[fellix]

CLBenchmark 1.1 Desktop Edition is an easy-to-use tool for comparing the computational performance of different platforms. It offers an unbiased way of testing and comparing the performance of implementations of OpenCL 1.1, a royalty-free standard for heterogenous parallel programming.

CLBenchmark compares the strengths and weaknesses of different hardware architectures such as CPUs, GPUs and APUs. The test results are listed in a transparent and public OpenCL performance database.

CLBenchmark 1.1 Desktop Edition supports any standard-compliant OpenCL 1.1 implementation and it is compatible with every major vendor’s solution.

Homepage

[Jawed]

First time I've actually ran anything OpenCL on my new system.

If you browse their results database you'll see my HD7770's scores.

[fellix]

I just wonder how many sub-tests of the suite are mostly affected by the NV's crooked OCL 1.1 run-time.

This thing simply makes a truly architectural comparison a dull task.

[Davros]

Originally Posted by Jawed

If you browse their results database you'll see my HD7770's scores.

are you the 4x7970 ?

[Lightman]

Originally Posted by Davros

are you the 4x7970 ?

Going from ATi X19x0 would be too much of a speed up to cope I think .
Glad to see Jawed moved from DX9 straight to DX11.1 era.

Now back to benchmark, time to run this on my HD6970!

[Jawed]

Yes, just the 7770. It plays World of Tanks at maximum IQ whereas the X1950Pro could only do minimum (except for texture filtering I think), both at 1920x1080.

The X1950Pro will continue one day when I get XP re-installed on the old PC and set it up solely as a video capture system. That will also enable me to finish Fahrenheit, a game that doesn't like post-XP Windows versions it seems.

fellix, I think it'll be a while before AMD's OpenCL will be "straight", especially for GCN. And I've recently seen results showing AMD OpenCL working better on Intel CPUs than Intel OpenCL (think they were linked around here somewhere). So "crooked" seems very much the norm and things are evolving.

http://www.kishontiinformatics.com/

implies mobile platforms too. Dunno if there are any handsets out there actually capable of running OpenCL though (i.e. runtime ready to go).

[CarstenS]

Great database (no clocks, no system info, nothing - or did I pay not attention?) My GTX480-results are all over the place - sometimes close to the Top10, sometime I need to scroll quite a bit down.

But it's an interesting Benchmark - and at last one, that does run to the end, not like the other CL-Benchmark from that asian HPC-site in the other thread.

[fellix]

Originally Posted by CarstenS

Great database (no clocks, no system info, nothing - or did I pay not attention?)

It's in "Coming soon..." status.

Originally Posted by CarstenS

But it's an interesting Benchmark - and at last one, that does run to the end, not like the other CL-Benchmark from that asian HPC-site in the other thread.

It looks like there's a room for more sub-tests. I hope they'll upgrade it in the future with things like basic arithmetic throughput.
And the units of measure needs to be more relative -- those "points" don't speak much for themselves, although it's good enough for the means of comparison.

[CarstenS]

Hm, I just tried to run it on a HD 6990, the world's fastest graphics card (tm), and the benchmark shows in the dropdown that you can run it on GPU 1, GPU 2 and 2x GPUs.

When selecting the latter, there's only one test to be run, raytracing. And that ones crashed immediately with a display driver reset.
Luxmark 2.0 works as expected though, so probably some quirk in the Benchmark itself.


Update:
WRT to multi-GPU being available only for the Raytracing part: That seems to be true generally, a GTX 590 behaves the same. It, however, finishes the RT test without error.

[CarstenS]

Great, kishonti website seems to be unavailable so I cannot run any more benchmarks now that I would need to.

If anyone at Kishonti reads this: That's a bulletproof concept of how to make sure, people working on a temporal budget will not use your products.

[kishonti]

Originally Posted by CarstenS

Great, kishonti website seems to be unavailable so I cannot run any more benchmarks now that I would need to.

If anyone at Kishonti reads this: That's a bulletproof concept of how to make sure, people working on a temporal budget will not use your products.

Sorry for the inconvenience, the problem is caused by ISP to ISP routing issue. We hope it will be solved soon.

[CarstenS]

That's good to hear, I'll keep my thumbs pressed for you to have it sorted out soon! But please consider letting the benchmark run without having to have Internet access all the time - maybe a timeout after a predetermined time without response from the servers?

[CarstenS]

Works again now - thanks for solving the ISP problems!
(7970 is really fast btw)