Comments on: Poor man’s SSD: No news is no news

By: ancientforest

ancientforest — Tue, 08 Mar 2011 20:56:26 +0000

I put a 400x Transcend 32 GB in my old ThinkPad T23 after reading about your experiences here. That was over three months ago. Not so much as a burp since then. Haven’t seen any write-blocking delays. I’ve installed Debian on the machine about half a dozen times as a result of tinkering, and it has been running 24/7 otherwise. The CF card is generally able to saturate this machine’s IDE interface. Finding it hard not to love this approach.

Per some other articles around the net, it’s a good idea to mount CF drives with noatime, to use GUID partition tables, because it helps with alignment, and to use ext4 if you want optimal read/write speeds. ext2 is of course a reasonable choice if one wants to wring every last write out of the medium. While this is all described elsewhere, here it is one place, as a footnote to kmandla’s opus.

Cheers!

By: K.Mandla

K.Mandla — Mon, 07 Mar 2011 22:03:47 +0000

We’ll see. Your comment is remarkably similar to some I heard before I started this experiment, so perhaps there are people who have had different experiences than the one I am having.

But like I said before, these cards have already surpassed my expectations. Even if all three fail simultaneously tomorrow, I will have gotten more than I expected from them.

And considering my track record with conventional hard drives in recent years, that will be at least as good as any other mechanical hard drive I’ve bought. In other words, if it fails, it fails … as all things must someday.

By: helf

helf — Mon, 07 Mar 2011 15:41:12 +0000

Newer CFs have wear leveling like their SSD cousins. I wouldn’t mind using a newer CF on an older machine, but I’d still be hesitant to use them on anything I had important data on without having many backups. I ran a laptop off a 2gb cf card many moons ago and it worked gloriously for months… then the card suddenly died. Couldn’t read or write to it. Nothing would even recognize it. SO, yeah, running CFs in machines has been easy and done for years, but I still wouldn’t recommend it unless you have a new fangled (and expensive, I might add) one with proper wear leveling. Otherwise it IS just a ticking time bomb once a segment that gets written to a lot finally goes.

GLad its working out for you at the moment, tho

By: Jean-Marc Liotier

Jean-Marc Liotier — Mon, 07 Mar 2011 09:51:25 +0000

I can confirm firsthand the freezes caused by the write-blocking problems described in this article. I wonder if Kmandla encountered them too.

By: Chubby Checker

Chubby Checker — Sat, 05 Mar 2011 17:50:16 +0000

just found this pretty detailed current article on LWN:

http://lwn.net/Articles/428584/?format=printable

I hope this is useful

By: Chubby Checker

Chubby Checker — Sat, 05 Mar 2011 11:52:04 +0000

Meh, forget to tell which card I use. It’s the 16GB 400x version from Transcend (http://www.transcendusa.com/Products/ModDetail.asp?ModNo=264&LangNo=0&Func1No=1&Func2No=13)

By: Chubby Checker

Chubby Checker — Sat, 05 Mar 2011 10:00:52 +0000

That certainly depends on a lot of factors. Just a quick test on my system:

writing:
$ time dd if=/dev/zero of=./zero.bin bs=10M count=100
100+0 records in
100+0 records out
1048576000 bytes (1.0 GB) copied, 34.665 s, 30.2 MB/s
real 0m34.672s
user 0m0.000s
sys 0m2.836s

reading (cached):
$ time dd if=zero.bin of=/dev/null bs=10M count=100
100+0 records in
100+0 records out
1048576000 bytes (1.0 GB) copied, 1.12598 s, 931 MB/s
real 0m1.130s
user 0m0.000s
sys 0m1.128s

just kidding, so clearing the cache:
# sync; echo 3 > /proc/sys/vm/drop_caches

and reading uncached:
$ time dd if=zero.bin of=/dev/null bs=10M count=100
100+0 Datensätze ein
100+0 Datensätze aus
1048576000 Bytes (1,0 GB) kopiert, 22,3152 s, 47,0 MB/s
real 0m22.332s
user 0m0.000s
sys 0m1.504s

By: x33a

x33a — Sat, 05 Mar 2011 09:30:58 +0000

Congratulations

I am curious though, how much transfer rates do you get?

Also, does it work well with heavy I/O, for example games?