home
about sensor
why sensor?
who's fastest?
p10 p50 p90
SensorPx
SPE CCUS
reserves
Bayes and Markov
Rd?
DCA?
frac conductivity
miscible
spe10
parallel?
Pc
gridding
FD vs FE
substantiation
map2excel
plot2excel
third party tools
services
publications
q & a
ethics
contact us
Dr. K. H. Coats

 

 

Industry Benchmarks

Spe1 through spe10 are problems from the SPE Comparative Solution Projects37-43*.   The formulation options shown are those found to run best in Sensor.

Machine 1 is a 2004 Dell Precision 470 workstation with 2.8 GHz Intel Xeon processors (Nocona, 1MB L2, 800 MHz FSB) running Sensor6k (32-bit) and Windows XP Professional with 2 GB RAM.

Machine 2 is a mid-2007 Dell Precision 690 workstation with 3.0 GHz Intel Xeon processors (Woodcrest 5160, 4 MB L2, 1333 MHz FSB) running Sensor64 and Windows XP Professional x64 with 16 GB RAM.

Machine 3 is a 2012 Dell 8500 Workstation with dual 3.4 GHz Intel Core I7 3770 quad core processors and 16 GB RAM.

The Sensor data and (Machine 1) output files are available in Downloads.  Our Sensor6k evaluation version will run all these cases to completion.  All CPU times are in seconds.

Case Formulation Blocks

Memory MB

Newtons Steps

Solver CPU (1)

Total CPU(1)

Total CPU (2)

Total CPU (3)

spe1 Impes, black oil, CFL 300 0.810 156 145   0.23 0.094 0.062
spe2 Implicit, black oil 150 0.889 34 10   0.09 0.047 0.015
spe3 Impes, nc=9, CFL 324 1.96 106 105 0.1 0.70 0.30 0.203
spe5 Impes, nc=6, CFL 147 0.784 718 716 0.1 1.1 0.47 0.296
spe7_1a Implicit, black oil 270 1.28 17 16   0.09 0.03 0.031
spe7_4a Implicit, black oil 270 1.28 45 36   0.29 0.09 0.062
spe9 Implicit, black oil, NF 9000 31.7 56 39 4.9 8.2 4.4 1.45
spe10_case1 Implicit, black oil, D4 2000 6.9 835 725 22.3 29.6 12.2 5.13
spe10_case2_75 Implicit, black oil 75 .463 23 22     0.06 0.02
spe10_case2_2000 Implicit, black oil 2000 4.1 112 111     1.5 0.3
spe10_case2 30x55x85 Implicit, black oil 140158 551 225 214 2452 2634 1803 490
The following case is from Reference 1 of our Miscible page:
spe79692x3 Impes, nc=7, CFL 1000 4.06 1011 1004 1.9 13.8 5.3 3.0

CFL - Impes stable step logic used.  NF - Nested Factorization solver option (default is RBILU(0)). D4 - Direct solver option.

Sensor's performance advantage of course varies by problem, but is not in general a function of problem size or reservoir complexity.  Also compare using your cases or their examples.  Run time comparisons of different models should be made on the same machine.

 

* Reference numbers are those of the Sensor Manual


2000 - 2017 Coats Engineering, Inc.