Running two virtual machines in Windows, each with two SMP clients, gets me a maximum of around 3,500 ppd, usually a lot less due to pesky 1385 pointers.
And because the nice Scottish bloke at Crucial told me today that I could just return the single faulty stick from each pair of my borked memory (since all four sticks were from the exact same batch and the two good sticks should be fine together), I've decided to carry on using native Ubuntu for a bit to see what sort of output is possible with four SMP clients.
Currently running two 3029 and 3050 WU's, each being worth 1440 points-
CPU1 (3029) -
[17:38:13] Completed 7900000 out of 10000000 steps (79 percent)
[17:38:13] Extra SSE boost OK.
[17:55:11] Writing local files
[17:55:11] Completed 8000000 out of 10000000 steps (80 percent)
[18:12:46] Writing local files
[18:12:46] Completed 8100000 out of 10000000 steps (81 percent)
[18:30:19] Writing local files
[18:30:19] Completed 8200000 out of 10000000 steps (82 percent)
CPU2 (3029) -
[17:38:27] Completed 8600000 out of 10000000 steps (86 percent)
[17:38:27] Extra SSE boost OK.
[17:55:53] Writing local files
[17:55:53] Completed 8700000 out of 10000000 steps (87 percent)
[18:13:25] Writing local files
[18:13:25] Completed 8800000 out of 10000000 steps (88 percent)
[18:30:56] Writing local files
[18:30:56] Completed 8900000 out of 10000000 steps (89 percent)
CPU3 (3050) -
[17:38:37] Completed 8300000 out of 10000000 steps (83 percent)
[17:38:37] Extra SSE boost OK.
[17:56:27] Writing local files
[17:56:27] Completed 8400000 out of 10000000 steps (84 percent)
[18:14:19] Writing local files
[18:14:19] Completed 8500000 out of 10000000 steps (85 percent)
[18:32:08] Writing local files
[18:32:08] Completed 8600000 out of 10000000 steps (86 percent)
CPU4 (3050) -
[17:38:56] Completed 100000 out of 10000000 steps (1 percent)
[17:38:56] Extra SSE boost OK.
[17:56:51] Writing local files
[17:56:51] Completed 200000 out of 10000000 steps (2 percent)
[18:14:45] Writing local files
[18:14:45] Completed 300000 out of 10000000 steps (3 percent)
[18:32:38] Writing local files
[18:32:38] Completed 400000 out of 10000000 steps (4 percent)
So the 3029's are averaging about 17.5mins per frame, and the 3050's are averaging about 18 mins per frame.
If my calculations are correct, I'm averaging ~1185ppd per 3029, and ~1150ppd per 3050. So we'll say an average of 1170ppd per client.
Making a total of ~4680ppd . That's a gain of almost 1200ppd over what I've been able to achieve running in VMware
And because the nice Scottish bloke at Crucial told me today that I could just return the single faulty stick from each pair of my borked memory (since all four sticks were from the exact same batch and the two good sticks should be fine together), I've decided to carry on using native Ubuntu for a bit to see what sort of output is possible with four SMP clients.
Currently running two 3029 and 3050 WU's, each being worth 1440 points-
CPU1 (3029) -
[17:38:13] Completed 7900000 out of 10000000 steps (79 percent)
[17:38:13] Extra SSE boost OK.
[17:55:11] Writing local files
[17:55:11] Completed 8000000 out of 10000000 steps (80 percent)
[18:12:46] Writing local files
[18:12:46] Completed 8100000 out of 10000000 steps (81 percent)
[18:30:19] Writing local files
[18:30:19] Completed 8200000 out of 10000000 steps (82 percent)
CPU2 (3029) -
[17:38:27] Completed 8600000 out of 10000000 steps (86 percent)
[17:38:27] Extra SSE boost OK.
[17:55:53] Writing local files
[17:55:53] Completed 8700000 out of 10000000 steps (87 percent)
[18:13:25] Writing local files
[18:13:25] Completed 8800000 out of 10000000 steps (88 percent)
[18:30:56] Writing local files
[18:30:56] Completed 8900000 out of 10000000 steps (89 percent)
CPU3 (3050) -
[17:38:37] Completed 8300000 out of 10000000 steps (83 percent)
[17:38:37] Extra SSE boost OK.
[17:56:27] Writing local files
[17:56:27] Completed 8400000 out of 10000000 steps (84 percent)
[18:14:19] Writing local files
[18:14:19] Completed 8500000 out of 10000000 steps (85 percent)
[18:32:08] Writing local files
[18:32:08] Completed 8600000 out of 10000000 steps (86 percent)
CPU4 (3050) -
[17:38:56] Completed 100000 out of 10000000 steps (1 percent)
[17:38:56] Extra SSE boost OK.
[17:56:51] Writing local files
[17:56:51] Completed 200000 out of 10000000 steps (2 percent)
[18:14:45] Writing local files
[18:14:45] Completed 300000 out of 10000000 steps (3 percent)
[18:32:38] Writing local files
[18:32:38] Completed 400000 out of 10000000 steps (4 percent)
So the 3029's are averaging about 17.5mins per frame, and the 3050's are averaging about 18 mins per frame.
If my calculations are correct, I'm averaging ~1185ppd per 3029, and ~1150ppd per 3050. So we'll say an average of 1170ppd per client.
Making a total of ~4680ppd . That's a gain of almost 1200ppd over what I've been able to achieve running in VMware
