Rhino 6 mac Beta

[Helvetosaur]

@Gijs

How about technical view?

Yep, Technical is very bad in V6 on this machine - thus probably any display mode that relies on the Technical pipeline like “Artistic”, etc.: V5 20 sec. / V6 130 sec.

But @Gijs - my Windows machine with a 5 year old GTX 780/3Gb does TestMaxSpeed in Rendered mode in 1.65 seconds, and Technical mode (without hidden lines checked!) in 5 seconds - so Technical mode is 3x slower than Rendered even on Windows here. With hidden lines shown Technical slows down by half again, takes 10.8 seconds. Artistic is also about that speed.

[Zsimon]

I tried using the magic mouse with Rhino and I think it caused zoom problems (and not only). Of course, each of us uses what he feels is right, but perhaps we could see if performance improves with a normal mouse.

[_val (Valentin Johannes)]

@Gijs

I agree it is a bit weak, but as long as this memory isn’t being used fully it shouldn’t matter. The better performance you find is not because of the memory but the 460 is just a faster card.
I don’t really use my MacBook for Rhino work in general, only for taking files with me to clients and for my work at the academy so don’t need it to be super fast. Rhino 5 is fast enough but 6 isn’t.
Also when opening multiple files it gets really really sluggish

Exactly! I don’t even expect Rhino to run smoothly with huge and overly detailed models; that’s why I made a really basic test model and it’s still not running at >30fps (when I would expect 60fps) Especially considering the render mesh has only 1112 polys.

@jeff_hammond

that said, if I recreate your example, I’m not getting the major slowdown after running TestMaxSpeed multiple times with your Outline display mode?

I’m also not getting it every time, I have no idea what’s causing it. When navigating in the viewport, everything seems to run smoother compared to using testmaxspeed.

Also, I did another test using gfxCardStatus to force Rhino to use the Intel HD card. Here are my results, average of 3 runs each:

            Intel HD 530     AMD Radeon Pro 450
Rhino v6  +----------------+--------------------+
Outline   | 5.42s          | 5.89s (!)          |
Shaded    | 1.54s          | 1.01s              |
Rhino v5  +----------------+--------------------+
Outline   | 1.11s          | 0.81s              |
Shaded    | 0.96s          | 0.70s              |
          +----------------+--------------------+

[diff-arch]

@Helvetosaur

The problem with AMD is not the hardware, it’s their drivers - on the Windows side at least. AMD drivers are optimized for gaming, not CAD - they don’t give a crap about the CAD market.

OK, I guess the Windows gaming community and cypto mining industry are the biggest markets for GPUs nowadays. CG is only a small one.

@robinp

Let’s not lose sight that the original point was that V6 seems much slower than V5.

That’s not my goal at all, but there are two sides to the coin. Like @Helvetosaur explained above, Rhino 6 is optimised for newer, high-performance GPUs and most mac users have older, less performing cards.

Here’s another test with 1000 random N.u.r.b.s. cubes:

V5 TestMaxSpeed:

Wireframe 0.96 seconds.
Ghosted 5.18 seconds.
Shaded 2.32 seconds.
Rendered 4.09 seconds.

V6 current Beta TestMaxSpeed:

Wireframe 2.06 seconds. (48.61 FPS)
Ghosted 17.27 seconds. (5.79 FPS)
Shaded 2.35 seconds. (42.63 FPS)
Rendered 14.50 seconds. (6.90 FPS)

Wow, the performance differences here are huge. I’m speechless! What a disappointment!

Here’s the file, if anybody else wants to give it a go:
test_cubes.3dm (17.8 MB)

[Helvetosaur]

@diff-arch

Rendered 14.50 seconds. (6.90 FPS)

If you change the lighting in Rendered to “Default” instead of “Scene”, does it go any faster?

[diff-arch]

@Helvetosaur

If you change the lighting in Rendered to “Default” instead of “Scene”, does it go any faster?

Here are the results:

V6 current Beta TestMaxSpeed (Default Lighting):

Wireframe 1.87 seconds. (53.36 FPS)
Ghosted 16.96 seconds. (5.90 FPS)
Shaded 2.02 seconds. (49.60 FPS)
Rendered 5.20 seconds. (19.23 FPS)

Everything improves a little, except Rendered, which gets much faster. How ever in comparison to v5, this is still barely tolerable.

[OXII]

I was sure that Rhino mac 6 is years away from its release since I was shocked by the slow performance which you could feel with every mouse move an click. And since I did experience it +/- on all macs we have I thought this is simply because we are still in early development.

[Philip]

@Gijs

Not necessarily that but I was always in the assumption that the slowness was due to things being in development

Me too.

@Gijs

Much has been changing during the wip phase, so I thought once it will be on par with the windows version, these things will be ironed out

Yes, that’s what I thought also.

@Gijs

What worries me though is that I get the feeling that McNeel seems to be surprised about these reports. It isn’t that difficult to compare v5 against v6 to find that the latter is much much slower

Yes, that seems a bit strange…

Philip

[Philip]

@Gijs

Also when opening multiple files it gets really really sluggish

Yeah, this is still quite bad (even if V6 doesn’t crash anymore)! Can’t imaging trying to use Worksession - when we get it on the Mac (V7 obviously)…

Philip

[jeff_hammond]

@Gijs

arctic:
Time to regen viewport 100 times = 50.98 seconds

hmm.. I think I’m going to not post any more comparisons in this thread because I’m getting way better performance than you all.. for example, using the test model in the thread with Arctic mode for me:

Command: SetDisplayMode
Display mode set to "Arctic".
Command: Testmaxspeed
Time to regen viewport 100 times = 5.06 seconds. (19.76 FPS)

---

using this computer

that’s a $4000 computer though.. surely can’t expect everyone to pay that for decent performance (imo).

---

edit-- hmm, actually, if I torture test it and set Arctic to Maximum Shadow Sharpness (which looks best to me), I then get super slow performance

Time to regen viewport 100 times = 96.96 seconds. (1.03 FPS)

---

edit2.. just realized I had 3 other models open when doing that so probably took a hit there.. I think it’s evident there’s a performance hit for everybody going to V6 vs V5.. so.. I’ll quit with the faulty tests

[Gijs (Gijs de Zwart)]

@jeff_hammond

Time to regen viewport 100 times = 5.06 seconds

that is actually very fast. On my workstation (windows) which has a quadro M4000 this still takes about 17
seconds

@jeff_hammond

I think I’m going to not post any more comparisons in this thread because I’m getting way better performance than you all

I think you shouldn’t stop as long as you compare it with v5. If it is slower then something is wrong imo.

@jeff_hammond

set Arctic to Maximum Shadow Sharpness

didn’t know you could adjust this. where?

[jeff_hammond]

@Gijs

didn’t know you could adjust this. where?

in the shadows tab of Display Modes

(in Preferences)

[Gijs (Gijs de Zwart)]

@jeff_hammond

in the shadows tab of Display Modes

ah, cool. Makes me wonder why this isn’t done dynamically/progressively. If you rotate at full setting, it disables the AO during spinning, but could just as well degrade the amount of samples instead and keep refining when ‘still’

[jeff_hammond]

@Gijs

@jeff_hammond

in the shadows tab of Display Modes

ah, cool. Makes me wonder why this isn’t done dynamically/progressively. If you rotate at full setting, it disables the AO during spinning, but could just as well degrade the amount of samples instead and keep refining when ‘still’

yeah, that would be a cool option.. allow you to set up your view with faster/degraded quality then switch to high quality after, say, one second of stillness

(if i understand you correctly)

[Gijs (Gijs de Zwart)]

@jeff_hammond

allow you to set up your view with faster/degraded quality then switch to high quality after, say, one second of stillness

exactly. It kind of does this already though, but not enough and not when it’s set to its max

[OXII]

What are you guys considering to do? Whom of you will upgrade?

[marlin (Marlin Prowell)]

Do not use TestMaxSpeed to compare Mac V5 and Mac V6 performance

TestMaxSpeed is a test command. I wrote both the Mac V5 and the Mac V6 versions. They are totally different implementations and cannot be used to compare V5 with V6. We can use it internally to see if code changes made within a single version have an effect on performance.

The command was originally written for Windows Rhino, where you can force the display to be drawn when you want. That is impossible to do on MacOS. The best you can do on the Mac is request that the display be updated. Mac V5 and Mac V6 use entirely different techniques to determine whether that has actually happened, so can have wildly different timings.

If the Mac code did not check that the display has actually been updated, then MacOS optimizes out a display update when quickly followed by another display update request. Without artificial delay checks, TestMaxSpeed always takes less than 0.1 seconds, because it only draws the first and last frame. All in between frames are optimized out.

The technique used in V5 to detect display updates mostly works, but the OS can still optimize out some of the updates. V6 does a better job of this and can appear slower because it actually draws every frame.

And, all this is Independent of display mode settings which also have a huge effect on performance.

[jeff_hammond]

@OXII

What are you guys considering to do? Whom of you will upgrade?

I will upgrade… I’m almost sure of that… it works ok for me.

plus, I like the fractions
(as well as other updates to annotations)

---

booleans work better too (even in simple situations that should just work from a user POV but didn’t in V5)

like, making this in V5 with boolean techniques fails… V6 can do it:

---

(just using these as the examples since they’re literally the files I’m working on at the moment )

[OXII]

Thanks for your opinion.

I did just use it sometimes for some screenshots since the rendered mode is more sophisticated that in V5. Unfortunately I was never really able to test V6 properly, which is a shame.

Edit:
I just tried to open my Rhino V5 Model (a project I am currently working on) but the layouts are gone. I V5 I had done ca. 10 A3 pages with detail views, but if I open the file in V6 the layouts are blank.

[Gijs (Gijs de Zwart)]

@marlin

TestMaxSpeed is a test command. I wrote both the Mac V5 and the Mac V6 versions. They are totally different implementations and cannot be used to compare V5 with V6

I will follow your advise, but: Fact is that V6 feels much slower in all display modes. Especially when zooming in and out it’s going to stutter where v5 keeps being smooth. The fact that we began comparing using testmaxspeed in the first place was because everyone notices the fact that v6 is slower.