OSDev.org

Is it true that increasing the hard disk size(replacing a 40GB with a 80GB) reduces performance?

yes, it reduces the performance, but only if you fill it up with data and dont defrag.

Far as I know, no. Not by definition at least, in some cases it'll speed it up (80GB is denser or has more platters, 40GB of data is then in half the seek time). It really depends a lot on what type of 40gb and what type of 80gb, as well as what filesystem(s) you'll be using and what type of usage you'll have.

I had read an article a few days ago, about hdparm, and ended up refreshing myself on IDE. I would have to agree with Candy based on the fact that more platters should mean more heads. If not the fact that the data is could be packed more dense. That would give higher potential data throughput. Potential because as I understand it, someone correct me if I am wrong, but after a quick google search most hard disks are already pushing the PCI bus throughput design limit. (SATA controllers sit on the PCI bus.).

http://en.wikipedia.org/wiki/Serial_ATA
"...a smaller power and interface cable plus the ability to hot-plug are the most practical SATA benefits to everyday computing."

The real deal should be with the newer technologies replacing the PCI bus that is used in almost all computers today, and many more advancements in the SATA technology or others in my opinion. So there should be no real reason not to plug in your 80 gig in favor of a smaller disk.

Do not forget that if a 80GB could produce y times the throughput that this could also become >y times near the outter edge of the disk by making your high accessed partition closer to the first sector on the disk.

Not exactly. The improvement of 133MB/s to 150/300 is pointless, since the usual PCI (33mhz 32bit) is limited to 133MB/s (33mhz * 32 bits / hz = 133M*8 bits). The harddisks themselves are however not actually transferring 133MB/s.

It's a weakest link idea, where the slowest link or the link causing the most delays is responsible for the speed. Your harddisk produces (values are for my harddisk in a test) 59MB/s max throughput for sequential read, with a more usable 12MB/s for Windows bootup throughput. That means, if I connect this to UDMA33 (however) the sequential throughput would be limited to 33MB/s immediately. If I connect it to UDMA66 as master, it'd be 59MB/s, if it's a slave it'll be tied up by bus overhead (or so I heard - need to get facts for this statement).

Older harddisks commonly have less density on the platter, less density per rotation (of your heads), but possibly more platters to allow the size with the then-current technology. The first two will increase throughput for newer disks (by a lot), the third will reduce it (also by a lot). However, if your new disk is in the same generic ballpark as the previous they should be pretty much equal in number of platters, especially since multi-platter harddisks tend to jump in price.

Generic advice is, if you didn't spend big bucks on your previous harddisk, any new one will be faster, pretty much. The bus speeds shouldn't be a problem, but given fast enough disks all operating at top speed they could clog up the bus.