• Please review our updated Terms and Rules here

running cpu faster then the math co processer

oblivion

oblivion

Veteran Member
Joined
Aug 28, 2010
Messages
1,003
Location
Apache Junction, AZ
my question is. if you run a cpu faster then the rating for the co-processor can that do any damage to the co processor?

I ask because I have a 5mhz / 10mhz 8086 system with a 8087 math co-pro. I accidentally ended up buying a V30 -16 so It seems my standard speed running is now 8mhz and 16mhz in turbo. so will running in 16mhz damage the math co-pro or does it just stop functioning when run over the rated speed?
 
The co-processor speed is independent of the processor. IOW, running a faster processor does not alter the co-processor's speed. IIRC, most co-processors ran slower than their processor counterparts, anyway.
 
You might want to double check the speed the system is running at. I think that unless you change the oscillator the V30 would run at the same 5 or 10 MHz that the 8086 did originally. The 8087 runs at the same bus speed as the the CPU so you would need the 8087-1.

Running an ISA bus at 16 MHz will cause problems with lots of expansion cards in addition to being very bad for the 8087. The 80287 normally was set to run at a lower clock speed then the 80286 but having separate clocks for CPU, NPU, and bus itself was uncommon on 808x systems. See what manuals for system say.

If you do have the system running at 16MHz, removing the 8087 would probably be a good idea. Even if the 8087 works at the high clock speeds, it will be pumping out a lot of heat while providing little benefit since few 808x programs used the 8087.

I think the above is accurate but the paucity of information about using a V chip with a 8087 means I don't have a reference to confirm it.
 
krebizfan said:
You might want to double check the speed the system is running at. I think that unless you change the oscillator the V30 would run at the same 5 or 10 MHz that the 8086 did originally. The 8087 runs at the same bus speed as the the CPU so you would need the 8087-1.

Running an ISA bus at 16 MHz will cause problems with lots of expansion cards in addition to being very bad for the 8087. The 80287 normally was set to run at a lower clock speed then the 80286 but having separate clocks for CPU, NPU, and bus itself was uncommon on 808x systems. See what manuals for system say.

If you do have the system running at 16MHz, removing the 8087 would probably be a good idea. Even if the 8087 works at the high clock speeds, it will be pumping out a lot of heat while providing little benefit since few 808x programs used the 8087.

I think the above is accurate but the paucity of information about using a V chip with a 8087 means I don't have a reference to confirm it.
Click to expand...

know any good diagnostic utilities that will let me know the cpu speed that will fit on a 720k floppy and run on 808x machines?

all I tried norton diagnostics but it gave me a "286 required to run this" error. so I ran MSD and under CPU it said nec v20 or v30/8. I just assumed the /8 was the speed since the previous 8086 ran at 5/10mhz I was assuming that this ran at 8/16.
 
Norton SI should do it--and you should find copies of Landmark SPEED test of various versions kicking around. I think that the SIMTEL20 archive has a copy.
 
Anonymous Coward said:
Do you know of ANY 808x systems where bus, CPU and NPU are NOT running at the same speed?
Click to expand...

Cromemco and probably most other CPU on a card designs. I don't know of one but I would not be surprised if there was a Turbo XT that kept the ISA bus at 8 MHz or less while letting the CPU stay at a higher clock.
 
In PC/XT computers FPU is always using the same speed as the CPU. They are very closely coupled.
I was trying to overclock an 8087-1, and the fastest I got it working was 11 MHz (or 10.66 MHz?!). I was not able to get it running on 12 MHz.
Now, my system with 8087 installed would run faster than 11 MHz, but bad things happened if I was trying to execute FPU instructions. (It worked fine after switching CPU speed to "normal" 4.77 MHz).

krebizfan said:
Cromemco and probably most other CPU on a card designs. I don't know of one but I would not be surprised if there was a Turbo XT that kept the ISA bus at 8 MHz or less while letting the CPU stay at a higher clock.
Click to expand...

ISA is asynchronous bus, there is no bus clock per se. (There is 14.31818 MHz OSC signal, and CLK signal that on XT's is usually equal to CPU frequency. But nothing says that other bus signals are synchronous to these signals).

Turbo XT clones (and AT clones, at least early ones) normally insert wait states, so that the ISA bus cycle timing is compatible/close to 4.77 MHz 8088 for 8-bit ISA, or 8 MHz 80286 for 16-bit ISA cards.
 
well after running some bench programs I still have no idea what things are running at as they all seemed to give different results, sysinfo ver. 6 said I was running at 12mhz and 15 in turbo while topbench reported 8mhz and 12mhz in turbo. I really don't need this machine to run that fast so next week I'll probably invest in a 10mhz v30 and hopefully that will fix things. I just don't want to fry my 8087 in the meantime if thats a possibility.
 
As Stone said in post #2, the cpu doesn't change the bus speed.. so if you had a 5MHz/10MHz system before you inserted the V30 then you still have a 5MHz/10MHz system. The tools probably have some trouble deducing the speed because the V30 is a different processor. And changing the CPU from a version rated at 16MHz to another rated at 10MHz will change nothing - that number only tells you at what speed it can run without failing.
 
The V30 is running at the same frequency as the 8086. Just because it's rated for 16 MHz doesn't mean it runs at that speed automagically after a drop-in replacement. Your FPU will be fine and you don't need to buy another slower processor. :xmas:

EDIT: Tor beat me to it. :)
 
Tor said:
As Stone said in post #2, the cpu doesn't change the bus speed.. so if you had a 5MHz/10MHz system before you inserted the V30 then you still have a 5MHz/10MHz system. The tools probably have some trouble deducing the speed because the V30 is a different processor. And changing the CPU from a version rated at 16MHz to another rated at 10MHz will change nothing - that number only tells you at what speed it can run without failing.
Click to expand...

okay, thanks. I suppose that's how it works with every other pc but my experience with PC/XT class machines is limited so I was wondering if things may be different. The odd readings from the bench programs didn't help.
 
The odd readings would be because the benchmark tries to figure out how fast the bus runs by running some CPU instructions and presumably trying to time that (there are some other members on this forum with real knowledge about exactly how those benchmarks work, but I don't think I'm too far off). As the V30 can do some operations faster than the 8086 (on the same clock speed) the benchmark comes to the wrong conclusion. If you could find a benchmark which would recognize or could be told about the V30 you would probably get a more accurate result.

The frequency marking on the CPU (and many other chips) is usually just what it tested good at - they may even come from the same batch. One fails at 16MHz, so label it as 10MHz. The other worked, so label it as 16MHz (and sell it at a higher price). No other difference. If they really came from the same batch or not may vary between fabs, but for the Zilog Z80, for example, it appears that modern 10MHz and 20MHz chips are actually the "same".. whereas the 8MHz ones are different. (Then you have things like the tri-core AMD in my father's PC.. that's just a quad-core CPU that failed the test. 3 cores passed, so it gets sold as 3-core[*]. And so on.)

So, the CPU doesn't set the frequency, except for some types which can alternatively use an internal PLL if you don't need the accuracy of an external oscillator, for example the Parallax P8X32A (aka Propeller). But the PC runs from a CPU-external oscillator all the time. The CPU has no saying there.


[*]although in that case I guess there's also a fuse or something that's blown after the test to disable the failing core.
 
krebizfan said:
Find the crystal and typically divide by 3. That will be more exact than the estimates done by software.
Click to expand...

Most "Turbo XT" systems use the regular 14.31818 MHz oscillator and just change the divider.

14.31818 / 3 = 4.77 MHz
14.31818 / 2 = 7.16 MHz
14.31818 / 1.5 = 9.54 MHz
 
vwestlife said:
Most "Turbo XT" systems use the regular 14.31818 MHz oscillator and just change the divider.

14.31818 / 3 = 4.77 MHz
14.31818 / 2 = 7.16 MHz
14.31818 / 1.5 = 9.54 MHz
Click to expand...

That might apply for newer XT clones built around a chipset ASIC. Older ones normally have a 8284 clock generator IC with a 14.31818 MHz crystal (for "normal" 4.77 MHz CPU frequency, as well as for PIT, DMA, and producing OSC signal that some CGA and audio cards are using), and an oscillator connected to EFI input of 8284 (this is to produce turbo frequency).

Also it is kind of difficult to divide clock by 1.5. I've seen systems with 28.63636 MHz crystals, that gets divided by 6, 4, or 3 to produce the CPU frequencies you've mentioned above.
 
Mines a very late XT machine. Epson Equity E1. I see 3 of what look like crystals on the motherboard, 1 is 14.31818, 1 is 20.000 and one is 48.000000mhz

*edit* found anouther crystal marked 21.175mhz
 
You must log in or register to reply here.
Back
Top