Archive for December, 2009

Strange Mac OS X name resolution

Thursday, December 24th, 2009

Now on one Mac I have a strange issue I never had before: Sporadically, e.g. after a fresh boot, name resolution does not work. While nearly every kind of application is effected, Safari, Mail, etc. even ping. To my surprise, though, dig(1) does reliable work in this kind of stuck system state. After some review it turned out that Mac OS X provides different DNS resolution services of the classic Unix / Posix services - that apparently can get out-of-sync from what the classic resolution configuration would use. To flush them:

10.4: lookupd -flushcache
10.5: dscacheutil -flushchache
10.6: sudo killall -HUP mDNSResponder

For me, under Snow Leopard (Mac OS 10.6), killing the mDNSResponder did actually indeed help! I’ll have to keep an eye on the situation, if it’s getting too annoying I’ll have to try a clean re-install. Sigh!

The performance stagnation that is the Intel Atom

Tuesday, December 22nd, 2009

The first Pine Trail Atoms, N450 and the like, are trickeling leaking out. And guess what, the performance meter did not move a bit: Just some randomly picked Geekbench results:

score 916
ASUSTeK Computer INC. 901 6 days ago
Intel Atom N270 @ 1.60 GHz (1 processor, 2 threads)
Geekbench 2.1.4 for Windows x86 (32-bit)

score 994
Hewlett-Packard HP Mini 311-1000 5 days ago
Intel Atom N280 @ 1.67 GHz (1 processor, 2 threads)
Geekbench 2.1.4 for Windows x86 (32-bit)

score 934
Acer AO532h 7 days ago
Intel Atom N450 @ 1.67 GHz (1 processor, 2 threads)
Geekbench 2.1.4 for Windows x86 (32-bit)

I wished Moore’s law would still be in effect as the Atom performance is barely endurable. At least the new Atoms finally come with the AMD64 (x86_64, EM64T) enabled, and might be slightly more performant in 64bit mode due to more general purpose registers being around, potentially shuffling more data per instruction, clock cycle (e.g. rendering your browser content, text, graphics et al.).

To my greatest amusement this is about exactly the performance of the Transmeta Efficieon, announced in 2003, and started to ship in 2004:

score 956
Linux PC (Transmeta Efficeon(tm) Processor TM8000) 20 months ago
Transmeta Efficeon TM8000 @ 1.60 GHz (1 processor, threads)
Geekbench 2.0.15 for Linux x86 (32-bit)

I wished more companies would switch to more competitive solutions, such as the AMD Neo and Neo X2 series that comes with a way higher performance envelope for some time, now, …

score 1687
Hewlett-Packard HP Pavilion dv2 Notebook PC 4 months ago
AMD Athlon Neo X2 L335 @ 1.60 GHz (1 processor, 2 threads)
Geekbench 2.1.3 for Windows x86 (32-bit)

Update:

While at it, the Via Nano does not appear to be that much of a screamer in this regard, likewise:

score 1064
LENOVO 20021,2959 7 days ago
VIA Nano U2250 (1.6GHz Capable) @ 1.60 GHz (1 processor)
Geekbench 2.1.4 for Windows x86 (32-bit)

Oh, and you are looking for decent reference score?

score 18996
Quad-Core AMD Opteron(tm) Processor 8384 5 months ago
Quad-Core AMD Opteron 8384 @ 2.69 GHz (8 processors, 32 cores, threads)
Geekbench 2.1.2 for Linux x86 (32-bit)

score 9777
iMac11,1 4 hours ago
Apple Inc. Mac-F2268DAE
Intel Core i7 860 @ 2.80 GHz (1 processor, 4 cores, 8 threads)
Geekbench 2.1.4 for Mac OS X x86 (64-bit)

score 1872
Power Mac G5 (Late 2005)
PowerMac11,2
PowerPC G5 (970MP) @ 2.00 GHz (2 processors)
Geekbench 2.1.6 for Mac OS X PPC (32-bit)