January 7, 2009 at 8:26 pm #160480
Vista or XP ?
1Gb or 2Gb Ram ?
Readyboost or no Readyboost ?
Then see the results from a little benchmarking exercise I carried out this evening.January 7, 2009 at 11:38 pm #178906AlfiharParticipant
Interesting results, it’s kind of what I would expect to see. With XP a bit faster than Vista and ReadyBoost not helping that much with this kind of test, and the same going for the extra RAM.
As Vista checks the SD card (or other memory device) to make sure it isn’t going to slow down the computer before allowing it to be used for ReadyBoost, I’m not surprised that SD Card 1 caused a drop in performance.
Now because I was curious I ran the PerformanceTest 6.1 Evaluation Version on Fedora 10 Linux using Wine (for anyone who doesn’t know it allows you to run Windows programs in other OS’s). I was unable to run the 3D and the optical drive tests (looks like I need a newer version for the 3D to work).
NC10, 2GB RAM, Fujitsu hard drive, Wi-Fi + Bluetooth active, Compiz 3D desktop effects enabled, running off battery.
PassMark Rating = 204.1 (Results Breakdown)January 8, 2009 at 7:56 am #178907
looks like Fedoras about on the halfway mark then.
In the real world Vista (with whichever SD card)seems slightly snappier overall than XP althoug XPx fgraphics performance for games (eg UT2004) is superior.
Hmm difficult call, hate to say this but Vista is growing on meJanuary 9, 2009 at 2:15 am #178910hufnMember
what program did you use? I’m running WIndows 7 with 2gb and would be curious to see what it would show 🙂January 9, 2009 at 3:15 am #178912January 9, 2009 at 8:39 am #178908January 10, 2009 at 3:01 am #178913
Comparing under Windows XP:
1. Samsung NC10: 1.6 GHz, 1GB RAM
2. Old Desktop: AMD Athlon underclocked to1.6 GHz, 1.5 GB RAM, video card Radeon 8500 (yeah aincient)
You have to click twice on the image that looks like a long table to enlarge it enough.
I also run an intensive program in Mathematica (same version installed on Sammy and Desktop).
Sammy finished it for 61s,
Desktop for 50s
so the computational speed of Sammy is 82% of that of Desktop [keep in mind speed ~ 1/time]. That is confirmed by the CPU benchmarks in the test: Sammy/Desktop = 241.8/301.7 = 80%. The separate CPU tests for integer and floating point math look bad for Sammy but it compensates at ‘image rotation’ and ‘string sorting’. I don’t know how Sammy succeeded getting 82% score on my Mathematica test then cause its mostly floating point calculations and some 3D plots.
The test shows in graphics both 2D and 3D Sammy is 50% of the Desktop. The RAM memory test is about the same for Sammy and Desktop. Finally Sammy has twice better Hard Drive than Desktop.
Now comes the BIG SURPRISE! I ran an even more intensive Mathematica calculation.
Desktop completed in 503s.
Sammy completed in 437s
so Sammy showed 115% of the Desktop speed. It’s FASTER LOL
I investigated and the ONLY option that seems to influence the speed of Sammy is the “Allow/Disable the Hyperthreading” iin the BIOS advanced settings. The above results are with Allowed Hyperthreading which is the default option. When its Allowed you will see two plots for the CPU Usage History in the WindowsTaskManager (CTRL+ALT+DEL) which means there are TWO virtual processors. When I disable the hyperthreading there is only one plot and the Sammy speed decreases by a factor of two. The speed of the intensive calculation is NOT influenced by what mode is chosen with Fn+F8 (silent, normal, speed). I guess these just control the power savings so they decrease the speed of the processor when it is not used heavily which is not the case in this calculation.
So with Hyperthreading the Atom processor probably beats any old 1.6GHz processor that doesn’t have it in long term intensive calculations.January 10, 2009 at 5:30 am #178914
Probably the reason for the poor Sammy performance test results compared to Desktop in integer and floating point math was that the tests did not engage the two virtual processors efficiently. I was watching them on the TaskManager and they looked as if they were doing exactly the same calculation – their CPU Usage History graphs were exactly the same during the CPU tests, which means they were cloning the work not dividing it..
During the real-world test, the long calculation with Mathematica, the processors distributed the work between them (the two CPU Usage History graphs were not the same) and that is probably the explanation why Sammy is more efficient at integer and floating point calculations than the performace tests suggested initially.January 10, 2009 at 9:22 am #178909
i guess its a little like a tandem cyclist racing a solo bike. The solo bike will have it in the shorter sprint but the tandem pair will do better over the long haul.
Out of curiousity if you run Mathemetica in speed mode does it make much difference to the timings and what setting was the power management on?January 10, 2009 at 11:45 am #178905dinoMember
yes, i suspect that there is a lot of re-engineering to the internal flow of the chip. the atom is only half as fast on paper as my old Pentium M 1.73, but in practical usage it is almost just as snappy. pleasantly surprised.
there was another thread where i was trying to figure out if there is specific software designed for hyperthreading processors that would handle encoding tasks (audio/video) any better. any ideas?January 10, 2009 at 3:25 pm #178915
Out of curiousity if you run Mathemetica in speed mode does it make much difference to the timings and what setting was the power management on?
EDIT: The processor computation speed does depend on the settings. First I left the default settings in BIOS allowing power management. Then the setting that influence the speed are the F8 (silent, normal, speedy) modes and the battery manager (Maximum Battery, Normal, Maximum Performance) modes and whether the Sammy is running with cord/ without cord. I ran my shorter Mathematica test (the 60s) and here are the times of execution depending on settings:
F8 BattManager……. MaximumBattery………Normal……….MaximumPerformance
silent………………………….. 114 / 114…………114 / 114………….115 / 114
normal…………………………..63 / 114……………63 / 62……………..61 / 63
speedy………………………….62 / 114……………63 / 63……………..63 / 62
The times are in seconds, the first number is running on cord the second is running on battery. For example 63 / 114 in second row first column means:
F8 at normal, Battery Manager at MaxumumBattery, running on cord -> 63 seconds
F8 at normal, Battery Manager at MaxumumBattery, running on battery – > 114 seconds
Make sure you set F8 and Battery Manager and plug/unplug cord right before starting the program cause the netbook has some internal logic and changing one of the settings may reset the other. Like plugging the cord usually leads to Normal F8 mode independent of what it was before.
The bottom line is: when you want maximum numerical power do not set F8 to silent mode(on battery and cord) and, if running on battery, do not set the battery manager on Maximum battery. If you make the mistake to chose those, the speed decreases by a factor of two. Any other settings work fine.January 10, 2009 at 9:28 pm #178911hufnMember
Windows 7 build 7000 2GB ram – 185.6
interesting… Not sure how it really fits in in the long run though 🙂
- You must be logged in to reply to this topic.