VOGONS


LuckyStar LS486E rev.C2 and Cyrix 5x86@133

Topic actions

Reply 80 of 110, by rad

User metadata
Rank Newbie
Rank
Newbie

Well during post the host processor type is displayed absolutely correct: Cx5x86-S CPU at 133MHz. It's only the configuration registers showing wrong CPU type. Otherwise I haven't noticed any side effects or differences because of that mismatch,

The ISA sound cards I've tried so far are Sound Blaster Pro 2.0 (CT1600) and Sound Blaster AWE 32 non PNP (CT2760). Both are behaving in the same manner in Windows. When there are no drivers loaded all is working fine and stable. When I load the drivers sometimes computer restarts when loading Windows (i.e. startup sound is being played). Or if I manage to get inside Windows and I start WinAmp, I can play the MP3 music just fine. The problem occurs only if I point with the mouse on the music control buttons during playback. So if I control WinAmp only with keyboard, everything is fine and stable. When I point the prev/next/pause buttons with mouse, the computer restarts.

Reply 81 of 110, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Probably a long shot, but, to isolate the issue, try using a) USB card with USB mouse b) change COM port c) change COM port IRQ d) try AWE64Gold

Did you try all the I/O Recovery times and wait states?

Plan your life wisely, you'll be dead before you know it.

Reply 82 of 110, by rad

User metadata
Rank Newbie
Rank
Newbie

Well I've managed to find some free time to tinker deeply with the system. Using the DEBUG DOS program for configuring the southbridge and the WPCREDIT program for northbridge (via PCI host adapter) I was able to test different combinations of both 8/16 bit I/O recovery times and 8/16 bit DMA wait states. I haven't tried all possible combinations, only 2-3 of them (like 4 CLK I/O wait states), but that didn't improve the situation. The only small difference that I've observed is that when I set some DMA wait state the computer just hanged in that case with the windows on screens... in all other situations and cases the system has been rebooting automatically. I've also tried to completely disable COM1 port but this time I've had the same problem with keyboard interaction during music playback. Either way I'm out of options for now and will keep PCI sound card until something else comes to my mind.

Finally the 10ns cache chips (for 512k cache) arrived! I've had only a half an hour this morning to try and play with them in the system. Definitely a progress compared to the old 15ns ones. QUAKE has been used for testing since it is the most demanding and heavily app for cache and memory. So far I'm not able to stabilize it in WB mode operation with 2x32MB EDO RAM, whereas with the 15ns chips (256k cache) it was fully stable. Anyway in WT mode (but again with 2x32MB RAM) everything seems good so far with only one exception - I've had to set DRAM write CAS pulse to 2 CLK, which was 1 CLK before. Otherwise the Quake hangs in the console graphic window in the beginning of demo. With 2 CLK several loops of Quake demo have been completed successfully. And what is more interesting, the results are faster than before with a tiny bit.

256k cache (2x32MB RAM) 10ns TAG, 15ns 4x 64kx8 chips:
18.8 fps full screen (health bar + ammo bar visible)
17.6 fps full screen (health bar only visible)
16.6 fps full screen (only 3D area)

512k cache (2x32MB RAM) 10ns TAG, 10ns 4x 128kx8 chips:
18.9 fps full screen (health bar + ammo bar visible)
17.7 fps full screen (health bar only visible)
16.7 fps full screen (only 3D area)

BIOS Timings (from the screenshot of BIOS Chipset options):
2-2-3 3-0-2-1-1 Slower
2 above is 2CLK DRAM write CAS pulse instead of 1CLK as before.

I'm looking to find out why I can't run the system in WB mode. It just hangs on "Locking 61Mb of memory..." when starting Quake demo. Right now in WT mode only half of the cache is actually used I guess (direct 1-way interleaving) and maybe that could be the reason the system is working fine. In WB it should use all 64MB of memory. So maybe there is some defective chip or a chip not working well at 10ns speed. I've bought a set of 10 pcs so I can try to swap them. Or I can also try to add another 64MB of memory (I've total of 4 32MB equal sticks) and try again in WT mode for full coverage. As I mentioned I didn't have enough time this morning to play around with more combos.

Do you know for a good model of working SRAM tester?

Reply 83 of 110, by feipoa

User metadata
Rank l33t++
Rank
l33t++

512 KB of single-banked cache is tricky to get working when pushing the system to the max. This is the reason I don't run systems with single-banked cache anymore. Some ideas,

1) Try to run a single stick of 64 MB and 512K-WB
2) Try to run a single stick of 32 MB and 512K-WB (just for trouble-shooting purposes)
3) Test all your SRAM cache, e.g. using a Wellon VP-390.
4) Try other more conservative settings, like RAS to MA Delay at 2 CCLK, RAS to CAS Delay at 3 CCLK, etc.
5) It is important to set your L2 cache tag bits to 8 when in WT mode, and to 7 when in WB mode, otherwise strange things happen.

On my DTK PKM-0033S motherboard, it will not function well with the cache (1024K) in WB mode when using 128 MB of RAM. I'm not sure why and I didn't try to troubleshoot it much. Maybe because the RAM & cache is maxed out and the chipset wasn't really tested for these extremes.

Plan your life wisely, you'll be dead before you know it.

Reply 84 of 110, by rad

User metadata
Rank Newbie
Rank
Newbie

Finally I was able to obtain a Voodoo 3 card. It is Voodoo 3 2000 PCI card with 16MB SGRAM (the one with 4 large RAM chips).
IBM 5x86c on 133 (2x66) with 32MB RAM, 256kB WB cache.

Quake I 1.06 timedemo1 with GLQuake 0.97, DX7, 3dfx drivers 1.04 on Win95c give me the following results:
640x480x16: 29.2 fps
800x600x16: 29.2 fps
This is with the default MiniGL driver provided with GLQuake. For higher resolutions I've to swap it with MiniGL newer version, I took 1.49 from VOGONS base. Results with it are a little bit slower, but constant on all resolutions:
640x480x16: 28.7 fps
800x600x16: 28.7 fps
1024x768x16: 28.8 fps
1280x1024x16: 28.7 fps

I couldn't imagine I'll be running Quake I on 486 class system on 1280x1024 and it'll be playable, perfectly playable!

Here is one of the screenshots I took:

IMG_1292.JPG
Filename
IMG_1292.JPG
File size
250.47 KiB
Views
2013 views
File license
Fair use/fair dealing exception

Reply 85 of 110, by feipoa

User metadata
Rank l33t++
Rank
l33t++

That is quite impressive. My system with Voodoo2 gets 29.2 fps with the MiniGL 1.49 at 800x600, but that is my max resolution. I'm not sure what my SiS/Cyrix-133/4x system gets with the Voodoo3-3000 at 1280x1024, but I think it was around 24.5 fps.

Plan your life wisely, you'll be dead before you know it.

Reply 86 of 110, by kanecvr

User metadata
Rank Oldbie
Rank
Oldbie

I'm having some strange issues with one of my LS486E boards. Again, this is an Acorp labeled board, but it's an LS486E rev C2 still. The issue I'm having is terrible memory performance. Regardless of what bios settings I use or how much ram I stick in, I can't get over 37 mb/s. That's about a quarter of what this board is supposed to get.... The board came DOA, it being the 3rd such board I've found so far. Of the first two, one is in my main 586 machine while the other I sent to a friend in trade - so I started working on this board witch would not post. Turns out the EEPROM chip was damaged and replacing it and flashing the LS486C2 bios on it brought the board back to life. It's currently using 128k of L2 cache and 16mb of FPM ram. Is it possible the poor memory performance is due to BIOS version I flashed?

rad wrote:
Finally I was able to obtain a Voodoo 3 card. It is Voodoo 3 2000 PCI card with 16MB SGRAM (the one with 4 large RAM chips). IBM […]
Show full quote

Finally I was able to obtain a Voodoo 3 card. It is Voodoo 3 2000 PCI card with 16MB SGRAM (the one with 4 large RAM chips).
IBM 5x86c on 133 (2x66) with 32MB RAM, 256kB WB cache.

Quake I 1.06 timedemo1 with GLQuake 0.97, DX7, 3dfx drivers 1.04 on Win95c give me the following results:
640x480x16: 29.2 fps

That's actually close to what I remember getting in GL_quake back in the day (I had a Cy5x86-133gp + Voodoo 2). I think it actually got about 26-27 fps but I didn't have any enhancements enabled (I did not know about them back then).

Reply 87 of 110, by kanecvr

User metadata
Rank Oldbie
Rank
Oldbie

Ok I figured it out on my own - it's the slower "9640" revision of the northbridge. Got it up to 66MB/sec with EDO tough - so it's not really THAT slow.

Reply 88 of 110, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Are you referring to the cachechk memory read speed? User rad reported 85.8 MB/s with a 66 MHz FSB. Is this what you get on your other LS486E at 66 MHz? What is different on the Northbridge markings besides the datecode? Have you gone through one by one to ensure the BIOS CMOS settings are identical for memory and cache? I only get 69.5 MB/s with my Biostar MB-8433 for memory read.

Plan your life wisely, you'll be dead before you know it.

Reply 89 of 110, by kanecvr

User metadata
Rank Oldbie
Rank
Oldbie
feipoa wrote:

Are you referring to the cachechk memory read speed? User rad reported 85.8 MB/s with a 66 MHz FSB. Is this what you get on your other LS486E at 66 MHz? What is different on the Northbridge markings besides the datecode? Have you gone through one by one to ensure the BIOS CMOS settings are identical for memory and cache? I only get 69.5 MB/s with my Biostar MB-8433 for memory read.

Not cachecheck, speedsys - my other board gets over 140mb/s in speedsys at 66mhz fsb and about 90 at 33mhz... The fast one is "9651" and is labeled Lucky Star inc. The slow ones are labeled Acorp and have the "9640" chipset.

Reply 90 of 110, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Could you provide the speedsys graph, cachechk results, and CMOS screenshots for both motherboards?

Plan your life wisely, you'll be dead before you know it.

Reply 91 of 110, by kanecvr

User metadata
Rank Oldbie
Rank
Oldbie
feipoa wrote:

Could you provide the speedsys graph, cachechk results, and CMOS screenshots for both motherboards?

There's a speedsys screenshot at the very bottom of this thread, on the first page. Unfortunately I can't replicate the result since my 5x86-100GP won't boot into dos at 133 anymore.

Reply 92 of 110, by feipoa

User metadata
Rank l33t++
Rank
l33t++

It would be nice to verify that there is a datecode specific SiS 496 chipset which has this significantly faster memory performance, however I have strong doubts.

What voltage are you running the chip at?

I ran one of my first IBM 5x86c QFP chips at 150 MHz pretty well for a few hours in DOS, then it died permanently.

Plan your life wisely, you'll be dead before you know it.

Reply 93 of 110, by kanecvr

User metadata
Rank Oldbie
Rank
Oldbie
feipoa wrote:

It would be nice to verify that there is a datecode specific SiS 496 chipset which has this significantly faster memory performance, however I have strong doubts.

What voltage are you running the chip at?

I ran one of my first IBM 5x86c QFP chips at 150 MHz pretty well for a few hours in DOS, then it died permanently.

I was running it at 3.45v

Regarding the "slow" board - I can get 55mb/sec using a 160MHz AMD and EDO after tweaking BIOS settings. Not bad.

Reply 94 of 110, by feipoa

User metadata
Rank l33t++
Rank
l33t++

3.45 V is way too low for stable and continuous 133 MHz operation on a Cyrix 5x86 chip. You will want to modify your motherboard's onboard VRM to up the voltage to 3.7 V. The modification is very simple and typically involves removing a single resistor and replacing it with a variable resistor.

Plan your life wisely, you'll be dead before you know it.

Reply 95 of 110, by Azriel_Strife

User metadata
Rank Newbie
Rank
Newbie

hi everyone, i have had 2 of these boards and both of them have had the same issue.

The first board posted ok for a while, but would not maintain cmos settings and after each post it would claim a bad cmos checksum and revert to defaults. The battery in the board was known good.

the second board is doing the same cmos checksum issue. again, batteries are good its not throwing any beep codes.

I have my 5x86 amd 133 running at 160mhz (4x40) mhz with the voltage jumper set to 4v, it wont post on the 3v setting.

Any help would be very greatly appreciated!

Reply 96 of 110, by lukas12p

User metadata
Rank Newbie
Rank
Newbie
rad wrote on 2016-05-17, 09:20:

The ISA sound cards I've tried so far are Sound Blaster Pro 2.0 (CT1600) and Sound Blaster AWE 32 non PNP (CT2760). Both are behaving in the same manner in Windows. When there are no drivers loaded all is working fine and stable. When I load the drivers sometimes computer restarts when loading Windows (i.e. startup sound is being played). Or if I manage to get inside Windows and I start WinAmp, I can play the MP3 music just fine. The problem occurs only if I point with the mouse on the music control buttons during playback. So if I control WinAmp only with keyboard, everything is fine and stable. When I point the prev/next/pause buttons with mouse, the computer restarts.

Hi,
Have You find the solution to use ISA sound card with FSB66?

Reply 97 of 110, by Zerthimon

User metadata
Rank Member
Rank
Member

Found this thread via a search in google, sorry for necro-posting.
Has anyone got this mobo (ls486e rev. c2) work with ps/2 mouse ?

ps: Thanks for the info regarding the 512k of cache!

Reply 98 of 110, by feipoa

User metadata
Rank l33t++
Rank
l33t++

I think vogons user maxtherabbit may be working on such a mod, but it isn't complete yet from the last I can recollect.

Plan your life wisely, you'll be dead before you know it.