Atom N270: difference in performance where no difference should exist

Fallingwater

Member
Nov 28, 2010
160
0
0
www.technfun.com
I have an EeePC 1005HA. I also just received a freebie EeePC 901, which has the very same CPU - Atom N270.

So why is it that all hardinfo benchmarks show the Atom in the 901 to be considerably faster? Internet browsing is affected too, the 901 loads websites noticeably faster - as it should be, actually, while the 1005HA feels really quite slow.

Here are the hardinfo results:

Test - 901 - 1005HA

Blowfish - 15.37 - 21.64 (lower is better)
Cryptohash - 56.19 - 36.35 (higher is better)
Fibonacci - 8 - 8.15 (lower is better)
N-Queens - 17.74 - 21.22 (lower is better)
FFT - 16.73 - 30.64 (lower is better)
RayTracing - 34.76 - 66.34 (lower is better)

If the difference was minimal I'd chalk it up to manufacturing tolerances and just shrug, but some tests show the 901 with twice the performance. I don't think this is normal.

Both systems have the latest BIOS, both are running the same system (CrunchBang Statler, and I did install the EeePC ACPI scripts on both). Both show the same frequency in hardinfo, and both show the same frequency (1.60GHz) and FSB speed (533MHz) in the BIOS.

As the 1005HA is my main netbook (the 901 is too darn small for me), it's really quite important I fix this.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,786
136
If you could benchmark using more common ones then you can compare your results with other N270 systems and see which system is the normal one. Then you'd be able to determine whether one is throttling or overclocked. Also avoid disk-intensive benchmark like PCMark.

Try CPUMark99, first with 1 instance then with two on both systems.
 

Fallingwater

Member
Nov 28, 2010
160
0
0
www.technfun.com
During the tests hardinfo reports the CPU speed, and it said 1600MHz in all tests. Unless some weird powersaving stuff is going on behind the scenes, but then I've no idea what's up - the distro running on both netbooks is exactly the same, down to the version number.

I can run other benchmarks, but they've got to be available for Linux - I don't like Windows and don't run it on anything other than my gaming box, for obvious reasons.
 

jhu

Lifer
Oct 10, 1999
11,918
9
81
Is hyperthreading enabled on both computers? It almost sounds like it's off on the slower one.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,786
136
Oh yea, forgot about Hyperthreading. That will have BIG impact on Atom chips. 30-50% gains should be quite common.
 

Fallingwater

Member
Nov 28, 2010
160
0
0
www.technfun.com
Is hyperthreading enabled on both computers? It almost sounds like it's off on the slower one.
Good sir, I gift you seven internets.

No idea why, but the distro I'm using (Crunchbang Statler) enables hyperthreading on the 901 but not on the 1005HA. Booted up Puppy and suddenly hardinfo sees two processors on the 1005HA as well, and the benchmarks have gotten close enough to the 901's Atom as to make no difference.

I'll do some investigation on the Crunchbang forums. Thanks, you've been a great help. :)
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,786
136
Blowfish - 15.37 - 21.64 (lower is better)
Cryptohash - 56.19 - 36.35 (higher is better)
Fibonacci - 8 - 8.15 (lower is better)
N-Queens - 17.74 - 21.22 (lower is better)
FFT - 16.73 - 30.64 (lower is better)
RayTracing - 34.76 - 66.34 (lower is better)

Looks like extra cores on Bulldozer, lol.
 

Fallingwater

Member
Nov 28, 2010
160
0
0
www.technfun.com
Kernel changed (it had the old -486 one for old computers, changed it to -686), hyperthreading enabled, benchmarks now almost the same. The 1005HA is still some 5% slower, for some reason, but I'm too lazy to hunt down whatever weirdness is causing this tiny loss of performance. It's not like it'd make a world of difference anyway.
Thanks again for your help. :)