@ Oldchap

I've entered my values for the HP Z600 computers. I'll add the values for my new E5-2673v3 computers when I have more runtime. Probably after next patch Tuesday, at which point I'll also measure the W/VA on the Z600.

Great initiative! Thanks.

Ok Oldchap, here are my numbers. I calculated a 14-day average on my two computers, 4 + 4vm and 5 + 5vm.

These are the numbers for the guest OS, Ubuntu 14.04 LTS, running 12 threads OET inside a VM on the first HDD, where the host OS is on a separate partition.

4vm= 974414 / 14 = 69601
09/01/2015 0:011:21:49:45 65,931 184
08/31/2015 0:011:16:29:35 67,954 193
08/30/2015 0:012:06:16:43 71,035 204
08/29/2015 0:011:15:36:39 65,964 190
08/28/2015 0:011:22:52:17 69,309 198
08/27/2015 0:011:20:32:46 70,380 200
08/26/2015 0:011:18:54:30 68,569 195
08/25/2015 0:011:18:55:32 68,833 196
08/24/2015 0:011:19:59:23 70,973 201
08/23/2015 0:011:22:30:24 68,154 191
08/22/2015 0:012:02:07:49 66,833 204
08/21/2015 0:011:21:32:30 94,254 321
08/20/2015 0:011:23:36:35 64,578 180
08/19/2015 0:011:11:04:52 61,646 177

5vm= 939540 / 14 = 67110
09/01/2015 0:011:14:48:33 64,816 181
08/31/2015 0:011:23:31:37 68,927 198
08/30/2015 0:011:15:22:22 67,204 192
08/29/2015 0:012:08:54:41 71,183 204
08/28/2015 0:011:17:53:29 67,174 192
08/27/2015 0:011:21:21:57 68,443 195
08/26/2015 0:011:12:49:23 65,398 186
08/25/2015 0:007:20:10:00 47,200 133
08/24/2015 0:011:18:16:37 71,068 200
08/23/2015 0:012:00:51:38 68,725 192
08/22/2015 0:011:21:10:40 65,347 201
08/21/2015 0:012:03:33:07 87,483 305
08/20/2015 0:013:19:41:02 74,967 211
08/19/2015 0:009:18:56:26 51,603 147


These are the numbers for the host OS, W7 PRO, running 12 threads CEP2, where the BOINC Data file is on the second short-stroked HDD. Note that this seems to be an inefficient solution, since the average daily runtime is closer to 11 days, instead of 12. From now on I'll mix the projects so that I have the CEP2 wus more evenly spread between the two HDDs. Later I'll also try RAM-cache, which should also be able to increase the low runtime efficiency for the host OS. RAM-cache seems to be inefficient when running VMs, so I wont use it for that.

4= 613004 / 14 = 43786
09/01/2015 0:013:12:14:25 49,598 40
08/31/2015 0:008:21:16:54 34,392 32
08/30/2015 0:010:14:05:18 51,248 51
08/29/2015 0:014:12:28:37 58,037 53
08/28/2015 0:011:13:16:54 47,378 44
08/27/2015 0:007:05:13:22 28,186 19
08/26/2015 0:011:15:21:36 42,318 36
08/25/2015 0:007:08:22:55 26,365 19
08/24/2015 0:011:23:55:35 40,859 33
08/23/2015 0:010:06:25:38 37,828 32
08/22/2015 0:011:07:32:06 51,100 50
08/21/2015 0:013:01:34:36 51,737 44
08/20/2015 0:011:22:34:54 48,393 43
08/19/2015 0:012:13:53:59 45,567 37

5= 604478 / 14 = 43177
09/01/2015 0:012:16:20:47 45,815 39
08/31/2015 0:012:02:15:09 46,388 46
08/30/2015 0:011:15:47:32 48,203 50
08/29/2015 0:009:07:42:16 36,675 34
08/28/2015 0:011:18:07:48 44,809 36
08/27/2015 0:011:12:05:55 42,062 33
08/26/2015 0:011:18:20:18 45,610 38
08/25/2015 0:011:17:14:48 40,442 37
08/24/2015 0:011:04:59:42 39,529 31
08/23/2015 0:013:04:44:16 49,052 45
08/22/2015 0:011:13:58:19 42,574 43
08/21/2015 0:009:22:01:25 37,998 29
08/20/2015 0:010:21:24:44 41,923 38
08/19/2015 0:011:15:58:52 43,397 34


4+4vm = 113387 / 7 = 16198 Bppd (138W)
5+5vm = 110287 / 7 = 15755 Bppd (141W)

Computer number 5 has a cramped case so it has two more fans, but everything else, except the PSUs, are the same. Computer 4 has a PSU with a 0,92W/VA and computer 5 has 0,95W/VA. The E5-2673v3 processors' stock frequencies are 2,4GHz, but all threads run simultaneously at 2,7GHz on these ASRock Extreme 4 mATX boards. They are also undervolted to 0,8Vcore. I don't know if there are any negative long-term effects related to the processors' longevity because of the undervoltings. Perhaps, as the processors age, the error rate will rise so I'll be forced to increase the voltages? I've also checked the accuracy of the Kill-a-Watt against 3,5,7 and 11W lamps and it reads it accurately.

I'll write computer four's efficiency value in your list since it's higher.

Oldchap, I've been running only OET in two VMs, on computer 4, for the past 14 days, to see what kind of numbers I could get. I've filled in the efficiency number in the VINA box in your sheet, so you can just add it into the same row as the mixed projects. I've also written in the comment box that the computer only needs 1 HDD when running only VINA and therefore is using 128W, 10W less than before.

4vm1 = 1'260'116 / 14 = 90'008
09/16/2015 0:011:13:42:38 90,867 927
09/15/2015 0:011:16:35:50 94,501 1,043
09/14/2015 0:010:17:24:44 129,741 1,285
09/13/2015 0:011:11:14:27 61,840 393
09/12/2015 0:011:17:34:56 59,355 604
09/11/2015 0:012:01:33:54 137,047 1,146
09/10/2015 0:011:20:00:18 76,489 359
09/09/2015 0:011:21:57:40 76,375 351
09/08/2015 0:011:02:49:47 154,763 757
09/07/2015 0:012:10:04:05 104,799 361
09/06/2015 0:011:19:39:28 66,289 185
09/05/2015 0:012:09:06:54 71,212 204
09/04/2015 0:011:13:50:45 66,271 189
09/03/2015 0:011:18:48:42 70,567 197

4vm2 = 1'260'625 / 14 = 90'044
09/16/2015 0:011:21:18:42 92,470 941
09/15/2015 0:011:16:40:51 83,609 968
09/14/2015 0:010:19:33:40 125,153 1,242
09/13/2015 0:012:06:32:41 65,864 413
09/12/2015 0:011:18:03:41 57,499 600
09/11/2015 0:011:19:05:08 147,518 1,218
09/10/2015 0:012:15:47:28 80,513 377
09/09/2015 0:010:23:21:23 70,973 331
09/08/2015 0:010:23:01:39 143,693 714
09/07/2015 0:012:01:22:58 116,756 391
09/06/2015 0:011:20:30:30 66,791 187
09/05/2015 0:012:00:54:42 68,680 196
09/04/2015 0:012:02:34:34 70,178 200
09/03/2015 0:012:06:07:37 70,928 199


180'052 / 7 = 25'721 Bppd (128W)


Even with the WCG server being down for 8 hours during this period, the numbers look too good to be true. Maybe WCG is giving this new processor model too many points? Or perhaps it has a slightly different way of processing data? I believe the model is an OEM only, and I've read that Intel is delivering custom processors to cloud companies. Perhaps this is one of those processor models?

Friendly reminder about the crunching efficiency page.

https://docs.google.com/spreadsheets/d/1ll78D...9Xazh6Bvo2xlII/edit#gid=0

Above you can see the best result I have seen so far for a xeon v3 rig.

Line 23 shows Anubis' dual xeon e5-2675v3 wiping the floor in the efficiency stakes

I have a feeling that a fair amount of vina units are in this but even so it's looking good

As a general rule I would hope people let their machine run for at least 14, preferably 28 days then take their average (RAC) over that time.