I have bought a new desktop which arrives next week - it's custom built. Which hardware tests would you recommend I run? Feel free to mention any appropriate software, preferably free
Plug it in. Turn it on. Does it boot? Are all fans spinning? Note the PSU fan may not spin until PSU warms up and/or demand exceeds some threshold. Are all drives recognized? Is total RAM recognized? If "yes" to all of the above, tests done. No other tests are necessary. I would, however, install Core Temp to monitor CPU temps in real time.
You can, but you don't need either. Windows knows SMART. And motherboards already control fan speeds.
MemTest86 for RAM testing It it true, but misleading. Windows checks basic attributes reported by S.M.A.R.T. but it does not i.e. run/dispatch diagnostic self-tests implemented by disk vendor.
It is not misleading at all. But more to the point, CrystalDiskInfo is not needed to test a new (and working!) drive in a new computer. Windows has its own Error Checking tool and if more in-depth analysis is needed, the command line tool, Check Disk (chkdsk), is already built in too. And for that matter, chkdsk will actually attempt to repair problems it sees, CrystalDiskInfo does not as it has no surface (block/sector) tests (as chkdsk does) nor does it test S.M.A.R.T - it only reports S.M.A.R.T status. It's a nice monitoring tool, but again, not needed here as it is not what the OP asked for. And frankly, I would rather use HD Tune, GSmartControl, or the diagnostics tools provided by and found on the drive makers websites, like SeaTools, for example, if the drive was demonstrating problems.
Oh, and I agree with MemTest86 for testing RAM - though again, the OP does not need it if all RAM is recognized when booted and the computer runs fine when booted.
I would recommend to test the CPU as well with something like LinX, OCCT or Prime95 (at least to test the stability of the system after all these BIOS and Microcode updates for speculative related issues) and especially if you overclocked the system and maybe the GPU with FurMark (but don't forget to monitor the temperatures and cancel the tests when needed). Then check the Event Viewer logs for any hardware related errors like Kernel-WHEA or nvlddmkm (for nVidia) - the last could be of course a driver issue and not hardware fault.
Oh, it's nice to have a brand new computer. Checking the Event Viewer--strongly agreed! You need to install any available BIOS updates and your chipsets and other drivers right away. But for longterm use, maybe a nice hardware monitor that's also free, like HWINFO64.
HWiNFO64 is, for sure, one of the best HW "information" programs - which, BTW, also provides S.M.A.R.T. information. But it does not "test" the hardware, which is what newbino asked for. Yes, Prime95 does some testing but all those benchmarking and stress testing programs are synthetic. They create artificial scenarios that don't prove your computer will support what you will be doing in the real-world. The only way to do that is to just do what you do.
Yes, you're right about that, I think it just displays and logs what the sensors are showing during benchmarking. I like what you said about formal benchmarking software--I don't care for it either. Furmark for the GPU has its detractors, for sure, some calling it a "thermal virus." I run Passmark now and then, just to see whether my aging and remarkably ugly computer is still relevant somewhere. To test my undervolt, I simultaneously ran System Maintenance, a Defender scan and opened Edge with like five tabs. Ran an intensive (100%) single core game, stable also.
They're supposed to create artificial scenarios. The idea of a stress test is to reveal hardware failures soon after you build your machine so you can pick the defective part, go to the shop, and replace it with a new one. Prime95 for the CPU and RAM and FurMark for the GPU are good choices. As to tools like CrystalDiskInfo and SpeedFan one of the advantages is to have real time monitoring that alerts you as soon as something goes wrong. SpeedFan gives you a granularity of control no motherboard will; it is also much more than a fan controller.
No they aren't. They just do - because they are synthetic. It's not by design. It's just like testing anti-malware software. As much as the people running the tests want to pretend their testing represents real-world scenarios, it doesn't. No its not. The idea of these stress tests is to establish a baseline to see how far you can push the hardware and maintain stability.That may be why you run them, but that is not what they are for. The fact of the matter is, real stress tests are designed to abuse the hardware and push the product beyond its designed limits until it breaks. Then they back off that point in the "published" specs. Sadly, many users run them just for bragging rights.
No, they aren't. I was a hardware guy in a major software development company for 10 years (400 developers at my location alone). So, yes, I know what stress testing is for. Yes, there are other purposes, but they are secondary. But I do agree we should just move on.