• Please review our updated Terms and Rules here

8087 on 8088 motherboard that have no socket for it

This is from page 23 of the Borland Turbo C Reference Guide Ver 2.0;

The _8087 variable is set to a nonzero value (1, 2, or 3) if the startup code autodetection logic detects a floating-point coprocessor (an 8087, 80287, or 80387, respectively). The _8087 variable is set to 0 otherwise. The autodetection logic can be overridden by setting the 87 environment variable to YES or NO. (The commands are SET 87=YES and SET 87=NO; it is essential that there be no spaces before or after the equal sign.) In this case, the _8087 variable will reflect the override, and be set to 1 or 0. Refer to Chapter 12 in the Turbo C User's Guide for more information about the 87 environment variable. You must have floating-point code in your program for the _8087 variable to be defined properly.
Click to expand...
 
I think disabling the coprocessor is useless. In a laptop one could have saved some energy but presently noone really uses such devices in the field.
 
george said:
I think disabling the coprocessor is useless. In a laptop one could have saved some energy but presently noone really uses such devices in the field.
Click to expand...
Right, useless it is. However, it is interesting to figure out what the performance gain is for certain applications. That is my main motivation. As the software switch does not seem to be reliable, or working with all applications.
 
The performace gain is not everything -- many CAD software simply was not running without a coprocessor. BTW: if the main CPU is 80186/8 or above there is a way to disable the coprocessor in software.
 
Yes, it can be a factor of 10. I understand that this can make a difference between usuable und unusable and a reason to not offer the software without 8087.

The main CPU is still an 8088 for me.
 
Usable or not is mostly subjective. There maybe other factors with expensive CAD software, you can never know for sure. For example if the user had not paid for a coprocessor then most likely they were using a pirated copy of several times more expensive CAD software. How can you measure the performace gain for example in AutoCAD?
 
george said:
Usable or not is mostly subjective. There maybe other factors with expensive CAD software, you can never know for sure. For example if the user had not paid for a coprocessor then most likely they were using a pirated copy of several times more expensive CAD software. How can you measure the performace gain for example in AutoCAD?
Click to expand...
Oh, I wasn't talking about AutoCAD... you were 😀 I measured the performance gain using this program (of course that's coarse, program specific and somewhat subjective, but it does a lot of FP)
 
With my 2nd board revision I am now indeed able to run at 9.54 Mhz - V20 and 8087-2. Interestingly, even my 8087-2 (I don't have a -1) works without issues at that clock speed, even though it was not "hand selected or certified" for 10 Mhz. That was what the 8087-1 version was apparently.

.. and thanks to @1ST1 for motivating me to do another round of layout.

20250214_090637.jpg
20250214_090705.jpg
20250214_090725.jpg
20250214_091032.jpg
 
Last edited:
george said:
Looks very rude to solder an 8087 directly to a PCB.
Click to expand...
I should probably have it in a shrine / glass vitrine in my living room... anyhow, I think that it is doing better than on a dump in China from where I got it, basically, even though if it suffers from my rude onto-board soldering. 😁
 
1ST1 said:
We need to check if it fits in M19 on CPU position.
Click to expand...
Yes, agree. I will have a look.

When I look at these images, there should be enough space in either orientation. I will take some measurements and verify orientation.

Good to see that the CPU looks as it is placed in a socket (blue), so no desoldering required.
 
Last edited:
You must log in or register to reply here.
Back
Top