Low Power War, Round 3: Performance on BOINC
"The proof of the pudding is in the eating", or as they originally said: "Jt is ywrite that euery thing Hymself sheweth in the tastyng".
Being an AMD User, writing for the AMD Users forum, one could think that I am biased towards a certain brand of CPU/APU/SOC products.
That would not be true however. I like AMD products, but i buy Intel as well when it suits my needs.
The need is for performance at a low price -both in purchase and power consumption.
I will try to generate a table with BOINC applications per project vs. the credit score per core/per day for all quad-core, low-power CPU/APU/SOCs I encounter WUProp data for -provided I do not crash their server with my query's-, starting with the X86/X86-64 platform:Project |
Application |
AMD
Athlon 5350 |
Intel
Celeron J1900 |
Intel
Celeron N2940 |
Intel
Pentium N3540 |
Intel
Pentium J2900 |
Albert@Home |
Gamma Ray Pulsar Search #4 |
389.1 (Win32) |
na |
na |
na |
na |
Asteroids@Home |
Periodic Search |
1,927.8 (Win64)
2,587.2 (Linux64) |
1,393.6 (Linux64) |
1,171.0 (Win64) |
1,267.8 (Win64) |
999.8 (Win64) |
Atlas@Home |
Atlas Simulation |
298.4 (Win64) |
na |
na |
na |
na |
BealF |
Beal Engine |
na |
na |
na |
na |
na |
Citizen Science Grid |
DNA@Home Gibbs Sampler |
757.2 (Win64) |
472.8 (Linux64)
790.0 (Win64) |
na |
517.2 (Linux64)
631.9 (Win64) |
na |
|
SubsetSum@Home Sum Calculator |
714.1 (Win64) |
813.8 (Linux64) |
na |
795.6 (Linux64)
597.2 (Win64) |
na |
CMS-dev |
CMS Simulation |
463.0 (Win64) |
na |
3,069.0 (Win64) |
na |
na |
Cosmology@Home |
CAMB |
na |
368.2 (Linux64) |
277.6 (Win32) |
698.6 (Linux64) |
284.5 (Win32) |
DENIS@Home |
Carro-Rodriguez-Laguna-Pueyo
Epicardial Model (Carro et al. 2011)
for human ventricular cells |
492.6 (Linux64)
418.1 (Win64) |
na |
na |
320.1 (Win64) |
211.7 (Win64) |
DistributedDataMining |
Biological Data Analysis:
Laryngeal Video Classification |
610.5 (Win64) |
na |
na |
na |
na |
Enigma@Home |
Enigma 0.76b |
1,011.9 (Win32) |
743.9 (Win32) |
na |
850.5 (Win32) |
na |
LHC@home 1.0 |
SixTrack |
2,664.2 (Win32) |
2,376.1 (Win32) |
375.8 (Win64) |
945.2 (Win32) |
na |
Malariacontrol.net |
Openmalaria |
386.6 (Win64) |
na |
na |
403.7 (Win32) |
299.7 (Win64) |
MindModeling@Beta |
ACT-R cognitive modeling
environment leveraging
Clozure Common Lisp |
508.9 (Win64) |
na |
283.3 (Win64) |
482.7 (Win32) |
407.3 (Win64) |
NFS@Home |
14e Lattice Sieve |
429.7 (Linux64)
430.2 (Win64) |
na |
na |
445.3 (Linux64) |
na |
|
15e Lattice Sieve |
690.9 (Linux64)
865.0 (Win64) |
na |
na |
717.5 (Linux64) |
na |
|
16e Lattice Sieve V5 |
1,321.6 (Linux64)
1,566.6 (Win64) |
na |
na |
1,389.6 (Linux64) |
na |
Pogs |
fitsedwrapper |
1,385.1 (Linux64)
1,348.7 (Win64) |
1,221.4 (Linux64) |
962.4 (Win64) |
1,007.4 (Win64) |
na |
PrimeGrid |
PPSE (LLR) |
778.9 (Win64) |
na |
na |
469.2 (Win64) |
na |
|
Sophie Germain (LLR) |
697.5 (Win64) |
na |
na |
571.8 (Win64) |
na |
|
The Riesel Problem (LLR) |
1,502.3 (Win64) |
na |
na |
na |
na |
SRBase |
Riesel Base |
1,425.4 (Win64) |
na |
na |
789.8 |
na |
|
Sierpinski / Riesel Base |
1,198.3 (Win64) |
na |
691.2 |
1,023.7 (Win64) |
na |
|
Sierpinski / Riesel Base - average |
1,172.9 (Win64) |
na |
412.1 |
na |
na |
|
Sierpinski / Riesel Base - average2 |
1,442.2 (Win64) |
na |
na |
na |
na |
|
Sierpinski / Riesel Base - short |
1,500.8 (Win64) |
na |
na |
na |
na |
|
Sierpinski Base |
991.1 (Win64) |
na |
na |
na |
na |
|
Sierpinski Base - short |
1,246.0 (Win64) |
na |
na |
na |
na |
TN-Grid Test Platform |
Gene Network Application |
321.5 (Win64) |
na |
na |
na |
na |
Universe@Home |
universe-xray sources v3 |
1,047.9 (Win32) |
1,204.5 (Linux64)
899.7 (Win32) |
na |
na |
937.7 (Win32) |
Yafu |
YAFU |
4,526.9 (Win64) |
na |
na |
na |
na |
na=not available
I have as yet no figures for boards fitted with the AMD A4-5000, a 15W tdp quad-core 1500 MHz SOC that is soldered onto the motherboard, like the Bay Trail/Cherry Trail boards.
Nor do I have results yet for the Intel Pentium N3700, Intels next claim to low power fame (6W tdp).
If the sometimes surprising results of the Celeron N2940 and Pentium N3540 -Bay Trail-M notebook CPUs- are anything to go by, we may soon be facing interesting times....and ditto credit levels.
Based upon the above I do now know for sure I will build myself another Athlon 5350 system, but running Ubuntu -as my present AM1 system is now running 64-bit Win10.
Low-Power war: Instruction sets
Feature |
AMD AM1 |
AMD since |
Year |
Intel Bay Trail |
Intel Brasswell |
Intel since |
Year |
MMX |
Yes |
AMD K6 |
1997 |
Yes |
Yes |
P5 Pentium MMX |
1996 |
SSE |
Yes |
AMD K8 Athlon 64
Clawhammer/Sledgehammer |
2003 |
Yes |
Yes |
Pentium III |
1999 |
SSE2 |
Yes |
AMD K8 Athlon 64
Clawhammer/Sledgehammer |
2003 |
Yes |
Yes |
Pentium 4 Willamette |
2000 |
AMD64/EM64T |
Yes |
AMD K8 Athlon 64
Clawhammer/Sledgehammer |
2003 |
Yes |
Yes |
Pentium 4 Prescott |
2004 |
SSE3 |
Yes |
AMD K8 Athlon 64
Venice/San Diego |
2005 |
Yes |
Yes |
Pentium 4 Prescott |
2004 |
AMD-V/Intel-VTx |
Yes |
AMD K8 Athlon 64
Orleans/Windsor |
2006 |
Yes |
Yes |
Pentium 4 Prescott |
2004 |
SSE4a |
Yes |
AMD K10 Barcelona
Phenom Agena |
2007 |
No |
No |
|
|
BMI1 |
Yes |
AMD K10 Barcelona
Phenom II Thuban |
2010 |
No |
No |
Intel Core2 Nehalem |
2008 |
SSSE3 |
Yes |
AMD Bulldozer |
2011 |
Yes |
Yes |
Intel Core1 (Xeon Woodcrest) |
2006 |
SSE4.1 |
Yes |
AMD Bulldozer |
2011 |
Yes |
Yes |
Intel Core2 Penryn |
2007 |
SSE4.2 |
Yes |
AMD Bulldozer |
2011 |
Yes |
Yes |
Intel Core2 Nehalem |
2008 |
AES(e.g. CLMUL) |
Yes |
AMD Bulldozer |
2011 |
No |
No |
Intel Xeon Gulftown
(Westmere-EP) |
2010 |
AVX |
Yes |
AMD Bulldozer |
2011 |
No |
No |
Intel Core2 Sandy Bridge |
2011 |
F16C |
Yes |
AMD Bulldozer |
2011 |
No |
No |
Intel Core2 Ivy Bridge |
2012 |
In the next round Intel will introduce its Apollo Lake SOCs, while AMD will hopefully have a new 25 Watt TDP (or less) Stoney Ridge SOC for Socket AM4
(or, preferably, still for Socket AM1, since memory controller, north- and southbridge are all integrated in the SOC)