Consider this comparison. A member of staff is paid $50000 per year. Allowing for overheads, the cost of that person is $100000 per year. You buy them a computer. A cheap computer with a reasonable screen costs $1400 and has a 1 TB (TeraByte) magnetic disk. The current retail price of a 1 TB magnetic disk is $79. A 128 GB (GigaByte) SSD (Solid State Disk) is $69. If you select an SSD instead of a magnetic disk, you save $10 and your $100000 per year person works faster.
Now lets make the comparison a little more realistic. Most companies buy computers from Dell. Dell charge you $325 to switch from the $79 magnetic disk to the $69 SSD. The $325 pure-profit-for-dell fee is a cost of only $1.35 per working day. Your worker costs $417 per day. If your worker saves only 34 seconds per day, the Dell ripoff price has paid for itself.
$135 will get you a fast SSD (not from Dell) and you can save more seconds per day. At a cost of $0.50 per day, a fast SSD could save you 5 minutes per day. At $417 par day, 5 minutes is $4.34 saved.
No matter how you perform the calculation, carefully selected hardware saves you money. Fast hardware is a good investment. The cost is irrelevant.
So what is a good choice versus a bad choice? You need to know a little bit about the work performed by the worker. A person editing photographs will require a more expensive screen than a person not editing pictures. A person working on large files might need an SSD larger than 128 GB. Very few people require the fastest processor that many people buy. But the difference in price is trivial so give them the overpowered processor and give them the bits they really need.
Modern quality computers last 5 years. They are obsolete in 3. You should replace notebook computers because the new ones are lighter, not because of failure or lack of speed. Companies have a a real problem replacing perfectly workable computers frequently yet they do exactly that with cars. You use a company car for 3 years then replace it. Replacing a car after 3 years is understandable with most American and European cars but not with a Toyota where 6 to 10 years of use is reasonable. Corporate Toyota's are replaced on exactly the same 3 year schedule as a Ford or a Kia.
Computers evolve at the same speed as cars. You get trivial changes each year plus useful changes every 3 or 4 or 5 years. USB 3 is more useful than USB 2. Companies upgrade cars and computers totally without consideration of those real improvements. They buy a new car with no real improvements in fuel economy or anything else. Computers are treated the same.
One difference is the cost of a mistake. If you overspend on the processor, a computer will cost you only $200 more or just $66 per year for the three years. If you make a similar overspend on a car, say a 6 cylinder car instead of a 4 cylinder car, the cost difference up front could be $5000 and the extra cost per year in petrol might be $1000 plus you will pay extra for insurance, servicing, and registration. Overspending on a car quickly adds up to the point where you question the value of extra power and features. The same degree of overspend in computers is trivial.
A jump from a $20000 suburban sedan to a high performance $80000.00 V8 is equivalent to a jump from a $1500 computer to a $2000 computer. Your extra spend is $500, not $60000. The extra $60000 spent on a car might save you 3 minutes per day, less the 5 minutes you spend talking to the traffic cop who pulls you over for speeding. The extra $500 you spend on your computer might save you 5 minutes per day. Why do companies happily paytens of thousands of dollars extra for cars but not a few extra dollars for a computer?
I admit to using Microsoft Excel for complex management reporting applications. Excel can perform magic when the data fits into the Excel range of data. A data warehouse summary report processed on a big server can take 20 minutes while the same data exported from the $40000 server to a $2000 desktop can consume over an hour. In one case the cost of enhancing a server to run a report under 10 minutes was over a $100000 while the cost of improving the speed of the Excel desktop process to a similar level was $4000.
When you look at big infrastructure upgrades, the cost can be staggering. the cost of desktop upgrades is truly insignificant compared to the improved productivity.
Servers are also underpowered
The typical server is also underpowered. Not because servers are slow, modern hardware is so fast that a server does not have to be slow. The speed loss is caused by the use of virtual machines.
Virtual machine software swallows from 10% up to 60% of the computer processing. When you do the maths, it is cheaper to run two servers than it is to squeeze the two systems onto one server with a 60% overhead.