Jump to content

Welcome to eMastercam

Register now to participate in the forums, access the download area, buy Mastercam training materials, post processors and more. This message will be removed once you have signed in.

Use your display name or email address to sign in:

Build a new PC and Mastercam isn't using all cores


gorsek1
 Share

Recommended Posts

Hello,

I have build a new PC.

-Intel i9 10900k

-Quadro rtx 4000

-128gb RAM

-Samsung 970 evo Plus SSD

-CPU is watercooled and never goes beyong 60 degrees even at full load.

 

So the problem that i have is that it doesn't utiilize all 20 cores @100%... The only time it uses them all is when i use the compare function in verify.

Any suggestions how i could increase the speed?

 

In fact its working slower than my previous i7 7700k. There i get 100% load in verify and in multi thread manager.

image.png.71f69da97aa8c201ab62a165f906665b.png

Link to comment
Share on other sites

Mastercam doesn't use all cores.  They're working on improving this, but most functions are thread limited, to one, two, or four threads.  This is why single core performance is far more important than core count.  Your new chip is 10 core 20 thread, and most of that will sit idle.  Your old chip was 4 core 8 thread, much easier to make use of.  You can use more if you have multiple things crunching at once.

Most software is not written to be massively parallel, and many tasks simply cannot be massively parallel; they must be calculated sequentially, or have a small number of sequential algorithms than can be run in parallel.  When building a rig I try to get a chip near or at the top of this list:

https://www.cpubenchmark.net/singleThread.html

So right now I'd probably get a Ryzen 7 5800X or an i9-11900K, but your i9-10900K isn't too far down the list.  Was your i7 overclocked?  I've been running my 6700K at 4.5Ghz for over five years (probably due for replacement).

Oh, I do seem to recall something about disabling hyperthreading; I don't remember if you want hyperthreading on or off to maximize performance.  Hopefully someone else will step in to clarify.

Link to comment
Share on other sites

Yeah the i7 7700k was overclocked to 4.8ghz.

But still with the i9 10900k i have 5ghz on all cores and 5.3ghz on single core. 

I really tought that this processor and gpu will speed up the process. Especially the verify part. Because i have large 3d parts that take alot of time to program... 

It takes me roughly 1-2weeks to finnish the programm for 1part working 8-12 hours a day... And the "waiting" part is killing me. 

 

For 2d programming mastercam is the best. but for 3d they need to optimize it so that it will crunch numbers faster. 

Link to comment
Share on other sites

Here I did a test for raster operation. 0.15mm steppover on a part that is 800mmx600mm.
i7 7700k: 6min56s
i9 10900k: 5min37s

Verify for this operation:
I7 7700k
processing NCI data: 53s
3d simulation: 3min18s
Compare: 15s

I9 10900k
processing NCI data :: 44s
3d simulation: 2min3s
Compare: 10s

When grinding the raster operation was only for a short time at 100% when doing Parralel passes:
https://prnt.sc/10duu1q
With linking parralel passes and all other operations it was 10-40%.
https://prnt.sc/10duvep
Where it was crunching numbers with a 100% load on CPUthe operation was almost immediately over.
Verify:
The processor was 100% loaded only during the compare operation.
In the simulation, it rarely went over 50%.

The Raster operation is one of the fastest out of the finnishing operations. And i have like 60 different operations per part. 

i7 7700k overclocked to 4.8ghz.
I9 10900k was on stock. I could still overclock it a bit so I still gain 5-10%. That's it. I don't think that Intel or Amd  can improve single core speed by too much. maybe 5% increase each year, But thats too slow.
If 3D worked like 2D it would be great. :)

Greetings G

Link to comment
Share on other sites

About a year ago, we had a brief discussion on what can and can't be multi-threaded in a 3d environment, here:

If you can describe your work a bit better, we can probably tell you whether it's benefiting from multi-threading or not. 

4 hours ago, Matthew Hajicek - Conventus said:

Mastercam doesn't use all cores.  They're working on improving this, but most functions are thread limited, to one, two, or four threads.

I'm not sure where you got this information, but that is not the case.   

Your second and third paragraphs are significantly more true, though, many tasks cannot be done in parallel when you think about them, but we've had test rigs that had 256 cores and we're able to use all of them.   A note though, the more cores & parallel processing you do, the more overhead there is on managing the job itself (dividing it up, handing out, receiving  the calculation back, collating it all back to a cohesive toolpath to be linked properly, etc.) so there is a point of diminishing returns that you pretty quickly reach. 

  • Like 4
Link to comment
Share on other sites
On 3/5/2021 at 5:12 PM, Aaron Eberhard - CNC Software said:

 but we've had test rigs that had 256 cores and we're able to use all of them. 

These testing rigs, can you provide any performance specifics about them? Do you have a testing file you can share that fired up all these cores?
I'm also looking for more hard numbers on video cards in Mastercam, there isn't much out there, anything you could share on that? 

This is something I'm sure has been brought up before, am I off base in assuming a GPU should be able to crunch numbers faster than a CPU? I'm just thinking since all this crypto mining is done by crunching equations and GPUs seem to be far and away the best option for this...is it just simply the nature of the math problems being solved that Mastercam wouldn't benefit? Maybe I'm not even realizing the GPU is being used for calculations in Mastercam?

  • Huh? 1
Link to comment
Share on other sites
51 minutes ago, mwearne said:

These testing rigs, can you provide any performance specifics about them? Do you have a testing file you can share that fired up all these cores?
I'm also looking for more hard numbers on video cards in Mastercam, there isn't much out there, anything you could share on that? 

This is something I'm sure has been brought up before, am I off base in assuming a GPU should be able to crunch numbers faster than a CPU? I'm just thinking since all this crypto mining is done by crunching equations and GPUs seem to be far and away the best option for this...is it just simply the nature of the math problems being solved that Mastercam wouldn't benefit? Maybe I'm not even realizing the GPU is being used for calculations in Mastercam? 

Sorry, I don't have the data handy, but I'm pretty sure they were using the Xeon Phi tech from a few years ago, maybe the year 2018?  They were in a big ol' server-box of goodness that was loaned to us by Dell.  My team wasn't directly involved in that effort, but I can try to see if anyone has something laying around from that specific project, give me a few days. 

In the mean time, you can replicate what we were doing by creating a complex pocket with a lot of the 3 axis high speed toolpaths (Opti, Waterline, Equal Scallop, etc). and embedding a LOT of them into an plate, like one of those 100+ part injection mold plates.    Something that has a lot of natural boundaries/pockets. 

And oh yeah, the CPU can do a lot of calculations faster, and we do leverage it when we can (search on here or our forum for OpenCL discussion).  Unfortunately, the types of computations GPUs can do faster is more limited in scope than you'd imagine, and for our sorts of data, the biggest bottleneck is the pipeline between the GPU & GPU RAM and CPU & system RAM.   Again, it's been a few years since I reviewed the data, but from memory, processing certain segments of the toolpath were something like 1800% faster, but, setting up the data, transferring it it, and reconstructing all of the data once we got it back from the GPU was a net loss of about 10%.  Unfortunately, the net loss increased as the complexity of the part increased even thought the core processing % was even better, as there was more external (to the GPU) data to manage, as even on the CPU, a lot of the time it's not the slice/contact point processing that's the slow part, it's creating the environment that you need to safely calculate the positions.

It's a really fun problem, and I always love talking through optimizations with them about it.  We're always working on improving the calculation times, so we're constantly seeing what we can do better.

  • Like 1
Link to comment
Share on other sites
1 hour ago, Aaron Eberhard - CNC Software said:

In the mean time, you can replicate what we were doing by creating a complex pocket with a lot of the 3 axis high speed toolpaths (Opti, Waterline, Equal Scallop, etc). and embedding a LOT of them into an plate, like one of those 100+ part injection mold plates. 

I did say you can use more cores if you have multiple things crunching at once.  But that's not typical workflow.  For me, I do a few ops, then a stock model to see the results, repeat until done.  I very rarely see all 8 of my logical cores engaged at once.  I'm sure part of that is that I'm still on X9, because the new interface is intolerable.

  • Haha 1
Link to comment
Share on other sites
4 hours ago, Matthew Hajicek - Conventus said:

I did say you can use more cores if you have multiple things crunching at once.  But that's not typical workflow.  For me, I do a few ops, then a stock model to see the results, repeat until done.  I very rarely see all 8 of my logical cores engaged at once.  I'm sure part of that is that I'm still on X9, because the new interface is intolerable.

Oh, hah.. Well, maybe there was a logical core limit back in X9?  I dunno, that was what, 8 years ago?  The threading would have been built to still support Windows XP then, probably..  7 and 8 came and went they both made major advancements to the thread management, and 10 really pushed it further.    My daily driver is in a version 8 releases newer, so I have no idea what the limits were then.

For anyone running modern software, my statements are all valid :)

  • Like 1
  • Haha 2
Link to comment
Share on other sites
56 minutes ago, Aaron Eberhard - CNC Software said:

Oh, hah.. Well, maybe there was a logical core limit back in X9?  I dunno, that was what, 8 years ago?

5 years.  I know 'cause it was the latest and greatest when I started here at Conventus.  I've been using Mastercam via access keys since '96.  When X came out, I waited four or five  years for them to fix the access keys.  I switched from V9 to X5 when they did.  Still waiting for 20XX to fix access keys to some system that makes sense or is user configurable, will happily switch over when it happens.  Or if I stick it out for another five years, if historical patterns hold, there will be another entirely new interface for everyone to relearn from scratch and lose productivity over.  And yes, this topic chafes a lot, as I've been told by CNC Software that I should love it the way it is, because everyone else loves it, and I'm the only one who's ever complained.  Meanwhile I know several people who have left for other software.  Since CNC won't approve the transfer of Conventus' licenses with the likely upcoming sale of the shop, I'll be making the decision from scratch again, and that other software is looking a lot more appealing.

I'm still waiting for the videos I was promised (three times) of someone, anyone, programming quickly and efficiently using the ribbon bar.

Link to comment
Share on other sites

In defense of PowerUsers everywhere;

[rant]...the ribbon IS a steaming pile of dog**** pure and simple. It's a cancer that has spread all throughout UI Dev. and it sucks. I hate it, but because literally every software I use has it I'm stuck with that dog**** and I've learned to reluctantly co-exist with it... at least until the millenials can figure out TidePods aren't for eating and not every application needs to look like it belongs on a tablet, phone or watch.

Look, menu driven software is INFINTELY more efficient for a great many application types (CAD/CAM/CAE being primarily in that category because there is so much input involved). IDGAF what studies, mouse tracking, blah, blah, blah, say. Constantly going back and forth between keyboard and mouse... It just jacks with work flow and it's just a hot mess. [/rant]

The above rant ABSOLUTELY WAS NOT directed at CNC Software or ANY of it's developers. It's aimed at the scum in Redmond that came up with this crap so the UI folks can unrustle their Jimmies now.

:coffee:

Maybe I should try Decaf......... nahhhhhhhhhhhhh. I might lose my edge. :rofl:

:coffee:

  • Like 1
  • Haha 3
Link to comment
Share on other sites
7 hours ago, medaq said:

This ^.   I have what I feel is a pretty solid system and this is infinitely frustrating.

Yes, I asked the same question.

Is it possible some of the latency is related to the actual message being displayed to the user?

Link to comment
Share on other sites
19 hours ago, Greg_J said:

What cures the tessellation message and wait time?

I believe this is due to the graphics work required to shade the solid models

Try leaving solid models unshaded when you don't actually need them (that's ALT S in pre ribbons days)

  • Like 1
Link to comment
Share on other sites

An issue I have that I have been unable to solve

I have had 32g of ram in this machine since I bought it.

RAM usage typically got up to 20g when doing heavy work.

I've added 64g for a total of 96G and now RAM usage is never more than 12g

I've tried just running the new 32G  sticks without the old 16g stick and nothing changed

I tried the suggestions of the a couple of internet tech support videos and ended up bricking the

machine,

Long term that was a good thing as it took a fresh Windows install to get running again, but the RAM

issue has not changed.

I believe this is a Windows issue, not Mastercam as I see similar results in Vericut.

I've been meaning to load some very large SolidWorks assemblies to see how much ram they eat, but have not had time 

 

 

Link to comment
Share on other sites
17 minutes ago, gcode said:

An issue I have that I have been unable to solve

I have had 32g of ram in this machine since I bought it.

RAM usage typically got up to 20g when doing heavy work.

I've added 64g for a total of 96G and now RAM usage is never more than 12g

I've tried just running the new 32G  sticks without the old 16g stick and nothing changed

I tried the suggestions of the a couple of internet tech support videos and ended up bricking the

machine,

Long term that was a good thing as it took a fresh Windows install to get running again, but the RAM

issue has not changed.

I believe this is a Windows issue, not Mastercam as I see similar results in Vericut.

I've been meaning to load some very large SolidWorks assemblies to see how much ram they eat, but have not had time 

 

 

This sounds suspiciously like an issue in your BIOS settings.

Have you tried updating to the latest BIOS available for your particular build?

https://appuals.com/windows-10-wont-use-full-ram/

Is Memory Map enabled in your BIOS Settings?

Is "Maximum Memory" checkbox enabled in (Windows Key + "R") > msconfig > System Configuration Dialog > Boot Tab > Advanced Options? (Turn this off, or set to a higher value for testing.)

Link to comment
Share on other sites
2 hours ago, Colin Gilchrist said:

Is "Maximum Memory" checkbox enabled in (Windows Key + "R") > msconfig > System Configuration Dialog > Boot Tab > Advanced Options? (Turn this off, or set to a higher value for testing.)

yes BIOS is the latest release

I don't see anything in my BIOS that looks like "Memory Map"

The BIOS does recongise the installed memory and the memory is all the same brand and on the manufacturers list of tested memeory

an msconfig edit is what bricked my machine .. I spent a whole day trying boot/restore/reinstall WIndows.. everything failed

I went to bed Friday night thinking it had all failed as a Windows reinstall had been hung for 6 hours. I shut it off and went to bed.

In the morning I powered by up and was greeted with the normal setup routines for a fresh WIndows install

 

 

Link to comment
Share on other sites
On 3/8/2021 at 11:51 AM, Greg_J said:

What cures the tessellation message and wait time?

On 3/9/2021 at 5:15 AM, Thee Byte™ said:

Yes, I asked the same question.

Is it possible some of the latency is related to the actual message being displayed to the user?

Sorry, I missed this the first time through.


There's a couple of places that can happen, are you talking about on File Open, or when you hit the green checkbox to start calculating a toolpath?  If it's when you're calculating a toolpath, it's probably because (depending on the toolpath) we may have to make many models to properly compute the toolpath. 

To walk through a hypothetical example,  You've chosen X number of faces as your drive geometry and you're using the whole model as a collision.   Because you chose individual faces as drive, we have to extract them from the solid (create surface > from solid) and then offset them based on the tool profile to ensure we have a safe and proper geometry to calculate the toolpath on, and all that has to happen well below your max tolerance that you told us the toolpath should be to ensure we don't violate that.   Then, the collision model gets checked to see whether it's the same as the cut model, and if not, we have to do the same process to prepare that for calculation.  In the case of multiaxis toolpaths like morph, we look and see if you have a different tolerance for your collision model(s), and if you do we have to calculate them separately.  And of course, they all have to be "puffed" up by the stock to leave against them. 

All of that has to happen before we can hand the models off to he toolpath engine to be calculated & collision detected against, and each individual model has to be processed single-threaded.

 

tl;dr:  send in an example of you specific part and some steps to reproduce where you're seeing the delay, and it'll get through to the right channels to see if we've missed optimization that would help you.  Those are the sorts of things we improve year-over-year, which is why the benchmark file times fall on the same hardware with a new release.

Link to comment
Share on other sites
5 minutes ago, Aaron Eberhard - CNC Software said:

tl;dr:  send in an example of you specific part and some steps to reproduce where you're seeing the delay, and it'll get through to the right channels to see if we've missed optimization that would help you.  Those are the sorts of things we improve year-over-year, which is why the benchmark file times fall on the same hardware with a new release.

It's a chook that calls silhouette boundary with the slice settings, everything is blazing fast except the slice part, and it slows down each slice.

It's rather complicated, I brought it up to the sdk Team, haven't had any progress there yet.

I was gonna try converting the Solid to a Mesh and see if that is supported, I know in 2022 it is, but I don't know if the chook stuff does the same thing as the functions in "vanilla" Mastercam do..

Link to comment
Share on other sites
56 minutes ago, Thee Byte™ said:

It's a chook that calls silhouette boundary with the slice settings, everything is blazing fast except the slice part, and it slows down each slice.

It's rather complicated, I brought it up to the sdk Team, haven't had any progress there yet.

I was gonna try converting the Solid to a Mesh and see if that is supported, I know in 2022 it is, but I don't know if the chook stuff does the same thing as the functions in "vanilla" Mastercam do.. 

Ah, if you're already in talks with the SDK team, you're in good hands, I probably can't help you much :)
 

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

 Share

  • Recently Browsing   0 members

    • No registered users viewing this page.

Join us!

eMastercam - your online source for all things Mastercam.

Together, we are the strongest Mastercam community on the web with over 56,000 members, and our online store offers a wide selection of training materials for all applications and skill levels.

Follow us

×
×
  • Create New...