Computer crashed. Have to get a new computer. I’ve done the research and I realize there’s a fork in the road:
Z97 & i7 4790
X99 & i7 5820
Does anyone here know whether more cores will benefit running Rhino over the long run? I’d like to future proof my purchase but I also don’t want to spend extra money for extra cores and DDR4 if it won’t benefit me.
Rhino is NOT multithreaded, which means that the bulk of its work (with few insignificant exceptions) must be done on a single core. So no, having more cores will not improve Rhino’s speed per se. However, if you routinely have many processes (applications) running at once, having more cores means that Rhino’s process will not be “intruded” upon by other programs and all your programs will run comparatively smoother.
And as jeff noted, some plugins might run faster if they take advantage of concurrency (such as some renderers)
I do render more and more (which I kind of hate compared to modeling) and I also tend to have numerous programs running. May have to flip a coin here. Feeling a little backed into a corner and having to make a quick decision (no computer = no income). I’m hoping that a little shop I found on the Lower East Side can get my existing computer back in shape long enough to limp into the Thanksgiving sales season.
Thanks for the replies. I kind of knew Rhino wasn’t multi threaded but I couldn’t remember for sure. That may give me a basis to decide.
There is reportedly quite a bit of work being done right now to multiprocess in Rhino where it makes the most sense, and the more hopeful among us are expecting to see the results in Rhino version 6. These days a 4 core machine with hyper threading is a general purpose machine. 6 cores aren’t that much more expensive, but they typically have lower clock speeds than 4 core chips. So for Rhino 4 or 5 the fastest 4 core with hyper threading that you can afford is probably best. On the other hand, slightly less clock speed in a 6 core would probably serve you OK out through the first release or two of Rhino 6 - if only we knew when that might be. Remember to get at least 16 gB of the fastest memory available for the CPU you select and a good video card.
[quote=“AlW, post:5, topic:13270, full:true”] So for Rhino 4 or 5 the fastest 4 core with hyper threading that you can afford is probably best. On the other hand, slightly less clock speed in a 6 core would probably serve you OK out through the first release or two of Rhino 6 - if only we knew when that might be.
heh… but if wonders do happen… going from 1core to 4 might be amazing in and of itself… it’s a different jump than doing something like 12cores to 16.
Hi. You also have to consider Hyperthreading.
My findings with holomark2 test is that with HT disabled I got about 20% more performance in the Rhino tests, but of course Rendering was about 30% slower.
With HT ON each core has 2 x threads so if the PC is busy with alot at once then Rhino is effectively only getting to use a single core for half the time.
With HT off and a busy PC it varies but Rhino gets full use of a core when it gets a chance.
Rendering needs as many cores and threads as you can throw at it, but this performance is not linear.
eg. just adding more threads with the same number of cores does not double the render performance because you effectively still have the same processing power available, you are just able to have fewer wasted CPU wait cycles.
So for Rhino use only, a higher clock speed i5 (without HT) can outperform a slower clock speed i7(With HT and more threads), but for rendering the i7(with HT) will rule. Michael VS
So far I know Rhino profit by a high speed per core, but more than two or four cores doesn’t help to speed up Rhino. If you think about rendering you shouldn’t ask you, how much time you can save with a faster Rhino and how much with faster Rendering. A faster CPU can be helpful for meshing the model, booleans … and you can keep some seconds/minutes over day. But for intensive rendering you can save hours by using more cores.
I would try to look for max speed for rendering, but try to keep the speed per core not to low.
Also don’t forget since Rhino doesn’t used all cores the few used cores are automatic overclocked by the CPU management. I’m not sure this automatic is used for all Intel CPU, but at my machine is see the effect. I have 32 threads at 3.1GHz, but if Rhino runs without rendering in the background the few cores are running at 3.4GHz.
On the question of how many cores… I’m researching a new Mac Pro system. My final output is using “Make 2-D Drawing” to create a hidden line illustration. Which is the better way to go… more cores or a faster clock speed with fewer cores help with Make 2-D Drawing? Options being… 3.7GHz 4 core, 3.0GHz 8 core or the 2.7GHz 12 core.
A few tasks in Rhino have been optimized for multiple cores(like ReduceMesh and…ummm…uhh…)but most tasks, they’re just trying to make them WORK nevermind fast, or they just aren’t really parallelizable.
Software rendering scales pretty much linearly with clockspeed & cores, except for say in Brazil there are ‘setup tasks’ getting the scene prepped to render that are single-core only, which depending on your scene could be a trivial amount of time or a huge bottleneck for an animation, which has led to the hilarious situation of me having 10+ instances of Rhino open in order to make sure my dual 10-core Xeon rendering box was actually being mostly utilized most of the time.
Conventional wisdom about Hyperthreading is that for parallelized tasks it’s of no benefit or even slows you down. On my Xeon (I guess it would be the previous-to-newest gen, got last year,) that’s not true, it actually somehow wrings more utilization out of the hardware, it renders a solid 15%+ faster with HT on.
but yeah, in rhino, you’ll benefit from a faster clock speed from a single core.
with the newer xeons though, they can pretty much hold their turbo speed indefinitely… if your system is mostly using only one core (ie- running rhino with no background applications doing things), the listed ‘up to’ turbo speed will be the potential clock rate.
Although it is true that Rhino does not use multi-threading for most operations at the moment, we recognise that if we’re to make use of the hardware people buy this needs to change. As we move forward, .NET will become a more important part of our development and the Task Parallel Library has some pretty good stuff for easily multi-threading stuff.
Although I hesitate to make a promise, it’s certainly a good bet that -over time- Rhino will become more and more multi-threading capable. If you’re buying this new machine for the long haul (several years) then multiple cores may not at all be a bad investment. If however you’re expecting to purchase a new machine just a short couple of years down the road, then it makes more sense to invest the fastest single/dual core machines you can find.
I render a ton with Flamingo as it is a delight to clients. I use a hexacore I7 3930K overclocked to 4.25 ghz and water cooled with two rads and a plethora of fans. Of course, for me, building the system and checking the temps was half the fun. But, the end product with 32 gig of ram is fast, cool and quiet and only dims my neighbors light during the worst of high res rendering when alll cores are at 100% for an extended period of time with temps in the high 50’s C. Cheers, Rob
Technically speaking, the late 2013 Mac Pro’s primary focus is in its dual workstation GPU configuration—which is unfortunately not utilized by Rhino (and is not very useful for most programs, as only one GPU is used for graphics while the other stays inactive unless a program is coded specifically to utilize it as an OpenCL processor [final cut pro, indigo renderer]; they can only be used simultaneously if you boot up in Windows).
The 12-core Xeon is mildly amazing for rendering software like Keyshot, but a user would benefit more from a dual-CPU setup, which isn’t available with the current Mac Pro.
If you need to be running Mac OSX and won’t be rendering, a 5K iMac with the top of the line CPU/GPU option would actually work out to be faster at running Rhino (as it has a faster clock speed than the xeons). It would be cheaper and come with a 5K display (which is absolutely amazing with Rhino). I definitely recommend this option as a Rhino for Mac user who has used both systems.
Or you could wait a month to see if they announce a new Mac Pro…
[quote=“96SS, post:12, topic:13270, full:true”]
Thanks for your response.
Didn’t mean to leave that one out the 6-core… Are you saying that the 6-core would be the one to get as it’s a good balance between processor speed and quantity of cores? [/quote]
right… i suppose most people would argue the 8-core is the sweet spot but when money enters the equation, the 6-core is the best deal…
to go from a quad to 6 core (1.5x increase in core count), it costs $500 extra.
to go from a 6-core to and 8-core (1.33x increase in core count), it costs $1500 extra.
Does ‘Make 2D Drawing’ use multiple cores?[/quote]
i don’t think so… i didn’t think any commands used multicore but jim is saying ReduceMesh does.
the Mac Pro that will complete Make2D the quickest would be the 4-core… depending on you current computer, it may not be enough of a speed enhancement to justify the cost of a new system… but if you need to replace your current computer anyway, the cost becomes easier to swallow.
benchmark geeks like to hang out online and argue about what computers put up the highest scores etc… just make sure the people advising to get an old mac pro then upgrading to a hacked gpu actually use their computers to do something productive… instead of trying to soup it up to get their name on the geekbench page… if they don’t actually know how to use any software in a production environment, take their advise lightly
also, about the classic mac pro vs the new one… there’s not an upgrade you can do to the old mac pro to make it run rhino faster than the new one will… you’re pretty much limited by the socket in the machine as far as what type of cpu upgrades you can do… you can’t put new cpus in old machines… and the older xeons aren’t as fast nor do their turbos work as well as the newer cpus.
OK, I’ve read all of this, an buying a new computer for Rhino 5, will worry about R6 later, and need some advice. It’s between a Xeon E3-1271 and an i7-4790 CPU. This is for an architecture firm. There well be one monitor running 1080p (27") with the tower. I’ve been running the software for years with a inferior setup, with no problems, until I started rendering a lot. It’s seriously slow going. The new tower it’s actually for my interns in the office, as I work from home, but I’d like to be able to use the tower remotely, by connecting to it over the Internet, and running Rhino off of it on my home displays (two 1080p monitors). Aside from the CPU decision, what else do I need to keep in mind? There’s going to be 32GB of RAM, and right now there’s no video card in the configuration, because I never needed one before with my older MacBook Pro. However, will the fact that my home setup had two monitors and the desktop in the office has one create problems? Both CPU’s are rated to run three monitors without the addition of an aftermarket GPU. What other configuration options are absolutely necessary? I read ask about Rhino using a single core for CPU processing, and both the CPU options between which I’m choosing have four cores with hyperthreading; but, I figure, if an intern is running Rhino at the office, and I’m running it off a different hard drive partition at the same time for rendering remotely, the extra cores will really help. Also, the graphics card in one if the processors has OpenGL 4.0, while the other has 4.3. Well that make a big difference? Can a GPU be added later to improve speed, our fits it have to be added at the time of the build? Finally, fits the GPU aid in rendering speed, or just help with display? (Is it that important for my desired “remote cloud rendering”?)