Hardware Recommendations and Certifications

View the most recent recommended graphic boards, laptops and desktop hardware configurations.

Recommended Graphics Boards

Recommended CAE/CAD graphic boards to use with HyperWorks applications.

The most recent vendor/manufacturer drivers should be used and all driver support for these cards should be addressed to the appropriate manufacturer of the graphic board.
Note: We no longer will support AMD graphics cards on Linux x86_64 operating systems in HyperWorks 2021 and higher products.

AMD Graphics Cards

Products GPU Model Driver Version
Radeon™ Pro VII

WX 9100

W5700(*)

WX 8200

W5500(*)

WX 7100

WX 5100

WX 4100

WX 3200

WX 3100

WX 2100

Windows 10 (64-bit)
20.Q3 (20.Q4*)
Linux (64-bit)
Not Supported
Radeon™ Pro Mobility WX 7130

WX 7100

WX 4170

WX 4150

WX 4130

WX 3100

WX 2100

Windows 10 (64-bit)
20.Q3 (20.Q4*)
Linux (64-bit)
Not Supported

NVIDIA Graphics Boards

Products GPU Model Driver Version
M

(Maxwell)

P

(Pascal)

V

(Volta)

Turing (RTX)
Quadro Series M2000

M4000

M5000

M6000

P400

P420

P600

P620

P1000

P2000

P2200

P4000

P5000

P5200

P6000

GP100

GV100 RTX 3000

RTX 4000

RTX 5000

RTX 6000

RTX 8000

Windows 10 (64-bit)
451.77
Linux (64-bit) ODE Long Live
450.57
Quadro Mobility M500M

M520M

M600M

M620M

M1000M

M2000M

M2200M

M3000M

M4000M

M5000M

P500

P520

P600

P620

P1000

P2000

P3000

P3200

P4000

P4200

P5000

P5200

N/A T1000

T2000

RTX 3000

RTX 4000

RTX 5000

Windows 10 (64-bit)
451.77
Linux (64-bit) ODE Long Live
450.57
Note:
Minimum OpenGL 3.2 and OpenCL 2.1 Requirement
Virtual server/clients and VirtualGL setups may work, but are not officially tested or supported.
NVIDIA Optimus or AMD Switchable Graphics
In order to ensure best performance, these options should be set to use discrete NVIDIA or AMD GPU and not the Intel GPU.
Power Options and Mobility Center
In order to ensure best performance, these options should be maximum performance for both GPU and CPU.
Graphics Driver Corruption or Installation Issues
In order to ensure best driver compatibility, it is recommended to use “Custom” and “Clean” install options instead of the general “Express” driver installer options.

Recommended Workstation Desktop and Laptop/Notebook Hardware

DELL Workstations - Desktops (1 of 2)

Product Workstation Model
Precision Workstation 3420T / 3430T 3440 SFF 3460 (mini) 3620T / *3630T

*3930 Rack

NVIDIA Quadro GPU P400

P600

P620

P1000

P400

P620

P1000

RTX 3000 P400

P600

P620

P1000

P2000

P4000

P5000

*RTX 4000

*RTX 5000

AMD Radeon™ Pro GPU WX 2100

WX 3100

WX 4100

WX 3200 N/A WX 2100

WX 3100

WX 4100

WX 5100

WX 7100

DELL Workstations - Desktops (2 of 2)

Product Workstation Model
Precision Workstation 3640T / 3930 Rack* 5810 5820 / 7820 / 7920
NVIDIA Quadro GPU P400

P620

P1000

RTX 4000

RTX 5000

RTX 6000*

P400

P600

P1000

P2000

P4000

P6000

P400

P620

P1000

P2200

RTX 4000

RTX 5000

RTX 6000

RTX 8000

AMD Radeon™ Pro GPU W5500

W5700

WX 3200

WX 4100*

WX 7100

W7100

W8100

W9100

WX 4100

WX 5100

WX 7100

W5500

W5700

WX 3200

DELL Workstations - Laptops (1 of 3)

Product Workstation Model
Precision Workstation 3550 3551 5530 5540 / 5740
NVIDIA Quadro GPU P520 P620 P1000M

P2000M

T1000

T2000

RTX 3000*

AMD Radeon™ Pro GPU N/A N/A N/A N/A

DELL Workstations - Laptops (2 of 3)

Product Workstation Model
Precision Workstation 5520 AIO 5520 5550 / 5750* 5720 7520
NVIDIA Quadro GPU N/A M1200M T1000

T2000

RTX 3000*

N/A P3000M

P5000M

AMD Radeon™ Pro GPU WX 4150 N/A N/A WX7100M WX 7100

DELL Workstations - Laptops (3 of 3)

Product Workstation Model
Precision Workstation 7530 7540 / 7740 7550 / 7750 7720 7730
NVIDIA Quadro GPU P2000M

P3200M

P4000M

T1000

T2000

RTX 3000

RTX 4000

RTX 5000

T1000

T2000

RTX 3000

RTX 4000

RTX 5000

M1200M

P3000M

P4000M

P5000M

P3200M

P4200M

P5200M

AMD Radeon™ Pro GPU WX 4150 WX 3200

WX 7130

N/A WX 7100 WX 4150

WX 7100

Lenovo Workstations - Desktops (1 of 2)

Product Workstation Model
Lenovo ThinkStation P320 SFF /

P320 TWR*

P330 SFF /

P330 TWR*

P340 SFF /

P340 TWR*

NVIDIA Quadro GPU (442.92 or higher) P400

P600

P1000

P2000*

P4000*

P400

P620

P1000

P2000*

P2200*

P4000*

P400

P620

P1000

P2200

RTX 4000*

RTX 5000*

AMD Radeon™ Pro GPU N/A N/A N/A

Lenovo Workstations - Desktops (2 of 2)

Product Workstation Model
Lenovo ThinkStation P520 P520c P620 (AMD Ryzen PRO 3975X) P720 / P920 NEC (mini)
NVIDIA Quadro GPU (442.92 or higher) P400

P620

P1000

P2200

RTX 4000

RTX 5000

RTX 6000

P400

P620

P1000

P2200

RTX 4000

RTX 5000

P620

P1000

P2200

RTX 4000

RTX 5000

RTX 6000

RTX 8000

GP100

P400

P620

P1000

P2200

RTX 4000

RTX 5000

RTX 6000

RTX 8000

P1000
AMD Radeon™ Pro GPU N/A N/A W5500

W5700

N/A N/A

Lenovo Workstations - Laptops (*Windows 10 support only) (1 of 3)

Product Workstation Model
Lenovo ThinkPad P1

Gen1

P1

Gen2*

P1

Gen3

P14s P14s

Gen1

P15

Gen1

P15s

Gen1

P15v

Gen1

NVIDIA Quadro GPU P1000M

P2000M

T1000

T2000

T1000

T2000

P520 P520 T1000

T2000

RTX 3000

RTX 4000

P520 P620
AMD Radeon™ Pro GPU N/A N/A N/A N/A N/A N/A N/A N/A

Lenovo Workstations - Laptops (*Windows 10 support only) (2 of 3)

Product Workstation Model
Lenovo ThinkPad P17

Gen1

P40

Yoga

P43s* P50 P50s P51 P51s
NVIDIA Quadro GPU T1000

T2000

RTX 3000

RTX 4000

RTX 5000

M500M P520 M1000M

M2000M

M500M M1200M

M2200M

M520M
AMD Radeon™ Pro GPU N/A N/A N/A N/A N/A N/A N/A

Lenovo Workstations - Laptops (*Windows 10 support only) (3 of 3)

Product Workstation Model
Lenovo ThinkPad P52 P52s P53* P53s* P71 P72 P73*
NVIDIA Quadro GPU (442.92 or higher) P1000

P2000

P3200

P500 T1000

T2000

RTX 3000

RTX 4000

P520 M620M

P3000

P4000

P5000

P600

P2000

P3200

P4200

P5200

P620

T2000

RTX 3000

RTX 4000

RTX 5000

AMD Radeon™ Pro GPU N/A N/A N/A N/A N/A N/A N/A

Acer Workstations and Laptops (*Windows 10 support only)

Product Workstation Model
Acer ConceptD

500*

ConceptD

700*

Veriton

K8

NVIDIA Quadro GPU RTX 4000 RTX 4000 RTX 4000
AMD Radeon™ Pro GPU N/A N/A N/A
Product Mobile Workstation Model
Acer ConceptD 3

Pro*

ConceptD 5

Pro*

(CN515-71P)

ConceptD 5

Pro*

(CN517-71P)

ConceptD 7

Pro*

NVIDIA Quadro GPU T1000 T1000 RTX 3000 RTX 3000

RTX 5000

AMD Radeon™ Pro GPU N/A N/A N/A N/A

Altos Workstations and Laptops (*Windows 10 support only)

Product Workstation Model
Altos

Computing

BrainSphere™

P130 F5

BrainSphere™

P530 F4

NVIDIA Quadro GPU RTX 2000 K420

P400

K620

K1200

P1000

P2000

P4000

P5000

P6000

GP100

RTX 2000

RTX 4000

RTX 5000

AMD Radeon™ Pro GPU N/A N/A

Comments

For NVIDIA GPU based laptops/notebooks the Optimus power saving option in the BIOS should be disabled and the NVIDIA drivers properly installed for optimal performance in HyperWorks.

For AMD GPU based laptops/notebooks; the Enduro/Switchable Graphics power saving option should be disabled and the AMD drivers properly installed for optimal performance in HyperWorks.

Optimus (Intel/NVIDIA) enabled drivers may create performance issues with notebooks/laptops compared to a dedicated non-shared GPU driver. Disabling the Optimus feature in BIOS, if available, will help give the best overall graphics performance.

Disable nView Window manager under NVIDIA drivers if you experience random crashes and/or issues.

All power saving modes, settings and governors for CPU frequencies and GPU performance should be set to maximum settings in order to get the optimal performance out of HyperWorks. This includes smooth graphics and high frame rates (FPS) on Windows and Linux platforms.

HyperWorks 2021 Solver Hardware Configuration Recommendations

Recommended hardware configurations for HyperWorks Solvers.

AcuSolve Solver

Problem Size Small Medium Large
Typical Workload

Steady State or Transient

Steady state: Up to 1M nodes

Transient: Up to 100K nodes

Steady state: Up to 10M nodes

Transient: Up to 1M nodes

Steady state: Greater than 10M nodes

Transient: Greater than 1M nodes

Throughput 1 Single job Single job Single job
CPU 2 Dual CPU socket

For example, Intel Xeon Gold “Cascade Lake” or “Skylake” or AMD EPYC 7002 series

Dual CPU socket

For example, Intel Xeon Gold “Cascade Lake” or “Skylake” or AMD EPYC 7002 series

Dual CPU socket

For example, Intel Xeon Gold “Cascade Lake” or “Skylake” or AMD EPYC 7002 series

Number of CPU / node 1-4 1-4 1-4
Number of cores / node 32 – 128 32 – 128 32 – 128
Number of nodes 1-8 8 – 48 >48
Minimum Memory Configuration / node 3 300MB to 3GB 3GB to 30GB More than 30GB

(3KB per CFD node)

Storage (minimum) 500 GB SATA or SSD 1.5 TB local storage 1.5 TB local storage
Network Interconnect Gigabit Ethernet Or Infiniband Infiniband or Intel Omni-path Infiniband or Intel Omni-path
Operating System Linux kernel 2.6.32 or higher

Windows 7 or 10

Linux kernel 2.6.32 or higher Linux kernel 2.6.32 or higher
GPU Yes Yes Yes
MPI Intel MPI 2018.4 or higher Intel MPI 2018.4 or higher Intel MPI 2018.4 or higher
Setup

(2000-3000 computational nodes per core)

Pure OpenMP or Hybrid OpenMP/MPI Hybrid OpenMP/MPI Hybrid OpenMP/MPI
Hyper Threading Not recommended Not recommended Not recommended

Feko Solver

Problem Size Small Medium Large
General recommendations given for MoM and MLFMM dependent on problem size in terms of number of unknowns / mesh elements. For other solution methods (FEM, FDTD, RL-GO, PO, UTD) many factors to be considered. Pure MoM: less than 50k unknowns.

MLFMM: between 100k and 500k unknowns

Pure MoM: between 50k and 100k unknowns.

MLFMM: between 500k and 5M unknowns

Pure MoM: >100k unknowns

MLFMM: >5M unknowns

Throughput 1 Single job Single large job or few jobs in parallel Single very large job or multiple jobs
CPU 2 Dual CPU socket

For example, Intel Xeon Gold “Cascade Lake” or “Skylake” or later

Dual CPU socket

For example, Intel Xeon Gold “Cascade Lake” or “Skylake” or later

Dual CPU socket

For example, Intel Xeon Gold “Cascade Lake” or “Skylake” or later

Number of CPU / node 2 2 2
Number of cores / node 32 – 56 32 – 56 32 – 56
Number of nodes 1 8 – 16 > 16
Minimum Memory Configuration / node 3 64 GB 128 GB 256 GB
Storage (minimum) 500 GB SATA or SSD 500 GB SATA or SSD 500 GB SATA or SSD
Network Interconnect Gigabit Ethernet Infiniband or Intel Omni-path Infiniband or Intel Omni-path
Operating System Linux kernel 2.6.32 or higher

Windows 7 or 10

Linux kernel 2.6.32 or higher Linux kernel 2.6.32 or higher
GPU Yes No No
MPI Intel MPI 2018.4 or higher Intel MPI 2018.4 or higher Intel MPI 2018.4 or higher
Setup Pure MPI Pure MPI Pure MPI
Hyper Threading Not recommended Not recommended Not recommended

Flux Solver

Problem type Small Medium Large
Typical Workload (depending on number of DOF, element type, and other factors) < 300 000 DOF Around 500 000 DOF Around 5M DOF
Throughput 1 Single Single Single
CPU 2 Dual CPU socket

For example, Intel Xeon Gold “Cascade Lake” or “Skylake”

Dual CPU socket

For example, Intel Xeon Gold “Cascade Lake” or “Skylake”

Dual CPU socket

For example, Intel Xeon Gold “Cascade Lake” or “Skylake”

Number of CPU / node 1 2 2
Number of cores / node 8 16 16+
Number of nodes 1 1 1-4
Minimum Memory Configuration / node 3 8 GB 16-32 GB 300GB
Storage (minimum) 500 GB SATA or SSD 1 TB local storage SSD 1.5 TB local storage SSD
Network Interconnect Infiniband or Intel Omni-path
Operating System Linux kernel 3.10.0-693 or higher

Windows 7 or 10 with SSD

Linux kernel 3.10.0-693 or higher

Windows 7 or 10 with SSD

Linux kernel 3.10.0-693 or higher

Windows 7 or 10 with SSD

GPU No No No
MPI Intel MPI 2018.4 or higher Intel MPI 2018.4 or higher Intel MPI 2018.4 or higher
Setup SMP SMP or Hybrid 2MPI/node SMP or Hybrid 2MPI/node
Hyper Threading Not recommended Not recommended Not recommended

Radioss Solver

Problem Size Small Medium Large
Typical Workload

Crash & Impact

Component tests, sled test, drop test, …

Less than 500K elements

Medium crash model, between 1 and 6 millions of elements model Accurate car crash model (rupture), very large model with size > 6 million elements
Throughput 1 Single job Single large job or few jobs in parallel Single very large job or multiple jobs
CPU 2 Dual CPU socket

For example, Intel Xeon Gold “Cascade Lake” or “Skylake” or AMD EPYC 7002 series

Dual CPU socket

For example, Intel Xeon Gold “Cascade Lake” or “Skylake” or AMD EPYC 7002 series

Dual CPU socket

For example, Intel Xeon Gold “Cascade Lake” or “Skylake” or AMD EPYC 7002 series

Number of CPU / node 2 2 2
Number of cores / node 32 – 64 32 – 128 32 – 128
Number of nodes 1 8 – 16 > 16
Minimum Memory Configuration / node 3 64-128GB 64-128GB 64-128GB
Storage (minimum) 500 GB SATA or SSD 1,5 TB local storage 1,5 TB local storage
Network Interconnect Gigabit Ethernet Infiniband or Intel Omni-path Infiniband or Intel Omni-path
Operating System Linux kernel 2.6.32 or higher

Windows 7 or 10

Linux kernel 2.6.32 or higher Linux kernel 2.6.32 or higher
GPU No No No
MPI Intel MPI 2018.4 or higher Intel MPI 2018.4 or higher Intel MPI 2018.4 or higher
Setup Pure MPI Pure MPI or Hybrid with 2 or 4 OpenMP threads per MPI Hybrid with 2 or 4 OpenMP threads per MPI
Hyper Threading 5 Yes, Hybrid with 2 OpenMP per MPI Not recommended Not recommended

OptiStruct Solver

Problem type Small or medium Large static Large dynamic
Typical Workload (depending on number of DOF, element type, and other factors) Nonlinear - less than 2M DOF; linear static - less than 5M DOF; NVH - less than 5M DOF Nonlinear - more than 2M DOF; linear static - more than 5M DOF NVH - more than 5M DOF
Throughput 1 Single Single Single or few jobs in parallel
CPU 2 Dual CPU socket

For example, Intel Xeon Gold “Cascade Lake” or “Skylake”

Dual CPU socket

For example, Intel Xeon Gold “Cascade Lake” or “Skylake

Dual CPU socket

For example, Intel Xeon Gold “Cascade Lake” or “Skylake

Number of CPU / node 2 2 2
Number of cores / node 8-24 24+ 24+
Number of nodes 1 1-8 1-8
Minimum Memory Configuration / node 3 16-64GB 128GB 256GB
Storage (minimum) 512GB local storage 1TB local storage 3 TB local storage, SSD and RAID0 recommended
Network Interconnect InfiniBand or Intel Omni-path InfiniBand or Intel Omni-path
Operating System Linux kernel 2.6.32 or higher

Windows 7 or 10 with SSD

Linux kernel 2.6.32 or higher

Windows 7 or 10 with SSD

Linux kernel 2.6.32 or higher

Windows 7 or 10 with SSD

GPU Yes No No
MPI Intel MPI 2018.4 or higher Intel MPI 2018.4 or higher Intel MPI 2018.4 or higher
Setup SMP or DDM hybrid DDM hybrid SMP or DDM hybrid
Hyper Threading Not recommended Not recommended Not recommended

Recommended GPU Computing Processor List for OptiStruct

Recommended graphic boards for use with the Altair HyperWorks solver suite of applications for high-powered GPU computing.

The following table lists the recommended graphic boards for use with the Altair HyperWorks solver suite of applications for high-powered GPU computing.
Manufacturer and Model Graphics Card Driver Version

(Minimum or Higher)

NVIDIA

(Tesla)

P100

V100

Linux (64-bit)

387.26

Windows (64-bit)

391.03

NVIDIA

(Quadro)

GP100

GV100

Linux (64-bit)

387.26

Windows (64-bit)

391.03

Note: The most recent vendor/manufacturer drivers should be used and all driver support for these cards should be addressed to the appropriate manufacturer of the graphics board.

Additional Information on Driver Installations

The NVIDIA Driver Update recommendation is to use the Custom installation option and select the Perform clean installation option to validate that there are no conflicts in DLL/drivers.

The same should be done with AMD hardware and drivers as well using AMD’s custom uninstall tools.


Figure 1.
1
Number of simultaneous jobs. Use of a workload management middleware like Altair PBS is highly recommended to insure optimal and dedicated usage of the CPU resource
2
Typical node configuration is based on dual CPU socket processors
3
It is extremely important to populate all the memory banks on the mother board.
4
In Hybrid mode, it is recommended to set a number of MPIs that is a multiple of the number of sockets and then set the number of OpenMP in a way that number of MPIs x number of OpenMP equal number of physical cores.
5
Hyper Threading (HT) may increase performance by around 10% on single node. In this case, recommended setup is to run 2 OpenMP per MPI, with a number of MPIs that matches the total number of physical cores on the node. On multi-node, it is better not using HT