Good buy!

Have you tested it yet?

Tetrium wrote:

Good buy!

Have you tested it yet?

I'm gonna really try and find some time next weekend. 😀 Hopefully I'll post back in under 7 days ...

What?? I missed another one!!?? 😁

Great find though! 😀

Congratulations again 😉

Just looked at the photo and something seems wrong to me... Please make sure the CPU is inserted the right way around. I mention this because on my board the writing on the CPU appears turned 90° to the right when the CPU is inserted correctly. I may be wrong, as I only have a simple 486DX to look at, and the lettering may be different on DX4 CPUs (can't look at mine any more, because they have a permanently glued heatsinks).
486 CPUs are easily inserted the wrong way round, they have no key pins to prevent this. The manual should have some info about the correct orientation.

5u3 wrote:

Just looked at the photo and something seems wrong to me... Please make sure the CPU is inserted the right way around. ...

Yikes - very well spotted! When I took that photo, I did glance at the CPU and thought that there was something a bit different about the wording written on it. At the moment, I am "out of town" and cannot examine the mobo. However, I can check it out this Friday. In fact, I'm determined to find the time and unpack my retro equipment and begin testing old hardware again.

So with that in mind, I'd like to make a list of as many ways to test this mobo as possible. Here's a list that I can think of -

Speedsys version 4.78
SuperPi
Memtest (versions 3.5 and 3.6)
Pcpbench (I'm not sure what graphics cards I will have available)
Superscape 3Dbench 1.0
Superscape 3Dbench 1.0c (I think)
Cache Check 7
CTCM7
CTCM
Doom (framerate test)
Disable L1 and also L2 cache, and redo all tests.

Some more tests:

* Boot up using Windows 98 to see if the desktop appears OK and Device Manager shows no conflicts.

* Replace the DX/4 CPU and try other CPUs. Eg AMD, and other slower Intel CPUs. Also, can this board be overclocked?

* Check to see what the BIOS date stamp is, and if necessary update it using the BIOS update file found on the Asus FTP server located at - ftp

I would really like to run some more tests if anyone can think of other utilities that I can download. Thanks!

retro games 100 wrote:

* Replace the DX/4 CPU and try other CPUs. Eg AMD, and other slower Intel CPUs. Also, can this board be overclocked?

Yes, but there is not much headway for overclocking, as the board lacks a PCI clock divider.
In fact, if you configure the board for a 40 MHz CPU (e.g.: AMD DX2/80), the PCI bus is overclocked by default, running at 40 MHz instead of the max. allowed 33 MHz.
This is what makes the board very fast in comparison with other 486 PCI boards, despite the rather poor PCI implementation.
There is also the possibility to run the bus at 50 MHz, but it's very difficult to get it stable at this speed. It's only possible by omitting all PCI cards, hand-selecting a VLB VGA that works at 50 MHz, exchanging the L2 cache for faster chips and lots of waitstates. In the end, I don't think it would be worth the effort.
I got the best results running an Intel DX4 at 120 MHz (40x3) or an AMD 5x86 at 160 MHz (40x4).

The board can be slowed down to low-end 386 speeds by running a 486DX at 25 MHz and disabling the caches, and there is still the turbo switch.

Thanks very much for the info, 5u3! And your observation about the incorrect looking CPU orientation was absolutely correct! I reseated the CPU in to its correct orientation, and also looked through the jumper settings using the supplied manual. Crucially, the "CPU type" settings were incorrect. The jumpers were set to a "regular/standard" DX4 CPU. I reconfigured these jumpers for a "write back" DX4 configuration, because the CPU has &EW written on it, and its CPU code is SK096, which is a "write back" and enhanced DX4 chip.

I am using a Diamond Riva 128 4Mb PCI graphics card. I power on the mobo, and see the POST info on screen. The onscreen BIOS information says -

#401A0-0306

06/25/96-SiS-496-497B-PVI-4SP3C-00

Looking at the first onscreen BIOS datestamp above, it looks like the BIOS ("0306") is the final version. BTW, the Asus BIOS (and other) files are kept here -

ftp://ftp.asus.com/pub/ASUS/mb/sock5/486sp3/

Before I max out the BIOS settings and begin testing, I attach an old 128Mb Compact Flash drive to a 40-pin IDE cable, and attach it to the mobo, then do the BIOS "auto HDD detection" operation. If I select the "LBA" option choice, I get a "Missing operating system" error when the Operating System is about to boot up, and so I select the "Normal" BIOS HDD choice, which allows me to boot in to DOS 6.2 OK. Also, I attach a DVD-ROM reader to the IDE ribbon cable, and this does not cause any problems when I boot up. It's very useful to use a CF reader device which does not directly attach to the mobo's integrated IO pins, but instead can be attached to the IDE ribbon cable, because you can then also use an optical drive unit using the same single cable (in slave mode.)

Tests:

1) Speedsys 4.78. See attachment screenshot. Overall score = 42.32
2) SuperPi. I'll do this test today, and report back...
3) Memtest version 3.5. Passed. (I will skip Memtest version 3.6.)

I'll do all the other tests today and report back...

I still have to get a PCI 486 motherboard. I can't find any that aren't selling for ridiculous sums of money though. 🙁

Very nice! 😉
Your board gets identical results to mine. According to the Speedsys results, the write-back-cache on the CPU works correctly (this does not reflect on the "Processor benchmark" score, but the L1 cache results rises from ~75 MB/s to ~95 MB/s).

A newer beta BIOS (v. 3.07) exists, it fixes the year2k bug. It does introduce a few more bugs however (at least on my board).

Congratulations, you found the base of the ultimate 486. 😁

Tests continued:

2) SuperPi, using Windows 98 SE.

16K = 24.280 seconds
64K = 2 minutes and 05.315 seconds
256K = 11 minutes and 25.570 seconds
1M = 56 minutes and 01.016 seconds. Score = 297.5290

4) Pcpbench, using a Diamond Riva 128/Viper V330 4Mb PCI graphics card. I am using Pcpbench at the windows 98 desktop. Currently, there is no Windows driver loaded for the V330 graphics card - it's just in VGA mode.

Mode 100 (640x400 8bpp LFB) = 8.5. The "help screen" tells me that the score range to be expected for a DX4/100 is 6.7 to 9.0, so I'm very happy with 8.5.

Mode 101 (640x480 8bpp LFB) = 7.6
Mode 130 (320x200 8pbb LFB) = 19.3
Mode 13d (640x400 16bpp LFB) = 6.8
Mode 111 (640x480 16bpp LFB) = 6.0
Mode 10e (320x200 16bpp LFB) = 16.1

I'll do some more tests in an hour or so...

Tests continued:

5) Superscape 3DBench VGA benchmark v1.0. Inside Windows 98, I selected Start -> Restart in MS-DOS mode.

66.6 Frames Per Second

6) Superscape 3DBench2 VGA benchmark v1.0c. Again, inside Windows 98, I selected Start -> Restart in MS-DOS mode.

62.5 Frames Per Second

7) Cache check version 7 (11/23/98, Ray Van Tassle). I run cachechk at the DOS 6.2 C: prompt, and the results are -

L1 cache is 16KB -- 103.6 MB/s 10.1 ns/byte (266%) (185%) 3.9 clks
L2 cache is 256KB -- 55.9 MB/s 18.8 ns/byte (143%) (100%) 7.2 clks
Main memory speed -- 38.9 MB/s 27.0 ns/byte (100%) [reading] 10.3 clks
effective RAM access time (read) is 108ns
Effective RAM access time (write) is 61ns

8 ) CTCM (another cache check program). I run ctcm.exe at the DOS 6.2 C: prompt, and the results are -

Best Time for 16K MOVSD (Cache /Page Hits) : 126 mcs - 129.6 MByte/s
average t. for 16K MOVSD (Miss + Hit) : 461 mcs - 35.5 MByte/s
average t. for 16K MOVSD (if clean) : 648 mcs - 25.3 MByte/s
average t. for 16K MOVSD (if dirty) : 1108 mcs - 14.8 MByte/s
Main Memory 16K MOVSD (Cache misses) : 1312 mcs - 12.5 MByte/s

average with 256 KB L2-Cache /DOS (640K) : 542 mcs - 30.2 MByte/s
average with 256 KB L2-Cache /WIN (4M) : 700 mcs - 23.4 MByte/s

And some more tests, this time Doom and also seeing if windows 98 boots up OK with this mobo:

9) Doom -timedemo demo3. This is using Doom 1 shareware version 1.9. I left the size of the screen at the default setting.

2134 gametics in 1778 realtics = 42 frames per second.

10) Doom -timedemo demo3. Again, this is the same version of Doom as above. Edit: Sorry, forgot to say that this test was Doom in full screen mode. End Edit.

2134 gametics in 2151 realtics = 34.72 frames per second.

11) Boot up with a Windows 98 HDD. OK, no problems.

More tests to follow...Please note that I'm breaking these tests up a bit, because if I lump them all together in to one large edited post, then it'll look almost unreadable.

I have set the mobo's jumpers from 33Mhz to 40Mhz, in order to overclock the DX4-100 CPU from 100Mhz to 120Mhz. I will now rerun some of the tests:

1) Speedsys 4.78. See attachment screenshot. Overall score = 50.79. (Was 42.32 using non-overclocked CPU.)
4) Pcpbench mode 100 (640x400 8bpp LFB) = 10.6 (Was 8.5)
9) Doom timedemo: 2134 gametics in 1491 realtics = 50.1 frames per second. (Was 42 FPS)

Just out of curiosity, how do you set the mobo to go from 40Mhz to 50Mhz? - just for laughs! Edit: The manual only mentions the CPU speed choices as being: 25, 33 or 44. End Edit.

5u3 wrote:

Congratulations, you found the base of the ultimate 486. 😁

😁

5u3 wrote:

A newer beta BIOS (v. 3.07) exists, it fixes the year2k bug. It does introduce a few more bugs however (at least on my board).

Bump! Where is v3.07 available for download? I did not find it in here, ftp://ftp.asus.com.tw/pub/ASUS/mb/sock5/486sp3/

My PVI-486SP3 came with BIOS version 3.05. I tried updating it to 3.06, but several different versions of the UUDFLASH and AWDFLASH say, "System was not new AWARD BIOS version ! Please update ROM BIOS first." I realise that 3.06 updates the AWARD version from 4.50PG to 4.51PG, but I had no problem with this type of update on my Biostar 486 motherboard.. Does anyone know a way around this aside from using a stand-alone flash programmer?

It was noted that this board has poor PCI implementation. What performance hits have been noted? Examples?

My Matrox Millennium G200 does not work in this board. On other SiS boards, I noticed that new BIOS revisions were required to get this graphics card working. Can anyone with v3.06 or 3.07 of the BIOS confirm if a Matrox Millennium G200 works or not? Thanks.

I've attached the v3.07 BIOS for Asus PVI-486SP3.

As for the flash software, try UniFlash, that worked about a dozen times on my board.

Thank you for sharing v3.07 of the BIOS. I was able to easily flash to this version using the uniflash programmer. Unfortunately, a Matrox Millennium G200 still will not function in this motherboard, however the original RIVA TNT will.

I layed in 512 KB of 15 ns cache, a 12 ns TAG, and inserted 128 MB of FPM RAM. For whatever reason, CTCM7 shows all 128 MB of RAM as L2-cacheable in both Write-back and Write-though modes. I've not seen CTCM mess up like that before, but regardless, we know it is all cacheable in write-though mode.

Another general observation is why did Asus put 8 DIP-28 sockets, with 4 being DIP-32 extensions? How much more would it have cost to just make all 8 DIP-32 to allow for 1024 KB of cache?

feipoa wrote:

Another general observation is why did Asus put 8 DIP-28 sockets, with 4 being DIP-32 extensions? How much more would it have cost to just make all 8 DIP-32 to allow for 1024 KB of cache?

I know! That would have made things much easier (and allowed more flexible cache configurations - using all DIP-28, all DIP-32, a mixture of both, etc).

How did your 512k upgrade compare in terms of speed to your old 256k config? My tests with speedsys and cachechk didn't show much speed increase other than increasing the cacheable area.

The Biostar MB8433-UUD board also did this same silly cache trick of 4 DIP-32 slots and 4 DIP-28 slots. You generally will not see any improvement when increasing from 256 to 512 KB cache, at least not in cachechk or speedsys. Some Windows benchmarks might, but I haven't done much of analysis on this.

I also don't see any speed increase of 10 ns cache vs. 15 ns cache. I think the wait states are the real bottleneck.

I'm not sure if any gaming benchmarks will catch this either. My only interest in 512 KB cache is to increase the RAM-cacheable limit.

Has anyone been able to use this board in Win98 with:

Cache Write Cycle set to 2
and
DRAM on Fastest

when using an AMD X5-160?

Otherwise, the memory read speed when DRAM is on Faster is the same as when using a standard AMD X5-133, that is, about 39 MB/s in cachechk.

I just bought one of these mainboards from eBay but it has a system password as well as a bios password applied. I removed the battery and its not reset them. I cant see a jumper on the board related to resetting it.

I have emailed the seller and I awaiting his response. But does anyone here have any suggestions on other things I could try to reset it.

Thanks

James