> Hallo,
>
> I have investigated more on the effect of decreased speed with EIDE drives
> at least with Triton I and II boards on the recent kernel.
>
> [skipped]
I can confirm that something happened after 2.0.22.
I said "dd if=/dev/hda of=/dev/null bs=1024k count=100" on a drive with
no mounted filesystems using a 2.0.22, a 2.0.27, and a 2.1.20 kernel. I
also said "hdparm -tT /dev/hda" as a check.
Here are the results:
kernel | time for 100MB | MB/s | MB/s (hdparm)
---------+----------------+------+------------------
2.0.22 | 88s | 1.14 | 1.14
2.0.27 | 166s | 0.60 | 0.65
2.1.20 | 87 | 1.15 | 1.16
| | |
With the 2.0.22 and the 2.1.20 kernels, the drive light appears to be
continuously lit. With the 2.0.27 kernel, it flickers noticeably and
regularly.
I purposely made these tests on a machine with an old 486 motherboard
that's had a Cyrix 586 brain transplant. The drive is a 1.2GB Western
Digital with a 128K internal cache.
Something definitely happened starting with 2.0.23.