What part of damaging your processor due to current draw, while not actually demonstrating any particular level of processor stability, sounds like a great test?

Nice contrarianism.

Yes, you can run it, and it should run successfully. Yes, sustained testing will rapidly damage your processor and cause notable performance degradation (reduced maximum overclocks/increase in required voltages) within a matter of days, and yes, you can still crash in Aida64 or other stress tests because it's not testing the whole processor very well.

It's far outside the normal current draw, even compared to other AVX2 loads like video encoding. It's an AVX load that fits entirely inside instruction cache and will hammer the AVX2 units nonstop at the maximum possible rate with absolutely no other instructions going to any of the other units. That's not how any other workload behaves, even video encoding. CPUs are not designed around that, they're designed for some reasonably normal mix of instructions.

Yes, it's a good idea to stress your processor a bit more than normal to prove it's stable, but if you make your 'proof charge' too large you will still damage the firearm.

I don't understand, when I buy a processor I expect it to work 100% load for all of the instruction set. Anything less is a defective product, and should be returned.

Furthermore, it should be able to do this 24/7 for years at a time.

This test, assuming it actually does kill the processor or stops executing, clearly indicates it should be returned.

It's fine at normal clocks. Not good for the chip, but it won't kill it within the normal operating lifetime (say, 3-5 years). That's why chips ship with a little more voltage than they need, so that they can suffer some electromigration and still function.

Overclocking? Yeah, it'll kill it quick, and that's your problem, because you're operating it outside normal specifications (more voltage than designed). If that's your standard, that you should be able to overclock and then redline the chip 24/7 for an unlimited period of time and it should never kill the chip, then you're going to be disappointed.

The fact that Threadripper is having problems is definitely a fault but it's something that you really shouldn't be doing anyway (and it sounds like the problem is the motherboard not the processor).

More generally: this is like arguing that you should be able to buy a camry and run redlined 24/7 and get the inevitable engine failures covered by warranty. If the manufacturer finds out you're doing that, they're not going to cover it, because it's not normal usage. Somehow people think it's a reasonable expectation for processors.

I agree Prime95 is overkill, but a processor should be able to handle it.

This is like the cooling argument all over again - "your cooling system is good enough for normal workloads". No, it must be good enough for permanent full load! Like a CPU under Prime95. If your heatsink lets it overheat, it's defective. You saved a few pennies on the copper and now it's a defective product.

Same for Prime95, the CPU must be able to run it for days at end and last ~5 years. Doesn't seem like too much to ask for, imo. Intel says their chips will run at 98 degrees for years, so why would a full load like Prime95 kill them in days?

One other thing, overclocking these days is actually undervolting combined with raising clocks. The old school "just raise the voltage" is a bad way to do it on any Intel chips post-Sandy Bridge...

If you overclock CPU, you lose warranty. Even if it's K-series which is made for overclocking. And stock settings probably are safe.

Yeah I did mean at stock settings.

If you overclock, you're responsible. Losing warranty sounds acceptable to me. Better than manufacturers doing everything to make overclocking impossible.

To me, it's a great way to get more performance out of an old chip, the warranty would be expired at that point.

It is hard to prove a CPU was overclocked though, so I can see some people taking advantage of that.

Did you notice that in the original article he is running the stock clock speed? If a CPU can't handle any instructions it supports at stock clock and voltage for any period of time it is defective, full stop.

Change the word "processor" for "car" and the absurdity of your argument should be fairly obvious. Most ordinary cars will provide reliable transport for many years, but will literally catch fire after a few hot laps on a racetrack. It makes no sense to design a passenger car with brakes, suspension, cooling and tyres that are massively over-engineered for normal use just in case the owner wants to go endurance racing.

3970x is a race car in your analogy, and thus should not catch fire

Damaging the processor? A CPU being unstable while overclocked doesn't mean it is damaging the CPU.