No announcement yet.

rendering hardware

  • Filter
  • Time
  • Show
Clear All
new posts

    rendering hardware

    we currently have some (i7 870, p55 chipset, 8gb ram, geforce 460gtx, 1tb wd, dvdrw, win7x64, revit 2011, 2012) workstations
    we currently have some (i7 2600, p67 chipset, 16gb ram, geforce 560gtx, 120gb vertex3 ssd, dvdrw, win7x64, revit 2011, 2012) workstations

    when we first got the i7 870's and were able to do a 11x17 rendering @ 150dpi in 2hrs or less, it was so much better than our older machines. but soon found that we were being requested to do 24x36 renderings, that looked better @ 300dpi, but would take 5+ hours to render. our current process is to model, add archvision entourage (trees, people, cars) in revit and render, then take it over to photoshop for some clean-up, a better horizon line, sky, etc. we are trying to see if there is a hardware related solution to speed-up rendering in revit. we have no plans to move to 3d studio max for renderings as this time.

    got a research question for pricing out a fast rendering machine for my office. as revit 2011 and on uses all the cores on a machine, i have the following questions:

    1. other than the cloud, is there a way that a revit end user can cluster machines to speed up rendering times, from what i can tell, there isnt?
    2. i cant find anything to do a network render, short of dividing up an view to be rendered and have multiple machines handle part of the view OR exporting to 3dmax, which we do not have and probably will not do.
    3. revit server has nothing to do with rendering, from what i can tell
    4. from what i remember when revit 2011 was released, there is no benefit from the gpu in terms of rendering time for revit 2011, 2012, probably 2013, but i havent had a chance to install or test it out yet)

    obviously, the newer ivy bridge i7 3770K (3.9ghz stock) will provide a bump in rendering performance over a i7 2600 (3.8ghz stock)
    passmark benchmark results of 10,443 and 8,894 respectively, a theoretical 17% increasel; though both are quad with HT, so it shows 8 cores in taskmanager)

    but would a xeon provide a noticeable improvement in rendering times?
    E5-2687W passmark benchmark results of 17872, a theoretical 71% increase over the 3770K (xeon is an 8 core, but has HT, so guessing it would have 16 cores in the task manager, not to mention the ability to put 2 on a board, for a total of 32 cores, perhaps bringing the raw benchmark power to 142% better, if revit can distribute over 32 cores that well) but was wondering if anyone had any advise, as i dont want to drop $3000+ on a guess.

    any other resources or research for this topic? i couldnt find that much other than blogs about updates in revit and everyone talking about the cloud, that appears to be not the best solution at this time as it is not finalized; dont want to get locked into a beta SaaS until autodesk has worked-out the kinks and clarified their business model for it.

    thanks for any help and advice.

    Autodesk - Autodesk Revit Architecture Services & Support - Rendering processor core limit for Revit 2011 products

    there are 2 posibilities:
    -dual xeons 8core each last generation (the fastest solution)
    -an overclocked 6core 3930k/3960x as much as possible ~4,5/4.8(i think is faster than a 1 xeon version because xeons are not overclokalable). 3930k has a big price advantage over 3960x with a speed allmost identical.
    I have it, it performs well in revit with all 12 threads.

    Anyway if you go dual xeon way, make a price scenario to see if a dual xeon station is cheaper than 2 stations with a 3930k overclocked. Because if it is around same price you have more versatility with 2 stations and the same speed in the end if you have more than one renderings to do( any station can render a different view so even if a station has half the speed of the dual xeon version, at the end, render time is the same)
    Last edited by gaby424; June 23, 2012, 07:04 AM.


      I wouldn't get into the "overclocking", but otherwise concur with what Gaby had to offer.

      You could look at the i7-3930K/3960X, which have 6 cores.

      The only reason to look at the Xeons is if you want to bump it up to 8 cores and/or dual (or more) CPUs, but yes, then you're looking at a lot of money.


        Originally posted by iru69 View Post
        I wouldn't get into the "overclocking", but otherwise concur with what Gaby had to offer.

        You could look at the i7-3930K/3960X, which have 6 cores.

        The only reason to look at the Xeons is if you want to bump it up to 8 cores and/or dual (or more) CPUs, but yes, then you're looking at a lot of money.
        we just got a new dual processor xeon server at the office; i wonder if i can talk our IT guy into letting me either test revit directly in windows server 2008 OR installing virtualbox and testing it in win7x64 to run the revit benchmark on it to compare the rendering numbers rather than a synthetic cpu benchmark utility like passmark.


          Frankly not sure what the point would be though? Besides, the VM is going to kill performance, and even if you run it natively under 2008, unless you shut down every service on the server, you have no idea how that's affecting performance.

          Adding more cores isn't going to scale 1:1, but it really should be "good enough" in evaluating the cost/benefit ratio. So, what's the question? Is it worth spending $4K-$8K on a rendering machine? I would respond to that with: how much do you bill for the renderings? Seriously, that's how you need to be looking at this. That E5-2687W might buy you a 75% speed bump, and two of them might be 3x faster, but they're funny-money expensive. Otherwise, just get an i7-3930 workstation for under $2K and call it good enough without breaking the bank.

          Wish it was "easier." Good luck.



            I think you can't sell revit renderings, at least without heavy post-production which will take almost same amount of time as revit render. So it'll be better to invest into another rendering software which will do the job faster and with better result.


              I have seen plenty of offices use Revit renderings with great success. Yes, for final, photo real, fund raiser kind of renderings there is a need for post processing, because RPC stuff looks horrid. But there are lots of good tricks in PhotoShop, et. al. that make this easier and faster. And honestly, most firms don't have someone with the eye to do this kind of work, and it is usually better to outsource it just like you would a large watercolor rendering.
              But there is a lot of good uses for Revit renderings besides "final". Monochrome materials with realistic lighting for early "form" studies, simple massing for urban context studies, "material board" renderings with no entourage focusing just on realistic materiality and lighting, sun studies with diagrammatic entourage (lollipop trees and pin head people), design studies with a notion of materiality, realistic lighting, and printed to a rough paper and (quickly) embellished by hand with a sharpie, marker and colored pencil, pastel, what have you, etc. None of which are necessarily "sold" as discreet rendering products, but all of which can and should be paid services, even if you just roll a certain amount of render effort in to a flat fee. And all of which can be done by the whole project team, rather than bottle necking with a single "render guy" in the office, which is what happens when renderings are done in 3DS Max or some other software with a whole new learning curve. I can get an interior designer up to speed on material creation and basic interior rendering in a few days, and every material board they ever do after that will be all the better for it. 3DS max, not ever going to happen.
              The office that realizes that designing in Revit means they actually save enough time to provide "hand" renders, walk through animations (probably in hidden line or consistent colors, not rendered), interactive exploration of the design model in real time, and a hundred times as many 3d views as the competition, all for a competitive fee with better profit, is the firm that wins in the market. The firm that uses Revit to replace AutoCAD as a "construction documents" tool and still uses SketchUp for "design", 3DS Max for "renderings" and PowerPoint for "presentations" misses the point (and the opportunities) completely.

              Pragmatic Praxis


                One small observation for the 3d max side. It can`t be placed in the same basket with autocad and sketcup.

                You can`t loose any minute and any info if you just do the render part in 3dmax (I realy mean only the part in revit where you want to push the render button). There is a hot link with revit, all the materials can be defined in revit, the cameras views, the artificial lights(which apers in the schedules), the sun, the orientation. In 3D max you just tweek some setting for lightings and some generic settings for rendering quality with any emploie that has some 3dmax light&rendering training. These settings are tweeked only one time in a model with a price of half an our, but the render time for all your views is cut drasticaly.

                Ok, you can do in revit almost everything you want for a bussines that is already contracted.
                But as far as I know, one good realistic image (with a good concept behind, of course), can worth more than a lot of words about the quality of your architecure explained over a so-so presentation if we talk about a lot of the clients which are not trainded to judge in an objective way in front of a good picture, in the precontract phase when they have a lot of offers. I experienced this kind of atitude, so why should I be scared about a tool that is in the same box in revit, that is the primary weapon for this kind of problems.
                Last edited by gaby424; June 21, 2012, 10:04 AM.


                  Originally posted by gaby424 View Post
                  with any emploie that has some 3dmax light&rendering training...
                  Ah, but that is exactly the difficult part. The vast majority of architects don't have that training, and given that the environment and process in 3DS is totally different, getting someone up to speed on Revit rendering takes a lot less time than getting someone to the same place in 3DS. In addition, everyone in the office already has Revit rendering installed, no extra software cost for the office, no limited license count bottle neck that frustrates users when multiple people want/need to render at the same time, etc.

                  The other things to add is that, in my opinion, making a decent render, especially an exterior render, is really easy in Revit, and much harder in 3DS. You really need much less of an artistic eye in Revit, because some basic decisions are made for you. If you have the eye to go beyond those basic settings, then you have to use 3DS because Revit doesn't expose the same amount of control. But, if you can't get a Revit rendering to look good, you will spend twice as long still making a bad (or worse) rendering in 3DS. Using 3DS makes great renderings possible, but it does not make them any more probably, only a really high level understanding of all the settings and what they do, and how real light and materials interact, can do that, no matter what the software.
                  So Revit offers a great, low cost opportunity for everyone to learn rendering to some level, and then if someone really has a talent for it, and there is a business case, upgrading to 3DS makes great sense. PowerPoint, on the other hand, never makes sense.

                  Pragmatic Praxis


                    thats for everyone's feedback

                    my initial thought for increasing the rendering speed was directed at wondering what the cost was for building a machine that could, well, render much faster than our current machines. our current workflow is to model the building and site in revit, then make sure the textures are correct and hit render, wait 5+ hours and then save out to photoshop and then paste the resulting image on a titleblock in revit and send the client a PDF and or print.

                    obviously, there is always the option of 3dsmax, but without purchasing, training and retraining new employees; it seems to be more of a time consuming option than we want to delve into at this time, but it will be an option to the princpals when i present my research.

                    if there is a 3rd party or even autodesk owned external 3d renderer that i can insert into the workflow that would not require additional training and or expertise, that would use the textures from revit, or at least not require someone to redo the textures on all of the surfaces the rendering view; that would be wonderful.

                    at this point, the cpu in our new server is an single board compatible xeon, and its individual synthetic benchmark specs are comparable to a i7 2600.
                    the suggested i7 3930K is about 52% faster (probably somewhat in part to the 6/12 cores verses the 4/8 cores, as well as the +4mb of cache) though the almost double chip price is daunting; though, compared to the xeons with similar performance, it is half the price.

                    pricing out a system with a 3730k, 16gb ram (upgradable to 32gb but probably wont, unless you all think it will help, doubt it will), geforce 560ti, 700w psu, thermaltake cpu cooler (doesnt come with one stock) runs me about $1500 or about $300 more than our typical workstations, which makes sense with the $250 cpu cost increase and added heatsink. not adding an ssd to the quote b/c i have a vertex3 120gb that i used prior to purchasing ones for our 2nd generation machines, so keeps it from going to $1600.

                    ironically a autodesk reseller did call me thursday at work, his ears must have been ringing.


                    Related Topics