FWIW, I just today installed a new computer on my CNC mill that happens to have an ASUS mother board also. The old computer got a gremlin in it and was doing very bizarre things.

The new computer is a budget build using an ASUS motherboard P5G41-M LX2 with an Intel E6500 dual core processor. I re-used the memory, PCI parallel port card, hard drives, case, etc. from the old computer. The new computer is getting along fine so far with Mach3 and is running stable at the full 100k pulse speed.

I am glad you had good luck with that Asus build.

On a side note I notice many folks talking about running at 100K. I am assuming this is the kernal speed. I am running at 35k. To be honest I don't know the advantages of running any higher or lower. Could someone let me know what happens if I bump it up? Do you need a monster processor to run at 100K? What are the advantages?

If you dont need to run any faster than 35KHz for your setup then dont as it will not give you any benefits and may actually cause issues. The only reason to use a faster kernel speed is if you cant get the required rapids from the kernel you are at now. With a stepper system 45KHz would likely be the normal max you would need, a servo system may, depending on encoder count, require faster to achieve its potential. Although its been quite a few years since I have used the parallel port with Mach I never had a computer that was 100% happy at 100KHz.

Hood

It is all in the numbers.

Figure out the RPM you want to run the motors and the numbers of steps required to achive that , for example (using round numbers);

Steppers  1,000 RPM x 200 steps per rev = 200,000 pulses (per minute) which is only about 3,300 pulses per second, so 25,000 kernel speed would be plenty, however, if you have 10 micro step drive, suddenly you need 33,000 pulses.

Servo     4,000 RPM x 8000 steps per rev = 32,000,000 pulses per minute which is 53,300 steps per second, so you yound need to run 60k kernel speed to get full rpm.

Alternatives to full kernel speed in Mach are pulse mulripliers in the drives or a pulse gerenator independent to Mach3 such as Smoothstepper or Kflop.

A few years ago is a long time in computer history. I just put together what I would call a 'budget' build using 2 gerations out-of-date technology (socket 775) and it is running Mach at 100k (so far). A good test of headroom is to open the task manager and set it on performance and then drag it rapidly around the screen as Mach is running a program.

Simpson36

that is one of the best descriptions of the pulse and a real good quick test.
the people that write the mach manual should use it.

bill

Thanks docltf

If they correct my rather significant boo-boo in the math (what's an order of magnitude between friends, eh?) then they are certainly free to use it.

The correct number of pulses per second for the servo example is 533,000 . . . slightly above Mach's maximum kernel speed.  

Even if one can run Mach at 100k, a 6x multiplier would still be needed in that example to get max RPM. This is  doable with many, if not most, servo drives . .  and also explains why I have to use multipliers . . .  

Another example for a realistic servo setup at the hobby level would be:

3,000 RPM x 2000 steps per rev (500 line encoder in quadrature) = 6,000,000 steps per minute which coincidentally is 100,000 per second . .  if I got it right this time  

Glad you added the "(so far)" bit
It is like you say that 2 yrs is a long time but I also thought mine was fine,  occasionally however it would get wee glitches and dropping the kernel down to 45KHz cured these problems. Worth noting was that 65 and 75k ran like crap for some reason and was apparent from the start that things were not right on these kernels.

As for electronic gearing, some like it others dont, I am in the latter camp. Real world it may not make a big difference buit I just didnt like the jerky motion on a clock (DTI) when I commanded a very slow move with an electronic gearing of 6 or so, put it to none and it was silky smooth Maybe that was down to my tuning or possibly even the way my drives deal with things but as Greg had sent me a test version of the SS plugin for  threading  I never really bothered looking into it further.
Hood

damn the math,the general direction of the concept was well put.

i am with you Hood,get the resolution done with the mechanical parts.that way you can put the motors on a full step.


bill

Glad you added the "(so far)" bit
It is like you say that 2 yrs is a long time but I also thought mine was fine,  occasionally however it would get wee glitches and dropping the kernel down to 45KHz cured these problems. Worth noting was that 65 and 75k ran like crap for some reason and was apparent from the start that things were not right on these kernels.

My last computer was as you describe. I could only run about 45 with the 2.6ghz processor. I switched only the processor to a 3.2ghz with hyperthreading and then I could run 75k. If I set it to 100k, it would only actually kick out 89k which I found odd. I would have though it would either work or not, but it was running as a reduced speed which made me nervous so I backed it off to 75, but it ran fine there . . . uh  . .  well , until the RPM readout went psyco on it. The new computer I just put in has a dual processor @ 2.93 (or there about) and dual channel memory so it is far more powerfull than the old one and SO FAR it is running at 100k no problem. I will leave it there for all of my testing, but probably drop it back to 75 if I cut anything expensive . .  

As for electronic gearing, some like it others dont, I am in the latter camp. Real world it may not make a big difference buit I just didnt like the jerky motion on a clock (DTI) when I commanded a very slow move with an electronic gearing of 6 or so, put it to none and it was silky smooth

There may be other factors, but one for certain is the way the drive handles the process. The Mistubishi MR-C series also is not very happy with multipliers. However, I would put a $100 bill on the bench and give it to you if you could tell the difference between a 6x an 0x multiplier on a Granite drive running a DC brush motor. Likewise the Mistubishi J2S series is smooth as silk running a HUGE multiplier . . .  the thing has outrageous resolution. "An up to 500kpps high-speed pulse train is used to control the speed and direction of a motor and execute precision positioning of 131072 pulses/rev resolution."The only time I have ever actually recommended the Smoothstepper was to use with Dugongs since that drive runs pretty badly with multipliers, especially at low speeds as you pointed out.  Curiously, the little brother Whale drive is fine with multipliers active. I still run those on my X and Y axis at 5x or 6x.  Mind you that when set at 1:1 the Dugong is a killer. Smooth and very powerfull with the ability to stop on a dime.

I couldnt physically see the jerky motion it was only if I commanded a move at something like 1mm/min and had the DTI against the axis, you could see the needle step round but with it at 1:1 it was smooth motion. As said real world it would likely not be an issue but if I know its there it bugs me
 The encoders that I have on the spindle motors on both lathe and beaver mill are what they call intelligent encoders and you can have, if I recall correctly,  the drive interpolate up to 2 million counts per rev(8million pulses per revolution   ), thats even beyond SmoothStepper territory I was happy just at the base 1024 x 4 for a spindle.

Hood