I can imagine a system at the very high end, dual xeons, dual quadros, 64 gigs ram, raid 5 ssds, dual 30'' monitors. Is that not a bit overkill though considering the amount it would cost when you could still manage it on much less.
At my studio, we only have a beast like that (32 core, tonnes of RAM) as part of the render farm... Well actually, the other co-founder loved the render-farm specs so much that he uses one as his desktop, but that's the first time I've seen a desktop like that
At the moment, my work PC is a AMD FX 8core/4GHz thingy, a Radeon 78XX(?) 16GB RAM, a small SSD and a huge HDD.
My home PC is an old Intel Q6600 quad core, a GeForce 4XX(?), 2GB of ram, a small SSD and a bunch of medium HDDs...
And for testing really low-spec stuff: an even older Athlon 64 2GHz, a GeForce 8 and/or GeForce 7, with 2GB of RAM and a small HDD.
At every job I've had, it's been standard for everyone to have two screens. Two 24" screens seems to be very common, though at my last job everyone had a 22-24" and a 19".
Currently at work I've got a 24" plus a 26" in portrait mode, and at home I've got a single 22".
It is nonsensical to pay so much and then provide inadequate tools.
Despite this, I've often seen developers struggle along on their 5 year old dev machines, fighting to get a $1000 upgrade... while many, many thousands of dollars worth of man-hours are going down the drain
How about Source control and backups, as far as I can remember, Raid 1 or 5 can be used for backup(if one HD fails), is source control such as GIT also used for backup, how periodically are backups made, I assume backups are also made on USB devices like portable flash drives and such?
RAID redundancy should almost always be used on the central servers, so that if a HDD fails, you don't lose any data.
GIT is nice because every single user actually has the entire repository on their own machines -- so if you did somehow lose the central server, you can recreate it from the most up-to-date user.
Despite this, I've actually worked for a developer who lost their SVN server once -- 3 out of 5 HDDs in a RAID system failed, causing it to become unrecoverable... We had to reconstruct the server from the data on everyone's PCs, but lost the version history of all files, which actually really hurt things a lot as we were in the middle of tweaking many different systems at the time.
I've never seen RAM disks used, but they are really effective at boosting the performance of certain kinds of tasks, assuming you've got enough RAM. SSD's are the next best thing -- every PC should at least have the OS installed on an SSD IMHO
As for render farms - at the last company I worked for, when you left in the evening, you would run a batch file that connected your PC into the render-farm network and allowed it to run rendering jobs overnight while you weren't using it. The development PCs were fairly crappy, but with dozens of them on the network you get a lot of compute power out of them.