VOGONS


First post, by majestyk

User metadata
Rank Oldbie
Rank
Oldbie

Currently I´m testing an old FIC 486-PAK-3 mainboard that was used in AT&T´s Globalyst 520.
I updated it´s BIOS from 1.00 to 1.06 (from the NEC´s website).
It´s now equipped with 64 MB of EDO memory and a 486 DX2-66 CPU. This CPU is correctly detected and identified by the BIOS.
The proprietary L2 cache stick is not installed - due to it´s rarity...

PAK_3_total.JPG
Filename
PAK_3_total.JPG
File size
1.84 MiB
Views
664 views
File license
Fair use/fair dealing exception

So I ran some initial tests like Speedsys and Sysinfo:

PAK_3_speedsys.JPG
Filename
PAK_3_speedsys.JPG
File size
1.23 MiB
Views
664 views
File license
Fair use/fair dealing exception
PAK_3_sysinfo.JPG
Filename
PAK_3_sysinfo.JPG
File size
982.27 KiB
Views
664 views
File license
Fair use/fair dealing exception

I couldn´t believe my eyes when I noticed CPU benchmarks. That´s a joke for a DX2-66 right?

There aren´t too many BIOS options and they don´t affect CPU performance. On the board there are no jumpers, just a microswitch to select between 25 MHz and 33 MHz FSB.
What could be the issue here?

Reply 1 of 14, by megatron-uk

User metadata
Rank Oldbie
Rank
Oldbie

Wow, that Norton result is slower than several of my 20MHz 286 boards. Even without level 2 cache fitted the scores should be way, way higher.

Obvious question though: is the turbo switch enabled? Does it use a keyboard combination to enable turbo rather than a physical switch perhaps?

My collection database and technical wiki:
https://www.target-earth.net

Reply 2 of 14, by majestyk

User metadata
Rank Oldbie
Rank
Oldbie

I found a picture of the frontpanel on the net, there´s no hardware turbo switch, just "power" and "power management". The latter being some energy saving feature I suppose.
Edit: In BIOS there are several options under "Power Management", also for slowing down or halting the CPU, but those are all disabled and should not apply.

PAK-3_frontpanel.jpg
Filename
PAK-3_frontpanel.jpg
File size
127.06 KiB
Views
649 views
File license
Fair use/fair dealing exception

I also tried "Ctrl + Alt + - / +" to no avail.

Reply 3 of 14, by majestyk

User metadata
Rank Oldbie
Rank
Oldbie

O.k., this seems to be CPU dependant.

1.) Intel 486SX -> the system won´t boot from floppy, CPU detected correctly
2.) Intel 486 DX2-50, Intel 486 DX2-66, AMD Am486 DX2-66 all detected correctly and all are extremely slow at 14.5 points in Sysinfo.
3.) AMD Am 486 DX4-100 falsely detected al DX2-100, slow at 14.5 points.
4.) Intel 486 DX4-100, detected correctly, in Sysinfo it´s at 216 points!
5.) AMD Am5x86-133, detected as DX4-100, same benchmarks as 4.).

PAK-3_dx4.jpg
Filename
PAK-3_dx4.jpg
File size
122.58 KiB
Views
612 views
File license
Fair use/fair dealing exception

I had to assemble a little self made VRM for the 3.xV CPUs to run.

PAK-3_vrm.jpg
Filename
PAK-3_vrm.jpg
File size
1.19 MiB
Views
612 views
File license
Fair use/fair dealing exception

But what´s the reason for this behaviour? Seems like everything in the DX2 class is being slowed down to a kind of turbo mode.

The Globalyst 520 was available with 486 DX2-50, 486 DX2-66 or 486 DX4-100 CPU. So it´s not surprising the DX4 works properly. But why don´t DX2 models?

By the way, this home made VRM fits perfectly on the "FIC 486 GAC-2". 3 jumpers need to be removed, because this board does not have the MOSFET switch for automatic 5V CPU detection.

GAC-2_vrm.jpg
Filename
GAC-2_vrm.jpg
File size
1.86 MiB
Views
576 views
File license
Fair use/fair dealing exception

Unlike the PAK-3 the GAC-2 detects ALL CPUs correctly no matter if Intel or AMD, DX, DX2, DX4 and all run at full speed. (There are 2 jumpers for SX models.)
I wish the 256K Cache Sticks could be built as easily as the VRMs.

Reply 4 of 14, by RockstarRunner

User metadata
Rank Member
Rank
Member
majestyk wrote on 2021-04-29, 13:16:

By the way, this home made VRM fits perfectly on the "FIC 486 GAC-2". 3 jumpers need to be removed, because this board does not have the MOSFET switch for automatic 5V CPU detection.

Have you shared this design anywhere?
I have a gac-2 and tried to use a dx4 100 in it, which is apparently supported, but the system typically freezes when trying to start DOS.
I was wondering if this was too do with voltage, but not sure, wouldn't it not start at all?

Do you have any idea what those three jumpers do that your vrm seems to be attached to?

Reply 5 of 14, by RockstarRunner

User metadata
Rank Member
Rank
Member

So, from I have now gathered, those pins are intended for such a vrm, to use 3v processors... I really wish that would have been clearer, when it was mentioned all over the place that this board supports dx4, but no mentions at all that a vrm module is required.
I guess my dx4 is toast.

Regardless, it would be greatly appreciated if you could share the vrm design.

Reply 6 of 14, by Hoping

User metadata
Rank Member
Rank
Member
RockstarRunner wrote on 2022-06-05, 17:29:

So, from I have now gathered, those pins are intended for such a vrm, to use 3v processors... I really wish that would have been clearer, when it was mentioned all over the place that this board supports dx4, but no mentions at all that a vrm module is required.
I guess my dx4 is toast.

Regardless, it would be greatly appreciated if you could share the vrm design.

You can check which pins of the connector go to the VCC pins of the processor socket and which pins possibly bring the 5v and GND with a multimeter and then easier, buy a modern DC-DC step down which are very cheap and much more efficient than a linear regulator and go from 5v to 3.3v easily. If the input is 5v, the step-down must be able to work with that clear voltage.
The processor only needs power it doesn't care where it comes from as long as it meets its requirements.

Reply 7 of 14, by majestyk

User metadata
Rank Oldbie
Rank
Oldbie

If the core voltage for the DX4 is not correct, the board might not start at all or run instable. With the VRM plugged in and with the lastest BIOS the DX 4 runs perfectly on a 486-GAC 2 and makes this mainboard far more exciting.
The design I used is the original design used by FIC. I can post the diagram here and the pinout of the 7-pin VRM-header when I´m back home which won´t be before June 18th.

It IS a buck / step down design by the way.
You can find everything you need in the datasheet of the 2576T:

https://www.mouser.com/datasheet/2/308/1/LM25 … 6_D-2314897.pdf

You can use the formula in Figure 15 to calculate the exact values of the resistors for 3.3, 3.4V or whatever you need.

Reply 8 of 14, by RockstarRunner

User metadata
Rank Member
Rank
Member

I have checked the J3V pin headers with my multimeter, and I believe they are as follows:
1 +5V from PSU
2 Voltage to CPU
3 Gnd
4 Gnd
5 Voltage to CPU
6 +5V from PSU
7 NC

I am not an expert or anything even close, so I don't know what the correct nomenclature for those are, but that's what I have regarding the headers so far.
If anyone who knows better, and thinks this is wrong, please let me know 😀

I have also got the components to build a VRM module, based on looking at the document attached, and repeatedly bugging a friend of mine who is actually good with electronics.
Next I have to figure out how i'm going to put it all together.

Reply 9 of 14, by dudulinek

User metadata
Rank Newbie
Rank
Newbie

Same problem here with FIC 486-GAC-2 (AT&T Globalyst 515) i cant use DX2 66MHz. Bios and software show everything is fine 2x 33FSB, 66MHz ... but cpu is extremely slow. Exact same result in sysinfo 14.5 🙁
Same computer here necroware restoration video and he has no issue change SX to DX2 i dont know what iam doin wrong 🙁

https://youtu.be/ZUt-3Y3vMso

Reply 11 of 14, by RockstarRunner

User metadata
Rank Member
Rank
Member

Built a vrm for the gak-2, and now the dx4 works, but I'm only getting 122 cpu score in sysinfo, and it seems like it's not quite as fast as it should be.
I've tried some bios settings changes, but it stays the same.

IMG_20220617_173256.jpg
Filename
IMG_20220617_173256.jpg
File size
1.61 MiB
Views
273 views
File license
Fair use/fair dealing exception

*Edit* got it up to 128 with some bios tweaks, still quite far off that 216 that was shown earlier in the thread. Still, the tweaks got some decent real world improvements, so I'm not so bothered, it's more a curiosity.

Speedsys gives me 39.27 for the CPU, which sounds correct, so I'm more leaning that sysinfo CPU bench is not reliable as a measure.

Reply 13 of 14, by majestyk

User metadata
Rank Oldbie
Rank
Oldbie

Today I ran some tests with a 486 DX2 66 (955) with write back cache on the FIC 486-PAK3. It´s detected by the BIOS correctly as DX2-66(WB) and all benchmarks are as they are supposed to be for a DX2-66.
Maybe the regular DX2-66 were not offered by NCR and they just didn´t include them in their BIOS?
There was no BIOS released by FIC for this board so we got to live with the NCR BIOS.

TLDR: If your benchmarks are awful with a DX2 or SX, try the WB-version.

I wish some cache sticks for these mainboards show up one day...