Luxmark results database

[inferrna]

There are a lot of garbage in top result tables. It may be good to show only best result for each user@device. Otherwise results of one user will fill all table.

http://www.luxrender.net/luxmark/top/top20/Sala/GPU/1
4 1805 2012-02-27 06:14:25 GPU 1 HD7970 zackmack
7 1611 2012-02-19 06:06:20 GPU 1 HD7970 zackmack
8 1557 2012-01-25 02:35:02 GPU 1 HD7970 zackmack
13 1202 2012-01-20 13:12:33 GPU 1 GeForce GTX 480 [15 units @ 1500MHz] nun1980
14 1141 2012-01-12 13:31:19 GPU 1 GeForce GTX 480 [15 units @ 1401MHz] nun1980
15 1139 2012-01-10 06:36:37 GPU 1 GeForce GTX 480 [15 units @ 1401MHz] nun1980

[Dade]

There are a lot of garbage in top result tables. It may be good to show only best result for each user@device. Otherwise results of one user will fill all table.

http://www.luxrender.net/luxmark/top/top20/Sala/GPU/1
4 1805 2012-02-27 06:14:25 GPU 1 HD7970 zackmack
7 1611 2012-02-19 06:06:20 GPU 1 HD7970 zackmack
8 1557 2012-01-25 02:35:02 GPU 1 HD7970 zackmack
13 1202 2012-01-20 13:12:33 GPU 1 GeForce GTX 480 [15 units @ 1500MHz] nun1980
14 1141 2012-01-12 13:31:19 GPU 1 GeForce GTX 480 [15 units @ 1401MHz] nun1980
15 1139 2012-01-10 06:36:37 GPU 1 GeForce GTX 480 [15 units @ 1401MHz] nun1980

Yup, I'm working on a new minor release of LuxMark (i.e. v2.1 http://src.luxrender.net/luxrays) with a couple of bug fixes and a minor new features (i.e. command line options for batch testing). We can probably develop a minor new update for the web side code too.

If anyone else as some idea for v2.1, it is a good moment to speak

P.S. my preferred LuxMark list, at the moment, is this one: http://www.luxrender.net/luxmark/search/avg/Sala

[whyoh]

would it be possible to record more details to help us pin down what's causing the differences between scores on apparently identical systems? i thought:

CPU details and usage during GPU-only runs.
overclocking details
cpu 'turbo' details (although i'm pretty sure we can assume "On")
motherboard make and model
memory channels and timings

i have no idea how much difference those things really make - or whether they're more significant in SLG, LuxMark or LuxRender.

on the web side i'd appreciate two or three more options on the search:

format=[table|csv|json|xml] so we can download complete results
after=[date] | [result ID] so we can fetch results received since we last checked

also on the web side it'd be nice to be able to search for "any of these words in the device name" so i can type in "juniper HD5770 HD6770" and get all the results for that GPU.

hope that's useful and not too late!

thanks for *lux*

[Dade]

would it be possible to record more details to help us pin down what's causing the differences between scores on apparently identical systems? i thought:

CPU details and usage during GPU-only runs.
overclocking details
cpu 'turbo' details (although i'm pretty sure we can assume "On")
motherboard make and model
memory channels and timings

The main problem is finding a way to extract the above information across all platform supported (i.e. all the versions of Linux, Windows, MacOS; hardware from AMD, NVIDIA, Intel, etc.). It is going to require a quite large amount of work and OS/Hardware specific know-how.

It would be nice to be able to access and merge the information collected by tools like GPU-Z and CPU-Z but I'm afraid they are closed source.

on the web side i'd appreciate two or three more options on the search:

format=[table|csv|json|xml] so we can download complete results
after=[date] | [result ID] so we can fetch results received since we last checked

also on the web side it'd be nice to be able to search for "any of these words in the device name" so i can type in "juniper HD5770 HD6770" and get all the results for that GPU.

hope that's useful and not too late!

They are good idea, I hope to add them in the future.

BTW, not only LuxMark is open-source but also the WEB side code is open-source too. It is available here: http://src.luxrender.net/lrmdb/

The repository includes the source code for LuxRender WEB material database and LuxMark result database. At the moment, no one has ever written a documentation about that code and on how to setup a test environment however Dougal2 and me can answer to related questions. Like in any other open-source project, everyone can contribute with good idea and extend existing code.