered-To: dera...@cvs.openbsd.org
>DKIM-Signature: v=1; a=rsa-sha1; c=relaxed; d=realconnect.com; h=subject 
>:to:references:from:message-id:date:mime-version:in-reply-to 
>:content-type:content-transfer-encoding; s=mail-key; bh=v6TAoSDi 
>qhR+xU4jzmiPMESBlnY=; b=YHdaKebjNunl691j7NzsWzplrmJmJYbGkqeGrslP 
>S6XoMsgDnVCEAE6Uj7/f3K7QPtUBMhtRliAGzg4RuY3gPJcAsBgJlaA3XiCOn17H 
>8Wg1h6JlguPdF/K/ov/7BFHEv/9vdWhCIOItGYs/vUaZ3vip3fHtyEUThpUq/scL Xvo=
>Subject: Re: Quick APU2 review
>To: misc@openbsd.org
>References: <20160415161241.ga80...@lorvorc.mips.inka.de> 
><57113bb9.858e1c0a.c6973.9...@mx.google.com>
>From: Daniel Ouellet <dan...@presscom.net>
>Date: Fri, 15 Apr 2016 15:38:16 -0400
>User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:38.0) 
>Gecko/20100101 Thunderbird/38.7.2
>MIME-Version: 1.0
>In-Reply-To: <57113bb9.858e1c0a.c6973.9...@mx.google.com>
>Content-Type: text/plain; charset=utf-8
>Content-Transfer-Encoding: 8bit
>List-Help: <mailto:majord...@openbsd.org?body=help>
>List-ID: 
>List-Owner: <mailto:owner-m...@openbsd.org>
>List-Post: <mailto:misc@openbsd.org>
>List-Subscribe: <mailto:majord...@openbsd.org?body=sub%20misc>
>List-Unsubscribe: <mailto:majord...@openbsd.org?body=unsub%20misc>
>X-Loop: misc@openbsd.org
>Precedence: list
>Sender: owner-m...@openbsd.org
>
>I don't have the APU2C4, I have the APU1C4
>
>and I can push 80Mb/sec of IPSec on it, way more obviously when I don't
>do the IPSec.
>
>My setup use ikedv2 from Rek@
>
>When I reach the 80Mb/sec, well it reach the full CPU utilization.
>
>When I do NAT only the CPU cores ( I have only 2 on that APU1) are use
>only at 45% each for 150Mb/sec real traffic.
>
>I wish I could test faster, but my line for now is 150Mb upgrading to
>300Mb/sec soon.
>
>If I do not do nat but use only fix IP's. it's even lower.
>
>And my PF rules have 37 active lines. Well my config is bigger
>obviously, but see the rules output for exact feedback.
>
>It run routing, pf, IKEDv2, NAT, unbound, dhcpd, ntpd, smtpd just for
>the local feedback, NOT for all my emails. I have a different server for
>that.
>
># pfctl -sr | wc -l
>  37
>
>I am upgrading it for the APU2c4 because if the AES-NI instruction set
>on the CPU to improve my traffic under IKED, NOT because it is not
>capable. I just want more traffic under encryption and the new CPU will
>improve that.
>
>But this one already can saturate the line I have already without IKED
>traffic, so I can't imagine that it woudln't do what you want assuming
>you are not running a fortune 500 company obviously.
>
>here is without IKED:
>
>http://www.speedtest.net/my-result/5253974103
>
>And if I push it via tcpbench OVER IKED, instead of normal traffic (
>from a server behind that APU1c4 that is not that box obviously and that
>need routing and all), it gets a bit lower, but here is the output
>anyway on average with the rest of the traffic running now. I stream
>Spottily and have a video running and about 9 ssh connection at the
>moment doing my work, and a few more stuff as well as my kid playing
>games League of Legends, etc.
>
>Conn:   1 Mbps:   55.166 Peak Mbps:   58.288 Avg Mbps:   55.166
>
>As you can see, plenty of capacity and the APU2C4 I am sure beat this
>hands down!
>
>It has 4 cores oppose to two and the encryption set on the CPU.
>
>Hope this help you.
>
>Daniel
>
>
>On 4/15/16 3:06 PM, Heine Lysemose wrote:
>> Hi
>> 
>> Can you give some real world through put? How much can you push through it
>> from a NAT’et device? And what is the device stats when doing so?
>> 
>> Best,
>> Lysemise
>> 
>> 
>> 
>> From: Christian Weisgerber
>> Sent: 15. april 2016 18:19
>> To: misc@openbsd.org
>> Subject: Quick APU2 review
>> 
>> I bought a PC Engines APU2 this week and thought I'd write up my
>> impressions.
>> 
>> TL;DR: Recommended.
>> 
>> The obvious point of reference is the Soekris net6501.  Now, that
>> comparison isn't really fair since the net6501 is several years old
>> and the APU2 is a new design.  Then again, Soekris canceled their
>> successor model (after stringing along potential customers for a
>> year), so they're without a competitive product now.  Tough for them.
>> 
>> http://pcengines.ch/apu2c4.htm
>> https://soekris.com/products/net6501-1.html
>> 
>> Here's what the APU2 lacks: It has only three Ethernet ports instead
>> of four, no front-side Ethernet LED

Re: Quick APU2 review

[Daniel Ouellet]

I don't have the APU2C4, I have the APU1C4

and I can push 80Mb/sec of IPSec on it, way more obviously when I don't
do the IPSec.

My setup use ikedv2 from Rek@

When I reach the 80Mb/sec, well it reach the full CPU utilization.

When I do NAT only the CPU cores ( I have only 2 on that APU1) are use
only at 45% each for 150Mb/sec real traffic.

I wish I could test faster, but my line for now is 150Mb upgrading to
300Mb/sec soon.

If I do not do nat but use only fix IP's. it's even lower.

And my PF rules have 37 active lines. Well my config is bigger
obviously, but see the rules output for exact feedback.

It run routing, pf, IKEDv2, NAT, unbound, dhcpd, ntpd, smtpd just for
the local feedback, NOT for all my emails. I have a different server for
that.

# pfctl -sr | wc -l
  37

I am upgrading it for the APU2c4 because if the AES-NI instruction set
on the CPU to improve my traffic under IKED, NOT because it is not
capable. I just want more traffic under encryption and the new CPU will
improve that.

But this one already can saturate the line I have already without IKED
traffic, so I can't imagine that it woudln't do what you want assuming
you are not running a fortune 500 company obviously.

here is without IKED:

http://www.speedtest.net/my-result/5253974103

And if I push it via tcpbench OVER IKED, instead of normal traffic (
from a server behind that APU1c4 that is not that box obviously and that
need routing and all), it gets a bit lower, but here is the output
anyway on average with the rest of the traffic running now. I stream
Spottily and have a video running and about 9 ssh connection at the
moment doing my work, and a few more stuff as well as my kid playing
games League of Legends, etc.

Conn:   1 Mbps:   55.166 Peak Mbps:   58.288 Avg Mbps:   55.166

As you can see, plenty of capacity and the APU2C4 I am sure beat this
hands down!

It has 4 cores oppose to two and the encryption set on the CPU.

Hope this help you.

Daniel


On 4/15/16 3:06 PM, Heine Lysemose wrote:
> Hi
> 
> Can you give some real world through put? How much can you push through it
> from a NAT’et device? And what is the device stats when doing so?
> 
> Best,
> Lysemise
> 
> 
> 
> From: Christian Weisgerber
> Sent: 15. april 2016 18:19
> To: misc@openbsd.org
> Subject: Quick APU2 review
> 
> I bought a PC Engines APU2 this week and thought I'd write up my
> impressions.
> 
> TL;DR: Recommended.
> 
> The obvious point of reference is the Soekris net6501.  Now, that
> comparison isn't really fair since the net6501 is several years old
> and the APU2 is a new design.  Then again, Soekris canceled their
> successor model (after stringing along potential customers for a
> year), so they're without a competitive product now.  Tough for them.
> 
> http://pcengines.ch/apu2c4.htm
> https://soekris.com/products/net6501-1.html
> 
> Here's what the APU2 lacks: It has only three Ethernet ports instead
> of four, no front-side Ethernet LEDs, no PCI-Express expansion slot,
> no LOM.  On the plus side, it has two USB 3.0 ports instead of a
> single USB 2.0 one.
> 
> Performance: Single-core speed of the APU2 seems to be comparable
> to the net6501-70 (the fast model), but the APU2 has four cores
> instead of two and it has AES-NI, which provides a big boost for
> many crypto applications.  A "make -j4 build" took exactly 120
> minutes.
> 
> Heat: The APUs have an innovative design where the CPU heat sink
> is coupled to the case.  Since this is typically assembled by the
> customer, a lot of attention is drawn to it and people obsess over
> the CPU temperature.  It's a nonissue.  Case temperature is about
> the same as for the net6501, where people are far less concerned,
> even a "make -j4 build" didn't raise the CPU temperature much (57C
> to 64.5C), and the design ensures good heat flow.  Ask me again in
> six months how it did in a 33C summer environment, but I expect no
> problems whatsoever.
> 
> The firmware is still being worked on; it's cobbled together from
> coreboot, a MemTest86 module (takes about 1h45 for one pass on the
> apu2c4), and iPXE.  It works.  I've booted via PXE, from an external
> USB key, and from mSATA.
> 
> Miscellaneous: The case is really compact.  The order of the Ethernet
> ports is reversed when compared to the Soekris and not marked on
> the case.
> 
> And yes, the APU2 is fully supported by OpenBSD 5.9.
> 
> Overall, I like it a lot.  Compared to the net6501, the APU2 is
> much cheaper and more powerful.  Compared to Intel Rangeley devices,
> it is readily available in small quantities (like, one) and, to
> pick the one that you can easily buy, again much cheaper than the
> RCC-VE 2440.
> 
> My APU2 is serving as my home gateway now, replacing a net6501.
> It f

Re: Quick APU2 review

[Heine Lysemose]

Hi

Can you give some real world through put? How much can you push through it
from a NAT’et device? And what is the device stats when doing so?

Best,
Lysemise



From: Christian Weisgerber
Sent: 15. april 2016 18:19
To: misc@openbsd.org
Subject: Quick APU2 review

I bought a PC Engines APU2 this week and thought I'd write up my
impressions.

TL;DR: Recommended.

The obvious point of reference is the Soekris net6501.  Now, that
comparison isn't really fair since the net6501 is several years old
and the APU2 is a new design.  Then again, Soekris canceled their
successor model (after stringing along potential customers for a
year), so they're without a competitive product now.  Tough for them.

http://pcengines.ch/apu2c4.htm
https://soekris.com/products/net6501-1.html

Here's what the APU2 lacks: It has only three Ethernet ports instead
of four, no front-side Ethernet LEDs, no PCI-Express expansion slot,
no LOM.  On the plus side, it has two USB 3.0 ports instead of a
single USB 2.0 one.

Performance: Single-core speed of the APU2 seems to be comparable
to the net6501-70 (the fast model), but the APU2 has four cores
instead of two and it has AES-NI, which provides a big boost for
many crypto applications.  A "make -j4 build" took exactly 120
minutes.

Heat: The APUs have an innovative design where the CPU heat sink
is coupled to the case.  Since this is typically assembled by the
customer, a lot of attention is drawn to it and people obsess over
the CPU temperature.  It's a nonissue.  Case temperature is about
the same as for the net6501, where people are far less concerned,
even a "make -j4 build" didn't raise the CPU temperature much (57C
to 64.5C), and the design ensures good heat flow.  Ask me again in
six months how it did in a 33C summer environment, but I expect no
problems whatsoever.

The firmware is still being worked on; it's cobbled together from
coreboot, a MemTest86 module (takes about 1h45 for one pass on the
apu2c4), and iPXE.  It works.  I've booted via PXE, from an external
USB key, and from mSATA.

Miscellaneous: The case is really compact.  The order of the Ethernet
ports is reversed when compared to the Soekris and not marked on
the case.

And yes, the APU2 is fully supported by OpenBSD 5.9.

Overall, I like it a lot.  Compared to the net6501, the APU2 is
much cheaper and more powerful.  Compared to Intel Rangeley devices,
it is readily available in small quantities (like, one) and, to
pick the one that you can easily buy, again much cheaper than the
RCC-VE 2440.

My APU2 is serving as my home gateway now, replacing a net6501.
It feels good to be running an AMD CPU again. :-)


PS: I bought mine from NRG Systems GmbH, Augsburg, Germany, who
sell convenient board/case/PSU/SSD kits.  Board and case were
already assembled.
--
Christian "naddy" Weisgerber  na...@mips.inka.de

Re: Quick APU2 review

[Timo Myyrä]

Otto Moerbeek  writes:

> On Fri, Apr 15, 2016 at 06:12:41PM +0200, Christian Weisgerber wrote:
>
>> I bought a PC Engines APU2 this week and thought I'd write up my
>> impressions.
>> 
>> TL;DR: Recommended.
>> 
>> The obvious point of reference is the Soekris net6501.  Now, that
>> comparison isn't really fair since the net6501 is several years old
>> and the APU2 is a new design.  Then again, Soekris canceled their
>> successor model (after stringing along potential customers for a
>> year), so they're without a competitive product now.  Tough for them.
>> 
>> http://pcengines.ch/apu2c4.htm
>> https://soekris.com/products/net6501-1.html
>> 
>> Here's what the APU2 lacks: It has only three Ethernet ports instead
>> of four, no front-side Ethernet LEDs, no PCI-Express expansion slot,
>> no LOM.  On the plus side, it has two USB 3.0 ports instead of a
>> single USB 2.0 one.
>> 
>> Performance: Single-core speed of the APU2 seems to be comparable
>> to the net6501-70 (the fast model), but the APU2 has four cores
>> instead of two and it has AES-NI, which provides a big boost for
>> many crypto applications.  A "make -j4 build" took exactly 120
>> minutes.
>> 
>> Heat: The APUs have an innovative design where the CPU heat sink
>> is coupled to the case.  Since this is typically assembled by the
>> customer, a lot of attention is drawn to it and people obsess over
>> the CPU temperature.  It's a nonissue.  Case temperature is about
>> the same as for the net6501, where people are far less concerned,
>> even a "make -j4 build" didn't raise the CPU temperature much (57C
>> to 64.5C), and the design ensures good heat flow.  Ask me again in
>> six months how it did in a 33C summer environment, but I expect no
>> problems whatsoever.
>> 
>> The firmware is still being worked on; it's cobbled together from
>> coreboot, a MemTest86 module (takes about 1h45 for one pass on the
>> apu2c4), and iPXE.  It works.  I've booted via PXE, from an external
>> USB key, and from mSATA.
>> 
>> Miscellaneous: The case is really compact.  The order of the Ethernet
>> ports is reversed when compared to the Soekris and not marked on
>> the case.
>> 
>> And yes, the APU2 is fully supported by OpenBSD 5.9.
>> 
>> Overall, I like it a lot.  Compared to the net6501, the APU2 is
>> much cheaper and more powerful.  Compared to Intel Rangeley devices,
>> it is readily available in small quantities (like, one) and, to
>> pick the one that you can easily buy, again much cheaper than the
>> RCC-VE 2440.
>> 
>> My APU2 is serving as my home gateway now, replacing a net6501.
>> It feels good to be running an AMD CPU again. :-)
>> 
>> 
>> PS: I bought mine from NRG Systems GmbH, Augsburg, Germany, who
>> sell convenient board/case/PSU/SSD kits.  Board and case were
>> already assembled.
>> -- 
>> Christian "naddy" Weisgerber  na...@mips.inka.de
>
> A dmesg! My kingdom for a dmesg!
>  ;-)
>
>   -otto

Here's one from my apu.

Timo

OpenBSD 5.9-current (GENERIC.MP) #1973: Tue Mar 29 19:42:47 MDT 2016
dera...@amd64.openbsd.org:/usr/src/sys/arch/amd64/compile/GENERIC.MP
real mem = 4261076992 (4063MB)
avail mem = 4127580160 (3936MB)
mpath0 at root
scsibus0 at mpath0: 256 targets
mainbus0 at root
bios0 at mainbus0: SMBIOS rev. 2.7 @ 0xdffb7020 (7 entries)
bios0: vendor coreboot version "APU2A_20150928-19-gbc96368-dirty" date 
02/11/2016
bios0: PC Engines apu2
acpi0 at bios0: rev 2
acpi0: sleep states S0 S1 S2 S3 S4 S5
acpi0: tables DSDT FACP SSDT APIC HEST SSDT SSDT HPET
acpi0: wakeup devices PWRB(S4) PBR4(S4) PBR5(S4) PBR6(S4) PBR7(S4) PBR8(S4) 
UOH1(S3) UOH3(S3) UOH5(S3) XHC0(S4)
acpitimer0 at acpi0: 3579545 Hz, 32 bits
acpimadt0 at acpi0 addr 0xfee0: PC-AT compat
cpu0 at mainbus0: apid 0 (boot processor)
cpu0: AMD GX-412TC SOC, 998.37 MHz
cpu0: 
FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CFLUSH,MMX,FXSR,SSE,SSE2,HTT,SSE3,PCLMUL,MWAIT,SSSE3,CX16,SSE4.1,SSE4.2,MOVBE,POPCNT,AES,XSAVE,AVX,F16C,NXE,MMXX,FFXSR,PAGE1GB,LONG,LAHF,CMPLEG,SVM,EAPICSP,AMCR8,ABM,SSE4A,MASSE,3DNOWP,OSVW,IBS,SKINIT,TOPEXT,ITSC,BMI1
cpu0: 32KB 64b/line 2-way I-cache, 32KB 64b/line 8-way D-cache, 2MB 64b/line 
16-way L2 cache
cpu0: ITLB 32 4KB entries fully associative, 8 4MB entries fully associative
cpu0: DTLB 40 4KB entries fully associative, 8 4MB entries fully associative
cpu0: smt 0, core 0, package 0
mtrr: Pentium Pro MTRR support, 8 var ranges, 88 fixed ranges
cpu0: apic clock running at 99MHz
cpu0: mwait min=64, max=64, IBE
cpu1 at mainbus0: apid 1 (application processor)
cpu1: AMD GX-412TC SOC, 998.11 MHz
cpu1: 
FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CFLUSH,MMX,FXSR,SSE,SSE2,HTT,SSE3,PCLMUL,MWAIT,SSSE3,CX16,SSE4.1,SSE4.2,MOVBE,POPCNT,AES,XSAVE,AVX,F16C,NXE,MMXX,FFXSR,PAGE1GB,LONG,LAHF,CMPLEG,SVM,EAPICSP,AMCR8,ABM,SSE4A,MASSE,3DNOWP,OSVW,IBS,SKINIT,TOPEXT,ITSC,BMI1
cpu1: 32KB 64b/line 2-way I-cache, 32KB 64b/line 8-way D-cache, 2MB 64b/line 
16-way L2 

Re: Quick APU2 review

[Otto Moerbeek]

On Fri, Apr 15, 2016 at 06:12:41PM +0200, Christian Weisgerber wrote:

> I bought a PC Engines APU2 this week and thought I'd write up my
> impressions.
> 
> TL;DR: Recommended.
> 
> The obvious point of reference is the Soekris net6501.  Now, that
> comparison isn't really fair since the net6501 is several years old
> and the APU2 is a new design.  Then again, Soekris canceled their
> successor model (after stringing along potential customers for a
> year), so they're without a competitive product now.  Tough for them.
> 
> http://pcengines.ch/apu2c4.htm
> https://soekris.com/products/net6501-1.html
> 
> Here's what the APU2 lacks: It has only three Ethernet ports instead
> of four, no front-side Ethernet LEDs, no PCI-Express expansion slot,
> no LOM.  On the plus side, it has two USB 3.0 ports instead of a
> single USB 2.0 one.
> 
> Performance: Single-core speed of the APU2 seems to be comparable
> to the net6501-70 (the fast model), but the APU2 has four cores
> instead of two and it has AES-NI, which provides a big boost for
> many crypto applications.  A "make -j4 build" took exactly 120
> minutes.
> 
> Heat: The APUs have an innovative design where the CPU heat sink
> is coupled to the case.  Since this is typically assembled by the
> customer, a lot of attention is drawn to it and people obsess over
> the CPU temperature.  It's a nonissue.  Case temperature is about
> the same as for the net6501, where people are far less concerned,
> even a "make -j4 build" didn't raise the CPU temperature much (57C
> to 64.5C), and the design ensures good heat flow.  Ask me again in
> six months how it did in a 33C summer environment, but I expect no
> problems whatsoever.
> 
> The firmware is still being worked on; it's cobbled together from
> coreboot, a MemTest86 module (takes about 1h45 for one pass on the
> apu2c4), and iPXE.  It works.  I've booted via PXE, from an external
> USB key, and from mSATA.
> 
> Miscellaneous: The case is really compact.  The order of the Ethernet
> ports is reversed when compared to the Soekris and not marked on
> the case.
> 
> And yes, the APU2 is fully supported by OpenBSD 5.9.
> 
> Overall, I like it a lot.  Compared to the net6501, the APU2 is
> much cheaper and more powerful.  Compared to Intel Rangeley devices,
> it is readily available in small quantities (like, one) and, to
> pick the one that you can easily buy, again much cheaper than the
> RCC-VE 2440.
> 
> My APU2 is serving as my home gateway now, replacing a net6501.
> It feels good to be running an AMD CPU again. :-)
> 
> 
> PS: I bought mine from NRG Systems GmbH, Augsburg, Germany, who
> sell convenient board/case/PSU/SSD kits.  Board and case were
> already assembled.
> -- 
> Christian "naddy" Weisgerber  na...@mips.inka.de

A dmesg! My kingdom for a dmesg!
 ;-)

-otto

Quick APU2 review

[Christian Weisgerber]

I bought a PC Engines APU2 this week and thought I'd write up my
impressions.

TL;DR: Recommended.

The obvious point of reference is the Soekris net6501.  Now, that
comparison isn't really fair since the net6501 is several years old
and the APU2 is a new design.  Then again, Soekris canceled their
successor model (after stringing along potential customers for a
year), so they're without a competitive product now.  Tough for them.

http://pcengines.ch/apu2c4.htm
https://soekris.com/products/net6501-1.html

Here's what the APU2 lacks: It has only three Ethernet ports instead
of four, no front-side Ethernet LEDs, no PCI-Express expansion slot,
no LOM.  On the plus side, it has two USB 3.0 ports instead of a
single USB 2.0 one.

Performance: Single-core speed of the APU2 seems to be comparable
to the net6501-70 (the fast model), but the APU2 has four cores
instead of two and it has AES-NI, which provides a big boost for
many crypto applications.  A "make -j4 build" took exactly 120
minutes.

Heat: The APUs have an innovative design where the CPU heat sink
is coupled to the case.  Since this is typically assembled by the
customer, a lot of attention is drawn to it and people obsess over
the CPU temperature.  It's a nonissue.  Case temperature is about
the same as for the net6501, where people are far less concerned,
even a "make -j4 build" didn't raise the CPU temperature much (57C
to 64.5C), and the design ensures good heat flow.  Ask me again in
six months how it did in a 33C summer environment, but I expect no
problems whatsoever.

The firmware is still being worked on; it's cobbled together from
coreboot, a MemTest86 module (takes about 1h45 for one pass on the
apu2c4), and iPXE.  It works.  I've booted via PXE, from an external
USB key, and from mSATA.

Miscellaneous: The case is really compact.  The order of the Ethernet
ports is reversed when compared to the Soekris and not marked on
the case.

And yes, the APU2 is fully supported by OpenBSD 5.9.

Overall, I like it a lot.  Compared to the net6501, the APU2 is
much cheaper and more powerful.  Compared to Intel Rangeley devices,
it is readily available in small quantities (like, one) and, to
pick the one that you can easily buy, again much cheaper than the
RCC-VE 2440.

My APU2 is serving as my home gateway now, replacing a net6501.
It feels good to be running an AMD CPU again. :-)


PS: I bought mine from NRG Systems GmbH, Augsburg, Germany, who
sell convenient board/case/PSU/SSD kits.  Board and case were
already assembled.
-- 
Christian "naddy" Weisgerber  na...@mips.inka.de