Meltdown and Spectre - Will patching effect us ?

As most will have probably be aware 2018 has brought about new security risks with our CPU’s.

Operating systems are getting patches, cpu’s will be getting new microcode etc to mitigate against the deviant behavior of some.   To most user’s this will be routine maintenance and will see no/little difference once patched. 

However there are reports of certain specialist workloads appear to be heavily effected by these security fix’s.  It appears mostly applications that have many small writes & context switching such as compiling, databases and some image manipulation.   Real-time applications and regular rendering appear to take a very small/insignificant hit if any.  It’s early stages, and will take a while for hardware vendors to release the patches.

I’m curious if and to what extent if any the performance impact of these patches may effect capturing reality.

Now a Shameless Bump.

If you wish to benchmark a data set before & after patching.  It would be interesting if you can share your results and the status of your patching.  As far as I am aware the OS patches will likely be forced upon us soon, however the bios/microcode may need to be manually applied when available.  Each may have a different impact. (as of the time of this post few vendors have released the microcode patches, and maybe a few weeks until so.)

Alpha 0.43 Can be downloaded here


The Benchmark requires CLI access, so either a demo, steam demo, or the full license. (promo is no go - sorry).
You can quickly install the steam demo alongside the promo, without issues.

Instructions:

1) unzip

  1. Choose your images and place them inside the newly created /CapturingRealityBenchmark/Images folder
    For now I’d suggest a smaller collection that you know work.  - None are currently included.

3) Run the  benchmark.bat  file
You will be asked to enter a identifier/nickname at the start, asked for a note if you wish, and to select the drive you use as the cache disk.

  1. Sit back and relax 

5) Once the benchmark has run it’s course you can review the results in the results.txt file which should pop up automatically.  And look similar to this.

Feedback on the benchmark tool itself is also appreciated.   If you have any, positive or negative please post in this thread.

https://support.capturingreality.com/hc/en-us/community/posts/115001227911-Hardware-Optimisation-Benchmarking-Shenanigans-?page=4#comments 

Many times ago i was made synthetic dataset that can be used for benchmarking, quality checking and other.

486 images, 24Mpx, Full 360 degree with angle not more than 10 degree between all loops and images. Perfect images with strong texture and without any DOF because synthetic dataset (3D rendered)

Images was specially randomized to prevent any kind of sequence align optimized algorithms!

https://drive.google.com/drive/u/2/folders/0BxVZA6GiS36gbEptcDB5QmhBdFE

*_mask.tif file specially added for other software that required masks (like photoscan that just can’t align this dataset without them)

some benchmark in RealityCapture made couple of months ago:
(Align and meshing settings was reset to default)

i7 6700 @ 3,4Ghz, Skylake
64Gb DDR4
Asrock Z170M
Gigabyte GTX 960 (2Gb)
from HDD

Aligning: 743,052 sec (12.38 min)
Meshing High: 19153.462 sec (5.32 hours)
(no region was defined)

i9 7620X @ 2.90Ghz
64Gb DDR4
Gigabyte X299 AORUS Gaming 7
Double (in SLI) Gigabyte GTX 1080 Ti
from Intel SSD

Aligning: 378.169 sec (6.31 min)
Meshing High: 9293.027 (2.58 hours)
(no region was defined)


 

And more interesting.

New tests made before KB4056892 but yesterday was faster (on i7 system) than couple of month ago.

 

And even more interesting after KB4056892 fix speed was INCREASED on i7 for both steps (Align and Meshing)

And only on i9 Align was slower than test couple of months ago, but meshing much faster.

 

Align: before: 661sec
after: 630sec

Meshing High:
Before: 11673sec
After: 10837sec
(i7 6700 3,4Ghz, Skylake 64Gb DDR4 Asrock Z170M)

I’ve not noticed a difference.