For a while now I’ve wondered how to build the most stable gaming/workstation possible. I’m sick of crashes, stutters, and general un-reliability. However, it’s a balancing act between price, performance, and reliability. (for example ECC memory is stable, but more expensive and slower)
Ideas I’ve had:
- ECC memory
- CSM sku motherboard
- Hugely overkill power supply, or even dual redundant PSUs
- RAID M.2 boot drives
- All air cooled
What do you all think? If you were to spec out a (realistic) ultra-reliable PC what parts would you use and why?
P.S. I’m looking less for specific recommendations as I am for general ideas, which is why I didn’t specify the use case or budget. I’m more interested in the concept and if it’s feasible.
Oh boy. This is kind of a land mine when it comes to getting recommendations.
The first thing that you have to address is EXACTLY what your goal is. If you are primarily building a gaming first workstation second computer that will have significant differences from if you are building a home NAS server where you one garunteed data accuracy/security.
In my personal experience the majority of crashes, stutters, and freezes I have experienced since I started building computers in the 90’s have been software related and nothing to do with hardware. I’ve still got a Pentium 4 running Windows XP SP2 that is bullet proof and never gives me any problems. It also doubles
as a great space heater. I would do your best to create a good system. UT understand that there may still be all the things you have experiences because it’s software not hardware.
As a general rule of thumb last generation hardware is going to be where you want to look. It is new enough that the performance will be good enough for the vast majority of things you want to accomplish and because of its age (in theory) has had enough time for bugs to have been ironed out and firmware/software fixes created.
Regarding CSM motherboards. They are not more stable than any other type of motherboard. It is a program designed to ensure companies can have a long lifespan on the same hardware because the manufacture gives a garuntee that they will continue manufacturing the part. The manufacturer also makes the promise that they will give a 6 month end of life warning for any SKU that they will be discontinuing so that enterprises have the opportunity to plan for upgrades to their systems if required.
ECC is great if your data is super important to you and you can’t have any loss whatsoever, even a single bit or two. I personally have never run ECC on my home rigs and the only issue I have ever had that I traced to my RAM was my computer crashing when doing large video renders. Turned out one of my sticks was functionally dead. Put any kind of load on it and it would just shut down and be unresponsive. No amount of ECC would fix that kind of issue.
There is evidence that ECC RAM can lead to overall high latency and power frame times during gaming but unless you are playing at 240hz at e-sports competition levels you probably won’t notice. Depending on what motherboard you choose make sure to check if it supports registered or unregistered Ram of you go the route of ECC. This will impact both your options available to you and the cost of the Ram.
If you can find ECC that’s compatible with your motherboard of choice and isn’t that much more expensive then go for it. Can’t really hurt anything.
Powersupply: The highest rated Seasonic power supply you can afford and a UPC. Check for the 80+ rating, in order from best to least (least here means still REALLY good but not as amazing) : Titanium - Platinum - Gold - Bronze. Try to get one larger than you system needs NOT because it means it’s more stable or efficient but rather so that if you upgrade down the line you don’t have to touch the power supply.
Dual redundant is only necessary if you are looking for as close to 100% uptime as possible and you literally cannot shut down your computer because you will lose money or data. Dual PSU will allow you to swap out a dead one for a fresh one with no downtime. That’s why they exist, no other reason.
Raid M.2 is a perfectly good idea but a good rule of thumb is to have your operating system on the boot drive and ABSOLUTELY nothing else. Have SSD’s for your programs and games. When running multiple SSD’s and M.2s and GPU’s (depending on your situation) you might need to keep an eye out on how many PCI lanes you are filling up.
I air cool probably the way to go for you. I’m a Noctua shill and I’m proud of it. I love their black editions of their fans and the NH-D15 is kind of the king of covering all your bases for colling needs. The silence is a major selling point as well.
Space heater :)
A quality power supply plus a good UPS will take care of any power-related instability. Other than that, buy reputable brands from reputable sellers, and you should be fine.
IMHO, “ultra-reliability” is kind of a trap. You can spend a lot more money on enterprise-class hardware to get another 0.01% uptime.
Your situation may be different than most consumers, so I hesitate to speak in absolutes, but for the vast majority of people, software configuration issues are going to be the cause of instability by orders of magnitude over hardware.
Any tips for a “good power supply”? I suspect a lot of the crashing I’ve had on my system lately is because my PSU is flaking out, but It’s a high-ish end Corsair one which I though would be solid. It’s an SFX unit since my build is SFF, but maybe that’s the problem? (Next build definitely won’t be SFF, too expensive and niche)
I used the PSU tier list while shopping for one.
https://cultists.network/140/psu-tier-list might help you.
It’s been a few years since I built my computer, so no specific recommendations. Today, I’d start at PCPartPicker.com and work through the wizard to build. It won’t (or shouldn’t) let you pick anything too small, so no worries about the wattage (but go a little more than you think you need). Once you have it narrowed down to a few options, read lots and lots of reviews.
That site has review scores. Amazon, Best Buy, Microcenter, New Egg, and others all have reviews from real buyers. Tom’s Hardware, IGN, AnandTech, and probably dozens more have professional reviews. Every brand is going to have an occasional stinker, so model numbers matter.
I’d say as a general idea, look into recent military spec hardware. I’m not sure how they do it these days, but back in the day their hardware would be more or less similar to some off the shelf hardware, but made a bit easier to service, frame built a bit more rugged, and underclocked around 20%…
Edit: Easier said than done to combine gaming specs and long term reliability, choose your battles…
Huh, hadn’t thought of that but it’s not a bad idea. I’ll look into it!
Warning! Mil spec hardware by definition is literally the lowest bidder on a government job. Just because something says milspec doesn’t mean it’s the best. Usually it means it’s pretty mid tier. I have had some good experiences with milspec gear, some OK experience, and more terrible experiences than the other two categories combined.
The most obvious answer is to use last generation hardware, it’s always more reliable with bug fixes and bios fixes.
Honestly raid boot drives sounds like it could introduce bugs, simply because it’s so much less common than a single boot drive.
Servers reguarly use raid as boot drives, so not that uncommon.
Yeah, I feel like I’ve heard that before. Maybe a small boot drive, and then raid to store all of the meaningful data for the system would work…
Hardware raid is always a significant failure point unless you have enough drives to do higher raids and good striping and redundancy, which you can’t do with 2 drives.
Software raid-like solutions like the ZFS filesystem gives you similar benefits with less downsides and instability.
A simple software drive mirror with a seperate periodic backup drives gives more stability than raid 0 or 1 ever would.
This chase for hardware “stability” in a non-server setting is mostly futile though. Probably >99% of crashes and system instability is due to software. Drivers, Windows updates, games, quality of life software, discord, Windows file system explorer bugs, upgrading to windows 11, any and all shitty RGB software, etc… will all crash themselves or your system 100 times before a hardware failure will ever crash your system or cause you to lose data.
Do you experience a lot of crashes now? What’s your current system? Did you build it?
To add to all the good points already, always look into the thermals of the case, as well as the components you’re using. Heat is a big killer in electronics. Air conditioning the room where the PC is may help too.
Think thats a great start, id do much the same. Would definitely do dual hotswappable psu’s. Have a reduntant mirror of the system as well. (Guess a raid setup on boot covers this to an extent)
Get a solid UPS to avoid sudden power down problems/ gives you time to properly shutdown your system.
Personally id try and mount this all in a 2U chasis with some high airflow fans as well.
deleted by creator
I would consider a passive CPU heatsink (Noctua makes one, for example) and a high quality thermal pad like the Kryosheet from Thermal Grizzly.
Fans are a point of failure, so reducing your reliance would be best, and a thermal pad wouldn’t dry out, like paste does.
Fans generally don’t fail, and if you use a cooler that has at least two fans, you have some redundancy.
Using a non-passive cooler is better I feel as reduced heat adds to the reliability.
They specified stable (which I think means “least prone to failure”), and I still maintain that a passive setup with a 65W or less TDP would do better than a fan-bound setup for that purpose, though you’d have to go with a mesh case or no case at all. The Noctua cooler I mentioned can handle 65W just fine, so as long as you don’t go overboard on the CPU, it shouldn’t be an issue.
You’re right that fans don’t often fail and usually give mechanical warnings that they will, and I would also recommend fans normally, but this sounds like a min/max dream build.
I’m trying to think which will go longer (part of the least failure bit), hence my emphasis on reduced life expectancy from heat.
Speaking of case, that’s another important point; I can’t bear the idea of going caseless, dust build up is going to cause problems as well.
You’d certainly have to keep up on your dusting regimen, but it would probably give better cooling performance overall, provided you do.
I agree with everyone else that last-gen should be more stable than the latest and greatest. Make sure your bios is up-to-date. I’m in the AMD camp and I wouldn’t jump to AM5 just yet. AM4 has been around for a while and is mature.
Have you considered under-clocking? Modern hardware is pushed to its limits pretty aggressively.
Do you have an Nvidia card? The Studio drivers are supposed to be more mature at the cost of cutting edge features and performance compared to the game-ready drivers.
Are you in Windows? Have you de-bloated and disabled as much telemetry as you can?
What’s a good/safe way to de-bloat windows 11, without re-installing? I’ve seen scripts and a “win-tool-something” whose name I can’t remember, but was looking for something easier/quicker to understand what it’s actually doing.
I’m not tech-illiterate, but not necessarily a super-user (depending on what I’m doing). I just haven’t had time to really get into it and wanted to know what the worst bloat-offenders were, that can be taken out easily enough
And to pre-emptively stave off the “install Linux” comments: I already have 2 machines running Linux, I’m well aware of that concept. I just want my windows laptop less boggy
Can’t help you there- my only interaction with windows is at work where machines on the plant floor have a Cortana process running along side industrial control software.
