Current best computer choices for Mastercam

[BenK]

If that ever becomes an option I will buy myself a server rack and start loading it up. 

Honestly though a lot of performance is left underutilized because they don't use GPU's for more than just the display. If they started taking advantage of Cuda technology you would see regen times plummet.

[Matthew Hajicek - Singularity]

You may need a $6000 (new) laptop for top of the line performance, but if you're not a road warrior you can get the same or better performance for about half the price on a desktop.  Even the same chips usually run slower in a laptop due to thermal constraints.  If I were spending $6k on a workstation I'd spend a lot of it on extreme cooling for overclocking.  It is theoretically possible to get more than 50% overclock under extreme conditions (liquid Helium):

https://www.extremetech.com/computing/250622-intel-core-i7-7740k-shatters-overclocking-records-hits-7-5ghz-thanks-liquid-helium

That's a much bigger difference than the ~12% improvement you get by waiting for the next generation of chips.

If you get a LN2 generator for cryogenic machining it could serve double duty cooling your CPU for maybe a 30% boost over stock.

 

[huskermcdoogle]

34 minutes ago, gcode said:

true... I was just referring to it's parallel computing technology...

 It would be sweet if Mastercam had the ability to operate off of a local computing cloud.  Have affordable user stations, coupled to a cluster computer via a high speed network, and the cluster does all the tool-path crunching.  If you have enough programmers I would think it would pay for itself pretty quickly if you could process tool-paths say 100x faster.  Then again, how many people within one company are crunching big tool-path files day in day out?

[crazy^millman]

1 minute ago, huskermcdoogle said:

  Then again, how many people within one company are crunching big tool-path files day in day out?

Raises hand.

[huskermcdoogle]

Just now, C^Millman said:

Raises hand.

You don't count....  Everyone knows you play with big files all the time, also how often are you working in a situation where you could have a server rack to process toolpaths with you?

[crazy^millman]

Just now, huskermcdoogle said:

You don't count....  Everyone knows you play with big files all the time, also how often are you working in a situation where you could have a server rack to process toolpaths with you?

I would have no problem rolling up to a job site and pushing out a sever rack and a swap cooler for the heat it generates and plugging it all into a 440 V outlet pulling about 600 to a 1000 amps. I have days of time processing files sometimes and if I could do some of the 71 minutes stock model processes down to 71 seconds I would be very happy to do so. I have people screaming for programs right now so not losing days of time having instant answers (Toolpath creation and Verification) would be worth it's weight in gold right now.

[huskermcdoogle]

10 minutes ago, C^Millman said:

I would have no problem rolling up to a job site and pushing out a sever rack and a swap cooler for the heat it generates and plugging it all into a 440 V outlet pulling about 600 to a 1000 amps.

Put it all in a special fabricated crate, and ship it LTL from site to site.  If need be have an on board generator....  Put it in the parking lot and drag a bundle of network cables inside.

[medaq]

52 minutes ago, huskermcdoogle said:

 Then again, how many people within one company are crunching big tool-path files day in day out?

Where does that line start? 10 megs files, 100 meg files, 500 meg files?  I am constantly using 100 meg files. I do not think it should matter really. If any programmer is sitting there waiting for something to process. Then they to would take advantage of faster computing. This whole thread has been informative but kinda a let down. I can only hope Mastercam is looking at how to fix, yes I am using the word fix, these long processing times.  

[crazy^millman]

12 minutes ago, medaq said:

Where does that line start? 10 megs files, 100 meg files, 500 meg files?  I am constantly using 100 meg files. I do not think it should matter really. If any programmer is sitting there waiting for something to process. Then they to would take advantage of faster computing. This whole thread has been informative but kinda a let down. I can only hope Mastercam is looking at how to fix, yes I am using the word fix, these long processing times.  

We agree, but I have taken some of my work and put it through other CAM and there was not much difference in the processing time for toolpaths. Some of the Majors struggle more than Mastercam on the work I and other do. I may seem like a rhubarb, but I do explore what is out there and every CAM has their strengths and weaknesses. One huge one for Mastercam IMHO is this forum. 

[Matthew Hajicek - Singularity]

I used to wait half an hour for a single-surface flowline operation for a guitar neck in the late 90's.  I'd do everything up with coarse tolerances and wide stepovers to make sure it's doing what I want, then tighten down the settings and let it crunch.  Everything seems so quick and easy now by comparison.

[huskermcdoogle]

29 minutes ago, Matthew Hajicek™ - Conventus said:

I used to wait half an hour for a single-surface flowline operation for a guitar neck in the late 90's.  I'd do everything up with coarse tolerances and wide stepovers to make sure it's doing what I want, then tighten down the settings and let it crunch.  Everything seems so quick and easy now by comparison.

I do this now with the multiax paths, except usually only to save a few minutes.  Longest regen times I usually have are like 3-4 minutes, and that would be because I screwed something up.  I don't do tight tolerance stuff, my stepovers are bigger, so it certainly lightens up the load.   But simulation, and stock models take forever in comparison.

[Matthew Hajicek - Singularity]

Stock models?! Luxury! In my day you had to REMEMBER what you cut! And we lived in a rolled up newspaper in a septic tank!

 

 

[huskermcdoogle]

I can't use stock model for the things it would actually be useful for in the work we do.  It crashes it certainly, every-time.  So I end up using mach sim, and then import an stl back in....   Bagged-tagged and logged as R-11614.

[mike93]

Well what should we build, we well be getting a desk top built in the next couple of  weeks for one of are primary programmers. Time is money so it needs to be fast. .01% faster for a extra grand ??? 10%faster for a grand!!!

[crazy^millman]

1 hour ago, Matthew Hajicek™ - Conventus said:

I used to wait half an hour for a single-surface flowline operation for a guitar neck in the late 90's.  I'd do everything up with coarse tolerances and wide stepovers to make sure it's doing what I want, then tighten down the settings and let it crunch.  Everything seems so quick and easy now by comparison.

I did a prototype part back in 2003 that took 4 days to process the toolpath. It was with a .03 Ball endmill with a .001 stepover. Time we vapor polished the clear Polycarbonate it looked like glass. It was a Waterfall shower head for a trade show so it needed to look like a production quality glass blown piece.

[crazy^millman]

1 minute ago, mike93 said:

Well what should we build, we well be getting a desk top built in the next couple of  weeks for one of are primary programmers. Time is money so it needs to be fast. .01% faster for a extra grand ??? 10%faster for a grand!!!

Did you look at what I spec'd out?

[Matthew Hajicek - Singularity]

I'd go with the i7-7740X to maximize single thread performance.  More expensive isn't always (and often isn't) faster.

[mike93]

Sorry, looked at a second time,  i was thinking it was laptop spec, I will print and send to IT builder, thanks!!

[Matthew Hajicek - Singularity]

It would be fun to try using a liquid cooling system, fill it with a water / antifreeze mix, and put the radiator in a freezer.  Or maybe hack apart a freezer and pipe the cold side into the CPU cooler.

Hey, Google tells me that sub-ambient cooling is now available off-the-shelf down to 0°C : https://koolance.com/

 

[huskermcdoogle]

Do those mount inside the case?

[Festus]

One thing I've never seen discussed is posting code. I deal with slower toolpathing speeds by trying to use high speed toolpaths as much as possible. this at least allows me to continue working. My bigger issue is when i go to post I may be shut down completely for  2-3 hours sometimes. any suggestions?

[medaq]

15 minutes ago, Festus said:

One thing I've never seen discussed is posting code. I deal with slower toolpathing speeds by trying to use high speed toolpaths as much as possible. this at least allows me to continue working. My bigger issue is when i go to post I may be shut down completely for  2-3 hours sometimes. any suggestions?

I have noticed this over networks and depending on your ping rate to the server the posting times will change dramatically. Have you tried to post locally? then upload your nc file to your network file?

[BenK]

40 minutes ago, Festus said:

One thing I've never seen discussed is posting code. I deal with slower toolpathing speeds by trying to use high speed toolpaths as much as possible. this at least allows me to continue working. My bigger issue is when i go to post I may be shut down completely for  2-3 hours sometimes. any suggestions?

 

You have something else going on. I have never seen a program take more than a few minutes to post. 

Are your posts on a network or locally on the computer itself? 

[crazy^millman]

I am posting 60 to 80 mb files in about 15 minutes. You need to post locally then save to the network. Never I repeat never try to post across a network with a post that is just asking for major trouble.

[Festus]

17 minutes ago, C^Millman said:

I am posting 60 to 80 mb files in about 15 minutes. You need to post locally then save to the network. Never I repeat never try to post across a network with a post that is just asking for major trouble.

Just tested a file where my post file ended up being 65.8mb, 1,816,625 lines of code. It took 10 minutes locally and 10 minutes to my server. No noticeable difference?