Has anyone replaced both 8284 and 8088 with NEC chips?

[Chuck(G)]

Swapping a V20 in place of an 8088 is common practice, but has anyone replaced the 8284 clock generator with a NEC 71011 chip (specific to the V-series CPUs)? That alone would cause a boost in CPU frequency. Instead of 4.77 MHz, it would be 7.2 MHz or thereabouts with no crystal changes.

Am I missing something important?

 

[Eudimorphodon]

I wonder if that might have any side effects, since the CPUCLK appears on the ISA bus at least. (Don't know what else on the motherboard might latch onto it.) Although...

Re: your comment in that other thread about the 33% vs. 50% duty cycle for the 8088 vs. the V20. (Although, actually, does the V20 actually *specify* a duty cycle in its datasheet? Scanning/searching the first datasheet I could find that covers the V20/V30 I couldn't find anything that actually mentioned "duty cycle" or said there'd be anything wrong with a 33% timing.) Here's a possibly interesting observation: according to the technical manuals for the Tandy 1000 EX and HX, which are dual speed 4.77/7.16mhz clones, they actually do run their 8088 CPU at a 50% duty cycle at the high speed. These machines use a custom ASIC for clock generation instead of a separate 8284.

I don't think my old tube oscilloscope is fast or accurate enough to make confirming that these manuals are telling the truth easy, but looking at the datasheet for the 8088-2 I think I can see enough weasel room in the AC specs for them to get away with it; the minimum "low" time for the 8088-2 is the same (well, within 1ns) of the minimum "high" time for the 5mhz 8088, and the "MAX" clock period for both is 500ns. That *implies* at least that whatever circuitry on the chip that relies on the longer low time will still be able to get the job done with the 50/50 duty cycle?

Assuming this is actually true I'm wondering if this hack might have been widespread in 4.77/7.16mhz "Turbo" XTs. The big plus is it saves you needing two crystals, which you otherwise would if the 8284 is hard-coded for the 33% duty cycle.

 

[Chuck(G)]

I assume that the duty cycle is computed from t[sub]cyk[/sub], t[sub]kkh[/sub], t[sub]kkl[/sub], t[sub]kf[/sub] and t[sub]kr[/sub] numbers.

 

[gepooljr]

I've never heard about this ever. My Leading Edge Model D has the NEC V20. If I did replace the 8284 without the crystal swap - would that not cause instability? Please elaborate.

 

[Chuck(G)]

I'm not sure, which is what I'm wondering about. The bus clock would rise accordingly; I don't know if that would affect peripherals. The main difference between the 71011 and the 8284 is that the processor clock is the oscillator frequency divided by 3 on the 8284 and by 2 on the 71011. The rest of the chip is the same.

 

[Eudimorphodon]

I assume that the duty cycle is computed from t[sub]cyk[/sub], t[sub]kkh[/sub], t[sub]kkl[/sub], t[sub]kf[/sub] and t[sub]kr[/sub] numbers.

I just threw an update into that other thread; FWIW, at least some of the V20 datasheets actually do specify minimum low times longer than min high times. (For instance, the PD70108-8 specifies 44ns high and 60ns low as minimum for a total minimum clock cycle of 125ns, with max rise/fall times of 10ns.) It never really says that the ideal duty cycle is 50%. The same numbers for the 8088-2 are 44ns/68ns, which is really about the same. So... I guess the question is at what point do you round to 50%?

Aha! I just found a datasheet for the 5mhz PD70108-5, and it says the up/down times are 69 and 90ns. (The numbers for a 5mhz 8088 are 69 and 118ns; the V-20s are consistent about being more tolerant of a shorter down clock, but it's not a *lot* different.) What's more interesting about that datasheet (Here's a link) is it specifically says either the 71084 or the 71011 will work for either the 5 or 8mhz versions, at least. So I guess that settles whether a 33% duty cycle is "okay" for a V-20.

I am curious, though, if the idea of swapping the 8284 for the 71011 would work to give you 7.16mhz in a 4.77mhz XT with no side effects. If the verbage in the Tandy manual is accurate it would essentially put out the same CPU clock cycle as a Tandy 1000 SX/EX/HX runs with in high speed mode, which means it *might* even work with an 8088-2.

 

[modem7]

Am I missing something important?

Refer to the diagram at [here]. Frequency changes shown in red.

- 8288 needs a faster rating ?
- Any fitted NPU needs a faster rating ?
- Change to PCLK ; impact to things relying on the 8253 (e.g. DOS clock, some BIOS delays, speaker frequencies)

 

[Eudimorphodon]

- Change to PCLK ; impact to things relying on the 8253 (e.g. DOS clock, some BIOS delays, speaker frequencies)

Oh, yeah, that’d be a problem. Pretty sure the Tandy ASIC I mentioned keeps PCLK constant regardless of CLK.

 

[Chuck(G)]

Refer to the diagram at [here]. Frequency changes shown in red.

- 8288 needs a faster rating ?
- Any fitted NPU needs a faster rating ?
- Change to PCLK ; impact to things relying on the 8253 (e.g. DOS clock, some BIOS delays, speaker frequencies)

All good points. I do recall early "Turbo" XTs that simply dumped in the output from a faster XCO into EFI. The 8253 not keeping good time was one of the major issues. You could tweak the ISR to bring that back into normal range, but the basic "tick" was still too fast. Yet the things sold hundreds of thousands. So there's that. That's the reason for the turbo mode switching uses "hotkey" combinations--so that the ISRs can be adjusted. Refresh on the DRAM was another side-effect (faster than needed).

I've got at least one motherboard whose BIOS traps INT 13h floppy calls and switches to the lower frequency.

Given all of this, you'd think that nobody would buy a "Turbo XT". TH99 says otherwise.