Tutorial 1: Hardware recommendations


Introduction

A professional neuroimaging workstation has the following characteristics:
  • The workstation is always powered on
  • It is intensively used for large and complex calculations
  • The system should be optimized for both speed and safety
  • The risk of hardware failure should be minimal
  • Down-time due to maintenance should be minimal

To conform to these workstation characteristics, enterprise class components should be used and preventive measures should be taken to cope with hardware failures. This guide addresses the issue which hardware to choose for which situation.

If you simply want a list of recommended hardware, take a look at the Dell configurations below (see Attachements). These specifications should easily be transferable to other vendors such as HP, etc. Note that these configurations contain no peripherals such as keyboards, monitors. All systems have at least RAID level 1 data protection, meaning that one disk can break down in the machine without losing data. These three system configurations will be updated regularly.

  • Entry level workstation (approx USD 1300)
  • Mid-range workstation (approx USD 1700)
  • High-end workstation(approx USD 2700)


Processor (CPU)

The choice of CPU has a large effect on the purchase cost of a workstation. However, once the price is payed it is quickly forgotten and what remains is the processor performance that affects your daily work. This section provides recommendation to help make a decision about what a reasonable investment is for you.

Chassis and Mainboard

The influence of the mainboard and power supply on a overall workstation performance and stability cannot be overemphasized. However, in out of the box computers, this component is often of an unknown source and its quality can therefore not be assessed before buying the machine. The recommendations in this section will be mainly relevant if you plan to have a custom workstation built.

Peripherals

A mouse, keyboard and monitor; these components make a workstation complete. There are many choices and as many personal preferences. Here the options that one could think of are discussed.

Random Access Memory (RAM)

How much memory should a computer have to run neuroimaging analyses. This section will provide a rule of thumb to answer this question. Also recommendations regarding the type of memory are given.

I/O storage subsystem

The Input/Output subsystem refers to how much information can be transferred to the CPU and RAM per second. This number has a great influence on overall performance. This section will focus on one of the weakest chains in modern I/O systems: the storage system or hard disk. The hard disk should not only be fast, but it should also safely store your data.

Graphics (GPU)

The choice of Graphics Processing Unit (GPU) has a similarly large effect on the purchase cost of a workstation as the choice of a CPU. Again, once the price is payed it is quickly forgotten and what remains is the GPU performance that affects your daily work. This section provides recommendation to help make a decision about what a reasonable investment is for you.

ċ
entry_level_workstation.pdf
(199k)
Niels van Strien,
Apr 9, 2011, 10:29 AM
ċ
high-end_workstation.pdf
(189k)
Niels van Strien,
Apr 9, 2011, 10:47 AM
ċ
mid-range-workstation.pdf
(190k)
Niels van Strien,
Apr 9, 2011, 10:41 AM
Comments