running 5x86 cpu in IBM 5150 with Inboard/386

[Cimonvg]

hello
Here is some news in my slowly ongoing hobby - of making the most out of the old 5150 motherboard.
In marts 2015 i did run 486 with inboard 386/pc in a pc-clone. (*1)
And cyrix 5x86 were the logic goal for this upgrade. But i had to acquire an cyrix 586 cpu and a VRM (voltage-regulator-modul).
Took some time to get the right product at the right price , and i did read some PDF's about the 486 cpu.(http://datasheets.chipdb.org/IBM/x86/486/)
My only notice were that Cyrix 586 do not like signal on J1 pin, and therefore made a socket extension with no connection on the J1 pin.
The hardware upgrade is made of :
IBM 5150 motherboard
Inboard 386/pc, 386+387 removed
extension sockets 132 pin (to make clearance for the RAM add on card)
transcomputer board (386->486 upgrade)
VRM , set to 3,45 volt and *2 clock
socket extension 168 pin with pin J1 removed
cyrix 586 GP100



But, it did not work in my IBM5150. Only the "VGA" logo appeared on the screen , then the pc halted.
The answer were of course the BIOS. The old original BIOS is not compatible with newer then the cyrix486-PGA132 cpu-class. The real Intel486 and cyrix 586 needs a newer BIOS.



Lucky i did buy a DIY kit with the "Anonymous" bios ,and here two
days ago it worked for the first time

Furthermore i have experimented with Netroom3 and the Inboard 386/pc. It runs wery sweet. You only have to edit in the CUSTOMIZE.INF file. The line "^inbrdpc" has to be added (otherwise rm386 has no extended memory to see when it
starts). Cloaked BIOS and VGA no not function, but the PC switches to protected mode and the PC feels more quick. Could be fun if the virtuel BIOS could run on the Inboard , and the old PC could support a lot of new features
When i try syscloak.exe the PC gives a message "BIOS system descriptor table missing.", but perhaps it is possible to cheat the Netroom to load anyhow ??

Now , i only have to get more juice out of the 5x86. At this moment at runs like 6 time a pc-xt
I did a Winstone94 test and it only concludes a score of 4,4. Not a lot, the original Inboard 386 with 80386 gives a score of about 6 !
I have not read all there is about the cyrix-586, with internal settings and so - perhaps i am lucky i can umph the speed a bit by editing setting internally in 586 ?

;Data collected by: TOPBENCH | Benchmark and detection stub | Version 0.97
;This file contains fingerprinting information about your computer. Please
;email this file to trixter@oldskool.org with a subject line of "Benchmark" to
;help test these routines and seed the TOPBENCH database.

[UID3336260]
MemoryTest=507
OpcodeTest=206
VidramTest=993
MemEATest=233
3DGameTest=198
Score=21
CPU=Cyrix 5x86-S (clock x2 mode)
CPUspeed=80 MHz
BIOSinfo=unknown
MachineModel=0000
BIOSdate=20120502
BIOSCRC16=3336
VideoSystem=VGA
VideoAdapter=VGA, unknown Chipset, 256kb Video Memory (BIOS)
Machine=PC





...and yes, dispite it is the same hardware pictured above - the different applications do not quite agree about the speed of the CPU. One remark, the CPU do not get hot...

thanks to everyone supporting this forum, and thanks to those quotes and refers to my posts.

other links in my upgrading the IBM5150:
(*1)http://www.vcfed.org/forum/showthread.php?46993-running-486-on-Inboard-386
http://www.vcfed.org/forum/showthread.php?49421-Running-Win32s-on-ibm-5150-with-Inboard-386-pc

 

[Chuck(G)]

6 times a PC XT was about what a PC AT ran at, no? You're probably doing pretty close to the limit.

 

[evildragon]

I swear, if someone manages to get Doom to run on a 5150...... lol

 

[Cimonvg]

@ Chuck(G), yes if the CPU depends on the BUS speed, then i am at the limit, but if some setting can open for more cpu speed ?

@ evildragon , have to try that one out some day

/cimonvg

 

[Cimonvg]

hello - well Doom 1.9 demo runs
...not fast, rather slow. I did count 200 frames and it took 86 seconds, it gives 2,3 frames per second



...AND the keyboard did not function, but with the mouse i did wander around in the castle....
/cimonvg

 

[Chuck(G)]

@ Chuck(G), yes if the CPU depends on the BUS speed, then i am at the limit, but if some setting can open for more cpu speed ?

It's not the CPU that you're fighting; it's the bus and supporting devices. After all, you have to get information to the CPU somehow and get it out somehow. Or, to quote an Irish saying: "Cuir síoda ar ghabhar agus is gabhar i gcónaí é."

 

[evildragon]

hello - well Doom 1.9 demo runs
...not fast, rather slow. I did count 200 frames and it took 86 seconds, it gives 2,3 frames per second

View attachment 37478 View attachment 37479

...AND the keyboard did not function, but with the mouse i did wander around in the castle....
/cimonvg

That is the most beautiful thing I have ever seen. Slow or not, it's running Doom!

 

[nestor]

Is the internal cache enabled? My 386 to 486 upgrade needs a driver to enable it. You can check if it is enabled with cachechk

 

[Cimonvg]

hello nestor - pretty sure that the cache is not enabled at this point - look at the ckcpu121.exe app on screenshot above .
- i have to look into the cache enabling
/cimonvg

 

[stynx]

6 times a PC XT was about what a PC AT ran at, no? You're probably doing pretty close to the limit.

There is still a lot of ways to go faster

Used Machine: PC XT (5160) with 8mhz bus, Inboard386 with clock-mod, Kingston 586 5V upgrade (x4 mode), 128k UMB, ET4000, SCSI (CF-HD, CD), 1.44mb floppy.





Doooooooooom

Aaaaand a bit faster now (no memory expansion though)


-Jonas

 

[Cimonvg]

hello
today i started out , thinking about writing , going back from Cyrix5x86 to IBM486BL due to the slow speed of the Cyrix.
And i did test the Cyrix (no cache) then the 486BL with cache and both at 50Mhz speed. The later did very well the first not so good.

And here a great thanks to stynx , for sharing the transcomputer driver tool , that enables the cache - it works and the result is a PC running at mind blowing high speed !

(most admit , did test the encache.exe for the first time today - had planed to test it earlier , but first go around to try it out today :ashamed2: )
Today i did test the Cyrix5x86(no cache) then IBM 486 BL and last the Cyrix 5x86 with cache on.
I did use the speed test from
http://www.philscomputerlab.com/dos-benchmark-pack.html

...it states:
"
At the end of the benchmark you will see a value for realticks. To workout the FPS you need to use this formula:

FPS = 74690 / realticks.
"
and the speed with 486BL in DOOM is notisable faster the Cyrix with no cache.



My DOOM test results:

ibm 486bl 50Mhz
small window 74690/3265 = 22,8 fps
maximum window 74690/11757 = 6,3 fps


cyrix5x86 at 50Mhz no cache

small windows 74690/11591 = 6,44 fps
maximum windows 74690/49193 = 1,52 fps

cyrix5x86 at 50Mhz with cache on

small window 74690/2757 = 27,01 fps
maximum windows 74690/10229 = 7,30 fps

( the demo runnning at 7,3 FPS looks very impressive )

I also did a Winstone94 test :

IBM 486BL at 50 Mhz : score 12,1


5x86 with cache on : Score,"13,33778"



other testresults:



and topbstub data , with cache on:

;Data collected by: TOPBENCH | Benchmark and detection stub | Version 0.97
;This file contains fingerprinting information about your computer. Please
;email this file to trixter@oldskool.org with a subject line of "Benchmark" to
;help test these routines and seed the TOPBENCH database.

[UID745019165]
MemoryTest=157
OpcodeTest=154
VidramTest=1180
MemEATest=163
3DGameTest=92
Score=36
CPU=Cyrix 5x86-S (clock x3 mode)
CPUspeed=80 MHz
BIOSinfo=unknown
MachineModel=0000
BIOSdate=20120502
BIOSCRC16=7450
VideoSystem=VGA
VideoAdapter=VGA, unknown Chipset, 256kb Video Memory (BIOS)
Machine=PC


Only thing about the encache.exe is , -i think- it makes the CPU unaware of what DMA does to ram. After enable cache the floppy "dir" did look like this :huh:



... have only to say, the 486BL is nice and fast, but the Cyrix586 at 50Mhz and the cache enabled, is just mind blowingly fast! And the Cyrix will not be replaced in near future



* a hardware note - my Inboard386 board were running at normal 16Mhz and not overclocked.

/cimonvg

 

[NeXT]

I am simply unable to believe that works.

 

[evildragon]

Now I definitely need something like this on my model 25. I don't care if I fabricate my own ISA riser with 3 slots (one inverted on the reverse side), I will make this happen now. Holy crap this is awesome.

 

[Anonymous Coward]

I think you tried Pentium overdrive before, correct? What was the story with that chip?

 

[Cimonvg]

I think you tried Pentium overdrive before, correct? What was the story with that chip?

hello
I know the Pentium overdrive exists , but i do not have one... But did Cyrix586 not do better the the overdrive , back in the days ?
The cyrix i have now, only supports *3 clockspeed, if i could get a cyrix 120Mhz/100Mhz with *4 , i would properly do better score (?)
...and i have to look into the heat exhaust , the cpu did not get that warm , but other chips did warm up...
/cimonvg

 

[lyonadmiral]

@ Chuck(G), yes if the CPU depends on the BUS speed, then i am at the limit, but if some setting can open for more cpu speed ?

@ evildragon , have to try that one out some day

/cimonvg

Way back in the day, I have a Zenith Z-200 PC, and in it, of course you had their proprietary but compatible backplane, but I had two video cards, a Zenith CGA card connected to via ribbon a Z-100 mode card even though the Z-100 mode card had its own external connectors.

I've thought about way back all of these processor upgrade cards, and of course, you can't make the 8-bit operate more than it was designed to do, reliably, but what if you somehow moved data off of the bus entirely in between compatible upgraded cards via ribbon or some other type of connector, kind of like how video cards are connected together via SLI, leaving the 8-bit bus to simply do initialization and base power?

 

[Cimonvg]

hello
sorry a bit late reply

@lyonadmiral,
I also like the idear of thinking out of the box, and persue the "impossible" sometimes to overcome the obstacles and make something that works

@evildragon , sorry to remind you, the Inboard386 only supports 8088 and not 8086 :huh:
- the SOTA386 supports 8086 , but no RAM extention there...

@next, today i did some test with the same hardware setup, but with DOS and himem.sys (no cdrom drivers , and no NetRoom3).
all of a sudden i rembered that the protected mode operation (Netroom) perhaps did influens the performance (and the accuracy of the speed measurements)

Boot into DOS then 5x86 register settings, and the enable of the cache (with encache.exe), but the score in DOOM is almost the exact same .

with HIMEM.SYS:

small window 74690 / 2716 = 27,5

big window 74690 / 10091 = 7,4016

and speed and checkit also look like the NetRoom3 scores:





TOPBENCH dos+himem.sys

[UID7450187E]
MemoryTest=156
OpcodeTest=119
VidramTest=1182
MemEATest=164
3DGameTest=93
Score=37
CPU=Cyrix 5x86-S (clock x3 mode)
CPUspeed=80 MHz
BIOSinfo=unknown
MachineModel=0000
BIOSdate=20120502
BIOSCRC16=7450
VideoSystem=VGA
VideoAdapter=VGA, unknown Chipset, 256kb Video Memory (BIOS)
Machine=PC


two notes about this setup.
1. i get a annoying error at warmboot. The PC starts to load dos then Inboard driver , but it do not stop at 4 Mb but carries on to 16Mb :huh:
- then i have to cut the power and start again. Hardware reset seems to go better
2. the read/write data to the floppy is unusable in DOS with the cache enabled with ENCACHE.EXE . BUT in windws 3.11 (with the cache enabled) i can read and write floppies

/cimonvg

 

[Cimonvg]

ISA bandwith limit

ISA bandwith limit

hello
here som follow up on the ISA bandwith limit.
Few weeks ago is was searching for Cyrix 5x86 capable of doing x4 , in my case 16Mhz x 4= 64 Mhz. I
did spot one on auction - but did not win it.
Thinking, thinking - perhaps my 5x86 120GP had the x4 option. I did some tests with evergreen
ET586.exe tool, and did read that cyrix586 does not enter x4 if booted in x3 mode. I then changed the
setting on my VRM from x3 to x2. Did some more test and did measured the speed. The x4 never appeared
on the data - but i noticed that x2 with Cyrix produced quite high results. And the story about the
goat and the oil in mind - i then did some more testing.
Started out with a few tests, but extended to cover : Winstone94, Doom_v1.9 demo, pimasm, intel
floatdos, lanmark, puzzle and windows Helix discover (memory test). In fact several pages of data.
I took some time to look at the data - and I chose to draw graph where x-axis cpu speed with data from
lanmark and y-axis with Doom demo (b) (big window) FPS.
And the result:





graph made of the following data:

cpu	Driver:	lanmark Mhz	Doom demo(b)	Doom -fps
386+387		19,77	41583	1,796
cyrix5x86x2	no cache	22,9	50027	1,49
Cyrix486 pga168	Driver: CACHEON.SYS	76,8	14653	5,09
ibm,486BL x2	Driver: revto486.sys	94,5	12989	5,75
cyrix5x86 x2	Driver: CACHEON.SYS	150	10690	6,99
cyrix5x86x3	Driver: CACHEON.SYS	200	10271	7,27
amd5x86 x3	Driver: CACHEON.SYS	223	10174	7,34
cyrix5x86x2	Driver:CACHEON.SYS and 5x86 register config	238	10522	7,09
amd5x86 x4	Driver: CACHEON.SYS	264	9932	7,52
cyrix5x86x3	Driver:CACHEON.SYS and 5x86 register config	299	10088	7,4

I read the data in this way - if i upgrade the IBM pc to more than (CPU Lanmark spec.) 200 Mhz - there is practically no additional benefit. Only more oil on the goat

NB ... all tests are made with same hardware / software, only difference is the CPU and cache driver tool
NB i would still like to try the cyrix586 x4 cpu - applications that do not use the ISA bus would still run faster
/cimonvg

 

[Trixter]

Oil on the goat indeed! Thanks for the updated data.

 

[rmay635703]

I do think a p24t “should” help due to better cache and floating point which are what you need on a bus constrained system

The trouble will be that the legendary Pentium overdrive was notoriously incompatible.

But when it did work with cache enabled it did things other 486 chips (including 5x86’s) couldn’t do

I have one in my basement