with the 7950 you have 717 GFlops DP Performance
but with the RX470 you get only 365 GFlops DP Performance
AMD has also reduced the DP Performance by the mainstream Cards.

And also reduced consumptions: 7950-200W, Rx470-120W.
717/200=3,58
365/120=3,04
Not a big difference.
And RX470 will be better with new drivers

It might be worth waiting for Vega 11.
http://wccftech.com/amd-vega-10-vega-11-magnum/

I just have read an article on tomshardware.de. AMD is working on a new firmware for the current gpu's, reducing power consumption and increasing performance.
http://www.tomshardware.de/polaris10-polaris11-radeon-rx-update-effizienzsteigerung,news-257105.html
Definitely too early to invest.in a new gpu.

with the 7950 you have 717 GFlops DP Performance
but with the RX470 you get only 365 GFlops DP Performance
AMD has also reduced the DP Performance by the mainstream Cards.

And also reduced consumptions: 7950-200W, Rx470-120W.
717/200=3,58
365/120=3,04
Not a big difference.
And RX470 will be better with new drivers

Let's see how the gpu load is.

with the 7950 you have 717 GFlops DP Performance
but with the RX470 you get only 365 GFlops DP Performance
AMD has also reduced the DP Performance by the mainstream Cards.

And also reduced consumptions: 7950-200W, Rx470-120W.
717/200=3,58
365/120=3,04
Not a big difference.
And RX470 will be better with new drivers

Let's see how the gpu load is.

I've just seen a mistake in my prior post the RX470 has only 309 GFlops. The RX480 has the 365.

With the TBP and TDP values you hav to be carefull. DP Calculation is not mainstream gaming ist goes to "Maximallast"
In Milkway the 7970 takes the full 350 Watts.

Im MW you can push up the GPU Load to 100 % when you calculate 2 - 4 Task in parallel. This is possible in many othe projects. (I've done this im MW and Moo Wrapper).
What elektric power an RX470 or 480 taks in MW mayby the MW guys know more.

Running 2x cards to get the same performance as one isn't ideal either. There is the system overhead cost and power consumption to think about. Motherboards and CPUs with full 16x lanes for 2-3 cards start to get expensive. Trade-offs either way.

Any idea how long before I can get GPU tasks to my AMD from Universe? All I've got left is SETI and Einstein, and some occasional Milkyway. It wants more :-)

In november this year, I will start the tests of main routine's code for my engine (clrunengine). I don't know how long that code will be tested, to be good and correctly work. Preliminary (current) version of tool to porting code to GPU was too much enough mature for this hard work. I need to fix and improve tools (mainly clrungen - generator) to make next leap. Ofcourse, I can compile OpenCL generated by clrungen to GPU (Radeon).However this code do not work correctly and give wrong results. In other words, just I need still maintain my tools, before make GPU application.

Take your time. But in the meantime I am wondering whether all the projects (BHSpin V2, Ultraviolet reionization, etc.) will be ported over to the GPU, or only some of them. Maybe it is too soon to answer that, or maybe you know already what the GPU app is supposed to do.

In november this year, I will start the tests of main routine's code for my engine (clrunengine). I don't know how long that code will be tested, to be good and correctly work. Preliminary (current) version of tool to porting code to GPU was too much enough mature for this hard work. I need to fix and improve tools (mainly clrungen - generator) to make next leap. Ofcourse, I can compile OpenCL generated by clrungen to GPU (Radeon).However this code do not work correctly and give wrong results. In other words, just I need still maintain my tools, before make GPU application.

Will the tests be run on your own equipment or on our computers? I have selected to allow the running of test applications.

These are internal tests of code of application. Before any alpha/beta test versions. And these tests will be done on my hardware. After successful fixes I will try to build the first version for open alpha/beta tests.

After some reconsiderations, I will be trying to port all code of application step-by-step with checking results and performance. I wrote some simple tools to testing code and intermediate results. Now, I am prepping simple tool to simplify writing program structure for my engine. I will update progress of that works.

The first small excerpt of the code has been ported to GPU. This is a part of main code of application. Moreover, we have tools to test that code and that code passed tests successfully. We are testing this code on the real data (workunits) with using original code with some modifications. More info about progress of the work coming soon.

The first small excerpt of the code has been ported to GPU. This is a part of main code of application. Moreover, we have tools to test that code and that code passed tests successfully. We are testing this code on the real data (workunits) with using original code with some modifications. More info about progress of the work coming soon.

Thanks for the update, keep up the fun/hard/exciting work :-)

After some works over clrunengine and extra stuff, I am ready to porting next code to GPU. I will write more details later.

Ready for beta testing!!! :-)

Ready for beta testing!!! :-)

And me!

I noticed that there is a beta GPU app for OpenCL. Is there work available for this?

Where did you find that?

Applications page...