8 replies to this topic

[Shahadet]

Hey guys - got all my stuff from NewEgg, and I'll be settling in over the weekend to start cracking open boxes and putting stuff together, but I was wondering.....

Before I built up the nerve to try and assemble everything myself as a First-Timer, I talked around to a couple of local computer build/repair places to get quotes, and they all talked about "assembly, 24 hour burn in" etc.

And when I read around the Internet, I see posts from people running MemTest for 9.5 hrs, and other benchmarking tests overnight to "stress test" their systems and make sure everything is ok.

Is all this testing necessary? I'm not exactly putting together an industrial server or a hard core rig that will be on and used 24/7.

Can someone explain this "burn in" stuff to me, and whether I should be doing it to my new gaming rig (once it is fully assembled and OS/driver loaded) ?

Thanks!

[Dymitry]

no...they are just trying to up-sell stuff. If anything is faulty (bad luck, but happens) you will see straight away, if anything has to go bad, it will go bad when it wants to.

Exactly as you say, you are not going to run a server or a "critical mission" pc, so do not worry.

Stress test are used to test the system behaviour when overclocked.

edit (in a "normal" pc)

Ah, and by all means, go and build it yourself. It will take few hours, but it good fun. Just follow the instruction on your mother board manual, and ask for any doubt.

Edited by Dymitry, 17 August 2012 - 04:36 AM.

[Aznpersuasion89]

also make sure you ground your self before handling the components. even a small arc is enough to destroy chips. good luck and have fun. i just built mine. easy peezy

[Major Cuddles]

Well not all problems are noticeable right away. I just ran OCCT for a couple of hours to test my max temp and to make sure i had no errors. I like to stress everything to 100% to make sure it's stable, if not RMA with newegg. Found a bad stick of ram and a video card that way, that normally didn't show any signs of silliness.

EDIT: Also the Video testing on OCCT has been known to well ruin graphic cards, but AMD and Nvidia drivers automatically detect said fur animation and downclock everything to prevent any damage. So the best way to test your card is to say max out graphic settings and play games and look for artifacts and geometry glitches and such.

Edited by Major Cuddles, 17 August 2012 - 08:41 AM.

[Waffles 2pt0]

Grounding ones self can be accomplished many different ways. An easy way to go with this is to have something large, massive, and metal (the kind magnets stick to) near by and place the pads of your fingers on it every few minutes or so. This will allow any static charge built up while you are moving around and such to transfer into the large metal object rather than the computer components you are working with.

for example: a refrigerator

Edited by S Morgenstern, 17 August 2012 - 08:43 AM.

[Zakie Chan]

If you are worried you can run memtest for 1-2 passes of total ram. Also Prime95 on blend for 20 min will show you any potential heat issues. Furmark is good for testing GPUs, run for 5-10 min if no artifacts show up youre good.

All of these tests are NOT crucial. If you do decide to do them make sure you are present and monitoring the system with RealTemp. That way if heat spikes or artifacts begin you can abort the test.

As mentioned these utilities are mostly employed by users pushing their components. Overclocking and changing ram timings.

Build your system, double check everything. If you have access to someone knowledgeable in pc building have them help you. Fire up your rig and if you have problems post here and we can direct you on what to check.

Goodluck

[FaustianBargain]

While you don't need to stress test for 24 hours, you should run short passes for your components: Memtest for your RAM and Prime95 for your CPU. And whenever you attach a heatsink to a CPU, you should run the CPU at high load for at least a couple of hours to actually "burn in" the thermal grease between the two.

[WardenWolf]

As others have mentioned on here, there are definitely some basic tests you can run. Do you *have* to? No - but it would suck to have system instability that you can't track down and then find out after the return period that a RAM module was bad. Therefore it is a good idea to run some basic tests with the goal of making sure the hardware is not defective and also that the system is performing in line with what its various components should be capable of.

What I would recommend (and check my signature link if you want to know my credentials) is as follows:

1) MemTest86 for a couple passes first, since having defective memory could negatively impact anything that comes after.
2) Some sort of drive test on your hard drive / solid state drive - ideally one from the manufacturer of the drive, or else something generic
3) Install Windows, drivers, updates, and antivirus (install AV before any web surfing, if you need to go online to get drivers)
4) Prime95 for a bit, watching your CPU temp to make sure it doesn't get too hot (Intel's official cap is ~95C, but I like to keep it a good bit lower: no more than 80C personally)
5) With Prime95 still running, turn on FurMark and watch the GPU temps *very* closely. AMD and nVidia GPUs can both go past 90C and be okay, but unless you have a really hot running card or have poor chassis airflow it shouldn't get that hot. What you mainly want to look for is that there is no artifacting (visual corruption) and that the GPU temp rises steadily and eventually flattens out. Once it reaches that point, turn off FurMark and check the CPU temp once more. It should have come up higher with the video card going too, but should still be reasonable.

At that point you should be good. If you want you can also run something like CrystalDiskMark to make sure the drive speeds are where they should be, or various real game benchmarks... but the above will pretty well ensure stability.

[Dymitry]

FaustianBargain, on 17 August 2012 - 08:49 PM, said:

While you don't need to stress test for 24 hours, you should run short passes for your components: Memtest for your RAM and Prime95 for your CPU. And whenever you attach a heatsink to a CPU, you should run the CPU at high load for at least a couple of hours to actually "burn in" the thermal grease between the two.

Mmh.. Burn in for a couple of hours.. depends a lot on the paste. Good 'ol artic silver 5 takes as much as 200 hours according to them, while other (newer) compounds need 25 hours or even any at all, plus no one specifies "high temp burn in" as a better alternative to "regular use" burn in; in mechanics term (think cars) that is also hardly the case. Baking your cpu without a specific need (as in testing an overclock, or a LCS) is a bit futile for me.

Also, and I am talking from personal experience, while memtest will indeed reveal bad modules, it can take as much as 12 hours of continuous testing. What memtest will not often reveal, are compatibility issues between ram and mobo. Prime is better in this regard, but again, it can take a while and will still not be 100% reliable.

All in all OP, I would rather:
- install all the components, the os, updates, drivers, all the sw (have you got a decent original AV?)
- install a very last gen game, maybe one that is still in beta? and start playing the poo out of it.
- Eventually, you will have to take a break, i.e. going to toilet, or eating.
- Fire a stress test at random (probably prime is the most comprehensive one) and check back in few minutes,
- assuming that smoke is not pouring out of the machine go back and leave it until you have finished doing your thing.
- get back to your new toy, and start playing more.

Edited by Dymitry, 18 August 2012 - 07:12 AM.