VOGONS


Reply 42 of 62, by Montgomery

User metadata
Rank Newbie
Rank
Newbie

Thank you for the quick reply. Unfortunately my CPU (an ADW variant) seems not to be beefy enough for faster than 160Mhz operation. I only managed 180Mhz for a couple of seconds, but the PC won't even POST further than the first screen. Of course it might be a motherboard issue, since mine seems to be factory crippled. The PCI divider was hardvired for 1:1 operation.

02.jpg
Filename
02.jpg
File size
159.98 KiB
Views
2140 views
File license
Fair use/fair dealing exception

So I had to solder in 3 pins for changing to 1:2 FSB. Luckily it didn't need anything else and worked perfectly after the hardware was installed.

03.jpg
Filename
03.jpg
File size
138.51 KiB
Views
2140 views
File license
Fair use/fair dealing exception

I downloaded some BIOSes from vogons and flashed a spare chip with the 1.72F. After fine-tuning the BIOS to the last bootable state I measured the following performance.

MAX.jpg
Filename
MAX.jpg
File size
361.81 KiB
Views
2140 views
File license
Fair use/fair dealing exception

Should I expect more from this setup or is it similar to your results with the AMD 5x86 CPU?

The board only takes FPM RAM (I'm using a pair of 16MB 60ns) so faster memory is unfortunately unusable. If I recall it correctly EDO is backward compatible and should work in FPM mode, but my 4DPS wont POST with any EDO stick.

I even tried setting the CPU to 5V (JP27 1-2; JP28 on-on) but it wont start at 200Mhz either set to 3x66 or 4x50. I guess I have to stay at 160Mhz, which seems mostly stable.

P.s.: It seems the floppy controller stopped working along the way. Maybe because of the new BIOS or the tight memory settings. Tomorrow I'll check it.

Reply 43 of 62, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Most Am5x86 chips will not run at 180 MHz. I have 8 or so of these chips and not of them run reliably at 180 MHz. I have ony one which will POST and run one DOS benchmark, but it will crash soon thereafter.

You can solder in a jumper for the PCI divisor. If you can't run chips with a 60 or 66 Mhz FSB, the PCI divisor doesn't help much. For Am5x86-160, you usually want the PCI bus running at 40 MHz.

I have two of these boards with dead floppy controllers. I'm not sure why these boards are more suseptible to this failure than others I have.

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

Reply 44 of 62, by Montgomery

User metadata
Rank Newbie
Rank
Newbie

I was lucky with the floppy controller. It was only a BIOS issue. After an exceptionally serious freeze - while playing with memory timings and FSBs - the BIOS self-reset itself and disabled the FDD. After switching it back on everything works fine.

Yes I know the divider weakens the video performance under 33mhz. This is why it would be interesting to run the chip at 2x66Mhz, like you do with your IBM 5x86. I know that the two CPUs are technologically very different but maybe - if the motherboard can handle the 66Mhz FSB - the processor would take the 133Mhz. Is it possible to set an AMD 5x86 to 2X FSB? I only saw options for 3X and 4X, the latter sharing settings with the Enhanced AMD DX2. Is there maybe a jumper on the 4DPS to set the 2X multiplier directly?

Reply 45 of 62, by feipoa

User metadata
Rank l33t++
Rank
l33t++

With the 160 MHz AMD and 4DPS, there is a non-ideal setting which must be set which cripples the RAM read and write performance. You'll be disappointed. Read here, Re: Help with SiS 496/497 Tomato 4DPS 486 motherboard

The Am5x86 chips can only run at 3x/4x. To run at 2x, you need to use a late Am486 DX2. See attached. However, I have confirmed that my 4DPS boards do not run at 66 MHz and it is unlikely that yours will. People have had success with the Lucky Star 486E, which is very similar to the 4DPS, but without the ps/2 mouse port.

Am486_DX2-66V16BGC.jpg
Filename
Am486_DX2-66V16BGC.jpg
File size
188.67 KiB
Views
2118 views
File license
Fair use/fair dealing exception

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

Reply 47 of 62, by Montgomery

User metadata
Rank Newbie
Rank
Newbie

I suspected that the 5x86 doesn"t have a 2x multiplier. It's unfortunate since these are the fastest chips of the era. I doubt that anyone ever was able to run a DX2/66 at 2x50 or 2x66Mhz, but please prove me wrong! 😀

I had the opportunity to tinker with three 486 systems in the last few days. I attached my test results to this post. Each of the three motherboards (PCChips M918i, Tomato 4DPS, Gigabyte GA-5486AL) has some advantages over the others but generally they seem to be in the same league. I guess I'll be satisfied with the 160Mhz CPU speed since it's still very strong for a 486 system. Unfortunately I haven't got access to quality mainboards or different variants of 5x86 chips, I only have AMD 5x86 133 ADWs.

I don't use Windows on these machines so I was lucky to be able to set the memory timings a bit tighter. There are some strange readings though in the performance results. The PCChips M918i seems to be the fastest with it's own buggy AMI BIOS. This BIOS reports write back cache however there isn't really any since I have the variant with the fake chips. I could only install a PnP sound card if I disable the COM2 and LPT ports and even then the MIDI synth of the AWE64 freezes the computer. There must be some interrupt manipulation behind the scenes which enable the BIOS to perform so well in benchmarks. I cannot discard the sound so I had to swap this BIOS with the Gigabyte GA-5486ALs 1.07 version.

5x86.jpg
Filename
5x86.jpg
File size
367.62 KiB
Views
2093 views
File license
Fair use/fair dealing exception

I'm still polishing the software package of these PCs but they run fine especially the one with the PCChips board which received a 3Dfx Voodoo for basic 3D. GTA and Tomb Raider are running fine. I even tried the Tomb Raider in S3D, but the Virge was true to its name it realy is a "3D decelerator". 😀

Reply 48 of 62, by feipoa

User metadata
Rank l33t++
Rank
l33t++
Montgomery wrote:

I suspected that the 5x86 doesn"t have a 2x multiplier. It's unfortunate since these are the fastest chips of the era. I doubt that anyone ever was able to run a DX2/66 at 2x50 or 2x66Mhz, but please prove me wrong! 😀

I have run the above posted Am486DX2 at 2x66 and 2x83 MHz. I probably still have the benchmark information somewhere if you really want it see it. The CPU pictured above was produced mid-2002. I suspect it is an Am5x86, but with a 2x and 3x option instead of a 3x and 4x option.

I had the opportunity to tinker with three 486 systems in the last few days. I attached my test results to this post. Each of the three motherboards (PCChips M918i, Tomato 4DPS, Gigabyte GA-5486AL) has some advantages over the others but generally they seem to be in the same league. I guess I'll be satisfied with the 160Mhz CPU speed since it's still very strong for a 486 system. Unfortunately I haven't got access to quality mainboards or different variants of 5x86 chips, I only have AMD 5x86 133 ADWs.

The 4DPS and GA-5486AL are generally considered of good build quality, they just do not have sockets for double-banked cache. The M918 is, on the other hand, is a real piece of _________, though it does have double-banked cache. But it didn't seem to work with L2 cache anyway,so your fake cache is about as good as it gets. If your 4DPS isn't out performing your M918, I suspect you are running into the issue I reported on, or you haven't optimised your BIOS settings.

I don't use Windows on these machines so I was lucky to be able to set the memory timings a bit tighter. There are some strange readings though in the performance results. The PCChips M918i seems to be the fastest with it's own buggy AMI BIOS. This BIOS reports write back cache however there isn't really any since I have the variant with the fake chips. I could only install a PnP sound card if I disable the COM2 and LPT ports and even then the MIDI synth of the AWE64 freezes the computer. There must be some interrupt manipulation behind the scenes which enable the BIOS to perform so well in benchmarks. I cannot discard the sound so I had to swap this BIOS with the Gigabyte GA-5486ALs 1.07 version.

5x86.jpg

I'm still polishing the software package of these PCs but they run fine especially the one with the PCChips board which received a 3Dfx Voodoo for basic 3D. GTA and Tomb Raider are running fine. I even tried the Tomb Raider in S3D, but the Virge was true to its name it realy is a "3D decelerator". 😀

GTA and Tomb Raider run fine for me on a 486 + Voodoo.

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

Reply 49 of 62, by Montgomery

User metadata
Rank Newbie
Rank
Newbie

I have run the above posted Am486DX2 at 2x66 and 2x83 MHz. I probably still have the benchmark information somewhere if you really want it see it. The CPU pictured above was produced mid-2002. I suspect it is an Am5x86, but with a 2x and 3x option instead of a 3x and 4x option.

I absolutely believe you and it's great news. The chances that I could find such a CPU here are marginal but maybe I get lucky or hunt one down at ebay. The high FSB really intrigues me.

As you have mentioned the memory timings and waitstates are hindering the performance. I realized I had to turn a few notches down on the 4DPS to get it stable so I will return finetuning it. What was the case with your DX2 at 66 and 83 Mhz? Were you also forced to set higher latency? If it's not a big trouble I would be very happy to see speedsys or other benchmark results of the setups. What motherboard was able to supply 83Mhz? Maybe you had modified it someway?

Reply 50 of 62, by Montgomery

User metadata
Rank Newbie
Rank
Newbie

Bummer! ... Sorry but I just realized some hardware defect which is responsible for the innumerable problems I was trying to solve for some time. Today I started to have issues with the 4DPS based 486. The AWE64 would play MIDI and MOD, but with MP3 it always froze the system. First I thought it might be some conflict with the freshly installed Voodoo card so out it comes ... the problem stays. Next I removed every lines of driver of the AWE64 and began benchmarking/testing. This time the floppy access made the PC freeze. I removed everything except the VGA but the floppy still made the PC stop. After that I tried rejumpering the motherboard to basic 5x86 speeds (133Mhz, PCI 1:1), and voila ... everything works.

During testing I realized there doesn't seem to be a difference whether I set Jumper 18 to 1-2 or 2-3, the graphic performace was always similar. I discarded this info believing that the overall system speed is too slow for the 20Mhz or 40Mhz PCI speed to have a great impact. Now I know I was stupid. 😀

If you take a look at one of my former posts where I complained about the missing JP18 pins you can see the long chip (right above JP18) is missing a leg. I didn't spot it that time but now it makes everything understandable. The PCI speed setting is unfortunately absolutely disabled on this board. That's why there was no difference between JP18 settings because the PCI speed is always 1:1 and the jumper doesn't change anything. And it is also the problem that made my system unstable. It seems my Virge and Voodoo doesn't take 40mhz too well. So for now I downgraded to mere 133Mhz and testing if there are any other issues.

Reply 51 of 62, by feipoa

User metadata
Rank l33t++
Rank
l33t++

I cannot say for certain if your issue is 40 MHz on the PCI bus, but there are other issues at 40 Mhz which affect stability and how well the board works. If the ISA is taking speed from a fraction of the PCI bus, make sure you have the ISA divisor set correctly in the BIOS. Also, did you read this in detail? Re: Help with SiS 496/497 Tomato 4DPS 486 motherboard Was your cache write cycle set to 2 or 3 when FSB is 40 MHz?

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

Reply 52 of 62, by Montgomery

User metadata
Rank Newbie
Rank
Newbie

Yes, my memory timigs were set to 2 with cache and DRAM too and the ISA bus at fix 7.xx Mhz. With old hardware one can never be certain of the cause of an issue. Yesterday I turned the PC off in good working order and today it won't boot. I swapped PSU, CPU, removed every card by the time it occured to me to check the flash BIOS. Of course it was corrupted. I put the original EPROM back and everything worked fine. I have to reflash the 1.72f BIOS for Y2K compatibility and better PnP and try my tests again. 😀 Maybe the board would eventually run stable at 40Mhz?

Reply 53 of 62, by feipoa

User metadata
Rank l33t++
Rank
l33t++

For 40 MHz, you should start by setting your cache write cycle to 3 and not to go above 256K cache. I have run into the corrupt EEPROM a few times on this board. I suspect something gets buggered up at 40 MHz with unstable timings at the time of writing ESCD.

You could also try with my version of the 1.72f BIOS. Attached.

Attachments

  • Filename
    4dps2013.rar
    File size
    80.04 KiB
    Downloads
    99 downloads
    File license
    Fair use/fair dealing exception

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

Reply 54 of 62, by Montgomery

User metadata
Rank Newbie
Rank
Newbie

Thank you for the updated BIOS. I think I will stay at 133Mhz for the sake of stability and my Voodoo card. I found the 1.72F the most stable version for the board. The 4.00A and 4.01E were unsuccessful in coordinating the two PCI boards and the AWE64 PnP.

Reply 55 of 62, by feipoa

User metadata
Rank l33t++
Rank
l33t++

That is how it starts out. Then some time passes and and the desire for more emerges. For 40 Mhz FSB, you really want a board with double-banked cache. If sticking with 33 Mhz, a Cyrix 5x86-133/4x is an attractive, but rare, option.

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

Reply 56 of 62, by alexh

User metadata
Rank Newbie
Rank
Newbie

I have a 4DPS with a non-functional PS2 port on 1.71. I'm wanting to try a few BIOS updates to see if I can fix it.
Had a go at flashing with AMIFLASH but it cannot erase the chip. Peeled off the sticker and it's a 27C010A-12, a UV erasable EPROM...
Hoping to not spend silly money to get a decent programmer so thinking I should replace it with an EEPROM that I can flash for free with flashrom and an old NIC or RPi.
Looking at the supported hardware on flashrom, found the 29C010A which, except for pin 1 (which I think is 12v on the EPROM but it's not connected on the EEPROM anyway) is pin compatible with the 27C010A-12.
Alternatively I thought that I might be able to boot with the existing EPROM, remove that whilst running, replace with the empty EEPROM then flash with AMIFLASH.

Do you think this will work? Looks like you guys are always flashing the BIOS on this board, how did you replace the original EPROM?

Once the EEPROM is in, I assume I also need to move JP4 to pins 1-2?

(btw my first post on Vogons after reading it for years thanks to Phil's Computer Lab)

Reply 57 of 62, by Chadti99

User metadata
Rank Oldbie
Rank
Oldbie

I’m also on this boat 🤣. I’ve tried a few different firmware versions and Windows isn’t finding a ps/2 mouse port to install. I can’t find an option in the bios to enable/disable the ps/2 port, not sure if there would be. But even without a mouse connected, Windows should be able to find a ps/2 port right? CTmouse tells me “error: device not found” when forcing the ps/2 option.

Thanks for any help!

Reply 58 of 62, by Chadti99

User metadata
Rank Oldbie
Rank
Oldbie

So I’ve wired the ps/2 leads directly to the board and I’m 90% sure I have it right. On the modified 400a bios it’s hanging after the memory test while a mouse is connected. The keyboard is also inoperable after entering the bios setup. So I feel like I’ve made some progress 🤣.

Reply 59 of 62, by Chadti99

User metadata
Rank Oldbie
Rank
Oldbie

I think I got it! Had to swap the clock and data connections. For anyone wondering, a PS/2 mouse port never shows up in Windows Devices tab.

Attachments