Jump to content
iaincognito

Workstation Build Feedback

Recommended Posts

Thanks in advance for the help. Below are the specs for a planned workstation. Would love any feedback in relation to missing components or wasted resources/cash. Primary use for this workstation is VW Spotlight. 50% Top/plan, 25% lighting/screen visualization, 25% renderworks of said visuals.

 

PCPartPicker part list: https://pcpartpicker.com/list/6FLgFt
Price breakdown by merchant: https://pcpartpicker.com/list/6FLgFt/by_merchant/

CPU: Intel - Core i7-8700K 3.7GHz 6-Core Processor  ($355.89 @ OutletPC) 
CPU Cooler: Reeven - Brontes RC-1001b 30.4 CFM Sleeve Bearing CPU Cooler  ($34.99 @ Newegg Marketplace) 
Motherboard: MSI - Z370-A PRO ATX LGA1151 Motherboard  ($99.89 @ OutletPC) 
Memory: G.Skill - Aegis 64GB (4 x 16GB) DDR4-2400 Memory  ($540.98 @ Newegg) 
Storage: Samsung - 970 Pro 512GB M.2-2280 Solid State Drive  ($229.99 @ Amazon) 
Storage: Samsung - 860 Pro 1TB 2.5" Solid State Drive  ($379.00 @ B&H) 
Video Card: Asus - GeForce GTX 1080 Ti 11GB STRIX GAMING OC Video Card  ($849.99 @ B&H) 
Case: Phanteks - Enthoo Evolv ATX Glass ATX Mid Tower Case  ($169.99 @ Amazon) 
Power Supply: SeaSonic - FOCUS Plus Platinum 750W 80+ Platinum Certified Fully-Modular ATX Power Supply  ($104.40 @ B&H) 
Operating System: Microsoft - Windows 10 Pro OEM 64-bit  ($126.88 @ OutletPC) 
Monitor: LG - 27UD58-B 27.0" 3840x2160 60Hz Monitor  ($296.93 @ Newegg) 
Total: $3188.93
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2018-08-17 12:48 EDT-0400

 

 

Thanks again!

Share this post


Link to post

The newly announced Xeon processors are based on the Coffee Lake processors. The 6-core 3.7GHz clock is basically an i7-8700K but supports ECC Ram.
When rendering, most RAM errors are negligible and result in a single misrendered pixel or the like. Only the very few really cause the rendering process to crash. With 64GB RAM an error happens 8-20 times per month according to Google study in 24h operation. If you use your PC only during working hours, you will have 2-5 RAM errors per month, most of which will have no consequences. For my purposes, Vectorworks runs absolutely stable without ECC RAM, but some users who render overnight and therefore depend on stable continuous operation rely on ECC. Of course, it's much worse for them when the work of an entire night is gone the next day. I understand that they use ECC. For normal rendering tasks, however, it is only a nice to have on which you will never really be dependent.
So I wouldn't wait until the new Xeon is on the market, but buy the i7-8700K you planned.

Edited by herbieherb

Share this post


Link to post

I used intel cpus for ages and tried the threadripper - not looking back! I would go threadripper with a Corsair AIO cooler - I have the 12 core OC’d to 3.9 ghz and it absolutely flies. I’ve read that you should be able to OC the 16 core cpu up to the same. They are dead easy to OC as well.

Share this post


Link to post

Yes, I would definetely also recommend the threadripper 2950x. But it's gonna cost you much more. You would need water cooling, the mainboard would be more expensive and to tickle the last one out of the processor you would need faster RAM.

 

That means to you:


Mainboard     +200$
Watercooling +100$
Processor     +500$
RAM              +500$
Total           +1,300$

 

So if you have a fixed budget, you'll blow it up. But if you figure out how much less time you have to wait in front of the PC until the rendering is finished, the investment is worthwhile anyway. When you say you're rendering 20% of your time. Then maybe you would be half busy with settings, the other half would be pure render time. So about 10%. That would be about 50 minutes at a normal working day. With the Threadripper 2950x you would be about 3x faster. You woul'd only need about 16 minutes, then. So you save 30 minutes per day, but you have additional costs of $1300. Depending on your hourly rate, you would have saved the additional expenses after less than a month.

 

Here a detailed discussion about the threadripper:

 

Share this post


Link to post

I'll second all that with regards to time spent drinking coffee looking at the teapot vs. hourly rate down the toilet. I don't do a lot of rendering - most of my work is top plan and drafting in viewport over model sections, but the update times for elevations, 3d's and model changes in edit section view re-renders were grating with an i7 4.2ghz quad core - its probably the biggest single issue I have with VW; coming from Revit there was no need to update anything and is the primary reason I can't move a couple of colleagues over the light. 

The reduced time in those process alone makes me more productive, which makes me feel better overall.

Plus given the VGM updates coming, more cores are worth considering. 

 

Share this post


Link to post

 

@Aspect_Design @herbieherb @P Retondo 

 

Thank you all so much for the valued feedback. I'll consider the input and post changes before we move to purchasing. To sum up and ask a follow up question, it seems to me the biggest (2) variances in both experience and opinion is

  1. ECC vs Non-Ecc
  2. Ryzen MC renderworks boost vs Intel's higher SC bench speeds

I remain convinced based on research and other's experience that ECC has it's place inside of a VW workstation, however isn't a must without marathon (>1-2 hrs) rendering sessions. My main concern will be if the MC performance boost from Ryzen is worth a lower SC speed when working with the model and not rendering. This assumes that we're comparing bench performance and not real world VW experience. It may be that the 2950x can keep up inside of the program just as well AND provide high MC performance. I'm asking if anyone has noticed a demonstrable difference between the two when working in the model and not rendering.

 

Thanks again to everyone for the input. 

Share this post


Link to post

1. ECC vs non-ECC


When you assemble a threadripper, you prefer higher RAM clock rates. The threadrippers like fast RAM with low latencies. Take the fastest RAM you can afford. Don't believe this video when you want to use ur RAM in a threadripper built: https://www.youtube.com/watch?v=D_Yt4vSZKKVk He didn't test the threadripper and he didn't test the OpenGL speed, the interaction between processor and graphics card. Here you are 15% faster than between 2133 and 3200 MHz. Between 3200MHz and 3600MHz there is another 15% speed boost. ECC doesn't give you such fast RAM. So you would have to give up up to 30% performance. Maybe if you want to use ECC RAM one of the coming out fast clocked Xeon processors would be better. But you'll pay a lot more for that system. As I said before: The probability that you will ever crash while rendering due to a memory error is in most cases negligible. ECC could only be worthwhile if your computer is to render reliably overnight on a regular basis without being monitored.

 

 

2nd Ryzen MC renderworks boost vs Intel's higher SC bench speeds


That would have to mean more expensive but better render computer vs low-cost general-purpose machine. In everyday drawing with Vectorworks, the lower clock rate of the threadripper 1950 is hardly noticeable. The 2950 is even faster. The difference is really small and in the future more and more processes will be multicore capable. So it's more a matter of budget than the choice between single and multicore performance.

Edited by herbieherb
  • Like 1

Share this post


Link to post

The assertions in this thread on the benefits of ECC are correct, but Vectorworks does not take advantage of ECC memory. If you have other applications that do, then that's a good choice, but don't opt for it just on our account.

  • Like 1

Share this post


Link to post

Jim, I don't understand your comment.  My understanding is that ECC detects and corrects errors in memory transfer between RAM and storage, which would be at the operating system level, not at the application level.  Is there something I am not aware of?

Share this post


Link to post

Here's my updated list for those who are interested

 

PCPartPicker part list: https://pcpartpicker.com/list/mtystg
Price breakdown by merchant: https://pcpartpicker.com/list/mtystg/by_merchant/

 

CPU: AMD - Threadripper 1950X 3.4GHz 16-Core Processor  ($748.37 @ OutletPC) will most likely switch this to 2950x if I get pricing approval


CPU Cooler: Noctua - NH-U14S TR4-SP3 140.2 CFM CPU Cooler  ($79.90 @ Amazon) will upgrade to liquid cooling CPU and GPU for OC


Motherboard: ASRock - X399 Taichi ATX TR4 Motherboard  ($283.98 @ Newegg) 


Memory: G.Skill - Trident Z RGB 32GB (4 x 8GB) DDR4-3200 Memory  ($515.98 @ Newegg) 


Storage: Samsung - 960 PRO 512GB M.2-2280 Solid State Drive  ($234.99 @ Newegg Marketplace) 
Storage: Samsung - 860 Pro 1TB 2.5" Solid State Drive  ($379.99 @ Amazon) considering changing this to a spinning drive or hybrid


Video Card:  EVGA - GeForce GTX 1080 Ti 11GB FTW3 GAMING iCX Video Card  ($709.99 @ B&H) depending on budget and timing, will switch this to 2080 ti


Case: Phanteks - Enthoo Evolv ATX Glass ATX Mid Tower Case  ($169.99 @ Amazon) 
Power Supply: SeaSonic - FOCUS Plus Platinum 750W 80+ Platinum Certified Fully-Modular ATX Power Supply  ($114.73 @ Newegg) 


Operating System: Microsoft - Windows 10 Pro OEM 64-bit  ($126.88 @ OutletPC) 
Monitor: LG - 27UD58-B 27.0" 3840x2160 60Hz Monitor  ($296.93 @ Newegg) 


Total: $3801.73

Share this post


Link to post

Yes, ECC corrects errors (flipped bits ?)

 

I don't mind if a rendered pixel has a slightly wrong color,

but I hate when my over night renderings stop with an error message

after a few frames when I have deadlines.

Which I had in the past with one interims non ECC machine and

3DS max. Maybe Modo, C4D VRAY or RW Renderings will never crash

without ECC RAM anymore - but I don't know because I always use

ECC since then again.

 

It is more expensive than standard RAM, often gets hotter (registered ECC ?)

and is often not available at latest frequencies and speeds.

But I saw speed comparisons between different rated RAM modules and

decided that the amount is not a real bottleneck for me.

https://techbuyersguru.com/does-ram-speed-matter-ddr3-1600-vs-1866-2133-and-2400-games?page=0

 

 

Share this post


Link to post

Please read the posts above again @zoomer. Your video is not relevant for a threadripper built for vectorworks:

22 hours ago, herbieherb said:

He didn't test the threadripper and he didn't test the OpenGL speed, the interaction between processor and graphics card. Here you are 15% faster than between 2133 and 3200 MHz. Between 3200MHz and 3600MHz there is another 15% speed boost. ECC doesn't give you such fast RAM. So you would have to give up up to 30% performance.

Also it seems that vectorworks doesn't use ecc at all:

18 hours ago, Jim Wilson said:

The assertions in this thread on the benefits of ECC are correct, but Vectorworks does not take advantage of ECC memory. If you have other applications that do, then that's a good choice, but don't opt for it just on our account. 

If you really want to use ECC RAM. Get one of the new Xeon not a Threadripper. But you'll get less power for the money.

 

@iaincognito I woul'd definitely wait for the 2950x. Don't try the Noctua cooler. I tried it myself and the threadripper got really hot. Get a water cooler. Make sure that your case is capable to fit the water cooler. If your water cooler has three fans your case must have room for three coolers next to each other. Get a case with good case cooling (same amount of fans as the water cooler). So if you get the water cooler with three fans your case has to have at least 6 positions for the case-fans, three of them next to each other.

Make sure you'll get the G.Skill Trident Z with the indication RGB F4-3200C14Q-32GTZRX. These are optimised for the threadripper. Also consider buying even 3600Mhz RAM as another user did. His cinebench openGL test was 15% faster then mine with almost the same system. But this did only affect OpenGL.

For the monitor i prefer more than one but smaller size and only 2k not 4k. With three 2k screens your gpu has to handle less pixels than with one 4k but you'll have way bigger screen size. Also the vectorworks pallets can thus be better distributed. But I guess that's a matter of taste.

On 8/8/2018 at 5:51 PM, herbieherb said:

Here is my battlestation:

battlestation.thumb.jpg.c9bf9eadea63937a3eb32844cf031b94.jpg

 

The screen layout:

1523025580_battlestationscreenlayout.thumb.jpg.ce4f2130fae94e4acc05d1d950510e4c.jpg

I recently added the third screen for the 2018s new multi view feature and it has really proved itself. I also sometimes use the left screen for other software like excel tables, pdfs and stuff.

Share this post


Link to post

I may look at the wrong Videos but I always see only 5-8% gains between

lowest and highest rated RAM. Something I would not notice.

 

 

Share this post


Link to post

Again: They testet the SC speed not OpenGL. You're right it doesn't matter much for renderworks but for OpenGL the right RAM configuration makes a huge difference up to 30% between a 2133MHz (most ECC-Modules don't go much faster than cheap normal RAM) and a 3600MHz (very High-End but non ECC RAM) configuration. I tested it myself with different RAM modules. So if you want to exhaust a high-end gpu, you have to use very fast ram with the treadripper. If you want to buy the cheaper ram you coul'd then also buy the cheaper gpu since it would'nt matter.

Example: I got a machine in the office with an i7-6700K (4cores at 4GHz) RAM at 2133 MHz and a Geforce GTX 1070. It has an OpenGL score of 140 frames.

The threadripper built i introduced even though he had a GTX 1080Ti

with 2x2133MHz RAM only had a score of 100

with 4x2133MHz RAM it had 105

with 4x3200MHz 116

and another user at this board had with 4x3600MHz 134 frames

 

So between the 4x2133 MHz and the 4x3600MHz there is about 30% speed difference in the OpenGL Mode (in SC, MC Tests the difference is only about the 5% as you mentioned). So when you render, you wont notice a difference. When you work in OpenGL mode with huge 3D Models, when you use multiview or use high screen resolutions (several 2k screens or a 4k screen) this definitely makes a huge difference.

Edited by herbieherb

Share this post


Link to post

OK, got.

I see some 15 % in some games but not much difference in CPU rendering

which doesn't seem to use RAM that extensive.

But you are talking about CPU Bottlenecks when filling GPUs.

 

But as far as I know for OpenGL GPU load in Pro 3D DCC software,

there is also less data pumping by keeping the whole scene in VRAM,

opposed to Games which constantly eject and reload parts of the scenes.

(Or why OpenGL in 3D DCC is still a good idea and no bottleneck)

Share this post


Link to post

When you work on big 3d models, you really notice every 10% increase in performance in OpenGL mode. The model loads faster, you can navigate more fluently and changes in class visibility etc. are faster. In Vectorworks its a good idea to also max the OpenGL score.

It doesn't seem to be only a bottleneck when filling the gpu, because the navigation in an already loaded model is also smoother on the other system with the higher OpenGL score.

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...