Two systems:
A is using a GTX 550ti and
B is using an AMD 6670 (that in windows is reported as HD 5700 Series (Juniper)

When BOINC starts in the event log, the for the 550ti it's reporting 670 GFLOPS peak and for system B, with the 6770 it's reporting 1360 GFOPS peak.

Yet the 550ti system, with half the GFLOPS peak is processiing WUs at about half the time.

Which of course doesn't make all that much sense and I hope someone can shed some light as to why.

Anyone...anyone...Bueler...

I presume you meant HD 6770 as it is a 'rebadged' HD 5770. The Juniper XT GPU in both is an older GPU which is less suited to GPGPU than AMD's current GCN-based GPUs (used in most HD 7000 series video cards and later).

Meanwhile, the GTX 550 Ti is Fermi-based which is generally good for both GPGPU as well as regular graphics work. NV GPUs will also have an advantage on Multi-Beam tasks as the applications available for those use NV's native CUDA libraries as opposed to the broadly-compatible OpenCL for AMD GPUs. While I understand it's possible to tune OpenCL applications to be as fast as CUDA ones, I don't think we have quite reached that with the applications for S@h.

Lastly, the GFLOPS reported in the BOINC messages is not a measurement of actual performance, which is why you may be confused as to the actual performance of your GPUs.

---

Soli Deo Gloria

Yeah..I was hoping I have something setup wrong but it's just a case of older GPU...
At least they keep the room warm :)
Thanks !

---