The SIlicon Image Chips are crappy on old systems 😒 i tried many times on different Systems and had mostly problems, it worked in some situations.. and had other Strange Issues. Windows look up/freezing etc.

Benchmark of Promise ATA133 Controller, Promise Ultra 133 TX2 PCI.

Running on Gigabyte 586TX3 with 430TX Chipset on 75Mhz FSB. (K6-3 450 @ 6x75)
Harddrive is a WD800EB 80Gb 7200 RPM Western Digital Drive at UDMA5.

carangil wrote:

edit: Later in this thread I have the necessary files attached. I get about 100MB/s in ATTO Benchmark at 2MB sizes and higher. K6-2+ 600 Mhz.

original message:
[..]Writing sucks bad: Write speed are 3-4 MB a second, and the mouse freezes when writing: [..]

Wrong alignment ?

matze79 wrote:

The SIlicon Image Chips are crappy on old systems 😒 i tried many times on different Systems and had mostly problems, it worked in some situations.. and had other Strange Issues. Windows look up/freezing etc.

Yup, the Sil3112/Sil3512 based cards are known to be troublesome.
Not so much because of the chips themselves, but the firmware.

Saying this, I've never used these cards in a PC but I tried to make them work in a Mac..
These cards are kind of famous for either supporting MacOS 9 or X, but not both (firmware/size limitation).

More about his can be read at 68kmla.org.

Anyway, I don't want to discourage you. It's fun to tinker with this stuff, nevertheless.

Maybe its better to invest a bit more and get a SATA FastTrak S150 PCI.
They work well, and have less Problems at all.

Well it seems that with a SATA magnetic disk it doesn't hang as the SSD. Performances are not clear anyway from Sandra I have strange results. Where to find Atto benchmark for windows 98?

next post

UPDATE: things are getting much better now; I switched from the 3512a to the 3112 with the same WDC not-ssd disk (16MB cache, 250GB) and now Atto benchmark give 56MB/S of write and 85MB/s of read!
Boost is visible in the usage even if not really "time correct" disk but I'm trying to build a sort of fastest K6-2 machine so...

Anyway the problem with SSD kind of disk remains and also the post-bios cdrom boot problem.

386SX wrote:

Anyway the problem with SSD kind of disk remains and also the post-bios cdrom boot problem.

I only once had this problem with an early 16GiB SSD and Windows XP. My problem was sort-of related to wrong alignment, I think.
Or maybe it just was the lack of Trim (no idea whether it already supported it or not).. Sorry, I can't be more precise about this. It's been a while.
Remember, we didn't have tools like GParted back then, which allowed us to align partitions easily to the megabyte boundaries via a few clicks.
Anyway, this very same SSD was very fast in Windows 7 - eventhough it was filled with data up to the max (!)
Still wonder what caused it to be so incredible slow in XP back then.

386SX wrote:

UPDATE: things are getting much better now; I switched from the 3512a to the 3112 with the same WDC not-ssd disk (16MB cache, 250GB) and now Atto benchmark give 56MB/S of write and 85MB/s of read!

Nice 😀

Would you be so kind and upload the driver and BIOS you are using? And a picture of your card maybe?

The right combination of these three seems to matter a lot.

vetz wrote:

Drivers: https://web.archive.org/web/20120502172712/ht … px?pid=29&cat=3 BIOS: https://web.archive.org/web/20101002122201/ht … 9 […]
Show full quote

Drivers: https://web.archive.org/web/20120502172712/ht … px?pid=29&cat=3
BIOS: https://web.archive.org/web/20101002122201/ht … 9&cat=15&ctid=2&
You need to flash to IDE mode if it's going to work properly in Windows 98 (atleast that is how 3112 works)
What is a shame is that archive doesnt have it all properly indexed, so it seems a great driver archive just got lost. I highly doubt Lettuce gives a fuck about Silicon Image now.

Just wanted to pop in here and give some thanks where it's due, that BIOS saved my bacon. (the "e" variant.)
I just flashed it onto an SiI3512 MiniPCI card made for embedded systems to rid myself of the RAID functionality I'll never use.

this post was the first result on google for information towards a BIOS, so you've done at least a couple people some justice on finding a BIOS and proper drivers for this thing on older OSes.

Easily getting 100MB/s each way off a ~2013 Toshiba MK3261GSYN 2.5" drive, but I recommend anyone with one of these cards, or even the 3112's, to pop a small heatsink on the chip when expecting that kind of throughput. in my past experiences with them inside G4s and even inside an old Sun Ultra 10 and 5 I have going, they tend to get a little toasty over time. just my two cents.

the board I have that installed in is an Aaeon HSB-811P. Pentium M 1.6 Dothan, 1GB RAM, 8GB CF card. 855GM-based, no speedstep.

Hey guys,

I just bought a Delock 70096 SATA/RAID controller which has 2 internal SATA ports and comes with the SIL3512. It had a RAID BIOS version 4.3.79 flashed when I got it. Since I just need to use the base IDE functionality I decided to flash a base BIOS from the Silicon Image site using the links vetz posted.

The controller detects my SATA-HDD and DVD-ROM fine, but I have difficulties booting up when I connect both, the HDD and DVD drive. Once I hook up the DVD, the BIOS always reports "missing operating system". When I only connect the HDD, DOS boots without any problems.

The second problem is under Windows 98 SE. I have tried out different combinations of base IDE BIOSes and drivers, including the latest BIOS and 9x drivers, the BIOS version it came with (albeit as base BIOS, not RAID) and BIOS version 4.3.84. No matter what combination I try, I could never get Windows to boot once I installed the drivers using the Device Manager. The card has its own IRQ, so no conflicts. Windows sometimes simply goes blank during boot and in other cases would display something like "System protection fault" or similar error.

Has anybody tried a combination of BIOS and driver that works in Win9x?

Thanks
locutus

I have a sil3512a and I put the 4.3.70 ide bios on it. I want to use it as the w98 boot drive in a new super socket 7 build I'm putting together with a gigabyte ga-5sg100 motherboard.

I probably have the bios settings wrong. On a p4 machine the card detects during boot and I even have had the primary OS running on it, but on the SS7 there's nothing. The p4 is a dell dimension 8200 with a pretty basic set of bios options. The SS7 has so many I'm lost. I set the pnp mode to auto and turned off the integrated peripherals default of irq 14 set to legacy/isa and made that auto. I also disabled the primary onboard ide controller.

I'd like to thank you all for this post.

Past year I got the sil3512 and I tried to install good windows 98 drivers with no success (freezing mouse while data transfering) ... but recently I've discovered this old post and finally I've installed the IDE BIOS on it.

I used the DOS bios flasher in windows 98 dos mode. It isn't very intuitive but I could make it!

With a regular crappy sata2 160gb drive I reach about 70/70mbps, and 98 starts in 3/4 seconds... Maybe I can post some pics of the result.

(Now I'm searching for more 3512s in local sales...)

I have been using SIL3112 drivers with 3512 (though I use RAID mode with 9x), I only had to add new VID and PID into the INF. Flashers never worked for me, I programmed the chips directly with a dedicated programmer.

Tiido wrote:

I have been using SIL3112 drivers with 3512 (though I use RAID mode with 9x), I only had to add new VID and PID into the INF. Flashers never worked for me, I programmed the chips directly with a dedicated programmer.

Yep, I tried windows and dos flasher and only last one worked for me. Unfortunately I don't speak the language of welders and programmers, so sometimes I can feel my hardware limitation...

But it was a nice retro-adventure...

Ah, I also tried a C&C SSD (chinese and cheap) Kingspec 60gb drive... in my Celeron 1300 Tualatin, with this result:

Cannot post w98 disk benchmarking because that disk hasn't a fat32 partition, but I think it would be similar...

Recently I got myself a Delock SATA controller which is SiI 3512 based. Here's a link to manufacturers page: https://www.delock.de/produkte/G_70096/merkma … ?setLanguage=en. I found the card on a Polish eBay alternative for about 10EUR plus shipping. The plan was to put it in a computer which runs Windows 98, connect a SATA SSD drive and enjoy Windows 98 with no HDD noise and relatively fast transfer rates. By default controller came with SiI 3512 SATARaid BIOS flashed so it required reflashing to SiI 3512 SATALink BIOS, as it was described in one of the first posts. So, on the still attached IDE disk I've placed a BIOS file 'B4370.BIN', 'UPDFLASH.EXE', with help of 'F8' key booted into DOS mode and flashed the BIOS. Process went nice and smooth.

Next step was to format the disk and put the Windows 98 installation files there along with all the necessary drivers, for that I used SystemRescueCD booted over PXE. Everything 'so far' went fine and time came for Windows 98 setup. Here the problems started. Setup crashed immediately after welcome screen, returning 'write error' message. Booting from Windows 98 CD, formatting disk and starting installation from CD led to the same issue. I've also experienced interesting thing - some attempts to create directories didn't produced any errors, but the directories were not there.

Since I noticed that '43.70' was not the newest BIOS for the controller, I googled a bit and found 'BIOS-003512-xxx-4384.zip' package for SiI 3512. Inside are BIOS ROMs for both SATALink and SATARaid. Long story short, 'b4384.bin' did the trick. After flashing it I was able to successfully install Windows 98 and the driver found in 'SWD-003x12-00W-1361.zip' attached to one of previous posts (first I've installed VIA Hyperion drivers for MSI KT4AV KT400A motherboard). No issues so far.

Attached to this message is a BIOS I used for flashing the controller, also an ATTO Disk Benchmark screenshot showing read rates going up to about 66MB/s and write rates around 37MB/s.

carangil wrote on 2016-04-19, 15:41:

The UPDFLASG utility said 'FAILED RETRY Y/N' , but the computer was frozen, so I couldn't attempt again. When rebooting, where the option rom would usually run the computer hangs. I can't find an option to disable the option Rom on this bios.

Put the card in my socket AM3 computer. Blocks the BIOS completely, again can't find a way to skip the option rom.

Had the same issue with SIL3112. Was able to overcome it by short-circuiting all contacts on the top of card BIOS chip with wire. After Win98 booted, I've removed it with pliers. And successfully flashed card BIOS with updflash from there.
Not that I recommend that solution, as there is probably a chance to damage card. But if there is no other way, think about it as a last resort

Ok here is what i found out so far:

I have a Sil3512 pci Raid card here. The newest bios is apearently b4384.bin. I have the card in a 440BX Board. And Win98 installed on an ide harddrive.
I put the bios files and the flash program on the ide harddisk and flashed it in MD-Dos mode. He didn't accept the bios file at first but then just flashed it....

As a original raid card its not even possible to install a driver in 98.
As soon as its flashed to the non Raid bios it installs fine.

My plan was to use a ssd with it. So i connected a 120GB Sandisk and tested with Atto Disk 2.2
result about 30mb/sec
then i formatet the ssd on a win 10 computer because i heard this would take care of Alignment issues.
Funny enough it only offers Fat32 for partitions up to about 32GB. So i made such a partition for testing.
after this i connected it back to the 3512 controller.
slightly more performance but not much.

Tested a stanard 160GB sata Notebook harddrive with the contoller. 70MB/sec Read/write speed.

Win98 installs fine on the said notebook drive. Although fomat c: shows it as a 24MB Drive instead of 32GB.
But windows install immediately finds the hard drive as if it is ide connected to the mainboard.

in contrast the ssd brings quite some trouble. windows install finds it and checks it after starting setup. result: "bad sectors". checkdisk offers to correct it but with enter key put down it corrects endless errors for half an hour.

tried another ssd. on this one i cant even write install files (98cab).

does it sound like an alignment problem ? and why is it so slow on atto Disk ?