Holomark 2 Released!


#779

Thanks @Holo for letting me test the beta Holomark V6 test. looking forward to version 3 :+1: . Hopefully others will find this interesting including the good folk at McNeel :beers:


Holomark V6 Milezee.txt (1.9 KB)


(Nathan 'jesterKing' Letwory) #780

Oi! No Raytracing to burn those cores on the CPUs and GPUs?

I propose GPU_22 and CPU_09.


#781

Nope, this is just Holomark2 for Rhino6 :slight_smile:
Holomark3 should have some raytracebenchmark where we let it render for 30 seconds and then score based on how many iteration it made. Holomark 3 will focus on running the test in a “fixed” time and gather data.

But I could add some tests to Holomark2_R6 for fun… do you have some code I could use to set number of iterations and number of cores to use, and a way to know when it is done, so I don’t have to monitor the CPU usage (I have code for that already, I used it for Neon, but not in Holomark)


(Nathan 'jesterKing' Letwory) #782

The best would be to do a Raytraced viewcapturetoclipboard or viewcapturetofile to bypass wasting computing cycles for updating the viewport(s) - and it would automatically block until done. But if you don’t want that you could use for instance code I showed to @Jarek here Help - Cycles settings

specifically the following code snippet (for those who are lazy enough to not click through to the discussion):

import scriptcontext as sc
import Rhino

rtdm = sc.doc.Views.ActiveView.RealtimeDisplayMode

# try-except block is needed because a non-realtime displaymode
# will give an internal, non accessable wrapper class, and throw an
# exception
# for any other realtime implementation this should work - like ProRender
try:
	print(rtdm.IsCompleted())
	print(rtdm.Paused)
	print(rtdm.Locked)
        if rtdm.MaxPasses < 1500:
	    rtdm.MaxPasses = rtdm.MaxPasses + 10
        else:
            rtdm.MaxPasses = 100
	print(rtdm.LastRenderedPass())
except:
	print("Active view not a realtime display mode")

Setting sample count I think can be done through MaxPasses property.

The function call LastRenderedPass on the rtdm would yield you the last rendered pass - in case you want to do it in the viewport and pause after your 30 seconds of patience span.

I have here a regression.py that creates for me 960 tests, raytraces them all (with viewcapturetofile), capture a rendered version still, and remember the times for each and everyone of them.

I’ll dig up some more code later to see if it is of any use for you. But I think with what I’ve showed you here, and in the thread with @Jarek you may have already enough to get started.


#783

And how long does that typically take? I’ll bet it’s a lot more than 30 seconds!


(Nathan 'jesterKing' Letwory) #784

Depends on the settings. The test run I did the other day was with resolution 2400x1200 @ 25 samples. The final report says 960 tests (1920 renders) completed in 14101.9095154 seconds. I run this with every major change to shading / core Cycles.

It’ll be interesting to see how @Holo is going to use the code snippets for the Raytraced benchmark (:

add: screenshot of the awful result page generated after running the tests


#785

You mean like in this rough test that I now have up and running where it first tests the CPU and then the GPU:
image
:slight_smile: I don’t know what I should do if more or less items are available though. If a GPU isn’t good to go, how to i know? I mean, I use rs.Command(“RhinoCycles_SelectDevice 1”) to set it to GPU.


#786

And how about if you detect more than one GPU? Would you test each individually and then test using them all? So far probably Nathan is the only one who might want this. :slight_smile:


(Nathan 'jesterKing' Letwory) #787

If you have the ability to parse the output of the command history you could do so for the command RhinoCycles_ListDevices.

I’d dare say all users want that info - a good collection of benchmarks for as many devices as we can get our hands on.


#788

I was just referring to the fact that you’re probably the only one with multiple high-performance GPU’s. I mean, who cares how an an Intel integral GPU performs on raytracing? :smile:


#789

Wow for those scores.
new i9 9900k flagship is getting my attention for good.
And if that P2000 is good i can imagine the P4000
:slight_smile:


(Nathan 'jesterKing' Letwory) #790

I do have an Intel HD Graphics 630 and a GTX 760 in my dev machine, too. (Besides the GTX 1060, and the Radeon Pro WX9100, but that’s not the point :wink: ).


#791

yeh we just had 3 new workstations custom built, I’m certainly no expert at specifying PC builds but did a little research, spoke with the company who build, and voila, they seem to be running our models fine, Rhino, Vectorworks, Autocad etc. We got these built and tested, with 3 year warranty for just under 2k each (£GBP) :+1:


#792

do you bought it from Puget?


#793

No bought from Scan , UK Bolton based specialist :+1:


#794

Do they have a web page?


#795

Yeh, https://www.scan.co.uk


#796

What about Holomark for Rhino 6.0?

any news?

Thanks !


#797

Have fun:


#798

Hi.
Is there any latest version of Holomark2 for Rhino5 link to download?
Cheers