386 DX 40 zu langsam

Auswahl, Einrichtung und Betrieb von Rechnern und Komponenten
Benutzeravatar
Duke_Nukem
HELP.COM-Benutzer
Beiträge: 28
Registriert: Fr 10. Jan 2020, 17:28

Re: 386 DX 40 zu langsam

Beitrag von Duke_Nukem »

Ich habe mir cache besorgt und werde dann mal schauen, ob das 1. überhaupt funktioniert und 2. ob es dann deutlich schneller wird. Dann sage ich Bescheid.

Mein Fokus auf 386 war schon beabsichtigt, weil ich viele Spiele aus der Ära interessant finde. Die untere Grenze sind dann Sachen, die einen 286 brauchen und die obere Grenze Spiele wie Pirates Gold oder One Must Fall. Bei 486 Spielen habe ich sehr gute Erfahrungen mit DOSbox gemacht.
Disruptor
Norton Commander
Beiträge: 124
Registriert: Do 2. Nov 2017, 12:59

Re: 386 DX 40 zu langsam

Beitrag von Disruptor »

Bitte die Module richtig einbauen.
Falls SRAM/Cache-Module schmaler als die Sockel sind, müssen diese mit der Seite, die der Nase/Einkerbung GEGENÜBER liegt, bündig an den Rand eingebaut werden. Die Nase/Einkerbung muss aber dennoch zur Markierung am Sockel/Board schauen.
Benutzeravatar
Duke_Nukem
HELP.COM-Benutzer
Beiträge: 28
Registriert: Fr 10. Jan 2020, 17:28

Re: 386 DX 40 zu langsam

Beitrag von Duke_Nukem »

Ich habe den cache jetzt versucht einzubauen. Es sind 15ns SRAMs. Auf dem mainboard sind jumper settings angegeben, aber wenn ich die jumper für 256KB setze, dann zeigt er nur 32KB an. Die anderen jumper settings für 32KB, 64KB und 128KB bringen entweder eine Fehlermeldung oder gar keinen cache. Dann habe ich die settings probiert, die vorher eingestellt waren als ich das board bekommen habe und die auch auf allen Bildern des boards im Netz immer konfiguriert sind. Dann zeigt das board 128KB cache. Das ist besser als nichts und ich werde es dann wohl so lassen. Ebenfalls rausgefunden habe ich, dass 0 waitstates mit cache nicht mehr funktionieren. Nur mit RAM (60ns) hat das funktioniert. Mit cache zeigt das board dann parity errors an. Im Ergebnis ist das Ding nun endlich so schnell wie es sein sollte. Auch OMF kann man endlich spielen.
Danke Euch!
386DX40 cache IMG_3462.jpg
386DX40 cache IMG_3462.jpg (256.18 KiB) 7317 mal betrachtet
386DX40 cache IMG_3461.jpg
386DX40 cache IMG_3461.jpg (275.44 KiB) 7317 mal betrachtet
386DX40 cache IMG_3458.jpg
386DX40 cache IMG_3458.jpg (236.16 KiB) 7317 mal betrachtet
Benutzeravatar
Duke_Nukem
HELP.COM-Benutzer
Beiträge: 28
Registriert: Fr 10. Jan 2020, 17:28

Re: 386 DX 40 zu langsam

Beitrag von Duke_Nukem »

Ich habe jetzt auch eine FPU bekommen, also einen IIT 3C87-40. Daher weiss ich jetzt, was es mit dem FPU jumper vs. dem Weitek BIOS setting auf sich hat. Der jumper auf dem mainboard ist notwendig, damit das System eine FPU erkennt. Die Weitek Option im BIOS sagt dem System, dass eine Weitek 3167 drin ist. Um also eine normale 387 FPU bzw. den IIT einbauen zu können, muss man nur den jumper setzen, aber die Weitek Option im BIOS ausgeschaltet lassen. Die FPU lässt sich also bei diesem board nicht im BIOS abschalten. Wenn man beides einschaltet, dann denkt das System es wäre auch beides installiert, also 387 + 3167. Das ist technisch im Prinzip möglich, weil beide parallel arbeiten können, aber das board hat natürlich nur einen FPU Sockel.
Wieder was gelernt...

Insgesamt bin ich damit jetzt sehr glücklich. Mit Turbo (40MHz) ist der PC nun schnell genug für OMF und Pirates, ohne Turbo (20MHz) ist es perfekt für Wing Commander.
80386SX
Windows 3.11-Benutzer
Beiträge: 6
Registriert: Mi 10. Jun 2020, 16:23

Re: 386 DX 40 zu langsam

Beitrag von 80386SX »

Duke_Nukem hat geschrieben:Ich habe jetzt auch eine FPU bekommen, also einen IIT 3C87-40. Daher weiss ich jetzt, was es mit dem FPU jumper vs. dem Weitek BIOS setting auf sich hat. Der jumper auf dem mainboard ist notwendig, damit das System eine FPU erkennt. Die Weitek Option im BIOS sagt dem System, dass eine Weitek 3167 drin ist. Um also eine normale 387 FPU bzw. den IIT einbauen zu können, muss man nur den jumper setzen, aber die Weitek Option im BIOS ausgeschaltet lassen. Die FPU lässt sich also bei diesem board nicht im BIOS abschalten. Wenn man beides einschaltet, dann denkt das System es wäre auch beides installiert, also 387 + 3167. Das ist technisch im Prinzip möglich, weil beide parallel arbeiten können, aber das board hat natürlich nur einen FPU Sockel.
Wieder was gelernt...

Insgesamt bin ich damit jetzt sehr glücklich. Mit Turbo (40MHz) ist der PC nun schnell genug für OMF und Pirates, ohne Turbo (20MHz) ist es perfekt für Wing Commander.
Hi,
sorry if I write in english cause it's the only second language I know, but I followed the translation of the discussion and I'm interested in doing some questions.
I have a mainboard that has the identical pcb layout of this, built by another manufacturer (MG Technologies) for which I don't find any manual too. I also found the other one mainboard (QD brand) that was already discussed above as an equivalent one but still no manual found of these boards.
So I found this discussion and I'd like to ask some things:
1) have you eventually tried a PGA cpu on that PGA132 socket instead of the already soldered 386DX-40? I ask because some says it may be a risk trying that not knowing if the mainboard would actually switch to the socket one or maybe would break something on the mainboard so I am not sure if I would try an alternative cpu (like a faster Cyrix) but just to know if it's possible).
2) I am trying to understand the jumpers: reading here the W4 jumper seems to be the one to enable any FPU beside the Weitek that needs the bios option that otherway must be disabled. What about the W7 and W6 jumper? I sort of think but I may be wrong just an idea, that W7 too may be related to the coprocessor signal to the cpu, looking where the trace end up in the cpu socket and in my case it's already factory shorted strangely different from any other equivalent layout board seen online where there's the option. The W6 is still not known, maybe something related to the possibility as said above by some user to not syncronize it to the 40Mhz freq I don't know, in my case is factory opened without possibility to change it.
3) Is there in the bios any option to enable or disable the CPU socket? Or anything related to it?
4) does the IIT fpu need to be inserted in the exact center of the NPU socket with one external line of pins left free, so the first pin of the fpu would be the B-2 on the socket matrix?

Thanks
Benutzeravatar
Duke_Nukem
HELP.COM-Benutzer
Beiträge: 28
Registriert: Fr 10. Jan 2020, 17:28

Re: 386 DX 40 zu langsam

Beitrag von Duke_Nukem »

80386SX hat geschrieben: sorry if I write in english cause it's the only second language I know, but I followed the translation of the discussion and I'm interested in doing some questions.
I have a mainboard that has the identical pcb layout of this, built by another manufacturer (MG Technologies) for which I don't find any manual too. I also found the other one mainboard (QD brand) that was already discussed above as an equivalent one but still no manual found of these boards.
So I found this discussion and I'd like to ask some things:
1) have you eventually tried a PGA cpu on that PGA132 socket instead of the already soldered 386DX-40? I ask because some says it may be a risk trying that not knowing if the mainboard would actually switch to the socket one or maybe would break something on the mainboard so I am not sure if I would try an alternative cpu (like a faster Cyrix) but just to know if it's possible).
I don't know. I have a mainboard with a soldered CPU. I have seen similar boards with a Cyrix 486 upgrade chip, but I imagine that you would have to remove the original 386 to make this work.
2) I am trying to understand the jumpers: reading here the W4 jumper seems to be the one to enable any FPU beside the Weitek that needs the bios option that otherway must be disabled. What about the W7 and W6 jumper? I sort of think but I may be wrong just an idea, that W7 too may be related to the coprocessor signal to the cpu, looking where the trace end up in the cpu socket and in my case it's already factory shorted strangely different from any other equivalent layout board seen online where there's the option. The W6 is still not known, maybe something related to the possibility as said above by some user to not syncronize it to the 40Mhz freq I don't know, in my case is factory opened without possibility to change it.
I haven't seen any documentation and I tried resetting them on my machine without any visible effect, so I kept them in the default position. My understanding is that W4 enables a 387, but a Weitek has to be enabled in the BIOS. Apparently, you could theoretically use both at the same time.
3) Is there in the bios any option to enable or disable the CPU socket? Or anything related to it?
I haven't seen a BIOS option for this.
4) does the IIT fpu need to be inserted in the exact center of the NPU socket with one external line of pins left free, so the first pin of the fpu would be the B-2 on the socket matrix?
You have to put it in the center. I have attached a picture of my machine.
IMG_4062.jpg
IMG_4062.jpg (287.25 KiB) 5629 mal betrachtet
Good luck!
80386SX
Windows 3.11-Benutzer
Beiträge: 6
Registriert: Mi 10. Jun 2020, 16:23

Re: 386 DX 40 zu langsam

Beitrag von 80386SX »

Duke_Nukem hat geschrieben: You have to put it in the center. I have attached a picture of my machine.


Good luck!
Thanks for the answers. My board has the 386 already soldered too but the fact it also have both fpu and pga cpu socket make me think that it could be "upgraded" and the only faster cpu out there are those rare Cyrix that were more 486 than 386 themself and interesting to test if they would work.
From the photo I think you've a difference IIT fpu of the one I have, cause mine has the metal plate above and the key point on the package is in the up-right of the fpu markings (when looking at the mainboard vertically) where the socket markings A and 1 pins (or better say B - 2) begins and the fpu marking on the package end up being vertical not horizontal.
The board boot, the bios post says "Numeric processor: present" and I suppose it's working but it's strangely too cold on the surface still I need to make every test on it.
In the photo there's my board and an example of the config I'd like to use (in that case the same cpu but let's say a Cyrix DLC 40mhz) but I'm not sure to try before having more info about it.
Also another thing I'd like to ask, I suppose the battery of the bios was gone just like in my mainboard: did you replace it on the board or used an external one on the four-pin connector?
Dateianhänge
1.JPG
1.JPG (140.8 KiB) 5613 mal betrachtet
Benutzeravatar
Duke_Nukem
HELP.COM-Benutzer
Beiträge: 28
Registriert: Fr 10. Jan 2020, 17:28

Re: 386 DX 40 zu langsam

Beitrag von Duke_Nukem »

80386SX hat geschrieben: Thanks for the answers. My board has the 386 already soldered too but the fact it also have both fpu and pga cpu socket make me think that it could be "upgraded" and the only faster cpu out there are those rare Cyrix that were more 486 than 386 themself and interesting to test if they would work.
I would really not power it on with both CPUs installed. My guess is that they are wired to the same signals. As far as I remember the Intel Overdrive CPUs sold at some later time were the only CPUs you could just install additionally to the old soldered CPU, because they were specifically wired to disable the original CPU. I don't think that was possible in the 386 age.
From the photo I think you've a difference IIT fpu of the one I have, cause mine has the metal plate above and the key point on the package is in the up-right of the fpu markings (when looking at the mainboard vertically) where the socket markings A and 1 pins (or better say B - 2) begins and the fpu marking on the package end up being vertical not horizontal.
Yes, there are markings on the FPU and near the socket to indicate how to plug it in. The printing on the top should not be important. It looks like you have a different revision of the IIT FPU.
The board boot, the bios post says "Numeric processor: present" and I suppose it's working but it's strangely too cold on the surface still I need to make every test on it.
You can easily check it with benchmarks or tools like Norton sysinfo. In my machine the FPU runs much cooler than the CPU. I eventually put a little heat sink on the CPU, just to be safe, even though you shouldn't need one.
In the photo there's my board and an example of the config I'd like to use (in that case the same cpu but let's say a Cyrix DLC 40mhz) but I'm not sure to try before having more info about it.
I really don't know what you need to make another CPU work on this board. BIOS updates weren't really a thing back in those days. My personal take on this would be that if you want a 486, just get a real 486 board. Right now you have a very nice 386. Putting a Cyrix CPU in there would just give you a slow 486.
Also another thing I'd like to ask, I suppose the battery of the bios was gone just like in my mainboard: did you replace it on the board or used an external one on the four-pin connector?
I desoldered the barrel battery and put in a holder for a coin cell. I got one off Ebay that was modified, so it wouldn't charge the coin cell. I am rather bad at soldering, but it's just two big pins.
80386SX
Windows 3.11-Benutzer
Beiträge: 6
Registriert: Mi 10. Jun 2020, 16:23

Re: 386 DX 40 zu langsam

Beitrag von 80386SX »

Duke_Nukem hat geschrieben:
80386SX hat geschrieben: Thanks for the answers. My board has the 386 already soldered too but the fact it also have both fpu and pga cpu socket make me think that it could be "upgraded" and the only faster cpu out there are those rare Cyrix that were more 486 than 386 themself and interesting to test if they would work.
I would really not power it on with both CPUs installed. My guess is that they are wired to the same signals. As far as I remember the Intel Overdrive CPUs sold at some later time were the only CPUs you could just install additionally to the old soldered CPU, because they were specifically wired to disable the original CPU. I don't think that was possible in the 386 age.
From the photo I think you've a difference IIT fpu of the one I have, cause mine has the metal plate above and the key point on the package is in the up-right of the fpu markings (when looking at the mainboard vertically) where the socket markings A and 1 pins (or better say B - 2) begins and the fpu marking on the package end up being vertical not horizontal.
Yes, there are markings on the FPU and near the socket to indicate how to plug it in. The printing on the top should not be important. It looks like you have a different revision of the IIT FPU.
The board boot, the bios post says "Numeric processor: present" and I suppose it's working but it's strangely too cold on the surface still I need to make every test on it.
You can easily check it with benchmarks or tools like Norton sysinfo. In my machine the FPU runs much cooler than the CPU. I eventually put a little heat sink on the CPU, just to be safe, even though you shouldn't need one.
In the photo there's my board and an example of the config I'd like to use (in that case the same cpu but let's say a Cyrix DLC 40mhz) but I'm not sure to try before having more info about it.
I really don't know what you need to make another CPU work on this board. BIOS updates weren't really a thing back in those days. My personal take on this would be that if you want a 486, just get a real 486 board. Right now you have a very nice 386. Putting a Cyrix CPU in there would just give you a slow 486.
Also another thing I'd like to ask, I suppose the battery of the bios was gone just like in my mainboard: did you replace it on the board or used an external one on the four-pin connector?
I desoldered the barrel battery and put in a holder for a coin cell. I got one off Ebay that was modified, so it wouldn't charge the coin cell. I am rather bad at soldering, but it's just two big pins.
Thanks again for the answers. I installed a Ni-Mh rechargable 80mah battery cause I don't have that. I also think that there's some config on the W1 jumper for external solutions but I still have to understand that well. The original config of that jumper gave to the original battery points something like 5 volts. At the end I preferred to solder a charged Ni-Mh rechargable one similar to the original and be ok with this but I'll look for that option as you installed.
Unfortunately I'm using an ATX case cause I've to find a good AT one and most I don't like. Did you have any problems with I/O controller cards? Of the one I have only a Winbond works, others not.. maybe it's related to the missing -5v from the psu using an ATX-AT adapter.
Benutzeravatar
Duke_Nukem
HELP.COM-Benutzer
Beiträge: 28
Registriert: Fr 10. Jan 2020, 17:28

Re: 386 DX 40 zu langsam

Beitrag von Duke_Nukem »

80386SX hat geschrieben:Thanks again for the answers. I installed a Ni-Mh rechargable 80mah battery cause I don't have that. I also think that there's some config on the W1 jumper for external solutions but I still have to understand that well. The original config of that jumper gave to the original battery points something like 5 volts. At the end I preferred to solder a charged Ni-Mh rechargable one similar to the original and be ok with this but I'll look for that option as you installed.
Unfortunately I'm using an ATX case cause I've to find a good AT one and most I don't like. Did you have any problems with I/O controller cards? Of the one I have only a Winbond works, others not.. maybe it's related to the missing -5v from the psu using an ATX-AT adapter.
I have a proper AT case, so I can't say what you need to do to attach an ATX power supply.
My experience with I/O controllers is that they are generally interchangeable. The biggest problem comes when your controller card has too much stuff on it and for example uses IRQs you want to have free for other options.
80386SX
Windows 3.11-Benutzer
Beiträge: 6
Registriert: Mi 10. Jun 2020, 16:23

Re: 386 DX 40 zu langsam

Beitrag von 80386SX »

Duke_Nukem hat geschrieben:
80386SX hat geschrieben:Thanks again for the answers. I installed a Ni-Mh rechargable 80mah battery cause I don't have that. I also think that there's some config on the W1 jumper for external solutions but I still have to understand that well. The original config of that jumper gave to the original battery points something like 5 volts. At the end I preferred to solder a charged Ni-Mh rechargable one similar to the original and be ok with this but I'll look for that option as you installed.
Unfortunately I'm using an ATX case cause I've to find a good AT one and most I don't like. Did you have any problems with I/O controller cards? Of the one I have only a Winbond works, others not.. maybe it's related to the missing -5v from the psu using an ATX-AT adapter.
I have a proper AT case, so I can't say what you need to do to attach an ATX power supply.
My experience with I/O controllers is that they are generally interchangeable. The biggest problem comes when your controller card has too much stuff on it and for example uses IRQs you want to have free for other options.
At the end I found a proper AT horizontal case too, a good one with a riser ISA card. I installed everything and I (still) didn't try the socket alternative option for the cpu but I'm not going to risk it until having more info about it and for a low end 486 Cyrix it has not much sense anyway. ;)
I've found the system to be strangely unstable. I sure used fast components (60ns 4x4MB ram and 15ns 256kb cache) that may be part of the problem but I noticed that both isa vga card I own with the VESA bios installed on board (GD5429 ISA and a TVGA 9000A) seems to hang the system random, graphical corrupted boot process, strange crash in Win 3.1... I also have an ATi Wonder card not having the vesa bios on board and it seems to be more stable. Have you noticed any of the above situations? Which video card have you tried? Also about the cache I think I've read somewhere the mobo should support 20ns ones, should these 15ns modules works outside specifications or simply run @ 20ns?
The problems seems variable, sometimes in WIndows 3.1 with its specific vga drivers the system and mouse hang, other times the system hang at boot.. I checked the ram with memtest+ 2.01 and no errors found on modules, I've been suggested to try the CheckIt program I've never used.
I suspect these may be the things:

1) the mainboard/bios doesn't like vga cards with vesa bios installed on board (even if with the cirrus logic I tried as suggested in the manual to exclude the emm386 memory area that's reserved for the vga)

2) the mainboard/bios doesn't like the ram/cache speed or the cache may have some problem itself but I don't know how to test, not their speed (already have and seems to read it ok) but for any errors

3) some memory/irq problems conflicts

In the bios there're not many options and all is @ default, no overclocking or whatever.
Benutzeravatar
Duke_Nukem
HELP.COM-Benutzer
Beiträge: 28
Registriert: Fr 10. Jan 2020, 17:28

Re: 386 DX 40 zu langsam

Beitrag von Duke_Nukem »

I haven't noticed problems with any VGA cards or any stability issues. So far I tried a Trident 8900, an old Oak Tech, a Cirrus Logic, and an S3 803. All worked in general, but the Cirrus Logic seemed the best compromise between compatibility, speed, and heat generation. There are DOS extensions you can load to have VESA compatibility.

You should look for faulty components, capacitors and such... Using faster SIMMs or cache modules generally works, but sometimes you have to tweak the system by setting wait states and so on correctly.
80386SX
Windows 3.11-Benutzer
Beiträge: 6
Registriert: Mi 10. Jun 2020, 16:23

Re: 386 DX 40 zu langsam

Beitrag von 80386SX »

Duke_Nukem hat geschrieben:I haven't noticed problems with any VGA cards or any stability issues. So far I tried a Trident 8900, an old Oak Tech, a Cirrus Logic, and an S3 803. All worked in general, but the Cirrus Logic seemed the best compromise between compatibility, speed, and heat generation. There are DOS extensions you can load to have VESA compatibility.

You should look for faulty components, capacitors and such... Using faster SIMMs or cache modules generally works, but sometimes you have to tweak the system by setting wait states and so on correctly.
The problems I suspect might be related to the GD5429 generic drivers that usually are found online with the 1.00B version. I begin to think this driver is not probably compatible for my ISA version of this VLB oriented chip and probably I'd need its own driver but I can't find them even in the old company webpage. But I've found that other brand did ISA card with this chip and I tried manually installing with the oemsetup of the STB Horizon VGA mostly similar card and it works and I've not seen any crash in Win. The only problem is that I can't install them with the setup.exe cause it probably detect the different card and not even its tool to change the refresh freqs so I've some strange vertical frequency for the monitor resolution. Tried with the fast WinBench 3.1 tests and results are almost similar to the generic GD5429 drivers (9.000.000 score vs 10.000.000 score of the generic drivers). I can't say if this driver make use of all the GD5429 features or just use it as a GD5426 or 5428 with lower clocks.

The rest of the system seems stable in dos, I suspect the other crashes were related to the loading of the autoexec/config.sys not in the right way/mem area. Also I changed ISA connectors; anyway I get quite high bench results with a 15,8 3DBench v1.0 score. I also tried installing Win95 on another disk and runs impressively well, almost incredible with the FPU I suppose it's using. No crash or whatever in the GUI there with the factory installed win GD5429/34 driver.

Meanwhile about the W6 jumper, not cause I'd need it now but maybe in the future or just to know more about it. EDIT: I suspect it might be related to the clock async of the FPU and the CPU and probably nothing to do with the PGA socket... :( I found there's a connection from the W6 to a specific pin of the FPU that might explain something..
Antworten