IBM-XT cpu upgrade?

[sergey]

The PC-SPRINT upgrade is best for a PCjr.

PCjr doesn't have DMA - no DMA no problem Well almost no problem... the real time clock will run faster, and sound pitch probably will change (PCjr uses SN76489, but I assume it is still fed with the frequency derived from CPU clock).

Sergey, I assume that your board maintains a 2.36MHz PCLK and 14.318MHz OSC signal regardless of the speed of CLK signal going to the CPU.

Right, that's what it was supposed to do... and you found the first bug! I made PCLK to be OSC/12.... and it actually should be OSC/6.

The frequency of 8284's OSC output is always the same as the crystal frequency (not EFI), so no modification needed there.

 

[Trixter]

And why would i upgrade? because it is fun to see old stuffs perform better then it was made for, not because i need it.

I share your view that this is fun to play with. In that vein, you can swap in an NEC V20 for about 20% more speed on average, and if you would like to see a "silly" speedup you can run FRACTINT before and after and compare speeds. You can then install an 8087 math coprocessor and run FRACTINT again and notice the 2-5x speedup on some fractals An 8087 will speed up CAD and spreadsheet software, and specialty math programs like MATLAB and FRACTINT. But that's about it, so it's not really necessary if you have trouble sourcing an 8087.

 

[maxtherabbit]

I've got a couple of clone boards with the (labeled) 5MHz 8237 DMACs. I can verify that they're run at 8MHz in turbo mode. I've wondered if it could be that it was common knowledge among the clone makers that 5MHz labeled parts were either from the same run as the 8MHz parts or simply all 8237s made since a certain date were capable of running in a PC at 8MHz.

The early "turbo" boards even had the defect that the time-of-day clock would run too fast and the tones coming out of the speaker would be too high.

This thread seems to have died on the vine, but I wonder - did anyone definitively determine if the PC-Sprint and the like actually run the DMAC at >5MHz in turbo mode?

 

[inmbolmie]

This thread seems to have died on the vine, but I wonder - did anyone definitively determine if the PC-Sprint and the like actually run the DMAC at >5MHz in turbo mode?

The PC-Sprint ignores DMA activity, so as the DMAC clock is generated directly from the CPU clock in the mainboard, it will run at whatever speed the CPU is running. With that you can go in most systems up in frequency to some point where DMA stops working and the system crashes.

With Sergey's design you can go higher as the CPU speed is throttled down when DMA activity is detected so the DMA chip always works at stock speed. I have tested Sergey's design http://www.vcfed.org/forum/showthread.php?70923-IBM-5160-overclock-Sergey%92s-way and I was able to go to 8,33Mhz in my 5160. Disabling the DMA throttling I can go up to 6.7Mhz before it crashes, but that swapping also the CPU (v20) and the 8284 chip for a faster one, so it is not that easy. The benefit of PC-Sprint over Sergey's is that you don't need to solder cables to get the required DMA signals for DMA throttling to work, but the maximum speed will be limited by the DMAC operation.