Threadripper for real work!...?

[JCBiggs]

Ltt...making videos,  transcoding, moving 8k files around..pffft..

 

We use our workstations for real work.  Writing programs for half million dollar cnc machines, 

 

..ok just kidding.. making videos is in depth in its own way. 

 

In mastercam (G CODE software / CAM)  there  is an option called dynamic machining. What this does is calculate a cutting tool's angle of engagement, its overlap,  and "missing material" from previous cutting operations, to give you what amounts to the fastest and smoothest cutting path possible.  (Removing the most material as fast as possible, without breaking your tool and maximizing its life as well) this is also one of the only things that's multithreaded and scalable. 

 

Im interested in supplying someone who owns thread ripper, with the software, license, and data to do some testing vs my 6 core 6850k.  This is one of the very computationally heavy and time consuming operations that cam programmers deal with daily; especially missing material calculations. (All you'll haveto do is install and click go basically) The industry will thank you.  

[OJC_GALANT]

If you say that it scales well with threads then surely Threadripper should be a no-brainer? I mean it offers the best price/multicore-performance of any CPU as of now, right?

 

Good luck

 

J

[W-L]

7 minutes ago, JCBiggs said:

Ltt...making videos,  transcoding, moving 8k files around..pffft..

We use our workstations for real work.  Writing programs for half million dollar cnc machines, 

..ok just kidding.. making videos is in depth in its own way. 

In mastercam (G CODE software / CAM)  there  is an option called dynamic machining. What this does is calculate a cutting tool's angle of engagement, its overlap,  and "missing material" from previous cutting operations, to give you what amounts to the fastest and smoothest cutting path possible.  (Removing the most material as fast as possible, without breaking your tool and maximizing its life as well) this is also one of the only things that's multithreaded and scalable. 

Im interested in supplying someone who owns thread ripper, with the software, license, and data to do some testing vs my 6 core 6850k.  This is one of the very computationally heavy and time consuming operations that cam programmers deal with daily; especially missing material calculations. (All you'll haveto do is install and click go basically) The industry will thank you.  

As long as MasterCAM has the drivers and support for the new processors and is able to take advantage of all the threads it would work straight out of the box. MasterCAM's dynamic or HSM toolpath calculation is fairly intensive but unless you have hundreds of separate toolpaths I don't see it being a huge leap and bound over your current hardware. 

[JCBiggs]

Not nessecarily.  Ipc  could be different or it might not be optimized well for this type of work. Once upon a time amd led nvidia in solidwork performance, even with much kess horsepower on paper.  That's no longer the case

[DocSwag]

11 minutes ago, OJC_GALANT said:

If you say that it scales well with threads then surely Threadripper should be a no-brainer? I mean it offers the best price/multicore-performance of any CPU as of now, right?

 

Good luck

 

J

For this type of workload stability might matter a lot though, they may want to consider xeons or epycs or something for stability reasons?

 

I do recall TR supports ECC though so there's that too...

[JCBiggs]

9 minutes ago, W-L said:

As long as MasterCAM has the drivers and support for the new processors and is able to take advantage of all the threads it would work straight out of the box. MasterCAM's dynamic or HSM toolpath calculation is fairly intensive but unless you have hundreds of separate toolpaths I don't see it being a huge leap and bound over your current hardware. 

Multipe "opti-rest"dynamic operations can take up to  5-6 minutes on my workstation. Using the "calculate tool path group" option is the heaviest. (For instance when you step down a tool size)  5 axis is even worse.  And if you use optirough as finishing it can take even longer. (Though that's not normal and only done for first pass finishing ops.)

 

Simulation takes a while as well. Even with a p4000 gpu.  

[JCBiggs]

5 minutes ago, JCBiggs said:

Multipe "opti-rest"dynamic operations can take up to  5-6 minutes on my workstation. Using the "calculate tool path group" option is the heaviest. (For instance when you step down a tool size)  5 axis is even worse.  And if you use optirough as finishing it can take even longer. (Though that's not normal and only done for first pass finishing ops.)

 

Simulation takes a while as well. Even with a p4000 gpu.  And i don't have a "slow" system by any means.  

 

[W-L]

1 minute ago, JCBiggs said:

Multipe opti-rest material dynamic operations can take up to  5-6 minutes on my workstation. Using the "calculate tool path group" option is the heaviest.  5 axis is even worse.  And if you use optirough as finishing it can take even longer. (Though that's not normal and only done for first pass finishing ops.)

Opt-rest is one I've never used but for sure with any 3D toolpaths such as opti-rough in multiple operations that can add up. I've always done it one the basis of calculating as I go but if you group it for different operations then that's a different story.

[JCBiggs]

9 minutes ago, W-L said:

Opt-rest is one I've never used but for sure with any 3D toolpaths such as opti-rough in multiple operations that can add up. I've always done it one the basis of calculating as I go but if you group it for different operations then that's a different story.

Sounds like your familiar with mastercam.

 

Try it sometime.  Just use a 3/4 tool to optirough a complex surface.  Then just copy the tool path  and change the rest material to all groups and a 1/4" tool.   Set stepovers  to 20% 

[W-L]

7 minutes ago, JCBiggs said:

Sounds like your familiar with mastercam.

 

Try it sometime.  Just use a 3/4 tool to optirough a complex surface.  Then just copy the tool path  and change the rest material to all groups.   Set stepovers  to 20% 

It was the first program I used and learned to program a CNC on, hmm might just give it a go and watch my system suffer  

[JCBiggs]

2 minutes ago, W-L said:

It was the first program I used and learned to program a CNC on, hmm might just give it a go and watch my system suffer  

Try it with a ballmill for even more heat. Lol. And .0004" linearization tolerance. (Yes we do surfaces that require that accuracy)

[W-L]

Just now, JCBiggs said:

Try it with a ballmill for even more heat. Lol. And .0004" linearization tolerance. (Yes we do surfaces that require that accuracy)

My goodness, I'mma hit cycle start and go for a super long coffee break then, haha!

[JCBiggs]

Lol its fast on the machine.  140-400 ipm feed rates...its just annoying waiting for mastercam to think.  You end up just staringat the screen waiting for the regens. 

[W-L]

34 minutes ago, JCBiggs said:

Lol its fast on the machine.  140-400 ipm feed rates...its just annoying waiting for mastercam to think.  You end up just staringat the screen waiting for the regens. 

Yeah one of the fellows I worked with used to call it mastercam TV since you just wait and wait. 

[JCBiggs]

Well I'm very interested whether it's going to use all that extra headroom where threadripper beats everything from Intel. (In the higher thread counts) If it does and it's significant I'll put together new workstations. 

[JCBiggs]

any chance you got to have a go with threadripper and mastercam?

On 8/11/2017 at 10:27 PM, W-L said:

Yeah one of the fellows I worked with used to call it mastercam TV since you just wait and wait. 

 

[W-L]

3 hours ago, JCBiggs said:

any chance you got to have a go with threadripper and mastercam?

No unfortunately I don't have access to anyone with a threadripper system that I could borrow. 

[JCBiggs]

On 8/24/2017 at 5:06 PM, W-L said:

No unfortunately I don't have access to anyone with a threadripper system that I could borrow. 

@LinusSebastian

[bdrescher]

JCBiggs,

 

I have just built a threadripper machine for our company.  I built it for Mastercam so I would not need anything but your toothpath's to regenerate and verify. Yes everybody thinks since portions of mastercam that the more threads the better but you are correct by saying that it's all about how the software is coded.  It has shown a 26 percent improvement in regen times over the previous 2600k overlcocked at 4.6.  I was very excited to see that it was also using all 16 cores to do so.  I did not expect this.  However I have this threadripper overclocked to 4Ghz.  It appears although it was using all 16 cores that pure speed is also quite necessary. If you want need my system specs just let me know.  I would.