A question about building my own PC

[Camacha]

Get the 21.5" with the 750m graphics card.

That hardly sounds like a cost effective solution. It might be a fine option in some situations, but I don't feel this is one of them.

[ZedNova]

8 gigs is plenty for gaming and abit of multitasking, you're not gonna use that much.

Rendering, Video editing, large File conversion, modding, photoshopping + Alot of multitasking; Those are the things that require massive amounts of RAM, 16Gb is the MINIMUM if you're doing two or more of these things on a regular basis.

[LaytheAerospace]

I think you misunderstand the purpose of a memory manager. It's not supposed to minimize the memory usage, but maximize the performance and minimize the delays. If there is enough memory available, a good memory manager often uses much more of it than the minimal amount required to run the same workload with a decent performance.

Yes, that is the goal of the memory manager. The problem is, your memory manager isn't actually speeding things up, it's just increasing your memory footprint. If I can run the same workload with half the memory and the pagefile disabled, then it seems pretty obvious that my memory manager is doing a better job. No fancy tricks needed, everything just fits in nicely within main memory.

And like I said, I have direct experience with both operating systems, both personally and professionally. Windows does more with less memory. You can keep the disk cache, I'd rather have no page faults.

There are basically three resources a memory manager handles. One is the memory used by applications. If memory is scarce, the memory manager can move inactive pages to disk, freeing physical memory for other purposes. Moderate swapping causes potential delays, while excessive swapping degrades performance.

And in my case, I just purchased the necessary amount of memory (8GB) and turned off the pagefile. Bingo, no paging, no delays. Memory is effectively unlimited, because 8GB is enough to run everything I need at the same time.

I've always considered the memory and the display the most important components of a desktop system.

You've got to be kidding. The memory, while it can be cost effective to increase performance, it a miniscule component of overall system performance, and has the sharpest diminishing returns of any component. I'd much rather have a top of the line processor and budget memory than a budget processor and top of the line memory.

Edited November 18, 2014 by LaytheAerospace

[Camacha]

I think you misunderstand the purpose of a memory manager. It's not supposed to minimize the memory usage, but maximize the performance and minimize the delays. If there is enough memory available, a good memory manager often uses much more of it than the minimal amount required to run the same workload with a decent performance.

I was merely responding to the suggestion that you need 16 GB for a smoothly operating system and easy multitasking. I stand by my statement that your memory manager is subpar if you need 16 GB or RAM to achieve that. Indeed, modern operating systems are generally designed to make most of available resources, but that does not mean that adding more RAM equals a better system. I fact, most are so good that adding more RAM beyond a certain point becomes useless, and that point generally comes sooner than later.

The idea that the performance difference between 8 GB and 16 GB seems to be confirmed by benchmarks. The biggest gains can be seen when you use a part of it as a RAMDisk to put temporary files on, but that can easily be done with 8 GB or RAM.

That's actually another rule of thumb I've been using.

It sounds like your thumbs have cost you a lot of money you did not need to spend over the years

And in my case, I just purchased the necessary amount of memory (8GB) and turned off the pagefile. Bingo, no paging, no delays. Memory is effectively unlimited, because 8GB is enough to run everything I need at the same time.

I have the same experience as you do. I have run a long time with 8 GB and Windows tends to do it with ease without going over the limit. Like I said before, I have far from an average workload. The things I do are normally considered work station tasks and work flow is the decisive factor.

You've got to be kidding. The memory, while it can be cost effective to increase performance, it a miniscule component of overall system performance, and has the sharpest diminishing returns of any component. I'd much rather have a top of the line processor and budget memory than a budget processor and top of the line memory.

The PSU and storage solution are the most important parts of a PC in my book. A bad PSU will cause all sorts of trouble and can potentially damage pretty much every component in your computer. The storage is important for obvious reasons: your data is on there. You need a reliable hard disk and a backup device with the same qualifications, preferably configured to do automatic backups.

Those two things are going to provide a good basis for a reliable system. They are going to save your hind quarters when things get ugly. On any budget you need to spend the money to get those things right and only what's left can be spent on luxuries like fast CPU's and fancy graphics cards. Of course, faster cards warrant more powerful PSU's, but buying an appropriate power supply should always take precedence over getting a quicker CPU.

Edited November 18, 2014 by Camacha

[trekkie_]

As fast as ram drives are (10GB/s or more), a good SSD is just as good if not better for most tasks.

You have to remember that RAM shares bandwidth between sticks. So using bandwidth on normal tasks, while also using a RAM drive, will reduce your overall available RAM bandwidth, your memory bandwidth for both tasks are effectively halved and not as consistent.

Many programs that use up system RAM are more reserving it than using it. For example, Photoshop will reserve a large portion of RAM by default, even if you aren't using that much RAM. It reserves 5GB of my 8GB. It can be lowered in the preferences though.

If you open up another RAM intensive program, like a renderer, you can get low memory issues....even if you're not actually using all that memory. But with an SSD, loading is soo fast that there really is no huge benefit to keeping tons of memory hog apps open at once. 8GB is RAM is definitely enough for virtually any task.

[magnemoe]

That hardly sounds like a cost effective solution. It might be a fine option in some situations, but I don't feel this is one of them.

Very far from cost effective, it has two benefits, it looks good and take little space.

Downsides is that its basically an laptop integrated on the back of a screen.

You pay premium for both the laptop size and the Apple brand and get an system who is as hard to upgrade or repair as an laptop.

[Elthy]

A imac is the worst possible thing for gaming:

Slow, expensive, Mac-OS, not upgradeable...

[Camacha]

Slow, expensive, Mac-OS, not upgradeable...

Hot. Those pretty cases are not really built for aspiration. OS X can be a drawback or a huge benefit, depending on who you ask. For some people it is ideal.

[Jouni]

You've got to be kidding. The memory, while it can be cost effective to increase performance, it a miniscule component of overall system performance, and has the sharpest diminishing returns of any component. I'd much rather have a top of the line processor and budget memory than a budget processor and top of the line memory.

CPUs used to be important in the 80s and the 90s, when every upgrade meant that the system became 5x faster. GPUs were likewise important in the 90s and the 2000s. These days, apart from some special cases like gaming, everything is fast enough. The performance that matters comes from programs starting faster and file management being smoother. For those, you need a large enough SSD and enough memory that the OS can have as large disk cache as it needs.

And like I said, I have direct experience with both operating systems, both personally and professionally. Windows does more with less memory. You can keep the disk cache, I'd rather have no page faults.

I haven't really used Windows since Win2k, because it's incompatible with the tools I use at work. Still, I understand that its default memory management has been optimized for low-memory systems. In OS X, the memory management has been optimized for situations, where at least 1/3 of memory is either free or being used as disk cache. If you have enough memory, you get better performance than in Windows, while Windows handles low-memory situations better.

Right now, my desktop and laptop have 24 days and 31 days of uptime. Neither of them have resorted to swapping since the last reboot.

It sounds like your thumbs have cost you a lot of money you did not need to spend over the years

Computers are so ridiculously cheap these days that their prices don't really matter, unless you're buying servers or high-end workstations. When you're buying a tool you're going to use for thousands of hours, you don't want to be cheap in the wrong place. Spending €2000 to €3000 generally gives you the best value for your money. These days you can get pretty nice stuff for that money, while in the 90s the same sum (considering inflation) only got you a basic desktop or a cheap laptop.

[cantab]

Computers are so ridiculously cheap these days that their prices don't really matter

Speak for yourself. For the rest of us a PC is a major purchase. (Well, maybe not if you're getting a £200 netbook or something.)

Spending €2000 to €3000 generally gives you the best value for your money. These days you can get pretty nice stuff for that money, while in the 90s the same sum (considering inflation) only got you a basic desktop or a cheap laptop.

2 grand seems like way too much to me, getting into the territory where you're paying massively more to get tiny performance boosts.

[Jouni]

Speak for yourself. For the rest of us a PC is a major purchase. (Well, maybe not if you're getting a £200 netbook or something.)

A PC is a major purchase, but it's still ridiculously cheap. Those two are not mutually exclusive. The price of a new PC should be compared to the thousands of hours you're going to use it. Only a few things in life are cheaper than that.

Besides, if you live in a major city, the costs of owning a desktop computer are mostly determined by the space required for the desk. Those few extra square meters cost real money.

[LaytheAerospace]

CPUs used to be important in the 80s and the 90s, when every upgrade meant that the system became 5x faster. GPUs were likewise important in the 90s and the 2000s. These days, apart from some special cases like gaming, everything is fast enough. The performance that matters comes from programs starting faster and file management being smoother. For those, you need a large enough SSD and enough memory that the OS can have as large disk cache as it needs.

You're overestimating the impact of your disk cache, especially on desktop workloads. Cache hits are fairly rare, rarer still for games which just load their assets into memory and avoid the disk like the plague. Disk writes aren't sped up at all by a cache, because it's write through, not write back (god I hope it's not write back).

Disk caches work great for things like databases, where you can keep frequently accessed data in memory for rapid access. But on a desktop, a few gigabytes of file cache isn't going to do much, if anything, because access patterns are much more diverse. And at any rate, an SSD has a typical seek time of much less than 1ms, which puts any potential difference far below the level of human perception (100ms). So now we're down to whether or not it will speed up applications which frequently read from the disk. And it will, with varying success depending on how consistent your access patterns are, and how much of the performance is dominated by disk reads. Generally, disk reads aren't a bottleneck, because programmers are well aware that disk reads are expensive and avoid them at all costs.

There's a lot less to be gained than you think.

I understand that its default memory management has been optimized for low-memory systems. In OS X, the memory management has been optimized for situations, where at least 1/3 of memory is either free or being used as disk cache. If you have enough memory, you get better performance than in Windows, while Windows handles low-memory situations better.

*facepalm*

Windows is not optimized for low memory, it's just not wasting it. I don't need a page file, because nothing is ever paged to disk. I don't have delays switching applications, because nothing is ever paged to disk. The operating system doesn't have to optimize my memory usage, because nothing is ever paged to disk. My OS X laptop, which has that disk cache you love so much, pages things CONSTANTLY in my experience. If I wanted to cache my disk in RAM (yep, Windows has had the feature for 15 years), I could. I just choose not to. You don't have that choice, which results in needing much higher amounts of memory for the same level of performance.

Your operating system is paging things to disk precisely because it IS dealing with low memory situations. It's running out, so it has to go use the disk. My computer never runs out, so it never does that. Every allocation succeeds, every read is for a block already mapped in memory. The memory manager has the easiest job ever, and making it more complicated would have no benefit.

You're trying to spin a glaring weakness as a strength, and Windows strength as weakness. Windows doesn't page anything out because it's memory manager is inferior, while OS X chugs on the same amount of memory because it's so much better? This is absurd.

Right now, my desktop and laptop have 24 days and 31 days of uptime. Neither of them have resorted to swapping since the last reboot.

And my computer has never resorted to swapping, ever. I know because it can't. At best, your computers could equal mine, while using twice as much memory.

Also, how were you able to determine that it had never used the pagefile? Does it show total page faults since boot? I don't spend a ton of time in the Activity Monitor, never looked for such a statistic.

Computers are so ridiculously cheap these days that their prices don't really matter

*double facepalm*

Spending €2000 to €3000 generally gives you the best value for your money.

Err, no. At that level, you're getting the worst value for your money, and every additional dollar spent is worse value still. Beyond about $1000, diminishing returns starts to kick in pretty hard. You need to step back and look at what you're paying for, and what you're getting for your money.

Besides, if you live in a major city, the costs of owning a desktop computer are mostly determined by the space required for the desk. Those few extra square meters cost real money.

Most people don't live in ultradense metro areas, and don't care about what fraction of their floor space the desk uses up. They do care about the real cost of the computer. Talk about amortizing the cost over its lifetime all you want, it doesn't make the sticker price any smaller. You're also ignoring that cheaper computers also amortize their cost, so the value proposition doesn't change when you consider amortization.

It's also worth noting that people don't actually get more expensive housing to accommodate their computers. That's a fixed cost, and using it to hide the cost of the computer behind a large recurring cost is just obfuscation. You could use the same argument to say that the cost of a car is unimportant, because you'll spend more on gas and maintenance in its lifetime. But clearly, the cost is important, or everyone would be driving fancy cars.

I don't want you to take this the wrong way, but it seems to me that Apple has done a very good job on you, convincing you that the cost of the computer is unimportant, that the most expensive computers are the most economical, that needing twice as much memory for the same tasks is a good thing.

Edited November 18, 2014 by LaytheAerospace

[Jouni]

You're overestimating the impact of your disk cache, especially on desktop workloads. Cache hits are fairly rare, rarer still for games which just load their assets into memory and avoid the disk like the plague. Disk writes aren't sped up at all by a cache, because it's write through, not write back (god I hope it's not write back).

Disk caches are more than just dumb caches trying to keep recently accessed files in memory. They're essentially automatic RAM disks that resize themselves when needed, and try to predict what you're going to need in the future. Loading a major piece of software may involve accessing hundreds of files, creating hundreds or even thousands of random reads. If the cache can avoid those reads by keeping the files in memory, because you already used the software hours or even days ago, it's a huge win.

The predictive performance of the cache depends, among other things, on the expected memory usage patterns, and whether the expectations hold or not. If the expected peak memory usage, excluding the cache, is 80% of physical memory, the cache is going to work in a different way than if the peak is expected to be at 60%.

Also, how were you able to determine that it had never used the pagefile? Does it show total page faults since boot? I don't spend a ton of time in the Activity Monitor, never looked for such a statistic.

I don't really know how the Mac-specific GUI tools work. I prefer to use the standard tools like top.

Err, no. At that level, you're getting the worst value for your money, and every additional dollar spent is worse value still. Beyond about $1000, diminishing returns starts to kick in pretty hard. You need to step back and look at what you're paying for, and what you're getting for your money.

You fail to see the whole picture.

In my country, three-year individual research grants are typically €250k to €300k. After overheads and research costs, that leaves you with a quite average middle-class salary. If buying twice as expensive computers increases your productivity by 5%, the return on investment is really good. In many other fields, the situation is even clearer, as professional services can easily cost €100 to €200 per hour.

[Camacha]

In my country, three-year individual research grants are typically €250k to €300k. After overheads and research costs, that leaves you with a quite average middle-class salary. If buying twice as expensive computers increases your productivity by 5%, the return on investment is really good. In many other fields, the situation is even clearer, as professional services can easily cost €100 to €200 per hour.

That is the whole point - you are not going to see an increase in productivity anywhere near those numbers, and probably none at all. Both LaytheAerospace and I have both reported that you can do pretty much anything you want on 8 GB which, according to you, would impair work flow somehow. So while you do have a very good point that you should not skimp on computer power reliability and productivity as a professional, it also seems to be moot in this discussion.

Additionally, we are mostly talking about personal private computers here, so talk of increasing productivity is a bit theoretical. But yes, I agree, when you talk about professionals the issue of equipment costs versus productivity generally is clear - you can go quite a bit overboard and still get good returns on your investments. There is a good reason Quadro and FirePro cards are so expensive but still sell like hotcakes - the seamless support usually easily earns itself back. Having an expensive employee sitting idle, while customers are waiting and contract deadlines might very well creep closer is not something you want to wager on saving 500 dollar over four years.

Then again, it does seem to depend on the type of work you do a lot. Highly paid engineers get state of the art workstations, but office workers will have to do with run-of-the-mill equipment.

[magnemoe]

A PC is a major purchase, but it's still ridiculously cheap. Those two are not mutually exclusive. The price of a new PC should be compared to the thousands of hours you're going to use it. Only a few things in life are cheaper than that.

Besides, if you live in a major city, the costs of owning a desktop computer are mostly determined by the space required for the desk. Those few extra square meters cost real money.

In this setting you buy an huge and expensive laptop, price don't matter much, its not designed to be carried but to be stowed away so weight is not important.

[Camacha]

I would argue that the space needed for a desktop computer is not really much larger than that you need for a laptop. Of course, you could use the laptop on the couch, but if you want to do some real productivity work, you would be wise to have some appropriate work space set up. Bad form when using a computer can wreak havoc on productivity too, not to mention your health.

I have had that experience with a bad mouse or a table that was just a tad too high - both were workable, but would result in sore hands and shoulders over an extended period of time. One trip to a medical specialist could buy you desk space for quite some time

[Jouni]

That is the whole point - you are not going to see an increase in productivity anywhere near those numbers, and probably none at all. Both LaytheAerospace and I have both reported that you can do pretty much anything you want on 8 GB which, according to you, would impair work flow somehow.

When you're involved in a creative field such as research, everything depends on your mood. On a good day, you can easily get 10x more done than on an average day, so a 5% improvement means just having one more good day in a year. There are also feedback loops: good days often follow good days, while bad days follow bad days. Even seemingly inconsequential things such as the delays when starting applictions may significantly affect your productivity in the long term.

There's also the placebo effect. When you're using tools you believe to be good, you are more productive than with tools that are just good enough.

Additionally, we are mostly talking about personal private computers here, so talk of increasing productivity is a bit theoretical. But yes, I agree, when you talk about professionals the issue of equipment costs versus productivity generally is clear - you can go quite a bit overboard and still get good returns on your investments.

Private individuals are actually quite rare after high school. When you're a student, everyone does productive work at home. After graduation, many become professionals or managers who bring their work home, as the distinction between work and free time is no longer what it used to be. Others become researchers, artists, freelancers, or entrepreneurs. Many get involved in some sort of volunteer work. For all of them, the quality of tools matters.

[Camacha]

Even seemingly inconsequential things such as the delays when starting applictions may significantly affect your productivity in the long term.

There is no delay, so there is no effect. I am not sure how else to put this to get through to you. It works like it should.

There's also the placebo effect. When you're using tools you believe to be good, you are more productive than with tools that are just good enough.

Then again, if you are not gullible and know what you are dealing with, a similar effect works without spending money needlessly. I know I have good and appropriate equipment, no need to spend money on things I am not going to use properly. Good tools are almost always worth their money, but a tool you don't use is money wasted.

Private individuals are actually quite rare after high school. When you're a student, everyone does productive work at home. After graduation, many become professionals or managers who bring their work home, as the distinction between work and free time is no longer what it used to be. Others become researchers, artists, freelancers, or entrepreneurs. Many get involved in some sort of volunteer work. For all of them, the quality of tools matters.

That sounds like a great story, but if you bring that to its logical conclusion it would mean that everyone needs a rather expensive computer after their student years. That just does not jive.

There also seems to be this misconception that you cannot make great work without the perfect and ideal tools. Especially in the creative industries, limits imposed can lead to much more impressive results. It is good to have good tools, but you should be careful not to make it a requisite for production. I feel that might be a different discussion though.

[Jouni]

There is no delay, so there is no effect. I am not sure how else to put this to get through to you. It works like it should.

And I have said that there are obvious differences. If you use 60% of memory for applications, 20% for disk cache, and 20% as RAM disk, you can cache the programs you use regularly but don't keep running all the time on the RAM disk. This obviously reduces the delays when starting those applications. If you use 60% of memory for applications and 40% for disk cache, you can get some or even most of the benefits of the RAM disk, if the cache algorithms are good enough.

[Camacha]

And I have said that there are obvious differences.

I have not found any benchmarks that indicate this - they generally point to the reverse. When running Windows, I must add.

[Mazon Del]

Good tools are almost always worth their money, but a tool you don't use is money wasted.

This is true. However computers are a bit different depending on the users overall behavior. RAM especially. Example, you have a fresh install of your OS, brand new hardware calculated to give you an extra 25% capacity as a nice safety margin, everything is great! Two years down the line, you find yourself shutting down extraneous programs that you normally left running for easy of switching back and forth. Certain programs start slowing down. New software needs to be set on lower settings. Etc.

Now, there is a lot you can do to prevent the latter eventualities. Regular defrags, system maintenance, occasional OS wipes, etc. Many of these things you can automate, some you cannot, some the user will CHOOSE not to. Every time I buy a gaming rig I build it as best I can without actually stepping on the bleeding edge of what is available. At the time of building, the extra 2-3K for the one or two better graphics cards isn't worth the end-time boost as an example. I get about 5.5 years out of each computer before it starts failing to meet my desires. In the early days of my current rig I could run the Mechwarrior Living Legends mod for Crysis on maximum graphics (main game), Second Life with max visual range and graphics (chatting with friends), Minecraft (automated skeleton killer machines), an offline max graphics Supreme Commander on a max level stage with a full complement of AI's set to max unit cap (1000 I believe) (what I played when waiting for a new game of MWLL to start), have dozens of tabs in Firefox and Chrome up and running displaying content of all kinds (I just bloat up with these), while running Solidworks simulations on several different parts (classwork and personal projects), all at the same time on four separate screens and my computer barely felt it. It was a wonderful time to be me.

Slowly over time that ram usage ticked up, that list of programs ticked down, the graphics eventually had to begin taking a few ticks if I wanted to maintain my minimum required set of programs active. Sure, some of this can be solved by wiping the OS for the first time in about 3 years. I'm even quite set up for that to minimally impact my data (OS on a solid state drive, all my data on high speed mirrored platters). But....I can't...just so lazy. I know its time to Old Yeller the PC once a brand new game can no longer be at max settings.

But regardless, the march of time dictates that program requirements WILL bloat. Case and point about how bad this can get. My laptop for college. It was quite excellent, nothing in the face of desktops by any means, but the second best Dell could manufacture (wasn't worth the extra 1K to nudge up to the next graphics card. I got this machine like a week before Dell bought Alienware...still sad about that). It could play the MWLL mod on Crysis at medium graphics (provided I had an ice water filled water bottle resting over the two internal graphics cards...), Minecraft, and several other programs running. I excercised the 'last days' clause in my warranty for that machine about 4 years later. Dell was required to replace all performance hardware (Ram, Processor, Graphics cards, etc. Screen and whatnot did not count) with factory fresh versions. New OS, etc. The laptop is as fresh as the day it was born for all intents and purposes. If I activate Firefox, I max out one core. If I also activate AIM, the last core is maxed out and my RAM is peaking. If I want to Reddit while using the laptop, I can't have AIM online.

So, what I am arguing for here is that PCs for your average user are not like industry machines designed with specific intended loads and whatnot. Whatever capability is beyond your needs right now WILL be useful later in keeping the machine usable longer.

If you are a linux power user or something akin to that will you be able to throw a bunch of effort and work into ensuring that you stretch out those lesser capabilities just as long as someone who paid for a more powerful machine? Yes, yes you will. But the average user like OP seems to be (apologies if that is not true, I simply assumed if you were asking for help on PC design that you weren't) will NOT put that effort into it. Either because they don't know how or because they just will not.

I will be building my new machine roughly a year from now. Currently if I stick to standard PC desktop style systems (without going to some crazy server powerhouse) my max RAM I can add is around 64GB or so (could be higher by now, frankly I haven't checked in a while). You can bet your copy of KSP that I'm going to load it up with every last GB I can. Will I need that anytime soon? ...technically yes, I have been doing some crazy software models of economic systems and the like lately....but ignoring those peak moments, will I need the RAM? Certainly not! Five years from now will I need it? Oh yes I will. Maybe not nearly as badly as my current machine (at 16 GB) needs more.

Really the next machine is going to be a bit of an experiment for me. With my 16GB of RAM right now there are actually only a very few situations where I get anywhere close to peak. Almost not enough to notice. But like the first gray hair on the love of your life's head, you pretend not to notice it. But you do. You do. I can conceive that in 5-6 years, I judge that 64 GB is all that I need on my next machine....but maybe it won't be. Maybe I want to do economics modeling while playing four different games and carrying on six research projects at the same time simultaneously. I'm like that sometimes.

[Camacha]

But regardless, the march of time dictates that program requirements WILL bloat.

This is true. However, buying expensive computer parts now because you might need them in the somewhat distant future is a rather counter productive strategy. Outside of specific cases, like the earlier discussed DDR3 becoming more and more expensive, you will spend too much money for technology that will be cheaper, better, more frugal and more quiet later down the line. Don't buy one 1000 dollar video card to last you 6 years, but two that will last you three years and use the money you will save on a RAM and maybe a SSD upgrade.

In computer land, you buy what you need now and in the foreseeable future. You try to pick up any new standards as much as you can so your gear might last a tad longer (think USB 3 or SATA600), but that is about it.

Edited November 18, 2014 by Camacha

[Jouni]

I have not found any benchmarks that indicate this - they generally point to the reverse. When running Windows, I must add.

Windows probably handles caching differently. Modern cache algorithms generally do a lot of stuff from heuristics to machine learning to predict better what to store and what to discard.

[Camacha]

Windows probably handles caching differently. Modern cache algorithms generally do a lot of stuff from heuristics to machine learning to predict better what to store and what to discard.

Since the vast majority of users here will be running Windows, I think we have a conclusion

Edited November 18, 2014 by Camacha

[Mazon Del]

In computer land, you buy what you need now and in the foreseeable future. You try to pick up any new standards as much as you can so your gear might last a tad longer (think USB 3 or SATA600), but that is about it.

This is sort of what I mean though. Yes, I acknowledge that THAT is the most efficient way to go about the whole thing. However people rarely actually go through with this in my experience. Again, power users who stick to their guns will do so. But the average person will just ride the computer down in flames because the act of spending money "now" (after the computer was bought) to make it better requires a new activity. Chances are by the time the computer reaches truly unusable levels and they decide to kick in for an upgrade, they aren't going to be imagining just a new graphics card, or a little more RAM. They'll be thinking about a new machine entirely. After all the attitude of "Look how bad it is now! Clearly one or two new components are not going to fix it!" is real. More then that, it ends up being to most people a 'devil you know' problem. They don't know for sure that changing out the RAM is going to actually solve their problems. And it is very easy to imagine buying new RAM, putting it in, and the computer being basically no different. In which case they would feel they have wasted money. Actually going out and buying parts piecemeal (which is what they WILL do. They will default to buying one part at a time, because if they are bulk buying...why not just buy a whole new one?) ends with negative confirmation bias. Unless the part in question has a rather major effect on the machine in question and basically solves ALL their problems, they will feel that their earlier theory that they were wasting their money was the right gut feeling to go with. Buying a better processor and making everything run faster doesn't make them feel like they have won if everything still runs slow or crashes because of insufficient RAM. Meanwhile they can know with absolute certainty that if they buy a whole new computer from scratch, when it shows up they WILL be satisfied with the speeds and responsiveness. They will feel positive confirmation bias, because after all, look how much faster it is running!

Plus you get a sort of inverse sunken effort fallicy going on with it too. After all, the three main aspects of the computer they are going to be upgrading are the processor, RAM, and graphics card. (I leave hard drives out, because space issues are a different beast.) And when you get right down to it, what more IS there in a computer that you honestly care about? The power supply, the case, the motherboard, the hard drives, and whatever internal peripherals you like (CD tray, etc). Most of those you can reuse...plus those are honestly the cheap components anyway. So even if they intend to just buy the main three, they will look at the others and go "hmm...with an extra $500 I can just buy all those, then everything is brand new AND I have a second computer...That is only a 25% increase in price anyway...Yeah! Let's do that! OOOH! Since I'm getting a new computer, I'm going to need a new monitor!" etc. I see it time and time again. After all, they have decided to sink 2000 on new hardware...and another 500 on new extras....why not toss in another 500 for a new this...and 300 on a new that....

What I am arguing here is NOT that your method Camacha is the wrong method of computer buying. I fully endorse it as the best method!

What I AM doing is arguing that, because of pyschological reasons, unless the average user in question has a reason why they MUST abide by it (I literally only have 500 to spend, I cannot afford a full 3K gaming beast even if I decided not to eat for a month) they will get more out of the wastefully extravegant computer in the long run. This is because if they have the ability to buy a whole new computer when their current rig starts slowing down, they WILL do so. And since they will do this, then the optimum cost strategy is instead to buy a computer that will last them as long as possible before they have to replace it. I don't have any hard numbers to provide, but comparing my experience to that of a friend of mine that economically HAS to upgrade piecemeal, my costs over time end up only being slightly more than his. However, this is dependent on his aging components NOT failing in a way that takes out other components. Unfortunately this happened to him about a month ago, his best guess is that a cap of some sort on the motherboard popped, caused a short that dumped supply power from the PSU into the processor and graphics card. Those are both DOA, the motherboard cap clearly trashed. But the RAM is useful and the PSU tests good at his lab. That said, I am staring at him as he goes through exactly the same logic I have described above.

Again Camacha, you are right from the perspective of a computer person. I am arguing from the perspective of the psychology of the situation rather than the technical side.