test your sd card read speeds on n800 - Page 2 - maemo.org

Active Topics

[Announce] genwall a simple iptables firewall (90)
to Applications by paoletto - 1 day, 4 hrs ago
N9 as WiFi repeater/extender? (1)
to Nokia N9 / N950 by paoletto - 1 day, 4 hrs ago
Sailfish OS flashing files (89)
to SailfishOS by monkeyisland - 2 days, 5 hrs ago
more...

Page 2 of 3

< Prev

Next >

Thread Tools

	daihard	2007-08-19 , 10:03
	Posts: 68 \| Thanked: 1 time \| Joined on Aug 2007 @ Lynnwood, WA	#11

Firmware: 4.2007.26-8
Kernel: 2.6.18-omap1 (stock)
Card: SanDisk SDHC (4GB)

Code:

Nokia-N800-26:/home/user/bin# ./hdparm -Tt /dev/mmcblk0p1

/dev/mmcblk0p1:
 Timing cached reads:   222 MB in  2.01 seconds = 110.62 MB/sec
 Timing buffered disk reads:   18 MB in  3.13 seconds =   5.76 MB/sec

I guess it's time to update the kernel, huh.

__________________
Registered Linux User: #281828
Fedora Core 6 / Nokia N800

Quote & Reply |

robmiller	2007-08-19 , 11:50
Posts: 64 \| Thanked: 9 times \| Joined on Jun 2007 @ abuja, nigeria (currently)	#12

Originally Posted by Milhouse

Before declaring that Class 4 is faster than Class 2 I think we need to see more Class 2 results from manufacturers other than Transcend.

good idea

The fact that I obtained two different results for the same card when accessing different partitions (and filesystems) suggests that hdparm isn't testing the raw performance of the card at all. Instead it appears to be testing the performance of the filesystem. Different block sizes in a VFAT partition may thus play a role in determining the figures returned by hdparm, and I'm sceptical about it's accuracy compared to dd.

hdparm uses the linux ioctls for reading blocks off disks, as I expect the filesystem layer does. it's written by the developer/maintainer of the (e)ide linux driver, and the man page is explicit about 'without any filesystem overhead' for the -t test. nonetheless, 'formatted filesystem' doesn't equal 'fdisk parameters' and I don't see hdparm or anything else yet as the ideal benchmark for this. I think others have posted here about logic in the cards to re-map memory blocks as they die, so there's probably a lot inside these black boxes that we won't be able to know about. there's also the point that both the benchmarking tools and the memory cards are 'imagining' a system of heads and spinning platters.

wonder if dd uses the same ioctls as well? how big were your dd tests?

both my cards are single partition ext2 filesystems tho, so can't do much testing on these different aspects. and I'm not that keen to dd a big block of data into the middle of either of them

rob.

Quote & Reply |

robmiller	2007-08-19 , 12:00
Posts: 64 \| Thanked: 9 times \| Joined on Jun 2007 @ abuja, nigeria (currently)	#13

Originally Posted by tolou

Testing internal Flash gives me a respectable result.
Code:
/dev/mtdblock4:
 Timing cached reads:   206 MB in  2.00 seconds = 102.83 MB/sec
 Timing buffered disk reads:   42 MB in  3.08 seconds =  13.66 MB/sec
How could that be? We all know that booting from mmc is boosting read performance.
Still on 51-6 with the SDHC kernel although I'm only running non-SDHC cards still.

perhaps the the sd card i/o is dispatched off to a separate bit of circuitry, while the internal flash and ram could be sharing access hardware (address and data lines)? just a guess of course...

rob.

Quote & Reply |

Milhouse	2007-08-19 , 12:40
Posts: 3,401 \| Thanked: 1,255 times \| Joined on Nov 2005 @ London, UK	#14

Originally Posted by robmiller

wonder if dd uses the same ioctls as well? how big were your dd tests?

About 240MB - benchmarking a dd *read* is non destructive.

Here's a link to the thread with my tests when using dd.

Quote & Reply |

jbenny	2007-08-21 , 21:24
Posts: 28 \| Thanked: 5 times \| Joined on Aug 2007	#15

I have the official SDHC firmware. Like others, I am getting about 5.5MB/S buffered read speeds on both internal and external slots. The internal card is an older SanDisk 1GB SD. The external is the A-Data 8GB SDHC that I just bought. This is a Class 2 card, but it performs like a Class 6 in a card reader in my PC. In my N800, both cards are performing about the same. Obviously, there is a severe bottleneck in the N800.

Quote & Reply |

	penguinbait	2007-08-25 , 05:50
	Posts: 3,096 \| Thanked: 1,525 times \| Joined on Jan 2006 @ Michigan, USA	#16

Originally Posted by penguinbait

sdhc 8GB - Class 6 - Kingston

Code:
Nokia-N800-26:~# ./hdparm -Tt /dev/mmcblk0p1

/dev/mmcblk1p2:
Timing cached reads: 242 MB in 2.00 seconds = 120.25 MB/sec
Timing buffered disk reads: 38 MB in 3.04 seconds = 12.50 MB/sec

updated kernel listed in this post

mounted on /usr/local

Ran again, 2GB Kingston MMCMobile, VFAT (setup inside winxp)

mmc0 is mounted on /media/mmc1

then I created a 1GB ext2 filesystem on a file in /media/mmc1
/media/mmc1/1GB_file

then mount -o loop to /usr/local

mount -o loop /media/mmc1/1GB_file /usr/local

/dev/loop0:
Timing cached reads: 216 MB in 2.01 seconds = 107.24 MB/sec
Timing buffered disk reads: 30 MB in 3.01 seconds = 9.98 MB/sec
Nokia-N800-26:~# ./hdparm -Tt /dev/loop0

/dev/loop0:
Timing cached reads: 234 MB in 2.01 seconds = 116.25 MB/sec
Timing buffered disk reads: 52 MB in 3.01 seconds = 17.30 MB/sec
Nokia-N800-26:~# ./hdparm -Tt /dev/loop0

/dev/loop0:
Timing cached reads: 242 MB in 2.00 seconds = 120.85 MB/sec
Timing buffered disk reads: 52 MB in 3.14 seconds = 16.56 MB/sec

Check this out, kinda interesting?

Quote & Reply |

robmiller	2007-08-25 , 08:42
Posts: 64 \| Thanked: 9 times \| Joined on Jun 2007 @ abuja, nigeria (currently)	#17

Originally Posted by jbenny

I have the official SDHC firmware. ... Obviously, there is a severe bottleneck in the N800.

yes, the bottleneck is in the firmware which does not implement the full speed the hardware is capable of. the view seems to be that this was probably done to ensure the widest compatibility for non-techy users. if you're concerned about measuring and getting the fastest speeds from your card then you have probably moved out of the initial target audience and should consider using the kernel with the speed enhancements.

rob.

Quote & Reply |

robmiller	2007-08-25 , 12:49
Posts: 64 \| Thanked: 9 times \| Joined on Jun 2007 @ abuja, nigeria (currently)	#18

Originally Posted by penguinbait

Ran again, 2GB Kingston MMCMobile, VFAT (setup inside winxp)

mmc0 is mounted on /media/mmc1

then I created a 1GB ext2 filesystem on a file in /media/mmc1
/media/mmc1/1GB_file
...
/dev/loop0:
Timing cached reads: 234 MB in 2.01 seconds = 116.25 MB/sec
Timing buffered disk reads: 52 MB in 3.01 seconds = 17.30 MB/sec
Nokia-N800-26:~# ./hdparm -Tt /dev/loop0
....

Check this out, kinda interesting?

it does seem like we can get different speeds from different regions of the card, tho really not clear what effect the ext2 on top of vfat has here

also, what class is the MMCMobile?

I do think Milhouse/Fanoush are right - dd and their 240mb read tests will probably give a more accurate picture of card speed overall than the hdparm 2sec/3sec tests.

ideally we'd all think to do dd write tests when we first get the cards before reformatting them and loading up all our goodies.

rob.

Quote & Reply |

niv	2008-01-26 , 01:56
Posts: 74 \| Thanked: 34 times \| Joined on Jan 2008	#19

A-DATA Turbo 8GB Secure Digital High-Capacity(SDHC) Flash Card Model TurboSD SDHC 8G - Retail (Class 6)
http://www.newegg.com/product/produc...82E16820211180
/dev/mmcblk1p1:
Timing cached reads: 186 MB in 2.00 seconds = 92.80 MB/sec
Timing buffered disk reads: 14 MB in 3.18 seconds = 4.40 MB/sec

second card:
/dev/mmcblk0p1:
Timing cached reads: 194 MB in 2.02 seconds = 96.26 MB/sec
Timing buffered disk reads: 14 MB in 3.10 seconds = 4.51 MB/sec

Last edited by niv; 2008-01-26 at 02:04.

Quote & Reply |

	Munk	2008-05-08 , 05:50
	Posts: 229 \| Thanked: 108 times \| Joined on Oct 2007 @ Sacramento, California	#20

/dev/mmcblk1p1:
Timing cached reads: 194 MB in 2.00 seconds = 96.42 MB/sec
Timing buffered disk reads: 10 MB in 3.18 seconds = 3.14 MB/sec

/dev/mmcblk1p2:
Timing cached reads: 192 MB in 2.02 seconds = 95.27 MB/sec
Timing buffered disk reads: 14 MB in 3.03 seconds = 4.62 MB/sec

OK, so why is my benchmark so darn slow!?! On my N810 I have a 4Gig Class 6 SD brand card. I have dual boot installed and currently booted into my SD card.

This has prompted me to look for the 48MHz mmc hack and I've installed the "Modified initfs with onscreen boot menu and usb network recovery" from http://fanoush.wz.cz/maemo/ which I thought had the 48MHz hack in there. When I got into "About" I see that I have version 2.2007.51-3 but on Fanoush's website it says "2.2007.50-2" so do I not have it installed then? If so, how am I dual booting?

I'm confused, any help would be great. Basically, I'm sick of how damn slow everything loads and was hoping that by having a faster SD card, booting from it, getting away from the jffs2 OS, and launching programs from my SD card, everything would load 25% faster or better.