• Please review our updated Terms and Rules here

8087 Not Detected?

Raven

Veteran Member
Joined
Mar 7, 2009
Messages
2,752
Location
DE, USA..
I just came cross the "Landmark System Speed Test" and ran it on my XT, which has a V20/8087 combo installed. I had assumed they had been working in Tandem fine, but upon running this software it detected no FPU. I opened the box, made sure the DIP switch was set properly for it (it was) and tried again - sitll no FPU. I tried the DIP switch in the wrong setting, and still no FPU. I have now pulled the 8087 and am curious as to if it were dead to begin with... Should I trust Landmark? Is there any way to tell if it's dead for sure?
 
Does Landmark explicitly say 8087: no ? (I can't remember if it does)

I'm not sure Landmark looks at a coproc at all.
What about Norton SI or Checkit?
 
Unless you do a lot of floating point calculations the co-processor doesn't buy you anything anyway.

Are you sure you have the co-processor switch on the motherboard in the right position? If I remember correctly, the switch in the "on" position indicates co-processor not installed.
 
The switch on the motherboard will only enable/disable the 8087 Interrupt line. If it's enabled, an 8087 interrupt will invoke a NMI.

Try this FPU diagnosis program by Intel. It will test any x87 from the 8087 to the 487, and it will run on any PC. MCPDIAG.EXE is a text-only version of the diagnostic program only, while RUNMCP.EXE is an EGA/VGA demonstration slideshow including the diagnostic check as one of the options.

Please note that INSTRUCT.LIB has been splitted into 2 files. It was too big to fit in one >97Kb zip file, so it has to be combined back into one file before it can be used.
 

Attachments

  • CoProcessorDiag5.zip
    91.9 KB · Views: 4
  • CoProcessorDiag1.zip
    87.1 KB · Views: 4
  • CoProcessorDiag3.zip
    84.5 KB · Views: 4
  • CoProcessorDiag4.zip
    66 KB · Views: 4
  • CoProcessorDiag2.zip
    87.3 KB · Views: 4
Sorry for not posting this sooner - had a USB fiasco with my main box and couldn't click my USB mouse. Fixed now.. anyway...

I put the original AMD 8086 back in on a hunch and then ran the software again, and it detects the 8087 and works fine.

Is it possible that the V20 is damaged in such a way that the only defect is lack of coprocessor compatibility? :p I pulled it from a box that died from corrosive damage, but the chip worked fine in all other respects in my 5160, being rated as equivalent to a 3mhz AT (not too shabby for 4.77mhz XT, heh).

If it comes down to needing a new V20 I can get one pretty easily at one point from Jameco or some such, but I'd like to know. My V20 is, by the way, an NEC 8750Y5 V20 D70108C -10, which I assume means it's a 10mhz rated model.
 
I put the original AMD 8086 back in on a hunch and then ran the software again, and it detects the 8087 and works fine.

The V20 is a 8088 replacement, if you have an 8086 you want the NEC V30 iirc. Perhaps thats the issue.
 
The V20 is documented to work with 8087. If it continues to be a problem for you, I'll check my NEC MicroNOTES--there's something about the 8087 in there, but it's been over 20 years since I looked at them.
 
Data width of both chips
V20 or 8088 >> 16-bit internal / 8-bit external
V30 or 8086 >> 16-bit internal & external

8087 >> 16-bit internal & external
 
Last edited:
Data width of both chips
V20 or 8088 >> 16-bit internal / 8-bit external
V30 or 8086 >> 16-bit internal & external

8087 >> 16-bit internal & external

Sorry, I've got to disagree here. The 8087 is a very different beast. If it sees BHE\ from the CPU high at reset, it reverts to 8 bit transfers (8088/80188 ). Otherwise, it uses 16-bit transfers. It maintains an instruction prefetch queue that exactly mirrors the one in the CPU, as the CPU has to do all of the address computation and manage all of the memory bus cycles.

Internally, the 8087 maintains an 84-bit data path in the numeric execution unit.

The 8087 is a lot more complex than most people think.
 
Last edited:
The V20 is documented to work with 8087. If it continues to be a problem for you, I'll check my NEC MicroNOTES--there's something about the 8087 in there, but it's been over 20 years since I looked at them.

Chuck, do you think my V20 is damaged in some way? If not, then perhaps it would be helpful if you perused those notes, otherwise I'll chock it up to a bug/chip damage and not worry about it - use the chip in a system that doesn't need FPU or somethin'.
 
My feeling is that your V20 is probably screwed up. I have a V30/8087 pair running at 10MHz...no problems there. However, is it possible that older revisions of the 8087 don't get along well with V20/V30? My first attempt at pairing an 8087-1 with my V30 was a complete failure. The 8087 would get really hot and the system would crash often. It was however detected by landmark. That was a C8087-1 (gold top). My current and second attempt is a D8087-1 (ceramic). It works just fine.
 
I have NEC V20 on a Tandy 1000 SX running at 7.16 MHz. I wanted to install coprocessor to it and i first tried this C8087-3. It didn't work well since usually the system froze before DOS prompt.

The C8087-3 is a 5 MHz part so i thought that 7.16 MHz would be too much for it so i never tried it with 8088 @ 7.16 MHz.

Then i tried this C8087-1 and it works fine as it should since the C8087-1 is a 10 MHz part.

Maybe the overclocking was't the problem since the C8087-1 seems to be later revision.
 
I'm underclocking my V20 and possibly my 8087 down to 4.77mhz. The 8087 works fine with the 8088, so it's happy at 4.77mhz. The V20 is a 10mhz part, but I'd imagine it should function at lower frequencies, right? Perhaps a different revision of 8087 will help, however, if I can find one sometime - good call on that.
 
I'm underclocking my V20 and possibly my 8087 down to 4.77mhz. The 8087 works fine with the 8088, so it's happy at 4.77mhz. The V20 is a 10mhz part, but I'd imagine it should function at lower frequencies, right? Perhaps a different revision of 8087 will help, however, if I can find one sometime - good call on that.

I read in the intel manuals that they could work with clock frequencies all the way down to 2MHz, past that, the registers gets unrelaiable (because the registers are stored as dynamic memory. I don't remember if there were a version using static memory, but in that case, it could be run by any frequency below the maximum rating).
 
In my 5160 I am running and NEC V20 and an 8087-3 (gold cap). I will check it ASAP to see if I have the same issue as you.

EDIT: Raven what tool are you using to ID the thing? I don't seem to have any software handy to check for you.

EDIT 2: I ran some Simtel Utilities I found, WhichCPU reports an NEC V20 w/ an 8087 NPU; SYSID reports the same; CPUINF agrees aswell.

Code:
WhichCPU V0.50 (c) by Michael Holin     //    Public Domain
WhichCPU ? for info

Looking for: 8088,8086,V20,V30,80188,80186,80286,80386sx,-dx,RapidCAD
             80486sx,-dx,-slc,-dlc,Pentium
CPU found  : NEC V20
NPU found  : 8087

Code:
CPU Information 1.1; Copyright by Michal Szokolo; 1995.01.02

CPU: NEC V20
FPU: 8087

CPUID: N/A Features: 0

CPU Bugs: -

Benchmarks:
 - Integer:    74 (0.0 * U5S/40, 1.7 * XT)
 - Real:       84
 - I/O:       795 (extremely slow)
 
Last edited:
Back
Top