René Rebe

According to an O2 employee I just spoke to (regarding my memory glitching device) the Nokia Booklet 3G is already on the list of soon to be End-Of-Life’d devices? Given that it just went on sale just slightly more than a week ago this makes a pretty short lifespan. Even for cellphone standards, … I wonder if there is more behind this, such as a batch of faulty memory chips, PCBs, high return rate in general?

My last days experience with the Nokia Booklet 3G were not too superb, the Google memory stress tester spotted some miscompares:

Hardware Error: miscompare on CPU 1(0×2) at 0×5bd98480(0×0:DIMM Unknown): read:0×0000000004000000, reread:0×0000000004000000 expected:0×0400000004000000
Hardware Error: miscompare on CPU 1(0×2) at 0×49363240(0×0:DIMM Unknown): read:0×0000000004000000, reread:0×0000000004000000 expected:0×0400000004000000
…

Could explain some things.

Question is who’s to blame? Memory, Chipset, CPU, OS?

Given this, and Google’s latest paper on memory errors I consider making an stressapptest run part of my standard system installation procedure, …

Update: With the same binary, (Linux) kernel and (Google) test-application, from the same USB stick the Google memory tester does not find an issue with any other machine int the office, from other Atom 330, over VIA C7, Intel Core 2 Duo, to 8-way Xeons. I even just walked over the the Sony Style Store in Berlin, checked a Sony Vaio VGN-X with 1.86GHz Z-series Atom - which did not report a glitch in some minutes of testing, either. So the “generic” Intel Poulsbo chipset option becomes less likely (well, you’d never know, I just say Intel F00F, FDIV bug).

While we are at it: The Nokia support is a huge joke, they claim no warranty if the OS was formatted, e.g. the Win 7 Starter updaeted, manually installed, even just Win 7 Home Premium / Professional, whatever. Forget about Linux. They stress not to remove the “hidden” recovery partition. I just had the worst support call, ever!

Update 2: The same happens on an US (mine is DE) Booklet in a Best Buy store over in the US. So it’s certainly not a single incident. However, given other server hardware shows the same in the google stresstest application list I have hope this is a bug in the tester :-)

Copyright (c) 1980, 1983, 1986, 1988, 1990, 1991, 1993, 1994
The Regents of the University of California. All rights reserved.

DragonFly v2.4.1-RELEASE (GENERIC) #14: Wed Sep 30 18:12:31 PDT 2009

The network configuration goes into /etc/rc.conf:

ifconfig_re0=”inet 192.168.2.20 netmask 255.255.255.0″
defaultrouter=192.168.2.1

(The “re” comes from RE(4) - RealTek … PCI/PCIe Ethernet adapter driver, and will certainly differ, see “ifconfig”.)

The nameserver as usual into /etc/resolv.conf:

nameserver 208.67.222.222

The services are enabled in /etc/defaults/rc.conf:

sshd_enable=”YES” # Enable sshd

Have fun.

So Nokia, or their contractor, did not fill the BIOS mode table too carefully, forgot the native resolution of 1280×720 pixels.

The Poulsbo X.org / Linux driver is still a bloody mess. Thus for using the VESA driver with the native resolution we need to get it somehow into the BIOS. Turned out the BIOS and mode-setting is so Intel i810′ish that adding GMA500 / Poulsbo support to 915resolution was easy :-)

915resolution 49 1280 720 16
915resolution 58 1280 720 32

Give the similarities for the RAMDAC / CRT / other pipe control and the i810, it might be easy to get GMA500 mode setting and external output control into the regular Intel driver. Of course without acceleration, which given the PowerVR IP core will be 100% different. But native mode-setting, including external output control might be a nice start for a non-messy X.org / Linux driver.