dc7600 Stage II - noise and speed

Some slow messing around with this machine has got me a bit further along:

* USB and storage working with USB.patches and the usb_scsi driver

* An Audigy2 has been ordered from eBay

* HLT is now enabled, so the processor fans are not running at full speed - but as I've no audio yet, I don't yet know if I need to do any of the workarounds require to stop it messing up audio

So the PC is working, or will be, to a usable standard. But I decided to run BeRometer, and found out that the perception that its working fairly quickly is likely due to the lack of seek time on the SSD - its actually hideously slow.

Copying a 608MB file from one partition on the disk to another took slightly over 8 minutes - and while this isn't an ideal disk I/O test, there's not a lot of others you can do. So sub 1MB/sec throughput on a SATA150 controller (the limiting factor in hardware here - the SSD is SATA600). But low throughput has to be expected, as when BeOS R5 came out, ATA33 was the newest tech around for IDE drives.

But there's the IDE Replacement Driver, which supports ATA100 and notionally ATA133, so lets try it. Significant performance improvements were achieved on my original contemporary machines anyway.

...it doesn't work. "unhandled cmd irq". There are IRQ settings for onboard devices - though, annoyingly, disabling parallel and serial does not recover those to be used - but all three fail to work. Disabling the IDE controller seems to make the output of the driver prior to that warning coming up a bit more sane, but with the same result.

I decided to look for BIOS updates, and it turned out that the version in use was the original release - and that there appeared to be significant uplifts in terms of features. Also, as this is a business machine, an ISO of the BIOS image is provided and can be upgraded from the BIOS itself - no Windows required.

This BIOS update did not allow me to change IRQ for the SATA controller to anything different. But it did provide some other tweaks - including merging the emulated IDE controller for the SATA controller to be primary and secondary channel on the same IDE as the DVD-ROM. And this did work. 

Now, the same 608MB file copy took NINE SECONDS. BeRometer shows it as the fastest system its ever seen (in its database that tops out at 700MHz Athlons, that is) for storage, by a factor of about 5. In fact, the only thing it isn't outright winner for in BeRometer is some graphics tests - where the VESA driver is letting it down. 

Now, my next problem is that the network driver is a bit dodgy and always stops net_server from closing cleanly. I've found an old enough Intel gigabit card here, along with various Realtek 8139s - but these were often nightmares for reliability - so will give these a go.