Best DC project for my Radeon?

Centauri

Golden Member
Dec 10, 2002
1,628
54
91
I've been doing F@H for a few months now and my Radeon does a decent job (4k PPD~) but it pales in comparison to mid-range Radeon discretes and even entry level offerings from nVidia. I suspect it's because the 7660D isn't double precision floating point, because its gaming performance would have you expecting better F@H numbers. And the new Core 17 hasn't done much other than to make my system unusable while running.

Where could I better focus my efforts? Something BOINC related? I like my system as it is, without adding a discrete, because at full CPU and GPU load it's only consuming 100-105w, so electricity cost is negligible.

Thanks.
 

networkman

Lifer
Apr 23, 2000
10,436
1
0
Not sure how it compares, but the Einstein@Home project does make use of Radeon cards of the 4xxx series on up. One of my rigs has a 6350 in it (work) and the card appears to do better than a more capable nVidia card that I had in it. YMMV.
 

Sunny129

Diamond Member
Nov 14, 2000
4,823
6
81
keep in mind though that Milkyway@Home requires FP64 (double-precision) performance, so the OP's GPU is not compatible w/ MW@H.
 

Centauri

Golden Member
Dec 10, 2002
1,628
54
91
Now that's the weird part, because I've dealt with MW before; in spite of the fact that my GPU isn't FP64, I still process GPU MW units. However, it takes closer to 30 mins per unit.

The guys over at Overclock were baffled by how I was even processing them at all.
 

Sunny129

Diamond Member
Nov 14, 2000
4,823
6
81
i found your GPU on this wiki page in the 2nd chart from the top. i see now that its the IGP for either an A10-5700 or A10-5800K APU. while the chart doesn't say whether or not this GPU is actually capable of double precision performance, it also doesn't contain a column for FP64 GFLOP figures, so my guess is that your GPU not FP64-capable. i have no idea how it was running MW@H tasks at one point. do you recall whether or not you were getting valid results? or did you witness many errors/invalids? it probably doesn't matter anyways...even if your card can participate in MW@H, 30 min. per task is a very long run time for a MW@H separation task...my HD 7970 completes a single MW@H task in ~30 seconds. with your particular AMD GPU, i would stay away from MW@H.
 
Last edited: