How much power to supercharge rhino

Rhino Windows Hardware
21

cfee. I feel like that lone wanderer, whose been traveling his entire life. who climbed up the huge steps up to some monastary. and you were there waiting for me. and i walked towards you as meditated on your pedestal. and you bestowed upon me the wisdom of your years.

I humbly thank you. I have no words for now. I must take your advice and go where it takes me. I need time to process all these things. Learn about them, I am sure I will be back with questions. Once I figure this out, I promise I will post my solution, my computer set up, so everyone and anyone else in my position can handle.

bows graciously

thank you, sir. very much. indeed. I was searching for this for a very long time. :grin:

22

Hey pardner -Happy to help. Sometimes the answer gets lost amidst the guesses. There IS an approach, however, to reaching any desired goal. I hope my reply to your question helps.
Knowing the issues and probable best answers without telling you exactly what “I’d do” can lead you to making your own best guesses. Doing so in an informed manner can significantly shorten your path to success.
Good luck, and we really are interested in how it turns out for you !
CFeeSr. Mechanical/Structural2D/3D CAD Designer, Modeler and AnimatorHouston, Texas.

23

Hi Mun144
Check out Holomark if you haven’t done so yet.
http://discourse.mcneel.com/t/holomark-2-released/8040/631
If you study the charts then you will see that the big differences can be seen with pure mesh setups, in the display mode where all fancy stuff is turned off (I named it RenderSpeed to indicate that it is a speedy render mode ;)). That’s the only area where you can supercharge Rhino as it is the only area where Rhino it self isn’t a bottle neck. So if you work in nurbs and not only meshes then you will waste a lot of money buying top of the line stuff.

CPU: i5 is as fast as i7 which is as fast as Xeon:
Xeon’s are not faster than i7. In fact they are slower, as they are clocked lower to prevent overheating as they are designed to be as reliable as absolutely possible. Ordinary modelling in Rhino does not benefit from multiple cores. (of course it benefits from four threads as background tasks in windows won’t affect modelling, but all cpu’s has that these days) If you render you will benefit from lots of cores, unless you use a cuda based render engine, then it doesn’t matter any more.

GPU:
Top notch Quadro’s are faster than Geforce when it comes to AA and massive meshes.
For ordinary modelling it doesn’t matter that much since ordinary modelling will use display modes that show curves (wireframe, shaded mode etc) and curves are calculated by the CPU as Rhino doesn’t use the GPU to accelerate that. (that of course is a simplified statement, but not too far from the truth)

SLI has no speedgain in Rhino, at least it has not in the past, so check that out thoroughly before you buy two monster cards!

HDD:
Running ramdisks won’t give you any speedgain what so ever. Unless you need to save everything veery often, and even then I doubt a ramdisk will be faster than an SSD as the cpu has to process what’s being saved. But an SSD for storage will faster than an ordinary hdd.

The best you can do is store your monsterproject on a disk, on the web, or what ever and then ask different people with different systems to test it for you. Store differend display modes and run TestMaxSpeed in different display modes to see where the systems start leveling out.

Edit:
PS! I’d LOVE nothing more than to be wrong, so if you choose to buy a top of the line machine and get a supercharged experience then let us know.

Personally I would buy a fast i7, a geforce 1080 or a k5200 depending on need (the 1080 as a few more cuda cores, both has 8 gb vram, the 1080 is better for VR, the quadro might perform better in Rhino, but I do not know as I have not tested them), 32 GB or 64 GB Ram (again depending on need, fill all memory slots, that’s important for the CPU to run as fast as possible when handling much data) two SSD’s, one for OS and one for storage. Get good cooling systems that are quiet, a noisy machine will likely irritate you or your colleges.

You can find tests on K5200 vs K6000 online, here is one: http://www.xbitlabs.com/articles/graphics/display/nvidia-quadro-k5200-k4200-k2200_7.html , and see that in some tests the K6000 is a bit faster, on other the K2200 is just as fast as the K6000. This indicates that software is the bottle neck and not the hardware.

3 Likes
24

I’ve been watching this thread with interest and want to add that some models that I have gotten from clients with lots of architectural components and especially furnishings from 3d Warehouse or the SketchUp resources can really bog down a Rhino file. I work with very large MegaYacht files and as long as I don’t load it up with mesh downloads for the outfitting still get good response. Be careful with imported ‘goodies’ especially plants!

25

Hi Jorgen, can you elaborate on that or link to some more info on this?

Thanks
-Willem

26

Willem, I read up on this regarding my dual xeon system an there it was correct to fill all, but I see that this isn’t nessesarily correct regarding an i5. It has to do with the memory bandwidth:

“Do I have to populate all CPU memory channels? What if I only put one or two DIMMs per processor?
Answer:
It’s ok to run with as little as one RDIMM/LRDIMM per system or per CPU (unbalanced configurations) but it will impact CPU performance. Socket R3 CPU’s have four memory channels and for best performance it’s recommended to have a minimum of one DIMM in per channel. A channel left unpopulated will reduce the memory bandwidth by 25%, so with only one RDIMM per CPU memory bandwidth performance is reduced by 75%.”

From:
http://www.supermicro.com/support/resources/memory/X10_memory_config_guide.pdf

On kingstons site it elaborates more and says to use one or two dimms pr channel, adding the third will lower speed: (unless I missunderstood something) http://www.kingston.com/us/business/server_solutions/best_practices/maximizing_performance

2 Likes
27

I use hp Workstations and this one has 2 x 16 G DIMM and the hp Performance Adviser gives the warning, confirming Jorgens information…

hp440_Memory.JPG759×825 90.7 KB
y

Barry.

1 Like
28

I’d still hold with getting a NVidia GTX 1080 for a driver release or two, since it is a very new card, and there are bound to be driver issues. From discussion with other Blender devs for Cycles (upcoming Raytraced viewport mode in Rhino WIP. I’m working on) the performance of that card isn’t as good yet as the hardware specs promise - but with time I’m sure the GTX 1080 will be a good card to have.

For interactive rendering personally I’d go for lots of high-speed RAM and VRAM (RAM on the GPU), and lots of CUDA cores. For good interactive raytraced view I prefer to have my GTX 760 not be the display card, for that I have currently my AMD R9 280x, which leaves all CUDA power for the Raytraced viewport and Cycles renderer). I suppose my dream setup would be a very good display card, and then a set of two or more CUDA cards that could be used in multi-CUDA setup. That’d probably be awesomely fast for interactive rendering in the viewport (or modal rendering for that matter).

That said, interactive rendering on GPU will be hard if you’re going to have scenes that don’t even fit in the GPU memory. OTOH, you probably don’t want to be interactive rendering in most cases, but who knows :slight_smile:

/Nathan

2 Likes
29

Hey Holo;

I haven’t truly researched this issue, but it appears that both Nvidia and AMD are offering rendering software soon. I believe that AMD’s Fire Render for Rhino V6 is going to be free. Those render engines are obviously unlike Flamingo that is CPU base; they will be GPU based and in the case of Fire Render it apparently will combine CPU and GPU? And just when I was getting truly handy with Flamingo…cheers, Rob

30

everyone, everyone, everyone who participated on this thread, i just wanted to say thank you, tremendously, from john to cfee to all you guys, i’m learning alot, i won’t forget you guys when i find out a solution, best of luck to everyone working, well wishes and good things for you all, adios!

31

Hi Robb, if Flamingo fits the bill then keep on using it. CPU based render engines will still be in the game. Their benefit is that they are not so hardware dependent (all they need is some kind of fast CPU) And the better you become in one engine the better you will be in other engines once you transfer.

32

I think the user community needs for the developers to take a more proactive role in answering these types of problems, rather than resorting to “define your requirement” type responses. It may not seem fair, or even reasonable, to expect developers to read user’s minds, yet that is EXACTLY what the user’s clients expect of them. I can hardly underestimate how badly a “ask specific questions about specific problems” response offered by an architect would be taken by the client. Needless to say, the client would be working with a new architect, and the former architect would be looking for a new client.

33

amen.
however.
i offer my empathy as the horizons they are dealing with are far newer and more more complex than architecture,
architecture has been existing for as long as humans have had houses. Architecture started when log cabins, or igloos existed. Whereas rhino has existed for as long as 3d modeling became a thing, which… at best is what… since the 90’s?

mid 90’s?
…

so. its easy to criticize and say HEY MAN. YOU"RE COMPLEX ALGORYTHMNS AND JUNK DON"T DO WHAT I NEED. but. i mean.
if you’re so well versed and shit, go make a new one. and call us when you get it to work.

we are at the frontier of technology man. respect the hands that feed you. peace.

1 Like
34

Fire render or any other freebe will not keep me from using Flamingo; I’ve found it eminently usable for my purposes and have created some pretty cool renders. I’ve attached one that uses a photo of our downtown Annapolis area as a backdrop…and keep up the great work, Holo, fun seeing you on the discourse, Robb

2 Likes
35

I fail to understand your rant, but in any case, enjoy Rhino; I do and it changed everything for me in the most positive possible way.

36

Your renderings and modeling are very nice—Mark

37

PCIE-SSD 50x read/write compared to FASTEST 10k HDD’s. Why is this important ? BECAUSE the “load” people experience is NOT file saves at the end of the day. Its their (data file’s) management in and out of transitory data the program uses as it goes along its working way. THIS is what really slows things down for conventional HDD users - the CONSTANT data swapping. LOTS of RAM can help, but more on that below. YOU don’t see the “phantom files” (data held in “limbo” while an operation uses it for calculations and swapping, etc.) but they’re there, being managed as “files” on the SSD as if they were in RAM, but with full RAM access operating speeds. On the HDD its the same, but with the -50x lag of a conventional HDD, even a 10k one.

My 2nd-gen I7 w/ 16Gb RAM and 500Gb PCIE SSD is FASTER (ESPECIALLY Rendering - as that’s primarily a main-processor function) than my colleauge’s latest-gen 1x XEON and 64 GB RAM. Yes, SSD, even a conventional one is a HUGE advantage over even 10k HDD’s. PCIE is even FASTER !

All the best -

-C.

38

I just finished up a recent server/WS build which, as of yet, has been unphased by just about everything I’ve thrown it’s way…

The key to making one of these things affordable is to both carefully select optimum componants based on use case, and wait to buy until truly good deals become available. For my build:

CPU- (2x) Xeon E5-2699V4 (Oracle OEM QS-Spec, SR2J0, 2699v4 = 2696V4 (actually a bit higher turbo clock than non OEM 2699)

Supermicro X10DAC Motherboard

128GB DDR4 (16GBx8, 2Rx8, PC19200 2400MHz)

750 1.2TB AIC

M.2 to PCIE x8 card with (2x) 960 PRO 1TB RAID0

Titan X Pascal for the time being… Quadro P6000 & newly released GP100 WS card are still too pricy.

Computer is fast and silent under heavy loads/modeling/rendering/numerous VM’s all running full instances of Windows Server 2016 in a lab enjoronment.

39

I"m a Sketch Up user right now but I seem to be approaching some sort of sluggish limit with my 160MB model. Is Rhino faster than SU? I don’t really want to switch, but this model will eventually grow to 4GB or so, even after splitting it off into other models for different parts of the building.

My iMac is 9 years old with just 4GB RAM and 256MB on video, but my co-worker has 24GB RAM and he says even his machine is getting slow with the 9m edges in this model. Please advise.

Hmm, at least your comment board is similar to SU. Interesting…

40

I found it very useful when dealing with large files: