Pretty ecstatic about these results! I tweaked the GPU profile and screen display settings to see how much more performance I could squeeze out of the system. Now to overclock the CPU 
New to this board and discussion, but I work with McNeel (Online Level 1 Training). Just purchased a new, max’d out Dell XPS 8700 Special Edition during Black Friday weekend and thought I’d compare this system and my prior system. Loving this ATI Radeon R9 270 (2 GB RAM). Tested with No AA / No MMF / High AF. Kicks some ass in “ultra” settings in Far Cry 4, too 
Hi there!
After quite a while without Rhino I put together a homemade workhorse, here is the result.
The cpu actually runs at 4.2 GHz, without any finetuning, and never above 45° Celsius within this test.
Which is cool, as it leaves headroom for further twisting;-. Motherboard is a ASRock OC Formula, and the GPU
is a 6Gb Sapphire HD 7970. I went for this slightly older one, it runs at stock settings, as I thought, the more RAM,
the better. Unluckily Rhino does not seem to recognize all of it, or at least the asking routine
within holomark. Anyway, now at least I can throw out some bigger meshes without going into stutter mode.
steff
Good total score, but … did you have seen your shaded mode score? The best results are around 30s vs. your 67s. I would stay on the GPU tweaking. Looks like your CPU pushed the result already. 
B-but the shaded score number relates to frames per second, not seconds, so 67 is better than 30…?
Are you joking? 
I’ll try this with the gpu too!..
Whereas, the behaving of Rhino with this cards is kind of strange, I have an object with 4.5 german millions polys, and when I copy these to five alltogether, the screen does not slow down to the same degree at all.
So maybe my 6 gb videoram do make a difference…
You are right, I have overseen that we have fps here. Great result. Did you test the old bike model speed test? I’m curious how the holomark score reflect the the power of difficult graphic situations.
If you like, open this model and disable advanced GPU lighting and shadows before run _testmaxspeed in shaded mode.
www.simulacrum.de/temp/Testmaxspeed.rar
Here my power machine result. Pro Hardware, bad results 
Holomark 2 v2,2,03
Total Score: 21363
Total Runtime: 380.61 sec
GPU scores: 19190
GPU_01 - 215.10 fps - Cube 4 tests
GPU_02 - 33.20 fps - UDT Shape
GPU_03 - 62.10 fps - Wireframe
GPU_04 - 30.40 fps - Shaded
GPU_05 - 17.50 fps - Rendered
GPU_06 - 16.80 fps - Block Rendered
GPU_07 - 17 units Nurbs @ 5 fps in Wireframe
GPU_08 - 9 units Nurbs @ 5 fps in Shaded
GPU_09 - 21 units Nurbs @ 5 fps in RenderSpeed
GPU_10 - 69.40 fps - RenderMesh Render
GPU_11 - 192.30 fps - RenderMesh RenderSpeed
GPU_12 - 92.60 fps - JoinedMesh Render
GPU_13 - 303.00 fps - JoinedMesh RenderSpeed
GPU_14 - 4 units mesh @ 15 fps in Shaded
GPU_15 - 5 units mesh @ 15 fps in Render
GPU_16 - 118 units mesh @ 15 fps in RenderSpeed
GPU_17 - 60.60 fps - mesh in Rendered Studio
GPU_18 - 8.30 fps - Nurbs in Rendered Studio
GPU_19 - 19.60 fps - Block Illustration
GPU_20 - 66.70 fps - 2D single
GPU_21 - 7.80 fps - 2D massive (20x)
CPU scores: 2173
CPU_01 - 10.73 sec - Booleans and Contours
CPU_02 - 46.67 sec - Twist and Taper (UDT)
CPU_03 - 4.74 sec - Meshing Mini
CPU_04 - 0.16 sec - Extract Render Mesh
CPU_05 - 0.06 sec - Join Render Mesh
CPU_06 - 15.46 sec - Reduce Mesh
CPU_07 - 146.62 sec - Calculating Technical display
CPU_08 - 5.63 sec - Making Silhouettes
ASUSTeK COMPUTER INC.
Z9PE-D8 WS
NVIDIA Quadro 6000 - 4095.0 MB
DriverVersion: 9.18.13.4121
Intel® Xeon® CPU E5-2687W 0 @ 3.10GHz
NumberOfCores: 8 NumberOfLogicalProcessors: 16
MaxClockSpeed: 3.1 GHz
Intel® Xeon® CPU E5-2687W 0 @ 3.10GHz
NumberOfCores: 8 NumberOfLogicalProcessors: 16
MaxClockSpeed: 3.1 GHz
TotalPhysicalMemory: 32.0 GB
Microsoft Windows 7 Professional
- Service Pack 1 - 64-Bit
Rhino 5 sr 10 64 bit
Hi @Micha, Looks like you have the same LONG technical Display calculation time problem I have. I have different CPU (i7 3770) but I am also using a Quadro 6000. Co-incidence? Or is the Q6000 have something to do with this. My CPU_07 scores were also always between 130 and 150 sec., whilst everyone else has <2sec times. Let me know if you figure this one out. Michael VS
Might have to look into overclocking the GPU then - Can only do so much with the profiles NVIDIA provides and the settings one can modify or look into a different card/dual cards
I bet i could squeeze a few more out if I overclock it. I read a few articles about an Intel employee promising lessons to people who couldnt overclock the CPU to 5ghz on air… and I am watercooled
Oh Man, yep, that’s a dreary result 
But I bet you bought that expensive pro card, 'cause you use other software like solidworks etc, that depend on it’s OpenGL arsenal? As for Rhino to be a hefty Alleinstellungsmerkmal that it does not need such a card is known for almost two decades
Anyway, why don’t you use two cards, one other for Rhino. I bought mine for 260 €, and it scores good enough here…
As with the Xeons, I was tried to buy into such a system, reliability, eg ECC, etc, but in the end I didn’t want to be without the chance of securely overclocking to even for much money unknown heights in Rhino’s performance. As I had read here, how poor it still behaves in multithreading, and thus depends on high speed numbers of the processor.
Meanwhile I tweaked the gpu a bit.
Merry Christmas to All!
Have a good time.
I did the bike with testmaxspeed, and it is 12,61 sek. Is there somewhere a chart for this test with the bike, so I could compare?
I did it with different gpu settings, and a gpu speed change of 33% yielded into a difference of drawing time of a half percent.
But a difference of cpu speed of 5% made a difference in time of testmaxspeed of 3%. So this test is heavily cpu dependant.
It seems like you have V-ray installed, and if so make sure you have upgraded it with the latest SR. It should fix it.
Why technical display mode is so slow I don’t know. I have only seen that on a few occasions, and never repeatable.
Have you tried restarting Rhino?
As far as I know this is CPU based and not GPU based, and I see you have two 8 cored Xeons (32 logical threads) so maybe @jeff knows if this is causing trouble (That is IF technical display mode is hypertreaded AND if “too many” cores can cause an issue with this… I have no idea how that works and presume I am wrong, but maybe it can inspire to find the real cause)
Good luck though, and try to unload (not uninstall) V-ray and run the test again to see if that helps.
Hi
i have a similar problem. I enter holomark 2 and some things are happening but i do not see a graph or test results. Yoy mentioned to hit F2 and then paste all the holomark stuff. Whereto would i have to paste this?
I run rhino 5 64 bits on win 7 64 bits.
I need to apologize. As I didn’t think it important for Rhino, I didn’t mention
I use this card with a modified driver, it’s called FirePro-9.003.3 modded, and it makes supposedly this
firepro driver use a consumer card from those days in 2012 like a pro one.
With great results for solidworks etc(which I didn’t test).
Now today I did Holomark2 with one latest consumer card driver,
and the results weren’t as good. So, this driver does even for Rhino make a difference.
One could download it from benchmark3d.com until a week ago or so, and now it’s gone.
The whole benchmark3d.com that is. Anyway, today’s results:
Lenovo W530 3rd Gen i7 @ 2.6GHz, 16GB RAM, nVidia Quadro K1000M
31,923 on nVidia Optimus
32,398 on nVidia Discrete
So, I’m back with the latest Nvidia beta driver installation. Shaded mode is a little bit faster and the calculating of the technical display is working now (interesting for @mvyess ).
@steff : My Quadro 6000 needs 17,6s for the bike test. If we compare the result with the Holomark test we see that the bike test is a more complex display test, the speed increase of the HD7900 is lower.
It’s a pity there is no test chart of this test. In the past there was a test result of a Quadro 4000 with 15s. So, in comparison my Quadro 6000 result is quite bad and the HD7900 isn’t a big step forward.
That’s why I like the bike test, my impression is this is the best and simplest real life model test. I have no problem with models where I get 30 fps, I have a problem with my models where I have 0.5fps. And it looks like this models will stay a low frame rates no matter which card is bought. No wonder since the GPU usage is realy low in this cases. No hardware can solve the software limit of Rhino 5, I hope Rhino 6 doe’s the job.
If Rhino 6 would use the full CPU+GPU power we would see a speed increase of several hundred percent.
(If I extract the render mesh and hide the NURBS, than I get 0.86 seconds - this is speed I expect by a modern card and software.)
Hi Micha,
GPU 08 replaced the bike test. Compare that score with other systems to see where the limitation is. No cards are faster than the rest on this test. Same as what you would see in the bike test because it checks how complex a scene can be before dropping below 5fps, instead of testing how fast the system is with a complex scene. (it’s the same but different)
I am considering replacing all the unit stress-tests with a fixed array test for SR2.
Hi Holo,
I see, the relation (HD/Quadro) of the GPU 8 test is 1,66, the bike test gave me 1,39 and the GPU 4 test was 1,8. So, the GPU 8 test goes in the direction of a complex scene, but the bike test seems to be the heaviest test scene we have. I suppose with my real project scenes (with a much slower frame rate) the difference between the high end cards will be not better. If a scene is really slow there is no way to get a big speed enhancement per hardware.
Thanks,
Micha