Hi,

What tools do you use when you get new hardware to test "hardware correctness"? We got a batch of hardware and we want to put it through it's paces. So far the tests that I did run were:
- memtest
- stress-ng
1. Watched CPU frequency and verified that CPU's preformed at 99.999% of capacity
2. Watched CPU temps to make sure they didn't reach any warning levels and stayed below.
Are there any other tools I can use that will put a system through it's paces? To be clear I am not expecting any issues but the client wants to document that all hardware is operating correctly.

The only other things I can think of are
> SMART test on the HD.
> Use BIOS to check the PSU voltages — or even use a multimeter.
I've never heard of any test for new motherboards other than a burn test.

As a hardware guy, let me say it's pretty impossible to test for correctness. Burn-in is it. Burn in at elevated or cycling temperatures is a way of aging electronic parts. A physical examination of large capacitors is worthwhile, because future failure is indicated if they're bulging. Get the customer to specify the work, and you do it. You can randomly 'scope around examining levels, but you're more likely to do damage. Be careful in any contract not to be the scape goat if something dies.

If the boxen do 3D, load up something that will make them switch into 3D output (does tuxcart still exist? if not I think Counter-Strike has a free offering) - more to make sure all the libraries/drivers/etc are setup correctly. Probably everything will work, and this is more of a test of 'is everything configured right' than 'is the hardware perfect' but still worthwhile imho. The other thing I can think of is checking every port - as in, if it has 10 USB ports on the back, plug and un-plug your keyboard from every port and make sure they all work, and again over with headphones (or something) for audio - a lot of newer machines will 'disable' their audio outputs if they don't detect something plugged in, and if there is a problem there you won't know until you hook something up. Again probably nothing will go wrong with new hardware, but I understand wanting to make sure everything 'works' before you physically install it (especially if its being racked up or otherwise put somewhere relatively awful to reach once its all done).

Back in the pre-Memtest86 days, compiling a kernel was a good stress test of the system. It hit memory, CPU, and storage (well, at least the disk where the kernel sources were located). I had a couple of instances where that process revealed a couple of substandard memory boards that wound up getting replaced---Signal 11 errors , if memory serves. Nowadays, Memtest86 would be my tool of choice.

For disks, a destructive badblock test on disk drives (or, back in my SCSI days, a low-level format) will let you know if they are worth using---especially if they're been obtained from other systems.


HTH...

From the man page of shred, I gather there is software for re-reading deleted deleted or even shredded (platter-based) hard drives and recovering the data, the files even. I presume that would work over most of the drive in the case of a failed hard drive. Does anyone know what that is?

The problem is that you may think every system is identical but even one chips version on the board can affect performance.

You'd need a system that is sold by a vendor and has some distro version guaranteed to work. Any deviation can create issues.

Hardware guys make an awful lot guaranteeing hardware, but there's nothing they can do either, except overspecify, and specify companies they trust.