Since it was eluded that Beta GPU is an additional opt-in [not going to show until GPU beta arrives], you'll have to be there when the announcement is made to sign up. This goes for Beta *and* production i.e. even if the "sign me automatically up for new project", there will be 1) an explicit sign up for the GPU production and 2), an unknown, but likely a "Sign me automatically up for new GPU projects". The old auto-signup would then get the CPU added to the line.

All to prevent that those that *use* their machines do not get an unwanted surprise in the office, the school benches or at home.

--//--

The Seti @ Home project allows for GPU crunching. Seti runs on the BOINC platform , just as Help Conqour Cancer.

There is a page on Seti @ Home that rates the best performing GPU cards. I wonder if the cards would get the same ranking on the future GPU project on HCC.

http://setiathome.berkeley.edu/gpu_list.php

In case you are wondering, the top 30 model ratings are listed below. The parethasis lists the performance, with 1.00 being the best and any other smaller value being less productive.

1.(1.000) GeForce GTX 480
2.(0.979) GeForce GTX 580
3.(0.871) GeForce GTX 590
4.(0.789) GeForce GTX 570
5.(0.709) GeForce GTX 470
6.(0.659) GeForce GTX 560 Ti
7.(0.599) GeForce GTX 460
8.(0.566) GeForce GTX 550 Ti
9.(0.519) GeForce GTX 465
10.(0.411) GeForce GTS 450
11.(0.386) GeForce GTX 560
12.(0.385) GeForce GTX 285
13.(0.281) GeForce GTX 275
14.(0.277) GeForce GTX 260
15.(0.276) GeForce 9800 GTX/9800 GTX+
16.(0.262) GeForce GT 440
17.(0.245) GeForce GTX 295
18.(0.222) GeForce 8800 GT
19.(0.184) GeForce GT 240
20.(0.183) GeForce 9800 GT
21.(0.150) GeForce 9600 GT
22.(0.141) GeForce GTS 250
23.(0.115) GeForce GT 220
24.(0.089) GeForce 8600 GTS
25.(0.075) GeForce GT 430
26.(0.062) GeForce 210
27.(0.060) GeForce 9500 GT
28.(0.060) GeForce 9400 GT
29.(0.032) GeForce 8400 GS
30.(0.031) GeForce 8600 GT

The bottom card on that list, the 8600GT, has OpenCL 1.0 compute capability and 256Mb of RAM. If I remember rightly HCC GPU is going to be OpenCL 1.0. If I'm wrong there, someone please correct me.
If it's capable of handling the HCC GPU units when they hit it's not exactly going to be a stellar performer, relatively speaking, but the amount of memory is going to be the main limitation rather than OpenCL compatibility.

Those SETI rankings may not be very accurate. Judging from my experience with the same statistics at POEM for ATI/AMD cards the way that performance index is calculated appears to be suspect. It does not take into account the number of concurrent tasks and so the lower performing models that are not as throttled and have less capacity run fewer concurrent tasks or even only one task at a time. So the runtime can be lower resulting in the performance index being higher. The higher performing cards do 4-6 tasks concurrently so the performance index rates them lower when the runtime is higher.

For example when my 5870 does one POEM task at a time the GPU load is low and the runtime is about 30 minutes. When it does 4 concurrent tasks the GPU load is almost twice as high and the runtime is about 39 minutes. The single task configuration reports the better processing rate even though it takes 30 minutes to complete a task compared to average computation time of less than 10 minutes per task when four tasks are processed at once.

GPU application performance reporting can be seen in Computers on this account>Details>Application details>Average processing rate. The number of concurrent tasks is not used in the calculation only runtime.

It is also unknown how often such rankings are updated and whether all recent models are correctly identified and included. For example the rankings for MilkyWay and POEM currently do not include any Cayman class cards at all.

Reference performances of some NVidia Compute Capable1.3, CC2.0 and CC2.1 cards (all OpenCL capable):

GTX 590 GF110 40nm Compute Capable 2.0 2488 GFlops peak
GTX 580 GF110 40nm Compute Capable 2.0 1581 GFlops peak
GTX 570 GF110 40nm Compute Capable 2.0 1405 GFlops peak
GTX 480 GF100 40nm Compute Capable 2.0 1345 GFlops peak
GTX 560 Ti 448 GF110 40nm Compute Capable 2.0 1311 GFlops peak
GTX 560 Ti GF114 40nm Compute Capable 2.1 1263 GFlops peak
GTX 295 GT200b 55nm Compute Capable 1.3 1192 GFlops peak
GTX 470 GF100 40nm Compute Capable 2.0 1089 GFlops peak
GTX 560 GF114 40nm Compute Capable 2.1 1075 GFlops peak
GTX 460 GF104 40nm Compute Capable 2.1 907 GFlops peak 768MB
GTX 460 GF104 40nm Compute Capable 2.1 907 GFlops peak 1GBk
GTX 465 GF100 40nm Compute Capable 2.0 855 GFlops peak
GTX 285 GT200b 55nm Compute Capable 1.3 695 GFlops peak
GTX 550 Ti GF116 40nm Compute Capable 2.1 691 GFlops peak
GTX 275 GT200b 55nm Compute Capable 1.3 674 GFlops peak
GTS 450 GF106 40nm Compute Capable 2.1 601 GFlops peak
GTX 260-216 GT200b 55nm Compute Capable 1.3 596 GFlops peak
GT 545 GF116 40nm Compute Capable 2.1 501 GFlops peak

It's worth noting that GPU performances depend on the app and driver. Expect an app that works well on CC2.0 to favor CC2.0 at the expense of CC1.3 and CC1.1 cards in particular. Don't expect the same performance from different drivers, and wait and see how different cards perform before buying, and don't buy before mid Apr.