Jump to content
JuanP

2021 - Teaser Tuesday: Performance Enhancements

Recommended Posts

@JuanP do you happen to know why these teaser posts don't show up in my unread forum posts? Are they originally hidden posts that you reveal? I'm not sure that many forum users are seeing them.......

 

Kevin

  • Like 1

Share this post


Link to post

@Kevin McAllister You're correct they are originally hidden post, I will check the forum settings to see if there is anything else I can do. If not, I will start posting the same day. 

 

Thanks for reporting the issue.

  • Like 3

Share this post


Link to post

@JuanP What is the total number of cores VW2021 will be able to utilise? Am I right in thinking up until now VW's limit was 4 cores?

Share this post


Link to post

@Josh Loy Cool! So theoretically if you had a fancy AMD processor with 64 cores and you had a theoretical process that could be parallelised across these cores (not including renderworks as I think that could already use any number of cores) there is nothing in VW stopping this from happening? It sounds great and hopefully results in a big improvement.

  • Like 1

Share this post


Link to post
Posted (edited)

This feels too good to be true... what’s the catch?

 

Edited by Mark Aceto
  • Laugh 2

Share this post


Link to post
Posted (edited)

Dug up this gem of a Teaser Q&A from 2 years ago which does an excellent job of explaining how multiple CPU cores "shovel" graphics to the GPU, so this really does sound like evolution of that development:

 

 

Edited by Mark Aceto

Share this post


Link to post
Posted (edited)

@Josh Loy thanks for clarifying. Improvements to Hidden Line are always welcome! Does this caching increase file size?
 

Edited by Mark Aceto

Share this post


Link to post

Looking forward to seeing the recommended hardware specifications for VW2021, along with an updated explanation for what component features will take advantage of this release.

  • Like 2

Share this post


Link to post
Posted (edited)
10 minutes ago, Josh Loy said:

@Samuel Derenboim improvements with hidden line this release but I wouldn't be surprised if it was next on the list.


@Josh Loy this is worded weird—are there improvements to Hidden Line or not?

 

Edited by Mark Aceto

Share this post


Link to post
1 hour ago, Mark Aceto said:


@Josh Loy this is worded weird—are there improvements to Hidden Line or not?

 

@Mark Aceto After I posted I realized I forgot the "no" and added it, sorry for the confusion.

  • Like 1

Share this post


Link to post
Posted (edited)
2 hours ago, Josh Loy said:

@Mark Aceto After I posted I realized I forgot the "no" and added it, sorry for the confusion.


Got it. And you already answered my question about caching adding file size.

 

Edited by Mark Aceto

Share this post


Link to post
Posted (edited)

Would love to see a followup video comparing performance side-by-side as it was presented 2 years ago:

 

 

Edited by Mark Aceto

Share this post


Link to post
Posted (edited)
3 hours ago, Clint Alderman said:

Looking forward to seeing the recommended hardware specifications for VW2021

 

Considering that adding cores lowers clock speed and increases thermals, I would still go for the fastest CPU base clock speed. For example, the Ryzen 3960 looks like the best performance (and value) in the table below.

 

@Josh Loy what are you thoughts? Also, how will the GPU/VGM factor into this?

 

phpThumb_generated_thumbnail.thumb.jpeg.bcdc14b367d9502de54fa70e725633f0.jpeg

 

Edited by Mark Aceto

Share this post


Link to post
2 hours ago, Mark Aceto said:

So, without being able to test drive this yet, this really doesn’t sound like true multicore performance in the sense that we’ve all been hoping for (as it was announced 2 years ago in a Teaser Tuesday). This sounds more like a workaround to speed up viewing vs drafting/modeling/rendering (at the expense of file size which could be an issue when project sharing).

Hey @Mark Aceto thanks for the feedback, I not exactly sure where you're going with this but any speedup for drafting/modeling/rendering seems like a win to me.  This system is independent of Project Sharing and will have no impact on its performance.

 

1 hour ago, Mark Aceto said:

This announcement reminds of last year when instead of moving more processes from the CPU to the GPU (VGM) as was the trend from 2015-2018, the workaround was to draw less detail in the background. Reading between the lines, it sounds like GPU and multicore optimization have plateaued, and we should manage our expectations accordingly.

I wouldn't say anything has plateaued or that there are any lines to read between.  We support a vast set of hardware and must provide support for a broad range of workflows while always providing the best experience.  The more tasks we can multi-thread and multi-task the better.

 

1 hour ago, Mark Aceto said:

I spend 80% of my time drafting and modeling (not just opening files and viewing the model). I spend 20% of my time plating (wireframe, hidden line, OpenGL). And I'm on a Mac, so I need the fastest single core base clock speed available.

Awesome and I'm sure most users fall into this same category, the performance gains will absolutely help your large model workflows while working in wireframe and OpenGL.  I would also agree and recommend higher clock speeds over core counts as long as you have at least 6 cores; I'm also on the Mac.

 

1 hour ago, Mark Aceto said:

Sounds like VW2021 will make use of more storage and faster broadband (to save and share larger file sizes). That said, I automatically uncheck all caching options to minimize file size, so I'll be passing on this feature.

It will use more storage 10 to 30% that I've seen up to this point, and if that's an issue for your broadband it can be easily disable just like "Viewport cache."  I wouldn't make any choices or pass judgement until you try it and are happily surprised. 🙂

 

1 hour ago, Mark Aceto said:

Considering that adding cores lowers clock speed, and potentially increases thermals (if VW is actually using more than 1-3 cores), I would still go for the fastest CPU base clock speed.

Some of this logic isn't quite correct.  As clock speed increases the power demand isn't linear, meaning the same 'n' amount of work can be done on 4 cores at lower clock speeds in the same amount of time and using less power than a high clocked processor.  Base clock speeds are pretty useless as the system will ramp up and down the clock speed depending on demand.  You'd probably be better served looking at the single and multi-core turbo speeds since they're generally the limiting factor and usually not far apart.  If you want to get really into it TDP will generally dictate the performance you'll see since cores can run at different speeds and in the end heat limits it all.  Feel free to direct message me if you want to talk about this more or have any additional questions. 

  • Like 1

Share this post


Link to post
Posted (edited)
46 minutes ago, Josh Loy said:

Hey @Mark Aceto thanks for the feedback, I not exactly sure where you're going with this but any speedup for drafting/modeling/rendering seems like a win to me.  This system is independent of Project Sharing and will have no impact on its performance.

 

I wouldn't say anything has plateaued or that there are any lines to read between.  We support a vast set of hardware and must provide support for a broad range of workflows while always providing the best experience.  The more tasks we can multi-thread and multi-task the better.

 

Awesome and I'm sure most users fall into this same category, the performance gains will absolutely help your large model workflows while working in wireframe and OpenGL.  I would also agree and recommend higher clock speeds over core counts as long as you have at least 6 cores; I'm also on the Mac.

 

It will use more storage 10 to 30% that I've seen up to this point, and if that's an issue for your broadband it can be easily disable just like "Viewport cache."  I wouldn't make any choices or pass judgement until you try it and are happily surprised. 🙂

 

As for you processor choices I wouldn't quite agree with your

Some of this logic isn't quite correct.  As clock speed increases the power demand isn't linear, meaning the same 'n' amount of work can be done on 4 cores at lower clock speeds in the same amount of time and using less power than a high clocked processor.  Base clock speeds are pretty useless as the system will ramp up and down the clock speed depending on demand.  You'd probably be better served looking at the single and multi-core turbo speeds since they're generally the limiting factor and usually not far apart.  If you want to get really into it TDP will generally dictate the performance you'll see since cores can run at different speeds and in the end heat limits it all.  Feel free to direct message me if you want to talk about this more or have any additional questions. 

 

Thanks, Josh. After rewatching the videos from 2  years ago, I deleted/edited a lot of my initial responses but thanks for taking the time to reply.

 

As for clock speed, the Turbo Boost speeds have proven to be little more than marketing BS for us Mac users, so the real world turbo usually splits the difference between the base clock and the turbo boost (until it gets too hot). The other point of reference is when my Mac maxes out with the message "Vectorworks 20XX (Not Responding)" and I look at Activity Monitor, and see 1 thread slammed against the ceiling, and all the other threads not even being used. That's why the CPU-constraint has always been the speed of a single core/thread. I don't care about the other 19 threads that aren't even being used if they're dragging down the performance of the one that is being used. The 2nd screenshot shows the "4.5GHz" turbo boost maxing out at 4.26 which is the fastest I ever saw it spike on that 2017 iMac Pro.

 

2109274385_ScreenShot2018-09-24at6_32.10PM(2).thumb.png.4fcb4e41d7fee7a06cd1b7c758e17c5b.png1065715544_ScreenShot2018-11-18at8_44.05PM(2).thumb.png.12da9e8618abf0d8357957206962d1d1.png

Edited by Mark Aceto
  • Like 1

Share this post


Link to post
Posted (edited)

And then there's Cinema/RW efficiently using all 20 threads @ 3.68GHz...

 

This is the dream (even just 4 cores/8 threads instead of 1 thread would be an 800% improvement):

 

1310638375_ScreenShot2019-01-23at2_18.05PM(2).thumb.png.35322b826077086be7e05fe21a9ea967.png

Edited by Mark Aceto
  • Like 1

Share this post


Link to post
14 hours ago, Mark Aceto said:

Dug up this gem of a Teaser Q&A from 2 years ago which does an excellent job of explaining how multiple CPU cores "shovel" graphics to the GPU, so this really does sound like evolution of that development:

 

 

 

It would be cool to bring back these Q&A sessions, any plans to @JuanP?

  • Like 1

Share this post


Link to post

It's a little disappointing to hear that there will be no improvements to hidden line rendering.

 

I find 3d performance not bad as it is. I can spin around my 3d models pretty smoothly most of the time. Most of my time spent waiting for VW to draw something, is spent waiting for it to draw hidden line viewports.

  • Like 3

Share this post


Link to post
16 hours ago, Mark Aceto said:

Dug up this gem of a Teaser Q&A from 2 years ago which does an excellent job of explaining how multiple CPU cores "shovel" graphics to the GPU, so this really does sound like evolution of that development:

 

 

I also thought it was cool in this video to see what looks to be the same example file being used to show speed improvements for opening large files (around the 6:50 mark) when demonstrating 2018 as they are now demonstrating 2021:

 

VW2018: 70 seconds

VW2019: 21.8 seconds

VW2020: not clear from the video, fades out after 10 seconds

VW2021: 3.14 seconds

 

On paper it looks very good, especially such a dramatic decrease in 3 versions, let's hope it translates well across other real-world examples.

  • Like 2

Share this post


Link to post

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  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.


 

7150 Riverwood Drive, Columbia, Maryland 21046, USA   |   Contact Us:   410-290-5114

 

© 2018 Vectorworks, Inc. All Rights Reserved. Vectorworks, Inc. is part of the Nemetschek Group.

×
×
  • Create New...