VOGONS


First post, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie

Here is the seller's photo and more text below for those who like their CPU porn with a story:

uVIDy6c.jpg

The story: I was looking to buy a 40MHz-rated 387SX chip. Turns out it's not easy to find one (most top at 33MHz and won't operate reliably at 40MHz) so I got this motherboard to pull it out from. Wasn't expensive at all, some 30 EUR, no big deal. And the mobo is very interesting as well:
- it's a 386SX class (16-bit bus)
- but it has cache
- and VLB slots!

Now, cache on SX is rare as hen teeth. VLB slots? First time I see one of those. And the BIOS sticker says "486SLC", not 386. The CPU is missing though and it wasn't so much desoldered as ripped off, with 2 pads now gone as well - but I've repaired worse.

I started investigating the missing CPU, and got stumped. I'm pretty sure this was a 486SLC from the factory and never had a 386SX soldered in there. The 0ohm jumper resistors don't make much sense though when looking at Cyrix 486SLC pinout. Like, KEN# is connected but FLUSH# is not? SUSP# is tied high. Why bother?
So here's me thinking this had an IBM chip in it. I've heard those have different pinouts but so far the power pins check out, the bus ones (as far as I can tell) do too, only the special ones (N/C on a true 386SX) are a mystery. Plus there is a 3V3 voltage regulator there, the PCB is actually designed for a low-voltage 5V tolerant CPU.

I've dumped the BIOS and it's a pretty generic AMI except seems to have Power Management Setup entry, and there's one string that says "Alaris, Inc." - and that's a clue: https://www.cbronline.com/news/ibm_oem_pacts_ … rom_eteq_alaris

So, TLDR: I need to ID this board and make sure it had IBM chip in it. If it did, I might try and hunt one down to maybe try and repair it - I think it's a very cool mobo and could still live. I am also very much interested in any IBM docs for their SLC series, even a confirmed pinout would be of great help to me.

BTW both the GREEN MATH 4C87SLC-40 and the RAM I got with the mobo are OK and there's no battery spill so I'm rather confident it will work if the CPU is replaced and the 2 pads rewired.

Reply 1 of 15, by keropi

User metadata
Rank l33t++
Rank
l33t++

I have the same mobo and indeed it has an IBM 486SLC2 66mhz soldered next to the fpu - it's fast and the vlb slots work fine.
I cannot tell you the exact model of the cpu because there is a heatsink glued on it.
The mobo is the ALARIS LEOPARD 486SLC2 -> https://stason.org/TULARC/pc/motherboards/A/A … SLC2-REV-C.html

🎵 🎧 PCMIDI MPU , OrpheusII , Action Rewind , Megacard and 🎶GoldLib soundcard website

Reply 3 of 15, by keropi

User metadata
Rank l33t++
Rank
l33t++

fast graphics , this is not a 386sx board I'll post some benches later

edit:

DOOM benchmark gives 16 fps

C4yDXXwl.png

J1lwWYtl.png

zWnIAONl.png

BIOS stock settings and I used a S3 928 VLB vga. What is bad about this mobo is that it only has 4 simm slots so you need to hunt down 4MB modules to install more than 4mb...

🎵 🎧 PCMIDI MPU , OrpheusII , Action Rewind , Megacard and 🎶GoldLib soundcard website

Reply 4 of 15, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie

Ah, "LEOPARD". Adding this word to the search phrase makes google not stupid again. So it is IBM mobo and the wire is factory job but all the photos have the CPU covered with a heatsink. I suppose that is why mine is missing the CPU - someone tried to get the heatsink off...

The only photos I found without heatsink are some russian one and the hosting expired so I'm looking at a post-stamp size miniature that doesn't tell me anything. But, funnily enough, I've also found Vogons posts about this mobo and it seems it more-or-less matches the TI 486SXLC performance? Well then I suddenly lost a lot of interest in it, I've already put an SXLC2 on a cache-less mobo in place of 386SX and at 50MHz it's pretty much the max of what can be had. Unless I could enable clock tripling or mod the clock gen on the IBM.

BTW I found that Trident 8900D, while not the fastest ISA card, is actually plenty capable in these SX systems. It's too bad it has some vertical banding (it's not very bad like on Realteks) but I got it cheap. So considering the SX would need 2 memory transactions to write to VLB slot anyway, I don't think it's that much of an improvement.

EDIT: My TI486SXLC2 at 50MHz (25MHz bus):

 CACHECHK V7 11/23/98  Copyright (c) 1995-98 by Ray Van Tassle. (-h for help)
CMOS reports: conv_mem= 640K, ext_mem= 3,200K, Total RAM= 3,840K
486 Clocked at 52.7 MHz
Reading from memory.
MegaByte#: --------- Memory Access Block sizes (KB)-----
1 2 4 8 16 32 64 128 256 512 1024 2048 4096 <-- KB
0: 27 26 26 27 74 74 74 74 74 74 -- -- -- us/KB
1: 27 26 26 27 74 74 74 74 74 74 74 74 -- us/KB
2 3 4 <--- same as above.

Extra tests----
Wrt 42 42 42 42 42 43 42 42 42 42 42 42 --<-Writing
Reading at different byte offsets, at megabyte #1. Byte offset of...
0= 26 27 26 28 75 74 74 74 74 74 74 74 -- us/KB
1= 32 32 32 34 111 111 111 111 111 111 111 111 -- us/KB
2= 32 32 32 34 90 90 90 90 90 90 90 90 -- us/KB
3= 32 32 32 34 113 111 111 111 111 111 111 111 -- us/KB
4= 26 27 26 28 75 74 74 74 74 74 74 74 -- us/KB
8= 26 26 26 28 75 74 75 74 74 74 74 74 -- us/KB
12= 26 26 26 28 74 74 74 74 74 74 74 74 -- us/KB
This machine seems to have one cache!? [reading]
!! cache is 8KB-- 40.5 MB/s 25.9 ns/byte (273%)
>>>> If you think you do have L2 cache, you might have FAKE CACHE chips! <<<<
5.2 clks
Main memory speed -- 14.8 MB/s 70.7 ns/byte (100%) [reading] 14.2 clks
Effective RAM access time (read ) is 282ns (a RAM bank is 4 bytes wide).
Effective RAM access time (write) is 161ns (a RAM bank is 4 bytes wide).
486 Clocked at 52.7 MHz. Cache ENABLED.
Options: -t0 -z

And some other benches:

** (RAM 4MB, 0 WS; ISA 0 WS)
JP1: 1-2

* NSSI (/SAFE):
20113 Dhry/s (fi 20188, 16-bit code)
3280 kWhet/s (fi 3292, real mode)

* SI 8.0:
73.8

* Speedsys 4.78:
22.24

-- Trident TVGA8900D (1024k x32):

* TOPBENCH:
118

* Speedsys 4.78 VESA mem:
3445 kB/s

* 3DBench (1.0, 1.0c):
22.7, 22.3

* PC Player Benchmark (320, 640):
5.7, 2.2

* Landmark 6.00:
203 MHz AT, N/A, 4147.85 chr/ms

* Doom (low, high):
2134/2182, 2134/7267

* FractBench:
Fixed: 0m:07s.19
Float: N/A

* Chris's 3D Benchmark (320, 640):
N/A, N/A

Reply 5 of 15, by keropi

User metadata
Rank l33t++
Rank
l33t++

the performance boost is significant, your DOOM benchmark is 10.2fps - the leopard one on my test scored 15.99
so that's +60% , not small keeping in mind this is still built around the 386sx limitations. I assume most of the speed bump comes from cache - vlb slot just lets you use better graphics cards without having to hunt the better ISA ones.
Certainly a more useful board than some upgraded cacheless 386sx one IMHO - provided it still had the cpu 🤣 🤣 🤣

Last edited by keropi on 2019-03-12, 11:43. Edited 1 time in total.

🎵 🎧 PCMIDI MPU , OrpheusII , Action Rewind , Megacard and 🎶GoldLib soundcard website

Reply 6 of 15, by Deksor

User metadata
Rank l33t
Rank
l33t

Yeah I know it's not really a 386SX, but a 32 bit bus with a CPU having only a 16 bit bus sounds kinda silly ^^

I wonder how that works (probably some weird things happening there, like with vlb pentium board's)

These benchmarks are quite interesting though as I plan to upgrade my 386SX25 to the Ti486SXLC2 and I've never seen benchmarks of it before. The speed looks very nice, it looks quite close to the 486DX2 despite the slower bus !

Trying to identify old hardware ? Visit The retro web - Project's thread The Retro Web project - a stason.org/TH99 alternative

Reply 7 of 15, by keropi

User metadata
Rank l33t++
Rank
l33t++

to my knowledge these were still budget boards, not meant to compete with fully fledged 486 systems so they made sense back then for that kind of market
Now they are just a curio for collectors or something 🤣

🎵 🎧 PCMIDI MPU , OrpheusII , Action Rewind , Megacard and 🎶GoldLib soundcard website

Reply 8 of 15, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie
keropi wrote:

the performance boost is significant, your DOOM benchmark is 10.2fps - the leopard one on my test scored 15.99

Eh, yes, but that's down to my ISA VGA and the fact that Doom just loves 486 cache so 16kB of L1 is probably making more difference than the extra L2.
For example, another mobo I modded to accept 486SXL this time, so true 32-bit but still with ISA and on 40MHz clock:

486SXL 40MHz + Cyrix CX-83D87-40-GP FASMATH

** (RAM 8 MB, 1/1 WS, cache W-Back 0/0 WS Burst Enabled, BARB)

* NSSI (/SAFE):
18619 Dhry/s (fi 18660, 16-bit code) - pretty much matches i486DX-33
5586 kWhet/s (fi 5603, real mode)

* SI 8.0:
65.7

* Speedsys 4.78:
10.25

-- Trident TVGA8900D (1024k x32):

* TOPBENCH:
126

* Speedsys 4.78 VESA mem:
4851 kB/s

* 3DBench (1.0, 1.0c):
23.2, 22.8

* PC Player Benchmark (320, 640):
6.4, 2.4

* Landmark 6.00:
168 MHz AT, 165 MHz 80287, 5146.81 chr/ms

* Doom (low, high):
2134/1768, 2134/6367

* FractBench:
Fixed: 0m:08s.84
Float: 0m:32s.35

* Chris's 3D Benchmark (320, 640):
19.6 (11.8 fps), 6.1 (3.6 fps)

Same VGA card, slower CPU but way faster bus and I still get an improvement in Doom. And I can probably make it better still by jury-rigging a proper FLUSH# circuit rather than use BARB option - the whole purpose of that mod, BTW. Still haven't got time to actually do it.
And here's a true 486 (U5S actually) at 40MHz in a lousy OPTi495SLC with very loose timings:

U5S 40MHz

** (safe: RAM 8 MB, 0/0 WS, cache 3-2-2-2, 1 WS)

* NSSI:
26524 Dhry/s (fi 26750, 16-bit code) - in the middle between Am486Dx-40 and i486DX2-66
3380 kWhet/s (fi 3392, real mode)

* SI 8.0:
111.8

* Speedsys 4.78:
21.12

-- Cirrus CL-GD5424 (VLB):

* TOPBENCH:
190

* Speedsys 4.78 VESA mem:
8883 kB/s

* 3DBench (1.0, 1.0c):
38.4, 37.7

* PC Player Benchmark (320, 640):
7.3, 2.8

* Landmark 6.00:
188 MHz AT, N/A, 9187.29 chr/ms

* Doom (low, high):
2134/1331, 2134/3862

* FractBench:
Fixed: 0m:09s.44
Float: N/A

* Chris's 3D Benchmark (320, 640):
N/A, N/A

It beats the 66MHz IBM Doom score. All thanks to 32-bit burst transfers on VLB. And that is way cheaper and easier to build so while I'm still intrigued by that mobo, I've just lost my drive to fix it. It's going to be shelved.

Deksor wrote:

These benchmarks are quite interesting though as I plan to upgrade my 386SX25 to the Ti486SXLC2 and I've never seen benchmarks of it before. The speed looks very nice, it looks quite close to the 486DX2 despite the slower bus !

Well I did that by putting a 3V3 chip with 5V tolerant I/O on a standard PC mobo. I did add a heatsink and I'm still below absolute max ratings for the TI chip but I can't guarantee it won't die prematurely. I don't mind though, it's only for testing. Here's a the original 386SX-33MHz scores with both Realtek and Trident to compare:

386SX 33MHz

** (RAM 4MB, 0 WS)

* NSSI:
6692 Dhry/s (fi 6718, 16-bit code)
591 kWhet/s (fi 593, real mode)

* SI 8.0:
25.4

* Speedsys 4.78:
6.45

-- Realtek RTG3105 (512k):

* TOPBENCH:
43

* Speedsys 4.78 VESA mem:
833 kB/s

* 3DBench (1.0, 1.0c):
7.8, 7.8

* PC Player Benchmark (320, 640):
1.9, N/A

* Landmark 6.00:
47 MHz AT, N/A, 1100.83 chr/ms

* Doom (low, high):
2134/5194, N/A

* FractBench:
Fixed: 0m:29s.66
Float: N/A

* Chris's 3D Benchmark (320, 640):
N/A, N/A

-- Trident TVGA8900D (1024k x32):

* TOPBENCH:
68

* Speedsys 4.78 VESA mem:
3448 kB/s

* 3DBench (1.0, 1.0c):
10.4, 10.3

* PC Player Benchmark (320, 640):
2.1, 0.9

* Landmark 6.00:
47 MHz AT, N/A, 4147.85 chr/ms

* Doom (low, high):
2134/4648, N/A
Show last 7 lines

* FractBench:
Fixed: 0m:29s.55
Float: N/A

* Chris's 3D Benchmark (320, 640):
N/A, N/A

Reply 10 of 15, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie

Haha, nice, the poor CPU-less mobo got immortalized. I wish I had a better photo to offer but alas it became a parts donor for other motherboards. So now it's even less complete, but at least didn't go to trash.

Reply 11 of 15, by stamasd

User metadata
Rank l33t
Rank
l33t

I have the little brother (or sister) of this motherboard, the IBM Opal LX (no VLB slots and no L2 cache). It is however complete and functioning. I see that it's not in the UltimateRetro database so I'll try to add it to that soon. I have the original manual too.

vYpmtsI.jpeg

(and no don't worry - what you see next to the keyboard connector isn't a barrel battery; it's been replaced by a bank of supercaps)

I/O, I/O,
It's off to disk I go,
With a bit and a byte
And a read and a write,
I/O, I/O

Reply 12 of 15, by Deksor

User metadata
Rank l33t
Rank
l33t

Awesome !

If you're not a fan of discord, you can upload the files in this thread [CLOSED] The Retro Web submissions thread: suggest edits and/or add new boards here

Edit : actually it looks identical to this one https://www.ultimateretro.net/en/motherboards/442
strange that they haven't the same name. Anyways, photos and manual scans are still welcome (it might help to figure out why they're named differently as well. Who knows ?)

Last edited by Deksor on 2022-02-08, 18:51. Edited 1 time in total.

Trying to identify old hardware ? Visit The retro web - Project's thread The Retro Web project - a stason.org/TH99 alternative

Reply 14 of 15, by Deksor

User metadata
Rank l33t
Rank
l33t

Yes, checkout the footer of ultimateretro 😀 (also I've edited my previous message)

Trying to identify old hardware ? Visit The retro web - Project's thread The Retro Web project - a stason.org/TH99 alternative

Reply 15 of 15, by stamasd

User metadata
Rank l33t
Rank
l33t

Yes, that is it. I can dump the bios and scan the manual too. Mine was sold originally under the IBM brand, but was made by Alaris.

I/O, I/O,
It's off to disk I go,
With a bit and a byte
And a read and a write,
I/O, I/O