EDO RAM on ErgoPro s450

[vp]

My son has an ancient Fujitsu-Siemens (orig. ICL Fujitsu) ErgoPro s450. I have upgraded its hard disks and doubled its huge memory. We still like to use the machine, but more memory is needed. I read the only manual I found about this machine (http://www.fujitsu-siemens.co.uk/rl/ser ... Ergo96.pdf). but couldn't really tell if it accepts EDO RAM (60ns). The manual in its way promises, though. It is impossible to get any official support from Fujitsu-Siemens (which is almost understandable).

Anyway, I found two BIOS versions on wimsbios.com pages and updated the machine with the "70ns" version, since the machine sadly has 70ns VRAM. I tried, but the EDO RAM (2x32MB in bank 0) cannot get the machine started

Is it possible that the BIOS version targeted for 60ns VRAM could also offer support for 60ns DRAM? Any experience anyone? Even memories

-vp-

[NickS]

What's puzzling is that the handbook says the BIOS ID is #53 and the links at Wim's say the BIOS ID is #50.
What is the AC number on your board ? Do you have a jumper block T109 ?
If you have the AC41655 (s450/133n) you should be using the #53 BIOS.

[vp]

Yeah, that detail only emphasizes the fact that the handbook doesn't really tell much about my son's machine! The BIOS ID has always appeared as #50 ("ICL System BIOS #50 Version 1.24" originally). So this piece of antique is, as we in Finland say: "karvalakkimalli", that is a fur cap model, stripped down version of s450/133n, s450/100.

The AC number is AC41644 and I see there is no jumper block T109, at least if you refer to the memory type selection jumpers (no labelling of jumper blocks on the PCB). That shouldn't hurt too much, because I think the memory type selection only allows you to mix FPM and EDO memory in the system!? My target is to use 4x32MB EDO sticks only.

I experimented with original 2x8MB FPM sticks in bank0 and 2x32MB EDO sticks in bank1 and the system comes up with POST recognizing all the installed memory (80MB). But I can only boot DOS . Windows98SE cannot pass its initialization! So I guess the BIOS can really only handle the first 16MB in this case. I can run memtest86, which sees all the same memory, too (80MB). Memtest86 passes all tests. It can be that memtest86 doesn't actually use BIOS which would explain the positive test result. Could that mean that if my son accepted Linux as his OS, he could have more memory?

I wondered if a BIOS change is needed to adapt a video card with 60ns VRAM (EDO?) into the system (it now has a 70ns ATI card), could that also allow the system to use 60ns EDO DRAM?

-vp-

[NickS]

This BIOS upgrade comes in two different versions, depending video RAM timing. B50_P supports video RAMs with 60ns access time, B50_PL supports video RAMs with 70ns access time.
* At Customer site, If onboard video (x250, x450, x650) is used:
* Enter RSU check from INFO screen which Video BIOS that is installed.
* If Video BIOS is B use B50_PL.LDB
* If Video BIOS is A or D,E use B50_P.LDB
* If onboard video not used in x or if s450, s650 use B50_P.LDB
* At Repair Centers, Service Offices etc.:
* For AC41633 with FW rev <10, use B50_P.LDB
* For AC41633 with HW rev E19 or later, use B50_P.LDB
* For AC41661, use B50_P.LDB
* For other AC41633, use B50_PL.LDB
* For AC41644 and AC41663, use B50_P.LDB (s450, s650)
* If both 60ns and 70ns VRAMs are used at the same time, use B50_PL.LDB (i.e. slower timings).
* The speed of the Video Memory is printed on the top the VRAM chips, too (-6 = 60ns and -7 = 70ns).
* If B50_P BIOS (i.e. faster memory timing) is upgraded to the System Board which have 70ns VRAMs (i.e. slower VRAMs than adjusted timings in Video BIOS), in very rare cases, some pixels could be missing on the screen (data is not lost).
* If B50_PL BIOS (i.e. slower memory timing) is upgraded to the System Board which have 60ns or 70ns VRAMs, some higher refresh rates are not supported

If you use an add-in video card it does not matter which BIOS you use.

[vp]

That's a good quote! I haven't seen that ever, thanks Nick!

So I'll check the situation once again with that P version, as suggested.

[vp]

Been there, done that, but unfortunately no change in the behaviour other than it kind of felt quicker or something

I guess I've found the dead end. Is there any way to affect the memory bus timing or whatever it is that's preventing the 60ns EDO from operating in this strange machine?!

[vp]

I am not quite sure if this message already belonged to some other forum, but this deals with the same problem still...

Happened to remember that I studied earlier on whether I should have parity memory or not. I ended up regarding the original FPM memory (KM48C2100AJ-7) as non-parity memory (it has four similar ICs on one side only). This was the conclusion also in a long topic in some newsgroup or box a while ago. The to-be replacing EDO memory is GM71C17403CJ6, which most likely is non-parity (it has eight ICs on both sides of the stick).

Today I found some web pages (e.g. http://www.embeddedlogic.com/TH99/m/I-L/31071.htm) suggesting that s450 should indeed have parity memory (x36)!? There is no evident way of instructing the BIOS or the motherboard about parity or non-parity memory. It seems that only the other is suitable. If the FPM is some sort of pseudo parity, does it then mean that only (pseudo) parity memory is usable? I wonder if 32MB (pseudo) parity EDO or even FPM sticks are available??

[KachiWachi]

Isn't that feature determined by the installed chipset?

I have an old Packard Bell here that says it can use x36, but it isn't required. It only accepts FPM...EDO will cause it not to boot. I happen to have a x32 stick in it...works just fine.

x36 is probably still available, but might be costly...unless you can find some laying around in some old board somewhere...same goes true for the "fake" parity modules.

The PCGuide has a good writeup on all this...

[vp]

That is most certainly true. I must check the chipset and try to find information about it. So far I have only followed the data regarding the particular computer model. Would have hoped, though, that the handbooks available would have revealed this information, that is, which chipset is in which model and which chipset allows the use of FPM, EDO, BEDO, SDRAM and so on.

Meanwhile I learned about 'fake parity', as it is most often called, and pity me, if my son's machine has been equipped with such cheat! All "official" data (ie. handbooks) handle this topic with vagueness. What puzzles me even more, is that these handbooks give the impression (=recommendation!) that EDO works in all s, x and e models of ErgoPro (see eg. page 85 in http://www.fujitsu-siemens.co.uk/rl/ser ... Ergo96.pdf)!?

Thanks for your link, it's a good reminder for all of us!

[NickS]

What puzzles me even more, is that these handbooks give the impression (=recommendation!) that EDO works in all s, x and e models of ErgoPro (see eg. page 85 in http://www.fujitsu-siemens.co.uk/rl/ser ... Ergo96.pdf)!?

But your motherboard is not in the Ergo '96 handbook - it's in the '95 one which, if I remember correctly does not hold out such hope.

That's a good quote! I haven't seen that ever, thanks Nick!

Written by someone in Finland, I believe ... that's where these machines were designed (ICL bought Nokia's PC division). I'll keep on looking as there may be someone around these parts who still has one.

[vp]

Indeed, it only says that 128MB can be installed, but nothing about the memory type. At least it settles the parity confusion: no parity works! That's anyway better than fake parity!!

I'd better find someone who could trade 4x32MB EDO for 4x32MB FPM

[NickS]

Update: V-P has the AC41644 motherboard.

Comparing the two handbooks;
1. All the SIMM modules listed for the AC41644 motherboard (by ICL part no.) in the earlier handbook are listed as being FPM in the later handbook.
2. On the AC41655 jumpers are used to tell the BIOS what types of SIMMs are installed.

This implies that the earlier board would only support FP mode, although the question remains why the EDO SIMMs do not work as FPM; possible chip organisation/refresh issue ?

3. On the question of CPU voltage, the inference I draw is that ICL Part No. AF33384 (P75), AF33705 (P90) and AF33380 (P100) were single-voltage Pentiums whereas the AF33386 (P120) and later were dual voltage Pentium MMX processors. This could have been stated more explicitly, but remember that the service manuals were written for the ICL service scenario and that the types of RAM and processors to be used would have been tightly controlled (in line with ISO9002). The single-voltage P133 may work (subject to current draw being acceptable) - awaiting feedback from V-P.

[vp]

Hello,
latest development in this issue:

133 MHz CPU works in s450/100! After some confusion with jumpers in jumper block 1, the cpu started correctly with jumpers in positions 3,5,6 and 10. At least it didn't start when jumpers installed in 3,5,6 and 9, which is still a mystery for me.

No breaking news regarding the EDO compatibility. ErgoPro Handbook '96 lets me understand that EDO DRAM works in s450/133n. On the other hand Handbook '95 only lists FPM DRAM in DRAM recommendations for s450/100. Now comparing the technical data of s450/100 and s450/133n reveals no significant changes in hardware (the same chipset, bus speeds etc.). Makes me believe that the EDO compatibility in buried in BIOS code. And the question is now: Could BIOS #53 actually work in s450/100, too? And, if it fails, is it possible to fall back to BIOS #50 or otherwise revive the system?

[Ritchie]

Documentation can sometimes be incorrect on CPU jumpers. Years ago a new main board and accompanying printed manual (actually Gigabyte) did not work with the documented settings for Pentium 100MHz CPU. We found the jumper requirements were different to documented. We were lucky to get the PC going because the monitor was also adjusted dim.

[NickS]

Hello,
latest development in this issue:

133 MHz CPU works in s450/100! After some confusion with jumpers in jumper block 1, the cpu started correctly with jumpers in positions 3,5,6 and 10. At least it didn't start when jumpers installed in 3,5,6 and 9, which is still a mystery for me.

J9 & J10 are described as "bootblock normal" and "bootblock reload". I speculate this means BIOS protect/BIOS update enable. Depending on how this is implemented, it could be that if a later BIOS tries ESCD/DMI update with a hardware protection on the flash, the update could fail.

No breaking news regarding the EDO compatibility. ErgoPro Handbook '96 lets me understand that EDO DRAM works in s450/133n. On the other hand Handbook '95 only lists FPM DRAM in DRAM recommendations for s450/100. Now comparing the technical data of s450/100 and s450/133n reveals no significant changes in hardware (the same chipset, bus speeds etc.).

Sort of the same chipset. The later board uses the "Viper-M" as did later versions of the early board. Can you identify whether your chipset is "M" ?

Makes me believe that the EDO compatibility in buried in BIOS code. And the question is now: Could BIOS #53 actually work in s450/100, too? And, if it fails, is it possible to fall back to BIOS #50 or otherwise revive the system?

I think there is a chance. To recover if it does not work you would need to use "hot-flash" BIOS recovery (see the "Collected Wisdom" forum posts on hoflashing and BIOS recovery) or a flashing service.

I am having difficulty identifying the organisation of the LG RAM chips listed above....