Remembering the Past

Post your tests, experiments and unfinished renderings here.
Post Reply
User avatar
FarbigeWelt
Donor
Posts: 469
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Remembering the Past

Post by FarbigeWelt » Thu Jan 31, 2019 7:58 am

Continuing an old story, based on a vague memory of a gadget made from a magnetic base a ferric spheres, I found this intermediate.
2019-01-31 (1).png
Path Render, or knows anybody pointing the right way?
160.8 | 42.8 (10.7) Gfp / Windows 10 Pro, intel i7 4770K@3.5, 32 GB | AMD R9 290x+R9 390x, 4 GB
17.3 | 19.0 ( 4.7) Gfp / macOS X 13.6, iMac 27'', 2010, intel i7 870@2.93, 24 GB | ATI Radeon HD 5750, 1 GB
#luxcorerender | Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Posts: 469
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Remembering the Past, updated

Post by FarbigeWelt » Fri Feb 01, 2019 7:55 am

Respecting first acceptable screen resolution of home PCs, I have chosen 800x600 base resolution and rendered a 300% 2400x1800 picture.

On my PC with two amd R9 x90X graphic cards and current recommended adrenaline 19.1.1 driver this took already a while, total path depth 40, clamped around 30'000.
Wiki lists for R9 390X, floating point single-prec. (IEEE-754 32 bit) 5,9 TFlops and for double-prec. (IEEE-754 64 bit) 0.74 TFlops, memory bandwith 384 GByte/s; the R9 290X is about as fast at 90% of 390X.
Which Way_You decide on your own_GPU Path_Sam 10'000_Sob 0.55_FBH 1.25_3200x2400_AutoBr.jpg
Which Way_You decide on your own_GPU Path 40_Sam 10'000_Sob 0.55_FBH 1.25_3200x2400_Lin AutoBr
Because I was not so happy with scenes draft version I added glass spheres, changed floor material form matte to glossy with some roughness, added four area lights, red, green, blue and amber (once a promising approach to mix white light for setable blackbody temperature) and moved spot light to the center above top of the sculpture.

The next picture rendered almost 7 hours on CPU intel 64 bit i7 4770K 3.5 GHz (4+4 cores) speed unclear yet ~0.17 TFlops, BiDir depth 80, Metropolis and tonemapper Reinhard.
Which Way_You decide on your own_CPU BiDir_Sam 3963_Metro 40,65536,5_Depth 60,80,40,80_FBH 1.25_2400x1800_6h40m.png
Which Way_You decide on your own_CPU BiDir_Sam 3963_Metro 40,65536,5_Depth 60,80,40,80_FBH 1.25_2400x1800_6h40m
160.8 | 42.8 (10.7) Gfp / Windows 10 Pro, intel i7 4770K@3.5, 32 GB | AMD R9 290x+R9 390x, 4 GB
17.3 | 19.0 ( 4.7) Gfp / macOS X 13.6, iMac 27'', 2010, intel i7 870@2.93, 24 GB | ATI Radeon HD 5750, 1 GB
#luxcorerender | Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Posts: 469
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Remembering the Past, the scene file

Post by FarbigeWelt » Fri Feb 01, 2019 8:03 am

I guess the are some tricks to improve rendering speed for the scene, with out loosing visible depth in reflections of the spheres in 4800x3600 pictures.

Feel free to download and tweak the scene.
Simple Remember Scene.zip
Free usable scene file of You decide on your own. (remarks: on your own path, there are some in the mirror, aren't there? Or is all illusion of choice only?
(398.74 KiB) Downloaded 22 times
160.8 | 42.8 (10.7) Gfp / Windows 10 Pro, intel i7 4770K@3.5, 32 GB | AMD R9 290x+R9 390x, 4 GB
17.3 | 19.0 ( 4.7) Gfp / macOS X 13.6, iMac 27'', 2010, intel i7 870@2.93, 24 GB | ATI Radeon HD 5750, 1 GB
#luxcorerender | Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Posts: 469
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Remembering the Past, GPU 4800x3600

Post by FarbigeWelt » Fri Feb 01, 2019 6:16 pm

Takes some time with total path depth of 80, doesn't it?
Which Way_You decide on your own_GPU Path_Sob 0.60_Sam__10'000_depth _4800x3600_VRAM 831 M_5h56m_Reinhard.jpg
Which Way_You decide on your own_GPU Path_Sob 0.60_Sam__10'000_depth _4800x3600_VRAM 831 M_5h56m_Reinhard
Both, test result discussion and conclusions follow soon, as well as some meta ideas behind the scene.
160.8 | 42.8 (10.7) Gfp / Windows 10 Pro, intel i7 4770K@3.5, 32 GB | AMD R9 290x+R9 390x, 4 GB
17.3 | 19.0 ( 4.7) Gfp / macOS X 13.6, iMac 27'', 2010, intel i7 870@2.93, 24 GB | ATI Radeon HD 5750, 1 GB
#luxcorerender | Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Posts: 469
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Remembering the Past, Resolutions and Relations

Post by FarbigeWelt » Sat Feb 02, 2019 1:52 am

My analysis requires a few intermediate steps, like this one, the collection of historical data.
Remember the Past_Resolutions and Relations_Info 2.png
Remember the Past_Resolutions and Relations_Info 2
The illustration is 50% reduced. Original resolution cannot be uploaded yet even with jpg size of 5 MB.
Remember the Past_Resolutions and Relations_50pc.jpg
EDIT 1: Due to layout issues, code part replaced with an image.
EDIT 2: bits per on screen image corrected and considered for theoretical speed comparision.
160.8 | 42.8 (10.7) Gfp / Windows 10 Pro, intel i7 4770K@3.5, 32 GB | AMD R9 290x+R9 390x, 4 GB
17.3 | 19.0 ( 4.7) Gfp / macOS X 13.6, iMac 27'', 2010, intel i7 870@2.93, 24 GB | ATI Radeon HD 5750, 1 GB
#luxcorerender | Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Posts: 469
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Remembering the Past, examples CPUs since 1982

Post by FarbigeWelt » Sat Feb 02, 2019 5:02 pm

Intels performance for Core ix's floating points/cycle ratio is impressive compared with AMD's current Ryzen's performance but AMD is probably in the be the better position with already released 7 nm products while intel postponed market launch of their 10 nm Core ix CPUs again. Moore's law or better intels hypothesis about numbers of transistors depens on structure size development and latter defines electrical resistance together with useable frequency what also means efficiency respective dissipation. As long as decreasing structures' size was 'simple' (besides to achivewith every step smaller particle sizes and counts in clean rooms, chemicals and also reducing abrasion of instruments and processes) companies' roadmaps followed intel's hypothesis as if there were a bounty to get. The cards are shuffled and freshly dealt. May the player with smarters moves get ahead of next round, the Dissipation Management.
Example CPUs since 1982.png
160.8 | 42.8 (10.7) Gfp / Windows 10 Pro, intel i7 4770K@3.5, 32 GB | AMD R9 290x+R9 390x, 4 GB
17.3 | 19.0 ( 4.7) Gfp / macOS X 13.6, iMac 27'', 2010, intel i7 870@2.93, 24 GB | ATI Radeon HD 5750, 1 GB
#luxcorerender | Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Posts: 469
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Remembering the Past, example GPU since 2006

Post by FarbigeWelt » Sun Feb 03, 2019 1:38 am

Before 2006 graphic cards for gaming did not support double precision (32 bit) floating point operation.
But since then the cards computing power increased rapidly and reached within a few years TFLOPs level much more oeconomicly than CPUs.
Example GPUs since 2006.png
Example GPUs since 2006
160.8 | 42.8 (10.7) Gfp / Windows 10 Pro, intel i7 4770K@3.5, 32 GB | AMD R9 290x+R9 390x, 4 GB
17.3 | 19.0 ( 4.7) Gfp / macOS X 13.6, iMac 27'', 2010, intel i7 870@2.93, 24 GB | ATI Radeon HD 5750, 1 GB
#luxcorerender | Gfp = SFFT Gflops

Post Reply