I would HIGHLY reccomend buying Intel. Always. If you're buying this GPU for the HCC project when it comes out, I would also wait to buy a GPU when the project starts.

New GPU's come out all the time which drive prices for others down, and since there is no project out yet, it will sit idle.

They seem to run quicker on most tasks. Knreed just stated that there were issues with VINA projects:

In this problem here: http://www.worldcommunitygrid.org/forums/wcg/...ead,32920_offset,0#372147 I'm actually referring to your computer. We had the AMD FX computers in their own HR class, but there weren't enough of them to keep from clogging up SN2S and preventing everyone from getting work. What is odd, is that in the test that I did, I split half of your results to be paired with Intel machines and half to be paired with AMD machines. The intel machines became valid. However, looking at your results still in the database, you have 99 invalid and 10 valid.

I'm going to go ahead and add the hr class for AMD FX back in for the VINA projects, but can I ask that you avoid selecting only SN2S so that we can avoid clogging up the project? We will have to think on whether there is another solution or not.

(me again) Further, they (Intel) produce better benchmarks and use less power. AMD's module design is a "bad" idea IMHO. They use marketing to get people to buy an "8 core" when it's really a 4 core (since they share floating points).

If I'm not mistaken, the i5 2320 is going for like $190 if you are building this system, upgrade to the 2500k, so you can OC. If this is going to be used only for crunching than save for an i7 (50% more "work" and 2x the runtime compared to i5). IB is also SUPPOSED to be released this week.

AMD CPU's just can't compete IMHO I find (literally) few extra bucks worth it.

As a COMPLETELY DIFFERENT note: I know the boards are more expensive, and if you are building it yourself, i just picked up a 3820 (4core w/ HT) for $250 from MicroCenter. The boards are more expensive, but it allows for MUCH higher upgradability, since it allows for 40 PCIE lanes. Remember it's always easier to add stuff to a system, than it is to have to build/buy another one. (In this case I'm talking future GPU's, which is why I'm switching to 2011 chips)

Disregarding the Intel vs AMD thing which seems to do nothing but inflame emotions. I own an FX and would not recommend getting that for crunching. If you go AMD, get a PII-X6. The points per day are within 6% and the cost is significantly less than the FX.

Personally if I could build anything I wanted for crunching on WCG ( excluding multi CPU Hex Xeon. I Wish $$$ )
I would go for...
i7 990x + 2x GTX 560 Ti [448]
Just for RAW HORSEPOWER
The 990x for Max output with a single CPU.
The 2 GTX 560 Ti [448] because from what I can tell they have the best Gflops/Watt ratio.

@nanoprobe
Mine ran the Beta's in 3 minutes...
That's plenty fast enough.
How about your's ?
Not CPU time. CPU time or GPU time is Meaningless .
Elapsed time is all that matters because that's how much
real work gets done.
From knreed's comment
http://www.worldcommunitygrid.org/forums/wcg/viewthread_thread,32990_offset,40
More information for results that were valid:

num_res = number of results
num_hts = number of hosts
avg_cpu = average cpu time
avg_elp = average elapsed time
video_card = type of video card

num_res num_hts avg_cpu avg_elp video_card
55 2 95.0 260.0 CAL|AMDRadeonHD(unknown)
21 3 156.0 1387.0 CAL|AMDRadeonHD6300series(Caicos)
193 19 92.0 323.0 CAL|AMDRadeonHD6800series(Barts)
803 22 86.0 256.0 CAL|AMDRadeonHD6900series(Cayman)
4 2 120.0 715.0 CAL|AMDRadeonHD6x00series(Sumo)
101 7 171.0 694.0 CAL|AMDRadeonHD6x00series(Turks)
686 8 81.0 138.0 CAL|AMDRadeonHD79x0series(Tahiti)
882 44 113.0 384.0 CAL|ATIRadeonHD5700series(Juniper)
2602 69 103.0 295.0 CAL|ATIRadeonHD5800series(Cypress)
137 14 205.0 631.0 CAL|ATIRadeonHD5x00series(Redwood)
496 21 112.0 301.0 CAL|ATIunknown
17 1 78.0 259.0 CUDA|GeForce9600GT
3 1 201.0 7760.0 CUDA|GeForceGT425M
95 12 664.0 1052.0 CUDA|GeForceGT430
7 3 234.0 884.0 CUDA|GeForceGT440
10 2 1157.0 1418.0 CUDA|GeForceGT525M
12 2 739.0 1029.0 CUDA|GeForceGT530
61 7 270.0 1125.0 CUDA|GeForceGT540M
31 4 320.0 737.0 CUDA|GeForceGT555M
4 1 512.0 1534.0 CUDA|GeForceGT630M
158 7 462.0 553.0 CUDA|GeForceGTS450
717 25 271.0 359.0 CUDA|GeForceGTX460
55 2 443.0 686.0 CUDA|GeForceGTX460M
84 1 409.0 411.0 CUDA|GeForceGTX460SE
138 2 271.0 286.0 CUDA|GeForceGTX465
94 4 172.0 315.0 CUDA|GeForceGTX470
279 12 179.0 233.0 CUDA|GeForceGTX480
75 4 357.0 430.0 CUDA|GeForceGTX550Ti
4 1 380.0 457.0 CUDA|GeForceGTX555
164 7 205.0 423.0 CUDA|GeForceGTX560
1 1 365.0 745.0 CUDA|GeForceGTX560M
1405 41 254.0 305.0 CUDA|GeForceGTX560Ti
905 19 178.0 200.0 CUDA|GeForceGTX570
1168 18 164.0 228.0 CUDA|GeForceGTX580
9 1 308.0 344.0 CUDA|GeForceGTX580M
311 2 166.0 168.0 CUDA|GeForceGTX590
145 1 202.0 238.0 CUDA|GeForceGTX680
18 2 905.0 911.0 CUDA|Quadro1000M
15 1 117.0 665.0 CUDA|Quadro2000M

My 560 448 ran average 196 seconds , so unless your willing to get the High-end ATI's then it seems the NVidia cards did just fine...
I could get 2 560's for the price of one high-end card either manufacturer.
Plus , I do use my rig for gaming too...