Good/bad SATA controllers for Win98?

[dencorso]

Post a link for a *good quality* pic of the actual card, and we may figure out which EEPROM does it have, if any. But the pic has to be good enough for us to be able to actually read what's written on the chips, OK?

[Cyker]

Nah, don't worry; The firmware isn't that critical (The card works, unlike the VT6421, so I can't really complain )

My main annoyance is the lack of an unmount ability, but my workaround for that is just to not unplug the drive until I turn the compy off

If you really want me to, I can pull the card out of my system to get a suitable highres pic, but otherwise I'm too lazy to (It's quite short and stuck in between my Radeon 9200 and the GF7950GT and I have to take both of them out to get at it! ). The card in question is this one:

http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=220570641708

[dencorso]

It's OK. I thought you were keen on updating the BIOS. Let it be.

[osRe]

You can't treat modern high-speed storage (esp. hard disks!) like you would a floppy disk; Floppy disks didn't use any buffering and didn't do stuff like delayed, out-of-order and batched writes.

It could auto-flush after a few seconds with no activity, or when there are no open handles (though at least on XP there seem to be remnant handles for unclear reasons). I suppose it does that anyway, or else power outages or crashes would be more catastrophic.

A little tray icon to indicate the write-behind status, and there, just like waiting for the floppy LED to turn off.

I tried updating it using the Silicon Image firmware updater which is available in the download area with the actual firmware update.

The Windows commandline one, the DOS, or the GUI one that's part of the driver?

You can get info on the chip using their DOS based UPDFLASH; in its cryptic menu mode, one of the options (don't remember which) outputs a cryptic number which is the flash chip ID. Then a web search to find what it is. Instead, possibly flashrom (http://flashrom.org/) can show the actual chip name with even less hassle. If it is flash you can probably update the BIOS regardless of the type.

The chip on the card I have is of the exact same family as some of those listed in the SI Readme, just with less capacity. I had to edit the 128K update to 64K (didn't try writing the 128K but I assume it wouldn't work right). Afterwards I flashed it with UPDFLASH choosing the closest chip type.

The firmware updates don't affect the BIOS hooks much; They're mainly for better support of SATA protocols (Presumably fixing some bugs?). Supposedly there is a speed boost too but I'm not sure if that's just for RAID'd drives or also singles.

I don't think it affects anything but the BIOS. Could be different for cards that do their own processing, but I believe this one delegates everything to the drivers.

[jaclaz]

OT but just for the record:

HotSwap!

http://mt-naka.com/hotswap/index_enu.htm

Maybe a Win9x backport of it or writing something similar is possible?

jaclaz

[Cyker]

Ooh! You, sir, are a diamond!!

I'll have to try that out on my Win2k side...

[Sfor]

I,ve installed a SATA drive to the ATA motherboard, a while ago.

I did tested two solution in the process:

1) SATA to ATA converter

At first it was not not a good choice, because the motherboard built in controlled detected the connection as 40pin cable. So, the mode was reduced to just UDMA2. But, I've learned, the 80 pin connection is recognized by the pin 34 of the IDE connector. If it is grounded, the controller complains about 40 wire connection, no longer. After modifying the adapter the UDMA5 mode kicked in, and the overal performance was higher than with the PCI SATA controller.

2) SATA PCI Sil 3512 controller with the built in BIOS.

At first I had some difficulties booting the system, but after installing the driver, everything was working fine. The controller driver was placed in the SCSI section, I saw no other problems with it. The performace was a bit worse than the built in motherboard controller with modified SATA to ATA converter. It could be related to slower performance of the PCI bus. Apparently the built in controller on intel 845 chipset can outperform the PCI cards.

Unfortunatelly, I've lost the HDD LED, in both cases.

Edited August 4, 2010 by Sfor

[jaclaz]

Unfortunatelly, I've lost the HDD LED, in both cases.

Well, if you look carefully with a small torch maybe you can find it....

Seriously, what do you mean you "lost" it?

You have nowhere to connect the HD led on the case?

You can (sort of) replace it by using a software:

http://technet.microsoft.com/en-us/sysinternals/bb896646.aspx

http://www.dirtcellar.net/software.php

For the SATA to IDE, whihc EXACT converter is it? (link if possible) most have an onboard LED, which you can replace with two pins for connecting the case led.

Which EXACT PCI card is it? (link if possible) Most do have pins for the led.

jaclaz

[osRe]

Which EXACT PCI card is it? (link if possible) Most do have pins for the led.

I've a card similar to what Cyker ended up getting. It has jumper-looking thing that, after following the traces back to the chip, I'm quite convinced is a LED connector. But according to the silkscreen it's missing a resistor or two just before the connector, and I'm not sure what might happen if indeed it's supposed to be there when a LED is connected.

Too bad HDDs no longer have direct LED connectors on them. I liked having one per HDD.

[Sfor]

Well, both devices PCI SATA card and SATA to ATA converter do not have HDD LED connector. Also, the motherboard controller HDD LED does not work when SATA drive is connected through a converter.

Software HDD LED emulation could be not a bad solution, but the software has to be Windows 98 compatible.

[jaclaz]

Software HDD LED emulation could be not a bad solution, but the software has to be Windows 98 compatible.

http://www.msfn.org/board/index.php?showtopic=131079

jaclaz

Edited August 5, 2010 by jaclaz

[pointertovoid]

Via:

- I've read this controller is bugged. It makes write mistakes, exactly what you don't want (lost data, OS deteriorating over time...)

- I've had a mobo with Via chipset before, its disk controller was bugged exactly that way - just as about each and every Via controller

- Via will never ever control again a disk at my home.

(But their Usb2 chip seems to work properly)

Adapters:

- I bought them dirt-cheap on eBay from Honk Kong or Shanghai

- The one hanging to P-Ata disks for S-Ata mobo is fast

- The one hanging to S-Ata disks for P-Ata mobo is slow. Slower than Silicon Image.

- Both worked from the beginning, and test programmes saw no write-read errors.

Silicon Image:

- I have a SiI3114 who explicitly proposes drivers for W98-Me but not W95.

It works without errors (except when flashing several times, when it needed repeated flashing before succeeding.

It is not fast: 77 to 93MB/s depending on the benchmark, the Bios, the driver. My platter achieves 134MB/s on a Sata300 ich10r, my Pci achieves 124MB/s. Maybe the disk (7200.12) itself is slow on Sata150.

- The Sata300 there, SiI3124, has no official driver for W95-98- nor Me, and I know no unofficial one.

- SiI3124 seems the only one on Pci that has Ncq. SiI3114 doesn't. On W98 (no parallel requests) it's unimportant, on Nt4-Nt5-Nt6 Ncq makes a huge improvement, experimentally.

- I've had many SiI0680a for Pata133 before, these ones were the absolute best, so the "slow" SiI3114 is a disappointment.

Eject a fixed disk:

- What you need is Hotswap http://mt-naka.com/hotswap/index_enu.htm

but unfortunately, it runs on

W2k (maximum v2.0.0.0) http://mt-naka.com/hotswap/file/HotSwap!%202.0.0.0.ZIP

Xp and more (v5.0.0.0) http://mt-naka.com/hotswap/file/HotSwap!%205.0.0.0.ZIP

No install, launch it, it adds a taskbar icon for the fixed disks.

- Beginning with v3, you can spin down a disk before removing its power... Refuses to install on W2k.

- Silicon Image tells explicitly their hardware (including Sata150) allows to eject Sata disks.

- There, a more complete discussion about hotswap:

http://www.tomshardware.de/foren/242374-10-esata-festplatte-rauswerfen

(forget the last contribution, full of mistakes)

Disk size can exceed 2TiB with SiI controllers but expect worries with W95-98-Me-Nt4-2k-(Xp-2k3)... Disks will have 4kiB sectors to try to cope with the 32b Lba limit, and then you'll know which programmes took notice that sector size may vary... Anyway, more than 128GiB should go easily with W98 and SiI, independently of your Mobo's Bios.

Enjoy Sata on your faithful machine!

Edited August 5, 2010 by pointertovoid

[Sfor]

Poor performance of the SATA HDD to PATA adapter could be related to 40 pin connection detection and transfer limited to UDMA2. It is possible to force the detection of 80 pin cable through grounding pin 34 in the IDE interface. After such a modification my adapter outperformed PCI SATA controller.

As for the Software HDD LED. It does seem the Rubber Ducky does not work for me. To be more specific, everyting is working, except for the HDD activity detection.

------------------

The pin 11 of the SATA power connector is used to provide HDD LED function in SATA docking stations. With pin 11 grounded drive works as a standard internal drive, in other case it can switch to "deferred spin" or some other features.

Edited August 6, 2010 by Sfor

[larrymoencurly]

My advice:

AVOID ANYTHING WITH A VIA 6421A CHIP IN IT!!!!!

I've wasted about £50 finding this out and I don't want anyone else to go through the same experience!

The Via 6421A is a bloody awful controller chip. It's not that it's a first generation SATA-I controller or even that it's limited to 1.5Gbit.

The problem with it is that it doesn't support auto negotiation!

Now, this is not a problem if you're using SATA drives with a 1.5Gbit limit jumper, but if you're using SATA-II drives at 3Gbit the controller freaks out and keeps trying to resync. This will cause weird slowdowns because the controller will be jamming everything with IRQs as it tries to sync with the drive.

The Silicon Image-based controllers are much better for one reason: They DO support auto-negotiation.

Even if you get a SiI that only supports 1.5Gbit, it will be able to tell the drive that and the drive will lower it's speed accordingly, without needing us to mess about with jumpers. For some reason, the Via doesn't and just keeps trying to connect at 1.5Gbit to a 3Gbit drive!

I experienced that problem with the VT6421, but recently somebody said VIA had been solved it, and I recently confirmed that with several brands of drives configured for 3.0Gbit. The VT6421A I tried was dated 4th week of 2010 (1004CD). But HDAT2 wouldn't run with that controller and hanged during the drive detection phase. OTOH Windows XP Home ran very sluggishly with the primary hard drive plugged into an Intel G41 based motherboard and the secondary drive into a Silicon Image SiL3512. This didn't happen with VT6421A and Promise 20378 cards.

[Sfor]

I've bought Unitek PCI-SATA-4r controller with SiI 3114CTU chip. It does have a HDD LED connector, so I was counting on an usable HDD activity LED. This particular feature works, all right. But...

1) The controller built in RAID BIOS does not work with just a single drive. I had to make a concatenation raid setup with a single drive selected. I was able to get the HDD working in DOS that way.

2) I'm unable to find a working driver for the Windows 98. The box says, the controller is Windows 98 compatible, but there are no proper driver on the CD. Apparently, the PCI device and vendor id does not match the INF file entries (VEN_1095&DEV_3114&SUBSYS_71141095&REV_02). It is possible to install a driver for the soft raid controller version, but it does not work correctly. It is enough to access a partition from the SATA drive for the Windows to hang completely. Forcing to install the driver from a proper folder does give the same result as the software raid version.