Luxmark v4.0alpha1 (Nov 2020) - OpenCL Benchmark

[eep357]

Well, you were partially right. I didn't modify the benchmark, but it appears as though when I reset my BIOS to stock yesterday while having my prior overclock values in Clover's config.plist it caused a mismatch between internal core clock and what OSX what expecting, results=inflated benchmark scores for CPU & GPU, with no real world performance gains (actually everything running slower, even video which gives a FPS boost cause seconds last longer allowing more frames, but it can be hard to see difference) Since I have no interest in cheating benchmarks, I would have to say all those benches from yesterday are invalid. 7970 still kills it without cheating

[k3nny]

To freshen up this thread a bit:

 

[gils83]

 

Maveriks 10.9.1  13B42

 

 

 

 

 

 

 

 

 

 

 

 

Edited December 18, 2013 by gils83

[blacksheep]

Managed to squeeze 1250 MHz from my 7950 and it kicks butt in Luxmark. Pretty close to R9 280x I owned before. Did it only for benches, normally it runs @1175 MHz – balanced performance/heat/loudness setting.

 

[gils83]

 

 

 

 

 

At the next ........

 

 

[Fangio]

So far 10.10 is a step backwards performance-wise.. lost ~170 pts in Sala with DP2, compared to Mavericks. But the jury is still out

 

[Fangio]

Some progress here... in DP5, for the first time Yosemite is slightly faster than Mavericks

 

[Derek12]

ATI HD5450 rulezz lol

 

In Windows I get 50

 

[adgenet]

Yosemite PB1
Radeon R9 290X

[npwski]

OS X 10.10 "Yosemite" DP5 build 14A314h

Nvidia GTX Titan @~1130Mhz

 

Luxmark 2.0 scene Sala = ~1650 marks

[RobertX]

asus hd 6570(low profile) and intel q9400 Mav 10.9.4

 

 

 

  

[gils83]

 

 

[gils83]

 

 

[mitch_de]

The devs of Luxmark working on new Luxmark 3.0 - is in eary stage and not comperable with LM 2.0 results (new written renderer, new scenes!)

You can try: http://www.luxrender.net/release/luxrender/dev/mac/LuxMark_3.0_alpha1_OSX_64.zip

 

For me with 1GB VRAM, only the small scene (ball) works - others give gpu memory error.

( RUNTIME ERROR: The InfiniteLight distributions buffer is too big for GeForce GT 440 Intersect device (i.e. CL_DEVICE_MAX_MEM_ALLOC_SIZE=268353536): try to reduce related parameters)

Would be interesting if its an bug or other GPUs with >= 2 GB or 3 GB  VRAM can handle the two other scenes.

 

 

LM 3.0 A1 : GT 440 1 GB  (GPU Only)

Luxball GPU

 

Both others CPU only (due GPU 1GB VRAM to less= error message running GPU only)

 

 

[gils83]

 

 

[gils83]

 

Yosemite 10.10.1 

 

 

[mitch_de]

Thanks! Hope we will see some with 2GB+ VRAM running the two more complex scenes

WIth 1GB VRAM only Luxball scene works (or you use CPU only)

[Riley Freeman]

4GB VRAM here. No memory error but with the other two scenes the kernels fail to compile on the GPU. They compile fine (and in a much shorter time) on the CPU.

 

 

 

2014-12-13 01:28:19 - [LuxCore] [618.214] [PathOCLBaseRenderThread::1] Compiling kernels

2014-12-13 01:30:44 - [LuxCore] [762.883] [PathOCLBaseRenderThread::1] PathOCL kernel compilation error ERROR clBuildProgram[CL_BUILD_PROGRAM_FAILURE]: Error getting function data from server

2014-12-13 01:30:44 - RUNTIME ERROR: PathOCLBase kernel compilation error

[fantomas]

 

 

[mitch_de]

Thanks! How much VRAM your GTX 570 have?

Luxball GPU mode much faster than my 430 GT  

I see you run both bigger scenes in CPU Mode (also error like me with GT 430=less VRAM?)

[fantomas]

it has 1,3GB VRAM  

 

yes,  the same error like you...

 

indeed, it's the first alpha release, maybe things will work better after but I think these bigger scenes are for 3GB / 4GB VRAM graphics cards  

[crusher]

 

 

 

[mitch_de]

V30Beta1 available (DL link first posting)

[Gumuskalp]

[Riley Freeman]

V30Beta1 available (DL link first posting)

 

No change here. Still crashes on compiling kernels.