Honestly, I’ve built a few PCs, and in my experience, it’s a bit of an excruciating process. Putting it together is pretty easy nowadays, but not so much debugging. The computer I use for games is annoyingly unstable.
Were I to do this again? Desktop computing performance has plateaued enough that a 10 year old computer is still viable for running the latest software. I’d just buy a 5-year old mid-to-high end business desktop coming off lease - that’d get you pretty good performance and a reliable system, and then just upgrade that. Put in a new GPU, RAM, and/or SSD, and you’re golden. Less quirks, and still affordable.
Huh; your experience is actually the opposite of mine. (Hooray for anecdotes!) My desktop, which gets heavily used for gaming and not much else, and desktops I built for my parents and brother, are all very stable. (Granted, the latter two are neither boundary-pushing nor running Windows, both of which help significantly.)
The challenge I find is catching up with whatever the latest and greatest technologies are when picking parts.
I tend to buy a new main PC every 5 years or so, so need to figure out pretty much from scratch all the comparabilities between things like AGP/PCI/PCI-E, CPU sockets, IDE/SATA/SATA2, molex/24+4/fdd/fan/gpu, VGA/DVI/HDMI/MHL/DP bandwidths of and power consumption each etc.
Gotta say I agree about the PC building experience. Admittedly I’m partially blind with fine and gross motor impairment, but I’ve had friends assert “No really ANYBODY can build a PC” but then when they tried to walk me through it the process was pain incarnate for all concerned. No thanks, I’ll pay the Apple tax and be happy about it :)
I love this post. Building your own computer can be a huge pain, especially when you’re assembling a bunch of parts from various sources and hoping they work out OK. That’s why it’s great to have a list of torture tests around.
Now, I’m going to nerd out a bit on this.
My only concern in this post is the use of Prime95. On Haswell and later Intel processors, later Prime95 versions use AVX2, which results in overvolting on Intel chips, leading to increased heat buildup. If you are already overclocking, then this potentially pushes it further than what you’ve set! (That said, the CPU should respond by downclocking if temps get too high.) In a sense, Prime95 is much, much worse than any real life scenario, to the point of being nearly useless. I totally understand wanting to put components through a test in order to try to force failures within the retailer’s RMA period, but this borders on silly.
You should still use Prime95 for thermal and stress testing, just stick to version 26.6. Know that later versions push your CPU needlessly hard, and thus are useless for determining the efficacy of your cooling.
In a sense, Prime95 is much, much worse than any real life scenario, to the point of being nearly useless.
It’s not useless. The whole point of of a stable overclock is that it can withstand any software workload for any period of time. You can’t claim that if you can’t run Prime95 for a day.
Maybe it’s fine to have an unstable overclock in some usage scenarios, just don’t put it in the same basket with the stable ones.
That’s fair. I admit I’m biased (aka: don’t think of other uses) in workloads towards games, which aren’t quite as demanding as other use cases, such as encoding.
Honestly, I’ve built a few PCs, and in my experience, it’s a bit of an excruciating process. Putting it together is pretty easy nowadays, but not so much debugging. The computer I use for games is annoyingly unstable.
Were I to do this again? Desktop computing performance has plateaued enough that a 10 year old computer is still viable for running the latest software. I’d just buy a 5-year old mid-to-high end business desktop coming off lease - that’d get you pretty good performance and a reliable system, and then just upgrade that. Put in a new GPU, RAM, and/or SSD, and you’re golden. Less quirks, and still affordable.
Huh; your experience is actually the opposite of mine. (Hooray for anecdotes!) My desktop, which gets heavily used for gaming and not much else, and desktops I built for my parents and brother, are all very stable. (Granted, the latter two are neither boundary-pushing nor running Windows, both of which help significantly.)
The challenge I find is catching up with whatever the latest and greatest technologies are when picking parts.
I tend to buy a new main PC every 5 years or so, so need to figure out pretty much from scratch all the comparabilities between things like AGP/PCI/PCI-E, CPU sockets, IDE/SATA/SATA2, molex/24+4/fdd/fan/gpu, VGA/DVI/HDMI/MHL/DP bandwidths of and power consumption each etc.
Gotta say I agree about the PC building experience. Admittedly I’m partially blind with fine and gross motor impairment, but I’ve had friends assert “No really ANYBODY can build a PC” but then when they tried to walk me through it the process was pain incarnate for all concerned. No thanks, I’ll pay the Apple tax and be happy about it :)
I wouldn’t take advice from those friends ;-p
[Comment removed by author]
I love this post. Building your own computer can be a huge pain, especially when you’re assembling a bunch of parts from various sources and hoping they work out OK. That’s why it’s great to have a list of torture tests around.
Now, I’m going to nerd out a bit on this.
My only concern in this post is the use of Prime95. On Haswell and later Intel processors, later Prime95 versions use AVX2, which results in overvolting on Intel chips, leading to increased heat buildup. If you are already overclocking, then this potentially pushes it further than what you’ve set! (That said, the CPU should respond by downclocking if temps get too high.) In a sense, Prime95 is much, much worse than any real life scenario, to the point of being nearly useless. I totally understand wanting to put components through a test in order to try to force failures within the retailer’s RMA period, but this borders on silly.
You should still use Prime95 for thermal and stress testing, just stick to version 26.6. Know that later versions push your CPU needlessly hard, and thus are useless for determining the efficacy of your cooling.
For more info on thermal testing Intel CPUs: http://www.tomshardware.com/forum/id-1800828/intel-temperature-guide.html
It’s not useless. The whole point of of a stable overclock is that it can withstand any software workload for any period of time. You can’t claim that if you can’t run Prime95 for a day.
Maybe it’s fine to have an unstable overclock in some usage scenarios, just don’t put it in the same basket with the stable ones.
That’s fair. I admit I’m biased (aka: don’t think of other uses) in workloads towards games, which aren’t quite as demanding as other use cases, such as encoding.
Are there really no real-world scenarios where you’ll be running AVX2 in tight loops? Don’t some video encoders do that, for example?
Installing and running Prime95 on Gentoo:
Oh come now; that’s not nearly as exciting as downloading unauthenticated executables directly over FTP.