Very easy question what's the difference?
 

What's the difference between server hardware and normal computers?

Does server hardware have a bigger fan, bigger hard drive and emergency electricity generator? And I know some servers come with different CPUs.

Is that right or wrong?

What other differences are there?

It all depends on what type of server your running and what purpose. You can have a 'regular' computer be a server... Though most of the time if I think server I think rack mount, because that's what I have dealt with. Though I do use my home computer for servers from time to time, infact we are working on a security project with some friends and are using standard hardware to run a video server. When it comes down to servers, it comes down to how much money you want to spend on it. Though HDDs don't change.... Well you can get SCSI but they are still the same size, just a different protocal that's faster.

[edit] I forgot, a lot of people are now using the electric backup units. If you work at home and a blackout hits you'll need time to save the project... It's really annoying when blackout hits and you loose all data. =(

Normally you can spec it more before purchase (for real servers), but loosely;

more cpu/cores, bigger RAM, more & bigger disks
also, faster cpu/cores, RAM, disks, faster/more bandwidth on backplane (I/O bus)
longer MTBF (mean time between failure) ie components last longer before dying
may have SCSI disks internally => longer warranty => more expensive

Its grey area though, you can use top end std boxes as servers and low end servers as std box.
It depends what you need vs what you can afford.
In some cases, only certain HW is supported by eg RHEL, Oracle etc.

There isn't necessarily a pat answer, and it will depend on your environment and needs. That said, a server will typically have more redundancy and more capacity for throughput-i/o. For example, my Sun T5220's have, on a single T2 chip, 8 core that can run 8 threads per core (looks like 64 cpus to the OS), 8 FPU, 8 encryption accelerators, and circuitry for two 10GigE connections. They have 6 PCI expansion slots (a mix of PCIE and PCIX), a DVD R/W drive, 8 internal hard drives with options for hardware raid, 4 GigE interfaces, dual power supplies, dual banks of multiple fans, a really high bandwidth bus architecture that ties all that together, . . .

The interesting thing is that the clock speed is maybe half what you could get on a laptop. So, if you had a single threaded application that was compute intensive, it would actually run faster on the laptop. However, if you had a thousand users connecting to multithreaded database, transaction processing, or web infrastructure, the laptop would choke to death, while the server would just truck along. Virtually all the redundant components are hot swappable, so, if something fails, you get error logs, indicator lights, etc., and the server keeps running. You swap in a replacement, and the server just keeps running.

The server typically has no need of a graphics interface. You drop all the baggage of the software required to run a graphical interface (use server edition OS), you don't have any graphics card, and you have no monitor. A laptop or personal computer is often focused on graphic performance, because that is the face that the user sees and feels.

A server would typically have a console interface that you could tie to and have a text only command line interface to receive error messages, alerts, etc., and to issue commands. On my T5220's, there is an ILOM (Integrated Lights Out Management) interface. I connect with a terminal session over a serial line to this interface. It is actually a computer within a computer (a single small board). It happens to be a PowerPC running a linux kernel with software for the ILOM interface and hardware connections to sensor and controllers for the server. I can shutdown or start the server from the ILOM remotely. I can check on the status of various hardware components. Even if the server were ever completely locked up, I could force it to power off and then start it up again, remotely.

All that said, it's not uncommon for people to run server applications off a box that was never intended to be a server. It's a budget approach. However, with the increasing availability of virtualization, it's not that hard for someone to budget for a virtual server that is hosted and managed by someone else. It is not really a physical machine, but rather a container running on a blade or physical machine in a farm somewhere, sharing resources with other VM's. Such can be had on a budget and is allocated resources according to that budget.

choogendyk's right. I forgot about the hotswappable stuff and everything. I usually don't work with the physical parts of a server, so normally I don't care. The most I've gotten with physical server work is plugging in cat5... lol But yeah that's a good description.

Another thing I didn't think of in my first post. Several years back, when I wanted to add memory to a DEC AlphaServer, I found that it would take standard PC memory. However, it had to be ECC memory. That's Error Correcting memory, which has extra bits for parity, and is somewhat more expensive. One document I found went over the probability of a memory bit getting zapped by a gamma ray. Pretty small. But, if it happened, it could cause a system to become corrupted, hang, or crash. For a desktop or personal computer, that's no big deal. They are going to reboot pretty often and get a clean start anyway. Some servers, however, run for a very long time without being taken down or rebooted. For those, creeping corruption from gamma rays can be an issue. Thus, ECC memory.

Yes I remember the ECC, some also required registered memory? I don't remember the details but PC's do it to. Registered vs Un Registered... If I remember right aren't you able to run more registered memory then unregistered?