FAH-GPU2 beta client is available!

Here's a great article/interview which goes into some depth around the performance and capability of the new GPU2 client.

http://www.extremetech.com/article2/0,1697,2284064,00.asp

 

Is there a chance of a Linux GPU2 client any time soon? I'm keen to get back into it with RV670, but I'm not really too keen on running Windows.

 

ATI Core v1.10
New larger WU (4742-4747 1254 atoms)

As bad as Fur Mark.

I get almost 100% GPU activity.

 

Here some interesting facts:
http://foldingforum.org/viewtopic.php?f=51&t=4879&start=45#p50698

It's really sad, that my X2 3600+ cannot feed my HD3850 (current project 4729), so the GPU activity is around 65 % and the HD3850 doesn't run with full speed (520/830 MHz instead of 670/830 MHz).

 

Grats on the large performance increase with big proteins.

I can't help but be a little curious though, why is the CPU so significant for the card with small proteins in GPU2 on ATI, when the competition runs stuff like p5005_supervillin_e1 at full steam (also quite a lot faster in general) with only about 15% CPU use on a typical modern dual core?

 

Are kernel launches in Brook/CAL asynchronous? I haven't looked at it in a few months, so I can't remember.

 

No offense intended, but the X2 3600+ was never a powerhouse of a CPU, even when it was released. It is an entry-level dual-core CPU. If you need more performance, you can overclock, or if it's in your budget, replace it.

 

Sorry to put you on the spot, Mike, and not truly expecting an answer, but here goes:

any chance we can expect to see those 1-1.2 (or even 2.4) TFLOPs put to good use soon and give similar returns by way of points per day compared to GT200? It's kind of embarrassing that even my o/c'd 8800 GT can out-produce a 4870 X2

 

So the gap will narrow signficantly. Well that's good. I hope to someday see greater performance out of Radeons, though.

 

On my 4870 I was getting 3600PPD according to FahSpy running a 4747 WU, currently on 2600PPD with a 4715 WU.

On the 1.04 & Cat8.7 it was only doing about 2000PPD on the small units.

Edit: got another 4747 WU & its doing about 3780PPD

 

You mean native support for the X2/GX2 cards? As opposed to runinng two separate WUs on each core? That would be a nice feature, allthough not absolutely nescessary.

AFAIK, currently the problem with the ATI 4870 X2 using only one core is at he driver/hardware level.

 

In the last four days I had two large proteins (4742&4743) and four or five small (eg 4715).

 

The other problem is one can't disable crossfire on X2 type boards. I believe Crossfire has to be disabled in order to get two clients working simultaneously.