Holomark 2 Released!

unhandled

#698

In my case results are ridiculous :smiley: slower and 6cored e5-2620 with older kepler quadro had better results…


#699

wouldn’t it be nice if holomark included a test of the GPU performance in the new raytraced mode / cycles ?


#700

Yes that would be nice. It would require that I can set and reset the preferred render (cpu vs gpu vs both) and control global quality settings with out messing with the users setup. Setting up a new floating viewport with a fixed size is the easy part.

-J


#701

ok,
so if not possible already, talk to @nathanletwory , right?
common, make it happen!
:grin:


#702

Can anyone tell me why my benchmark is getting so few points in comparsion to others with 1080ti’s ?


(Nathan 'jesterKing' Letwory) #703

If you use RhinoCommon you could manually add a ViewportSettings user data to the viewport you are working with, then set settings as you want. For this scenario you may want to set the allow override settings - first save the old values, change if necessary, do your thing, put back old values.


#704

RhinoCommon is allways an odd place for my simple mind to venture into :smiley:


(Nathan 'jesterKing' Letwory) #705

I can always add the odd command to do that with scripting


#706

@Holo can u say few words about those results? This time i made benchmark again on the same setup - i didnt touched any settings and suddenly bench has half more pts. Is it relieable with pascal gen cards ?


#707

I guess the latest test has such a peak score on GPU_16 is that you turned off AA.
The reason OpenGL AA is limited on Geforce cards is so they don’t compete with Quadro cards.

Why it appears that Rhino struggles with some cpu’s with more than 4 cores are unknown to me.
But you have other scores that are lower than the previous test so this baffles me, and these inconsistencies are one of the reasons I made the test.

Sorry I can’t help you more though.


#708

Actually AA is set x4 all the time i didn’t touched anything. I have bench from previous setup with slower cpu per core which actually get more points.


#709

Hm, that is strange.

Here is a comparison: older Dual 6 core xeon with a 1070 card.
I have hypertreading turned off in the bios since it affects VR framerates in some single core driven apps as the cpu is a bottle neck for the system.

4xAA:

No AA:

And 8x AA:
(As you can see the card doesn’t suffer more from 8x than 4x)


#710

it would be a huge improvement if holomark would make sure to always run with the same AA settings - always on and 8x for example.
right now there is a lot of confusion about the results because of the different AA modes used. scores are no good if they are not comparable.

by default the drivers should let the application decide which AA mode is used.

at least the used AA-mode should be included in the info part of the scoring-sheet


#711

I agree and I would have added it if I knew how it would be possible to extract the data :slight_smile:


#712

I’m sure you’ll find a way, just like you will with the cycles integration :grin:
great work so far, it’s a much needed addition to rhino

here are my scores with 8xAA and a gtx1060 btw:


#713

Surely @dale or @stevebaer know how to do this. Or if there’s no function to do it right now they could surely create one. :slight_smile:


(Dale Fugier) #714

https://mcneel.myjetbrains.com/youtrack/issue/RH-39091


#715

Hmm… Strange things happen that no ti 10x series performes better than ti - can it be that drivers aren’t fully optimized yet?

Strangest thing here is that 1060 gets 30K while 1070 only 20K. One thing i know now - nothing beats quadro in viewports in case of huge scenes.


#716

yes, that’s weird, i noticed that too.
also I had an RX460 (scroll up) in the same system before i swapped it against the GTX1060 and the RX460 had a little higher score than the GTX1060 - that’s also very weird.
especially because the GTX1060 feels like a noticeable improvement over the RX460 - of course that’s how it should be.


#717

AMD don’t limit their gaming cards OpenGL performance. But they do have a history of bad OpenGL drivers with much lower AA quality than what GeForce and Quadro has. (This might have changed lately though)