At FOSDEM 2009 I grabbed one Vortex86SX based device — Koala Nano PC which is available at Koan software. Device came with Ångström distribution and was running 2.6.26 kernel. I wanted to get something more fresh on it and after some fighting I booted 2.6.29-rc5 kernel today.
Device use Vortex86SX SoC which is based on 486SX core. Yes — this is x86 machine without FPU. Overall speed of that is… nearly not existing.
First benchmark which came to my mind was “hdparm -T /dev/sda”. Results were dramatic: 9-13 MB/s for cached reads (with 133MHz memory and 300MHz cpu). I decided to compare against other devices:
| Device | Arch | CPU | CPU speed | memory type | cached reads |
|---|---|---|---|---|---|
| Koala Nano PC | x86 | vortex86sx | 300MHz | DDR2 | 9-13MB/s |
| old PC | x86 | pentium2 | 266MHz | EDO DRAM | 33MB/s |
| ATNGW100 | av32 | AT32AP700x | 130MHz | SDRAM | 35MB/s |
| ASUS WL-500gP | mips | bcm3302 | 266MHz | 46MB/s | |
| Freecom FSG-3 | arm | ipx42x | 266MHz | 43MB/s | |
| iBook G3 | ppc | g3 | 300MHz | SDRAM | 65MB/s |
| AT91SAM9263-EK | arm | at91sam9263 | 180MHz | SDRAM | 62-63MB/s |
| Compulab | arm | pxa270 | 312MHz | SDRAM | 74MB/s |
| NSLU2 | arm | ixp420 | 266MHz | SDRAM | 74MB/s |
| Koala nano33 | x86 | vortex86dx | 1GHz | DDR2 | 74MB/s |
| Nokia N810 | arm | omap2420 | 400MHz | SDRAM | 82MB/s |
| AT91SAM9G20-EK | arm | at91sam9G20 | 400MHz | SDRAM | 96MB/s |
| Linkstation pro duo | arm | ixp4xx | 266MHz | 147MB/s | |
| BeagleBoard | arm | omap3530 | 500MHz | mDDR | 152MB/s |
| Alix.1c | x86 | geodelx | 500MHz | DDR | 209MB/s |
| kirkwood reference board1 | arm | kirkwood | 1200MHz | DDR2 | 209MB/s |
| BUG | arm | i.mx31 | 533MHz | DDR | 294MB/s |
| my desktop | x86 | core2quad | 2400MHz | DDR2 | 3300-3500MB/s |
| Cliff’s desktop | x86 | Core-i7 920 Quad | 2.67GHz | DDR3 | 6400-7200MB/s |
But remember that this test is not so good for benchmarking — I am preparing better set of tests to really compare speed of devices. So far it contains openssl speed and MP3 encoding/decoding.
But device has also few nice things. Everything is integrated so 12x12cm box is enough to keep everything inside. It has ATA controller, FastEthernet, graphics based on XGI core (with accelerated framebuffer able to do 1680×1050 resolution) and working USB. There is a place to put 2.5″ HDD inside (normally it boots from CompactFlash card), second Ethernet or WiFi are available options…
But what is use for such slow device? There are lot of ARM based ones which offer similar (or better) functionality and are faster… But wait — there is one use: event displays as this machine has nicely working framebuffer (I got even 1680×1050 resolution).
UPDATE: added results from iBook G3 300MHz and some other machines.
-
kirkwood reference board uses same cpu as SheevaPlug device. ↩
Related posts:
I’ve tried one of them, and the only interesting feature is the low price. Speed is quite disappointing, but it’s cheap and have a lot of connector (and 24 gpio port easy to reach).
I wonder if it is really a 300Mhz cpu…
And i wonder if vortex86dx processor have better performance.
@diego: BIOS say that it is 300MHz but can it be trusted… I also saw that lot of Ebox-2300 devices are listed as 200MHz ones.
What configuration did you use to run kernel version 2.6.29-rc5? I’m trying to use Debian Lenny un a Vortex86SX box but I cant boot the X server. Maybe with that kernel I can.
Some more results:
ASUS WL-500gP | Broadcom BCM3302 V0.6 | 266 MHz | 46 MB/s Freecom FSG-3 | XScale-IXP42x Family rev 1 (v5l) | 266 MHz | 43 MB/s
Another stat for your table:
Core-i7 920 Quad 2.67GHz, DDR3: 6400-7200MB/s
http://groups.google.com/group/beagleboard/msg/c54dde7b9d55cf99 shows the BeagleBoard can handle 741MB/s on cached reads. Do you have reasoning behind the large discrepancy to your numbers?
Jason Kridner: thx for information. Results which I put in table I got from one of BeagleBoard users (I am waiting for C2 version). But according to post which you link to that 741MB/s is in L2 cache only.
I plan to check some benchmark utilities and do better tests.
Although final result may be interesting, I think that you should specify the storage type and model in this table, because “hdparm -T /dev/sda” is a test of IDE/SATA performances and not only a CPU/Motherboard one.
mck: Let me quote hdparm manual:
This is without disk access…
@ Cliff I’m astonished by your results. Could you please post your machine details (cpu, motherboard, ram, hd brand and model) ? thanks
AT91SAM9G20-EK arm at91sam9G20 400MHz SDRAM 96MB/s
Koala nano33, x86, vortex86, 1GHz, 256M DDR2, 74MB/s
mck, details on the i7 system are available here: http://bec-systems.com/site/302/getting-linux-on-a-dx58so. I don’t think the HD model affects these results as we are talking about cached data.
useful comparision. would be interesting to see where an atom based eeepc/eeebox would fit in here. wonder if Asus will expand on their low end and come up with an arm based “plug” with an own optware package feed. the i.mx31 sure is nice, but a headless nas/router box probably needs no video in/out.
Hi, Please could you post your xorg.conf, I can’t get the 1680×1050 resolution.
Thanks.
@Damian: I got 1680×1050 resolution on framebuffer. Did not tried it under X11.
I wonder how the 1GHz Vortex86DX (i586 core with FPU) performs.
There is an entry for such system in table – “Koala nano33″. And Vortex86DX is still i486 core not i586.
Vortex86sx/dx are great for these reasons:
That said, sx is too slow for e.g. youtube or ipython startup, and ide driver is slightly broken so dma must be disabled, thus the meager io results.
Dx feels 10 times faster (although only 3x in clock speed).
Sure, they have some value — I used vortex86sx for some time to keep papers from flying on my desk.
Vortex86DX is much faster mainly due to no need for FPU exception handling. Maybe you know how good/bad does MX graphics works? With SX/DX it was usually XGI chip used and I do not remember which gfx chipset MX got.
It is more useful than a paperweight, we have a few hundred of these deployed. Very few come back cooked, compactflash on the other hand is less reliable. DX supports microSD which is great improvement.
I only use compactpc’s in headless applications. I didn’t get native vga driver that to work, because XGI drivers are too old for modern X.org server and I didn’t have that much time.
3300DX uses the same chip as SX. MX integrates gfx on chip, no idea what it is.