Ansys 145 Intel SSD Benchmark

Distributed ANSYS 14.
5 benchmark on Intel SSDs
Hunter Wang
ANSYS, Inc.
March 2013
1 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential
Contents
Basic Hardware Configuration
Benchmark models Information
Basic benchmark results, 4XSSD RAID 0, 2XSSD 0, 1XSSD, 1XHD
CG-2
SP-3
SP-5
SP-6
Memory size effects
64/32/16GB, SP-5
128/64/32GB, SP-6
6Gb/s vs. 3Gb/s SATA RAID Controller
SP-5
SP-6
Working Directory on 4XSSD, Virtual Memory on HD vs. on 4XSSD
SP-5
SP-6
Working Directory on HD, Virtual Memory on HD vs. on 4XSSD
SP-5
SP-6
Observations & Conclusions

Single Intel W2600CR Workstation
CPU: Dual Intel Xeon E5-2687W(3.1Ghz, 8-core), total 16 cores
Memory: 8 X 4GB DDR3-1333 MT/s, Reg. ECC, total 32GB
OS: Windows 7 Professional x64 SP1
RAID Controller: Intel Embedded Server RAID Technology II (ESRT2), 3Gb/s SATA
Storage scenarios (Working directory and virtual memory):
4 X Intel SSD DC S3700 400GB, RAID 0, SATA 3Gb/s
2 X Intel SSD DC S3700 400GB, RAID 0, SATA 3Gb/s
1 X Intel SSD DC S3700 400GB, SATA 6Gb/s
1 X 1TB 7.2K RPM HD, SATA 6Gb/s

Benchmark models
Name Solver DOFs Max. Memory Analysis Type

(16-core DANSYS)
CG-2 PCG 11.8 M ~25 GB Static structural, linear
SP-3 SPARSE 2.3 M ~18 GB Transient structural, nonlinear
SP-5 SPARSE 2.1 M ~ 48 GB (in-core) Static structural, nonlinear (contact)
SP-6 SPARSE 4.9 M ~35 GB (out-of-core) Static structural, nonlinear
CG-2 SP-3 SP-5 SP-6

CG-2 Benchmark Results, Rating
120
113
4XSSD-RAID-0-SATA-3Gb/s 110
103
2XSSD-RAID-0-SATA-3Gb/s
100
94 93 93
1XSSD-SATA-6Gb/s
HD(7.2K RPM)-SATA-6Gb/s
80
Rating (jobs/day)
64 63
61
60
43 44
41 42
40
40 36
34
26 25
24
22
20
0
#Machine X #Core 1X1 1X2 1X4 1X8 1X16
Memory 15.2GB 17.6GB 18.8GB 20.7GB 24.9GB

SP-3 (in-core) Benchmark Results, Rating
700
4XSSD-RAID-0-SATA-3Gb/s 635 631
617
600
1XSSD-SATA-6Gb/s
517 514 514
500
Rating (jobs/day)
400
336 338 336
300
231 223 215

193 193 193 187
200
109 110 109 109

100
0
Memory 10.4GB 11.9GB 13.5GB 16.5GB 17.9GB

SP-5 (in-core) Benchmark Results, Rating
450
4XSSD-RAID-0-SATA-3Gb/s 419
400 384
SSD-SATA-6Gb/s 368
350
301
300 283
275
Rating (jobs/day)
250
200
180 180 180
145 146 144

150
124 118
95
100 89 89 88 88
52
50
0
Memory 29GB 33GB 35.6GB 40.8GB 47.8GB

SP-6 (out-of-core) Benchmark Results, Rating
250
SSD-SATA-6Gb/s 208
200
168
164
157
Rating (jobs/day)
147
150
137
111
102
97
100
67
63 62
50 43 41 40
28
25 23 22
18
0
Memory 10.2GB 13.5GB 16GB 21.9GB 34.4GB

Contents
CG-2
SP-3
SP-5
SP-6
Memory size effects
64/32/16GB, SP-5
128/64/32GB, SP-6
SP-5
SP-6
SP-5
SP-6
SP-5
SP-6

so the load balance was drastically altered and is the
64 vs. 32 vs. 16 GB Memory, SP-5, Rating
Memory eater program could only run on 1 core,

500
SP-5 477
4XSSD-64GB-Memory
450 (32GB & 64GB, in-core)
4XSSD-32GB-Memory
(16GB, out-of-core) 419
reason for this run being so slow.

400 4XSSD-16GB-Memory
350
320
Rating (jobs/day)
301
300
266
250
200 190
180
167
144 145
150 136
99
100 87 89 84
50
0
1X1 (29GB) 1X2 (33GB) 1X4 (35.6GB) 1X8 (40.8GB) 1X16 (47.8GB)
#Machine X #Core (Memory Used)

64GB vs. 32GB Memory, SP-6, Rating
350
4XSAS-15K-128GB-Memory (in-core)
4XSSD-64GB-Memory (in-core) 302
300 4XSSD-32GB-Memory (out-of-core)
260
250
212 208
200
Rating (jobs/day)
168
150 134
111
100
80
67
49
50 43
0
1X1 (56.1/10.2GB) 1X2 (59.5/13.5 GB) 1X4 (63.9/16 GB) 1X8 (70.1/21.9 GB) 1X16 (81.5/34.4 GB)
#Machine X #Core (Memory Used in-core/out-of-core)

Contents
CG-2
SP-3
SP-5
SP-6
Memory size effects
64/32/16GB, SP-5
128/64/32GB, SP-6
SP-5
SP-6
SP-5
SP-6
SP-5
SP-6

6Gb/s vs. 3Gb/s SATA, SP-5, in-core, Rating
500
4XSSD-RAID-0-3Gb/s SP-5 (32 GB Memory)
443
450
4XSSD-RAID-0-6Gb/s 419
400
350
301 302
Rating (jobs/day)
300
250
200 180 183
145 144
150
100 89 88
50
0
Memory 29GB 33GB 35.6GB 40.8GB 47.8GB

6Gb/s vs. 3Gb/s, SP-6, out-of-core, Rating
250
SP-6 (32GB Memory)
200
183
168
Rating (jobs/day)
150
111 112
100
71
67
50 43 45
0
Memory 10.2GB 13.5GB 16GB 21.9GB 34.4GB

Contents
CG-2
SP-3
SP-5
SP-6
Memory size effects
64/32/16GB, SP-5
128/64/32GB, SP-6
SP-5
SP-6
SP-5
SP-6
SP-5
SP-6

Virtual memory on 7.2K HD vs. 4XSSD RAID 0
SP-5, Rating
500.0
SP-5, 32 GB Memory
SWAP on 7.2K RPM HD, in-core Working directory on 4XSSD RAID 0, 443.1
450.0
SWAP on 7.2K RPM HD, force out-of-core 6Gb/s SATA
400.0 389.2
SWAP on 4XSSD RAID 0, SATA 6Gb/s, in-core
350.0
302.1
300.0 288.0
Rating (jobs/day)
250.0
200.0 183.1
175.3
149.7 148.7
150.0 139.1 144.2
100.7
100.0 88.0 85.5 88.1
65.5
50.0
0.0
1X1 (4.1/29 GB) 1X2 (6.4/33 GB) 1X4 (9.3/35.6 GB) 1X8 (13.3/40.8 GB) 1X16 (20.6/47.8 GB)
#Machine X #Core (Memory Used out-of-core/in-core)

Virtual memory on 7.2K HD vs. 4XSSD RAID 0
SP-6, Rating
250.0
SWAP on 7.2K RPM HD, out-of-core SP-6, 32GB Memory 229.8
SWAP on 7.2K RPM HD, force in-core working directory on 4XSSD RAID 0,
6Gb/s SATA
SWAP on 4XSSD RAID 0, SATA 6Gb/s, out-of-core
200.0
SWAP on 4XSSD RAID 0, SATA 6Gb/s, force in-core 183.1 185.4 185.8
181.5
Rating (jobs/day)
150.0 140.3
112.2 112.4
100.0 94.1
70.5 71.3
59.0
50.0 44.7 45.0

34.9
18.0 15.1
12.9 14.1 11.3
0.0
1X1 (10.2/56.1GB) 1X2 (13.5/59.5 GB) 1X4 (16.0/63.9 GB) 1X8 (21.9/70.1 GB) 1X16 (34.4/81.5 GB)

Contents
CG-2
SP-3
SP-5
SP-6
Memory size effects
64/32/16GB, SP-5
128/64/32GB, SP-6
SP-5
SP-6
SP-5
SP-6
SP-5
SP-6

Virtual Memory on 7.2K HD vs. on SSD RAID 0
SP-5, in-core, Elapsed Time
1800
1673
SWAP on SATA 6Gb/s 7.2K HD SP-5 (32GB, in-core)
1600
Working directory on 7.2K
SWAP on 4XSSD RAID 0
1400 RPM HD
Elapsed Time (Seconds)
1200
987 981
1000
912
800 746 732

697 693 691 674
600
400
200
0
1X1 (29GB) 1X2 (33GB) 1X4 (35.6GB) 1X8 (40.8GB) 1X16 (47.8GB)
#Machine X #Core (Memory Used in-core)

Virtual Memory on 7.2K HD vs. on SSD RAID 0
SP-6, Elapsed Time
6000
SWAP on SATA 6Gb/s 7.2K HD SP-6, 32GB
SWAP on 4XSSD RAID 0 (force in-core) Working directory on 7.2K RPM HD
5000
4704
Elapsed Time (Seconds)
4000 3848
3747
3507
3078
2929
3000
2696
2561
2447
2325
2000
1000
0
1X1 (10.2/56.1) 1X2 (13.5/59.5) 1X4 (16/63.9) 1X8 (21.9/70.1) 1X16 (34.4/81.5)

Max. disk I/O rate, virtual memory on 4 X SSD RAID 0 6Gb/s
Peak I/O rate reached ~1.8GB/s, captured ~1.5GB/s

Observations and Conclusions
SSDs helps DANSYS reduce elapsed time and scale better with multi cores
Max. 33% performance improvement from 4XSSD RAID 0, 3Gb/s SATA to 1
SSD, 6Gb/s SATA (SP-6, 16 cores, out-of-core)
Up to 10% performance improvement from 3Gb/s to 6Gb/s SATA RAID
Controller (SP-6, 16 cores, out-of-core)
More memory, more chance to run in-core to improve DANSYS performance
Set both virtual memory (1~2X physical memory) and working directory to
SSD RAID 0 to achieve best performance.
SSD RAID 0 or PCI-E SSD with high sustained read/write rate and low latency is
highly recommended for ANSYS Mechanical users.

Ansys 145 Intel SSD Benchmark

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ansys 145 Intel SSD Benchmark

Uploaded by

Copyright:

Available Formats

Distributed ANSYS 14.

5 benchmark on Intel SSDs

2 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

3 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

Name Solver DOFs Max. Memory Analysis Type

CG-2 SP-3 SP-5 SP-6

4 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

5 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

231 223 215

109 110 109 109

6 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

145 146 144

7 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

8 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

9 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

Memory eater program could only run on 1 core,

reason for this run being so slow.

10 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

11 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

12 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

200 180 183

13 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

14 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

15 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

16 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

50.0 44.7 45.0

17 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

18 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

800 746 732

19 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

20 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

Peak I/O rate reached ~1.8GB/s, captured ~1.5GB/s

21 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

22 2013 ANSYS, Inc. March 25, 2013 ANSYS Confidential

You might also like