Step out of the Box, Step Into a Cube
Computers for CFD and FEA Simulation

PADT offers a line of high performance computing (HPC) systems specifically designed for CFD and FEA number crunching aimed at a balance between cost and performance. This concept is called High Value Performance Computing, or HVPC. These systems have allowed PADT and its customers to carry out larger simulations, with greater accuracy, in less time, at a lower cost, than is possible with name-brand solutions. This leaves you more cash to buy more hardware or software.

Assembled by PADT's IT staff, CUBE HVPC systems are delivered with the customer's simulation software loaded and tested. Each system is configured specifically for simulation, where choices are based upon PADT's extensive experience using similar systems for the same kind of work. Only features that simulation users need are included, and the focus is put on the hardware and setup that delivers performance. Keeping with the "no frills" approach, each system comes with 4 hours of phone support and a 1 year parts warranty.

CUBE HVPC Systems are optimized to deliver the most performance per unit of cost, resulting in systems that may be 20% to 40% slower than similarly configured HPC systems from larger suppliers, but at a cost that is 50% to 75% less. There are better known brand names out there for those that want maximum performance, 24/7 support, or a recognized name. However, if you know what you are doing and need the most bang for your buck and that "corporate mandate" solution is just not working – CUBE systems are designed just for you.

Flexible Configurations to Maximize ROI
PADT offers a line of base systems so users can get gauge the pricing compared to other solutions. However, customers are encouraged to contact PADT to configure a custom system to meet specific needs.

Updated 2012 pricing and base systems! Click on the image below and download our latest CUBE HVPC brochure.

Call 1-800-293-PADT, 480-813-4884, or e-mail sales@padtinc.com to Configure your system now.

How the heck are you making money today!

Optimized Component Selection

PADT’s IT staff has spent considerable time picking each component used in a CUBE HVPC system, choosing the components as an optimal balance between cost and performance.  Learn more about which components we use by visiting the component page. (Click Here)

Benchmarks

Please visit our benchmark page to see the performance of these systems. (Click Here)

This page includes ANSYS R13 and ANSYS R14 benchmarks.

Accessories

PADT has pre-selected some of the more commonly needed accessories.  Visit our accessories page to find out more. (Click Here)

FAQ

CUBE High Value Performance Computer FAQ's (Click Here)

Base Systems

Summary of CUBE HVPC Base System

Updated CUBE HVPC Systems for 2012! now with AMD Opteron 32nm Interlagos Processors

Base System Details

C128SE-IB
Mini-Cluster

128 Core (2x4x16 2.3GHz AMD 6276) / 2 x 2U Rack / 512GB DDR3 1600 ECC RAM / 256 GB SSD SysDrive / 6 TB Data Drive, RAID 0, 6 x 1TB SATAII / 2 x GigE / 40Gbps Infiniband / Matrox 16MB / Linux 64-bit / Mobile Rack / KVM / Keyboard / 17” LCD Monitor / 8 port GigE Switch / 2 x 1.4 KW UPS’s

This system is perfect for large CFD or Explicit Structural jobs that can take advantage of all those cores and memory for a single run. 

The C128SE-IB is an incredible amount of power and flexibility in a very practical package.  We start with two 2U enclosures that each contains 4-16 core AMD Opteron 32nm Interlagos processors. These enclosures use the latest Supermicro motherboards that deliver great performance, and have a couple of special features that are great for simulation.  The most important is that fact that, with their Infiniband cards, we connect the two boxes over Infiniband without an Infiniband switch.  Saving thousands on the switch and extra cables.  The next is that they use a cheaper form of memory with a 8 slots per CPU, allowing us to use relatively inexpensive DDR3 1600 MHz 8GB RAM sticks.

We package the two boxes in a small, 12U rack on wheels, add a GigE switch, a keyboard (in a tray), a small monitor and 2 UPS’s.  This leaves a 3U space for you to add anything else you need in your new, portable mini-cluster.  At 100% CPU utilization the UPS’s last about 15 minutes so you can actually move your system while keeping your jobs running! (See Video)

We pack each box with 3 TB of fast (SATA II) disks to give you enough room to solve at high speed and store a result file or two.  The system drive is a 256 GB solid state drive (SSD) to speed up the OS and swap space.

If you have big problems that you have trouble fitting on your current hardware, cannot seem to get time on your corporate cluster, or need something hefty for your secure area, this is the system for you!

c48
Compute Server

48 Core (4x12 2.6GHz AMD 6238) / 1U Rack / 256GB DDR3 1600 ECC RAM / 256GB SSD SysDrive / 3 TB Data Drive, RAID 0, 3 x 1TB SATAII 7.2K/ 2 x GigE / Matrox 16MB / Linux 64-bit

 

Designed for CFD and Explicit Structural users who have large, but not huge, problems, the C48 is a very cost effective way to increase job turn around for less than $15,000.  FEA users might find it a very cost effective way to run multiple jobs in parallel, utilizing 4-8 cores per job which is usually a sweet spot for FEA solvers.

The system starts with a Supermicro 1U barebones system that has four CPU sockets and 8 ram sockets per CPU.  This lets us put in the affordable 2.6GHZ AMD 6238 CPU’s and 256 GB’s worth of affordable 8GB stick memory.  We then add 3 1TB SATAII drives in RAID0 to give good I/O performance and lots of scratch place to use for solves and to store data on.  The OS and swap then get loaded onto a 256GB GB solid state drive (SSD) to give that little bit of extra performance without breaking the bank.

The whole thing takes up one slot in your sever rack and can plug into your GigE network.  See the accessories page for cabinets and other server add-ons

c32
Compute Server

32 Core (4x8 3.0GHz AMD 6220) / 1U Rack / 256GB DDR3 1600 ECC RAM / 256GB SSD SysDrive / 3 TB Data Drive, RAID 0, 3 x 1TB SATAII 7.2K/ 2 x GigE / Matrox 16MB / Linux 64-bit

 

For most users a 32 core system is a nice sweet spot.  CFD and Explicit Structural users give up a few cores compared to a 48 cores system for some considerable savings, but they still get the RAM and disk space they need. FEA users might find this an attractive system because it gives them access to 128 GB of RAM for very large single jobs or running multiple jobs on 4-8 cores each. 

The system starts with a  Supermicro 1U barebones system that has four CPU sockets and 8 ram sockets per CPU.  This lets us put in the affordable 3.0GHZ AMD 6220 CPU’s and 256 GB’s worth of affordable 8GB stick memory.  We then add 3 1TB SATAII drives in RAID0 to give good I/O performance and lots of scratch place to use for solves and to store data on.  The OS and swap then get loaded onto a 256GB GB solid state drive (SSD) to give that little bit of extra performance without breaking the bank.

The whole thing takes up one slot in your sever rack and can plug into your GigE network.  See the accessories page for cabinets and other server add-ons.

PADT really feels that this system sits at a nice price vs. performance vs. capacity point.

w16
Simulation Workstation

16 Core (2x8 3.0GHz AMD 6220) / Full Tower / 64GB DDR3 1600 ECC RAM / 256GB SSD SysDrive / 3 TB Data Drive, RAID 0, 3 x 1 TB SATA II / 4 x GigE / AMD ATI V5900 / Windows 7 64-bit or Linux 64-bit

 

Many users need some horsepower on their desktop. The W16 is designed for three kinds of users. The first need capacity to mesh and post process their large CFD, Explicit Dynamics, or FEA models. The second run medium to large sized CFD and Explicit Dynamics problems that will fit in 64GB and 16 cores. The last group makes up the vast majority of FEA users. With 16 cores they can take advantage of parallel solving. With 64 GB they can fit even larger models into in-core memory solvers, and they will have the bandwidth to solve small and medium problems whiles still doing CAD, pre/post processing, or productivity work.

The Supermicro based workstation motherboard is put into a generic workstation tower with lots of room to breathe. We use two (3.0GHz) AMD 6220 eight core chips to give 16 cores. We then use the SSD drive from our server line to give 256 GB of space for Windows to use. To top it all off we add in 3 x 1TB SATA II  drives. They are put into an RAID 0 array using the integrated LSI 2008 SAS2 Controller. This offers approximatly 3TB of disk space to make models, solve, review results and even store a few models.

Since this system is designed for interactive work, we add in a AMD ATI V5900 graphics card. We also load Windows 7, under the assumption users will want to run their CAD, math, and office productivity software on the same machine. Linux 64-bit is also an option.

This is an excellent workstation for $7,400. Yet another affordable way to solve many CFD or FEA capacity problems.

**Optional Add-ons/Configurations

• Linux 64-bit Operating Systems
• Quadro 6000 with 6GB of GDDR5 RAM
• Nvidia Tesla C2075 GPU
• Infiniband 40 Gbps interconnect or 10GigE Ethernet
• Additional RAM configuration - 128GB DDR3 1600 MHz

w12i
Simulation Workstation

12 Core (2x6 3.47GHz INTEL x5690) / Full Tower / 96GB DDR3 ECC 1333 RAM / 256GB SSD SysDrive / 1.7 TB Data Drive, HW RAID 0, 3 x 1 GB SATA II / 2 x GigE / AMD ATI V5900 / Windows 7-bit 64 or Linux 64-bit

 

The W12i is a full tower workstation that uses two INTEL XEON X5690 processors. It was designed for three kinds of users who desire high frequency core clock speeds in addition to a large amount of RAM. The first need capacity to mesh and post process their large CFD, Explicit Dynamics, or FEA models. The second run medium to large sized CFD and Explicit Dynamics problems that will fit in 144GB and 12 cores. The last group makes up the vast majority of FEA users. With 12 INTEL XEON 3.47GHz cores they can take advantage of parallel solving. With 96 GB they can fit even larger models into in-core memory solvers, and they will have the bandwidth to solve small and medium problems whiles still doing CAD, pre/post processing, or productivity work.
The Supermicro based motherboard is put into a generic workstation tower with lots of room to breathe. We use two (3.47GHz) INTEL X5690 six core chips to give 12 cores along with 18 memory slots. Those slots are then filled with 96GB of DDR3 ECC 8GB memory sticks. We then use the SSD drive from our server line to give 256 GB of space for Windows to use. To top it all off we add in 3 x 1TB SATA II drives. They are put into an RAID 0 array using a hardware raid controller. This offers approximately 3 TB of disk space to make models, solve, review results and even store a few models. The system also comes with USB 3.0 and a top mounted hard drive docking bay.
This system is designed for interactive work, we add in a AMD ATI V5900 graphics card. We also load Windows 7, under the assumption users will want to run their CAD, math, and office productivity software on the same machine. The base system with the AMD ATI V4900 graphics card is $11,750.

Yet another affordable way to solve many CFD or FEA capacity problems.

**Optional Add-ons/Configurations

• Linux 64-bit Operating Systems
• Quadro 6000 with 6GB of GDDR5 RAM
• Nvidia Tesla C2075 GPU
• Infiniband 40 Gbps interconnect or 10GigE Ethernet
• 72GB DDR3 1333 ECC RAM

w12a
Simulation Workstation

12 Core (2x6 3.3GHz AMD 4238) / Tower / 64GB DDR3 1600 ECC RAM / 256GB SSD SysDrive / 1.5 TB Data Drive, RAID 0, 3 x 500GB SATAII 7.2K/ 2 x GigE / AMD ATI V5900 / Windows 7 64-bit or Linux 64-bit

 

The W12’s is our most economical model in the CUBE HVPC line. This workstation is a multi-socket many core workstation for under $6,300. Three users are likely to use this economical workstation. The first need capacity to mesh and post process their large CFD, Explicit Dynamics, or FEA models. The second run small and medium sized CFD and Explicit Dynamics problems that will fit in 64GB and 12 cores.  The last group makes up the vast majority of FEA users.  With 12 cores they can take advantage of parallel solving.  With 64 GB they can fit even large models into in-core memory solvers, and they will have the bandwidth to solve small and medium problems whiles still doing CAD, pre/post processing, or productivity work.

The dual socket workstation motherboard is put into a generic workstation tower.  We use two of the faster (3.3GHz) AMD 41238 six core chips to give 12 cores and 16 memory slots.  Those slots are then filled with 64GB of affordable 4GB memory sticks. We then use the SSD drive from our server line to give 256 GB of space for Windows to use.  Then 3 500GB SATAIII drives are put in a RAID0 configuration to offer 3 TB of disk space to make models, solve, review results and even store a few models.

Since this system is designed for interactive work, we add in a AMD ATI V5900 graphics card. We also load Windows 7, under the assumption users will want to run their CAD, math, and office productivity software on the same machine.
This is an excellent workstation for under $6,300