VLB IDE Controllers

From Vogons Wiki
Jump to navigation Jump to search

Full VL-bus IDE chip list

Software-configured

Brand Chip Earliest FAB Date IDE Ports Fastest Mode IO RDY Internal Strobes Config Method MR BIOS Support XUB Support Commandline Parameters Notes
ADD2 DW729 Late 93 2 PIO-3 Driver Example Parameters: D0:S/20 R/68 W/FF M/FF P/11
Dx --> x is the drive number starting a 0
S/xx --> hex sectors per transfer for the drive <=32
R/ar --> Read cycles. a = active - 1; r = recovery - 1. Example: R/68 = 9 active cycles, 7 recovery cycles
W/ar --> Write cycles, same format as Read
M/ar --> Other I/O cycles, same format as Read
P/xx --> Bit 0: Enable read-ahead, Bit 1: Enable Post Write, Bit 2: Disable read waits, Bit 3: Disable write waits, Bit 4: Enable 32-bit transfers, Bits 5-7: Undefined. Example: P/11 means Enable read ahead and 32 bit writes
/X --> Skip speed test
Notes: Driver code is XOR-encrypted
Other names: Dataworld DW729
Detection: Bases 100h-1C0h: out base,85h unlock, register 0Ah reads 41h ('A'). Genuine part: register 17h = B5h
Timing registers: Registers 80h-87h at base+1 after each 85h unlock
Examined: ADD_IDE.SYS 1.03
Appian ADI/2 Mid 93 1 PIO-4 Yes Yes Driver 3.21 verbose --> Prints debug output while loading
floppy --> Changes chip register 0Fh (handling of port 3F7h) so the chip works alongside a floppy controller
initonly --> Programs the chip but leaves the disks on the BIOS; also turns multiple mode off
noinit --> Skips the chip set-up routine; the driver still loads
io16 / io32 --> Forces 16-bit or 32-bit data transfers instead of auto-detecting
b=n --> Bus clock period in ns: b=30 = 33 MHz, b=25 = 40 MHz, b=20 = 50 MHz
c=m --> Enable multiple-sector (block) transfers for C:
c=s --> Disable multiple-sector (block) transfers for C:
ctime=320,320 --> set PIO 0 for Drive C:
ctime=300,270 --> set PIO 1 for Drive C:
ctime=210,150 --> set PIO 2 for Drive C:
ctime=80,100 --> set PIO 3 for Drive C:
ctime=70,50 --> set PIO 4 for Drive C:
Notes: flexidrv.sys from Adaptec seems like the best driver. FLEXIDRV rejects the DMA options (,a) on this chip, and using one drops both drives to mode 0
Other names: ALi M5213; CL-PD7220; AIC-25VL01
Detection: Ports 35h/B5h must differ from the PIC masks at 21h/A1h. Then an ID echo at 0AD4h, 34h or B4h: ID below 38h = AIC-25VL01 (38h+ is the AIC-25PC10)
Timing registers: ID at base, index at base+4, data at base+8 (write the ID first). Registers 3/4 (drive 0) and 8/9 (drive 1): active clocks in bits 7-4, recovery clocks in bits 3-0. 7/12: address setup. 1: config; 13: revision
Examined: FLEXIDRV.SYS 2.14
Adaptec AIC-25PC10 Not seen yet Not seen yet MWDMA-2 Yes Yes Driver 3.21 All of the ADI/2 options, plus:
c=m,a --> Multiword DMA for C:
c=s,a --> Single-word DMA for C:
c=a --> DMA for C:, multiword if the drive supports it, otherwise single-word
ctime=60,40 --> PIO 5 timing for C:; only faster than PIO 4 at 40 or 50 MHz
Notes: The AIC-25PC10 seems to have additional functionality. Its DMA is on the IDE cable only: the driver sends READ/WRITE DMA, but the CPU still moves the data through the data port
Detection: VL-bus mode: same probe as the ADI/2, with an ID echo of 38h or higher at 0AD4h, 34h or B4h. PCI mode: Appian 1097:0038
Timing registers: As the ADI/2, but the data port is a word wide. DMA timing in registers 23h/24h (drive 0) and 28h/29h (drive 1); bit 4 of 7/12 set for DMA
Examined: FLEXIDRV.SYS 2.14
Atronics IDE-2015PL Late 94 2 PIO-4 Driver /Dx:Px where x= 0-4. /T:25/33/40/50 picks the bus-clock table Notes: 2015PL.SYS /Dx:P4
Other names: DTC 803
Detection: Index E0h/data E4h, then B0h/B4h; register echo with 3F6h reads between accesses
Timing registers: One register per drive through the same index/data pair. 2015PL.SYS: per-clock tables for PIO 0 / 3 / 4 at 25-50 MHz; /T:66 reads past the tables (bug). Linux dtc2278.c writes A0h/C3h for PIO 3+ at B0h/B4h
Examined: 2015PL.SYS 3.27, DOSEIDE.SYS 2.1A
CMD PCI0640B Late 94 2 MWDMA-1 Yes Yes Driver 3.21 /PIOMODE:X where x= 0-4. CMD640X.VL 2.09 uses /A<MHz>, /O<port><drive><mode>, /H Notes: Advanced features cause data corruption in multitasking environments.
Hidden switches: /CMD, /FOO:BAR
Detection: Index 178h/data 17Ch (else 78h/7Ch): register 50h not 00h/FFh, bit 5 set; register 5Bh echoes BDh
Timing registers: Registers 51h-58h; per-drive clocks computed from /A<MHz>
Examined: CMD640X.VL 2.09
Promise PDC-20230c Late 93 1 PIO-3 Yes Yes Driver 3.21 Yes 0(slow), 8 (fast), A (iordy). The PROMISE.SYS 3.3 README lists levels 0-8. Config-mode 1F3h speeds 0-7: 630 to 240 ns at 33 MHz; jumpers give 2, 4, 6. Levels 0-8 confirmed ; PROMISE.SYS has 9-entry tables, one per level Other names: VG3, TurboChip C
Detection: Config-mode unlock on the IDE ports: in 1F5h, set 1F2h bit 7, in 1F2h, in 3F6h ×2, in 1F2h ×3; bit 7 reads back clear = Promise
Timing registers: Index 1F8h/data 1F9h: registers 2n/2n+1 get a word from one of four tables chosen by the measured bus speed. Register 7 = transfer status. No chip ID read; same code for every Promise chip. Four 9-entry tables (one per measured bus class), shared by all three Promise chips
Examined: PROMISE.SYS 3.3
Promise PDC-20630 Mid 94 2 WDMA2 Yes Yes Driver 3.21 Yes 0(slow), 8 (fast), A (iordy), M (mwdma) Notes: READ/WRITE DMA (C8h/CAh) enables MWDMA to the chip, but data still moves to memory by PIO from 1F0h. Driver only seems to set MWDMA-1 (SET FEATURES 21h, DMA levels 7-8); it never sends 22h
Other names: VG4
Detection: As PDC-20230c, then 1F4h bit 3 makes 1F8h read back its index (bits 0-2)
Timing registers: 1F3h speed as PDC-20230c, plus four timing registers through 1F8h/1F9h (enabled by 1F4h bit 3)
Examined: PROMISE.SYS 3.3 (same code path; no chip check)
QDI Vision 6500 Mid 93 1 PIO-2 Driver Yes Setup configures per drive. Driver picks a timing class from its drive table Detection: Register 31h (or B1h): upper nibble C0h; bit 3 set = disabled
Timing registers: One timing byte at 30h/B0h (active bits 0-2, recovery bits 4-7)
Examined: QD6500.SYS 2.4
QDI Vision 6580 Late 94 2 PIO-4 Card Half Driver Yes Setup configures per drive Notes: PIO timing based on bus clock makes the controller less robust on some systems
Detection: Register 31h (or B1h): upper nibble A0h or 50h
Timing registers: As QD6500, plus secondary-channel timing
Examined: QD6580.sys 3.7
SIS 83C601 Early 95 1 PIO-4? Yes Yes Driver 3.21 Waitstate 1 (fast) -7 (slow) Notes: 83C601 has newer fab dates than the 83C611.
Other names: Might also work as a PCI controller PCI ID 1039:0601
Detection: in 1F3h; in 1F2h ×4; out 1F2h,F3h; 1F2h reads D3h
Timing registers: 1F2h ← F0h + (wait state − 1) after the same unlock
Examined: S134XM1.SYS 2.0, S134XD1.COM 3.5; SIS611.SYS
SIS 83C611 Late 93 1 PIO-4? Yes Driver 3.21 Wait state 0-7 for each channel Notes: Older than the 83C601
Other names: Driver banner says "VL/PCI-IDE"; the PCI check is unused
Detection: Same 1F2h sequence (0C14h)
Timing registers: 1F2h: F3h, B5h, then computed wait states
Examined: SIS611.SYS 2.11
UMC UM8672F Late 94 1 PIO-2 Driver Speeds 0 (slow) -11 (fast) Notes: PIO-4 with Modded Driver. Driver calls speed 10 the "proper speed" for Enhanced IDE drives
Other names: UM82C872F
Detection: None: 108h/109h unlocked with 5Ah, no ID read back
Timing registers: Registers D0h-DDh through index 108h/data 109h; all set to CBh (slowest) first. Six 12-entry tables in UM8672.SYS (chip ID A0h/A2h × 3 bus-speed classes).
Examined: UM8672.SYS 2.1
VIA VT83C461 Mid 94 2 PIO-2 Yes Driver PIO_TIME=, DMA_TIME=, SET_DMAM= Notes: Has a BIOS / AV140 / AV150 / AV300
Other names: Driver also covers the VT82C496G
Detection: Index A8h/data A9h: register 00h upper nibble 70h/90h, register 71h bit 0 set
Timing registers: Registers 71h (IDE enable), 93h (timing/mode)
Examined: VIAIDE.SYS 4.02
Winbond W83759AF Early 95 2 PIO-3 Yes Yes Driver /D0:13 (speeds 0-13) Notes: software speed config and "advanced mode". Measures the bus clock with the PC timer. Might support PIO 4; "A" version is software configurable
Other names: DTC 2278E Rev W.
Detection: Bases 4B0h/5B0h… then 430h/530h…: 63h written at base+0 reads back at base+0Ch; register 87h low nibble = revision; register 80h bit 7 set
Timing registers: Index base+4, data base+8
Examined: WBIDE.EXE, DOSEIDE.SYS 2.1A (at 5B0h only)
Micronics MIC491 Mid 93 1 PIO-2? BIOS Setup IDE Speeds: Normal / Medium / Fast
(register 104h = 69h / 67h / 63h)
Notes: No driver; but a Gateway BIOS seems to set the controller speed
Other names: Onboard on the Micronics GJX30G; BIOS says "Gateway 2000 Local Bus BIOS"
Detection: None. The BIOS only checks that no other IDE controller answers at 1F0h
Timing registers: out 9Ch,7Fh opens ports 102h-106h; 104h = speed, 102h bit 0 = IDE on
Examined: GJX30G BIOS 09-24-93

Jumper-configured

Brand Chip Earliest FAB Date IDE Ports Fastest Mode IO RDY Config Method Speed Settings Notes
Chaintech CT9000 Late 93 1 PIO-3 Yes Jumper 0-3 Notes: Not found yet: ct9000.zip
DTC DTC 805 Early 94 1 PIO-2 Jumper 3.9MB/s - 7.2MB/s Notes: DOSEIDE.SYS has no code for it; it only shows the drive's reported rate
Other names: DTC 2278VL, jumpers W1-W5
Examined: DOSEIDE.SYS 2.1A
GoldStar GM82C711 Early 93 1 PIO-3 Jumper 0-7 Notes: GS_DOS.SYS driver does not set timing; Not a PC Chips 82C711
Examined: GS_DOS.SYS 1.21 (driver doesn't name the chip)
GoldStar GM82C712 Mid 95 1 PIO-4 Yes Jumper 0-7 Notes: Same driver as GM82C711
Other names: CompassLab CL3202.
Examined: GS_DOS.SYS 1.21
Holtek HT6560A Late 94 2 PIO-2 Jumper 0 (slow), 3 (fast) Notes: Driver Not found yet
Holtek HT6560B Early 95 2 PIO-3 Yes Jumper + Driver 0 (slow), 3 (fast). Driver /Dn=rt,at: recovery 2-16, active 2-15 bus clocks Notes: Driver allows AT:2 RT:2
Detection: 3E6h handshake: after three out 3E6h,0 / in 3E6h rounds some of bits 0-5 read back clear; on the fourth all six are set
Timing registers: Four reads of 3E6h, then config byte to 3E6h and timing byte (recovery << 4 OR active) to 1F6h. Linux pata_legacy 5.15 agrees
Examined: HTIDE.SYS 2.41
Lion Lion 3+ Mid 95 2 PIO-3 Jumper + Driver 0 (slow), 2(fast). Driver will lower speed if certain drives are detected Notes: Driver doesn't support ECHS or LBA
Detection: Config window on 1F0h/1F1h (opened by reads of 1F3h, 1F2h, 3F6h); register 5 & E0h = 60h
Timing registers: Per-drive registers through the 1F0h/1F1h window
Examined: LION3.SYS
OPTI 82C611 Early 93 2 PIO-3 Jumper Notes: Probably programmed the same way as the 82C611A; to be tested
OPTI 82C611A Late 94 2 PIO-3 Card Jumper + Driver 0 (slow), 3 (fast). SETUPVIC: read/write pulse, recovery, address setup in ns; 25/33 MHz Detection: 1F3h echoes C0h; then in 1F1h ×2, out 1F2h,03h (config mode), 1F3h changes, out 1F2h,83h
Timing registers: In config mode: 1F3h selects the drive bank, 1F1h = write timing, 1F6h = setup and read bank, 1F0h = read timing, 1F5h bits 0-1 = bus-clock strap. Linux pata_legacy 5.15 uses the same sequence for the 82C611A
Examined: optivic.sys 1.0, opti-pio.exe, opti-set.exe
PiC 1288A-125 Late 93 1 PIO-2 Yes Jumper 0 (slow), 3 (fast) Notes: No driver found
Other names: VIDE-1
Promise PDC-20230b Mid 93 1 PIO-3 Jumper 0 (slow), 2(fast) Notes: Works with the PROMISE.SYS 3.3 drivers.
Other names: CompassLab CL3201 (fab 9314); TurboChip B
Detection: Write 55h to 1F2h, read it three times back to back: a 20230-B returns 00h on the third read (Soucek; also unreached code in PROMISE.SYS at 2D3Bh). Needs the same three reads as a "VLB sync" before 32-bit I/O
Examined: PROMISE.SYS 3.3
UMC UM82C871F Mid 93 1 PIO-3 Jumper Best speed: AT:3 RT:4 Notes: Driver only uses standard IDE commands; its F5 setup toggles multi-sector mode
Examined: UM82C871F.SYS
UMC UM85C418F Late 93 1 PIO-2 Jumper Best speed: AT:5 RT:9 Notes: Combo IDE & VGA controller; The driver writes register D1h when certain drives are detected. Stability issues at fastest speeds
Detection: None: 108h/109h unlocked with 5Ah, no ID read back
Timing registers: Register D1h through index 108h/data 109h (one value for both drives)
Examined: UM85C418F.SYS 1.07
Winbond W83759F Early 94 2 PIO-2 Jumper 600ns, 500ns, 400ns, 240ns Notes: winbond_port (Linux pata_legacy parameter; default base 130h, alternate 1B0h)
Detection: Same probe as the W83759AF? WBIDE reads a revision letter; DOSEIDE needs revision B or later. Linux only programs the W83759A, after turning off jumper mode (register 81h bit 7).

No configuration

No driver or setup utility has been found for these chips.

Brand Chip Earliest FAB Date IDE Ports Notes
QDI Vision 5500 N/A 1 The existence of this chip was inferred by the existence of a QDI5500 VLB IDE controller in the TH99 repository. No photos found.
Tans TS8310 Mid 94 2 References to ts8310.zip exist on BBS sites, but no uncorrupted versions exist.
TANS BIOS files have been found but are controller agnostic.
Three option ROMs from TANS cards checked:
TANS "Enhance IDE driver" v2.00 (1997)
SIIG v3.10, which resembles v2.00 with a name change
TANS "Enhance IDE driver" v2.10 ">8.4GB" (1999)
All use only the standard IDE ports at 1F0h/170h/1E8h/168h:
No chip detection, no timing writes, and 16-bit rep insw/outsw only (no 32-bit I/O)
Tans TS8310A Late 94 2 Same as above
Mach Speed GTK6189 Early 94 1 No driver found.
Frequently found paired with a Cirrus CL-GD5428 VGA;
The 40-pin IDE header has never been seen fitted, some traces run to the VGA chip.
The card ROM is a standard Cirrus VGA BIOS with nothing about IDE. Jumper positions are unpopulated except one that is hard-shorted, probably to disable the IDE function. Possibly IDE functionality was dropped late in the design
Opus Op VIC1 Late 93 1 No driver found.
"Op VIC" = Opus Vesa Ide Controller ?
Opus Op VIC3 Early 94 2 No driver found.
"Op VIC" = Opus Vesa Ide Controller ?