How Long Should You Keep a Computer?

tim wrote in post #17538722
My i7 2600K machine was built in 2011, so 4 years, and it's still going strong. When I check out cpu benchmarks the performance of new CPUs is around 30-40% better, though I'm not sure about real world performance. I figure in a few years I might upgrade.

2600K CPU score. 4790K CPU score.

We're in exactly the same boat. My i7-2600 replaced an early Core 2 Duo so I got a 4-5x increase in CPU performance which made a big difference in Lightroom and Premiere Elements. Now I look around and even if I spent something like US$2500 on a Xeon CPU (never mind the more expensive motherboard, etc) I'd still only get 2x the performance I've got now. No thanks.

Since building this machine, I've upgraded the video card to support larger monitors, doubled the RAM to 16 GB, replaced the internal hard drives with SSDs, and moved the bulk of my data to a Synology NAS. So I'm still able to improve performance and expand storage without junking the whole thing.

And that old Core2 is still in use. I used to along with a few recycled parts to upgrade an old P4 my dad was using. It still works fine for web browsing and other light duties at 8 years old.

Mark0159 wrote in post #17538850
You upgrade when you start having this problem. You come across a new app that you can run because you your OS needs an upgrade and there is no upgrade for it.

the time to upgrade is when the computer is too slow and you can't get parts for it. If it's got a fault and the cost of the repair would be the cost of a new computer.

but like anything it up to you. You may find that a new computer could last you another 7 years.

As for myself I build my own so replacing parts is what I do. It gives me the flexibility to upgrade bits as I need to.

Even if you can and know how to upgrade, sometime after doing cost/benefit analysis, it may be better/cheaper to buy a new PC than doing multi-stage upgrades.

mike_d wrote in post #17538874
We're in exactly the same boat. My i7-2600 replaced an early Core 2 Duo so I got a 4-5x increase in CPU performance which made a big difference in Lightroom and Premiere Elements. Now I look around and even if I spent something like US$2500 on a Xeon CPU (never mind the more expensive motherboard, etc) I'd still only get 2x the performance I've got now. No thanks.

Since building this machine, I've upgraded the video card to support larger monitors, doubled the RAM to 16 GB, replaced the internal hard drives with SSDs, and moved the bulk of my data to a Synology NAS. So I'm still able to improve performance and expand storage without junking the whole thing.

And that old Core2 is still in use. I used to along with a few recycled parts to upgrade an old P4 my dad was using. It still works fine for web browsing and other light duties at 8 years old.

In the past few years/generation of CPU, most of the technology goes into lower power consumption and not drastic performance incrwment. So for a desktop CPU going from high-end 2nd gen core i-series CPU to 3rd or 4th gen usually doesn't make much sense. I too have the 2600k cpu and there isn't a reason to upgrade. On the other hand, the older core duo/quad CPU are not very power efficient compares to most recent one. I still have the Q6600 CPU that I no longer use , it is now boxed up on the shelf because it eat too much power and heat the room up very quick to be use as a workhorse PC. Now I use my laptop with 3/4th gen i7 for most of the work, and leave the 2600k only for transcoding. It is much more efficient for me this way.

My rule of thumb............... until it can't do what I need it to, anymore. Machine before this one, lasted me 8 years! Current one is 4 years old now, but still does everything I need it to.

someone0 wrote in post #17538908
Even if you can and know how to upgrade, sometime after doing cost/benefit analysis, it may be better/cheaper to buy a new PC than doing multi-stage upgrades.

Perhaps but I have been doing this for 20 years so it's habit for me. My last upgrade would have been motherboard and cpu only. the extra ram was paid for by selling the old kit to a friend. That was over a year ago and would have cost me about $400. To buy a machine with the same spec would have cost me a lot more. Due to the upgrade I don't have to get any new for years to come.

I too have never seen a case where incremental upgrading or even total rebuilds have ever been anywhere near the cost of a pre-assembled PC unless you toss aside specifications. Like for like, a self assembled PC will always be less expensive.

Until the cost of upgrading is more attractive than the frustration of delays. But the first thing I do is wipe the computer and lay down a fresh install of the OS and all the software I use, and look at cost-effective upgrades such as more memory or faster HDD (now SSD). Often this will yield another year or more of use.

I agree the decision between buy and build isn't straightforward. There are intangibles to consider, such as support and potential repairs. Mac has become very popular by selling an essentially closed-box at a premium price in exchange for a better user experience. Conversely, if you are willing to take responsibility for learning about your hardware and software options and doing your own troubleshooting and support, you can get much better performance for less money by building your own PC or Hackintosh.

For example, I built three dual-boot Win7/Mountain Lion towers for $1500 each with dual IPS displays a few years back, less than a single Mac tower would have cost at the time. But I've also had to deal with some problems every time Apple updates OS/X. I also could have configured a Lenovo tower at about 50% more than what I eventually paid to DIY, but then I could rely on Lenovo for support.

It's not rocket science, but it can be briefly nerve wracking when your editing system won't boot or boots but won't recognize the audio or mouse. Fortunately, a little bit of searching the web always produces a solution. On the PC end my son accidentally deleted his WLAN drivers and it took a bit of effort to find the right ones to get it working again (which, of course, also required access to a second computer with a working network connection).

A high powered/well spec'd computer (like Mac Pros tend to be) will get me 5+ years. My old system was a home built Core 2 Quad that I had mildly overclocked on below rated volts (on the CPU). Other than a stick of RAM going bad at around year 3.5 or 4 it never gave any trouble. Other than replacing old/small hard drives and one video card upgrade (for better performance) I replaced it after about 5 years though because 8GB of RAM was beginning to be a limitation.

I have a homebuilt system with an i7 3770K and 32GB of RAM now that's about 2.5 years old. I don't see it going anywhere any time soon, and it's not giving me any trouble.

Would either system last 7, 8 or 9 or more years? Probably. I leave them on all the time but shut them down the clean dust as needed and everything is always running smoothly other than freak random things like the RAM, but that can happen with any system.

Would I use a system for 7, 8, or 9 years? Whether it's your mac pro or any of my systems, that's unlikely for a number of reasons. I like upgrading but I also tend to push for more and do more with the system, and take advantage of newer software. Even the best old system isn't going to give you the same experience as a good new system. It's particularly limiting with video cards which have a shorter lifespan and if you do any gaming (I do some) you'll not be able to get a current card jammed in to a system that old or if you can, you probably won't get as much out of it.

You should keep a computer for as long as it does its required job reliably. I still use an old 186 or 286 based system from time to time. (it is connected to some scientific hardware that the lab is way too cheap to replace, but it still works, and the only people who ever complain about it are new people who haven't used the system before, and sometimes whoever it is that is stuck teaching said new people.)

My personal rigs tend to get heavily upgraded every 1-4 years, depending on the need of projects I'm working with. Last upgrade was replacing a dead UPS, before that it was replacing a dead graphics card. Last serious upgrade was a few years ago when I went from a Core 2 Duo with 4GB ram to an i5 with 16GB ram, which cut the calculation time on a project from an estimated 1-6 weeks to an afternoon.

Luckless wrote in post #17539794
You should keep a computer for as long as it does its required job reliably. I still use an old 186 or 286 based system from time to time. (it is connected to some scientific hardware that the lab is way too cheap to replace, but it still works, and the only people who ever complain about it are new people who haven't used the system before, and sometimes whoever it is that is stuck teaching said new people.)

I'm not aware of any suchthing as 186. I know there was the 8086 and 80286 CPU. and those don't have floating point processor integrated. Those must be very limited supply. As for me, I would avoid using such old system and try to migrate to something more powerful and less power hungry, but then your usage may require differ.

someone0 wrote in post #17539894
I'm not aware of any suchthing as 186. I know there was the 8086 and 80286 CPU. and those don't have floating point processor integrated. Those must be very limited supply. As for me, I would avoid using such old system and try to migrate to something more powerful and less power hungry, but then your usage may require differ.

Intel 80186 known as the 186 was common in 1982.

My first was a 286 and I wasn't even in the US back then.

Yep. 80186 or 80286, but I've never known anyone to refer to them as such outside of exact specification sheets.

Most of the equipment in the setup is actually older than I am. We've replaced some of the parts in the actual machine, and I think it is on its second or third controller computer, but no one has ever wanted to sit down and completely port the program onto something newer. (I do have a mostly finished rebuild of it running on a raspberry pi, but there is still a bunch of coding to finish.)

But the system is still chugging along, doing everything, we need it to, and we have no pressing need to replace it as long as the controller computer keeps powering on every morning. There are a few spares kicking around in the back closet that we took out of an accountant's office for the price of a few coffees. (As in he bought us coffees to haul the stuff away for him.) We can do without the system for awhile if we do have to jump into crunch time and bang out replacement software that can run on a newer computer.


But my point is: Computers are tools. If your tool is working, if it is doing its job as it is suppose to in a manner that is acceptable to you, then why replace it? Do you replace a hammer just because a new model comes out? No. You replace your hammer when it breaks, or when you realize you need a proper framing hammer instead of a smaller carpenter's hammer, or when you decide that the cheap one you bought awhile back is killing your elbow and you want to try a new one with a better haft on it.

Personally I would replace the old 286 with something of the same architecture instead of RPI which do not require porting the application. There are many x86 that is low power, while not as low as ARM, but close enough and significantly lower then the current system.

As far as seeing them as tool, yes but I don't replace them only when it break, but also when it would be more efficient/ save future time/money. Of course every one threshold in the regard would be different.

Unfortunately the software in question for the machine isn't just a user level app which we can simply drop on a new system and run, but rather it is an ancient design that runs directly on the system in place of an operating system. It has fun things, like cycle based timing assumptions. We've fiddled around with emulators, but we could never get it to reliably run through a whole job. And frankly, rewriting it from scratch based on design specifications on a modern development environment sounds so much more interesting than digging into nearly 30 years of assembly code (Mostly written by a chemical engineer...)