How does everyone evaluate stability now?

[ChronoReverse]

I agree. Why bother with the heat and noise.

I'd rather undervolt (which also has potential instability issues).

I undervolt and overclock at the same time.

escrow4's statement sounds clever but isn't really useful. There are modern Intel CPU's that could overheat or have problems even on stock if you're in a nice warm climate while running AVX2 Prime95 for instance but that's a contrived scenario after all.

Practically speaking? Run whatever burn test makes you feel comfortable whether it be AVX2 Prime95 or just IBT. Then use your computer and if you don't encounter random crashes, great!

 

[WhoBeDaPlaya]

Overclocking is never stable. Never was, never is and never will be. If the chip you buy needs more puff buy a faster model. You can scream till you collapse that all those tests are 100% stable but 100% stable is stock out of the box.

If every uArch generation had a 4790K-like part (where it is already factory OCed not too far from the limit), then yes, there would be not much point.

 

[StefanR5R]

With Haswell-E and Broadwell-E, the question of whether or not to include AVX workloads in stability tests is moot again, because you can configure different multipliers for AVX and non-AVX workloads. [...]
I wonder when this useful feature will trickle down to the mainstream socket.

The answer is: now.
http://www.anandtech.com/show/10959...ration-kaby-lake-i7-7700k-i5-7600k-i3-7350k/8

 

[Rifter]

Overclocking is never stable. Never was, never is and never will be. If the chip you buy needs more puff buy a faster model. You can scream till you collapse that all those tests are 100% stable but 100% stable is stock out of the box.

Not true at all, Intel bins their chips and not all chips binned for top tier parts get sold as top tier parts due to the fact that lower tier parts sell much more in volume than higher tier parts. You could very well buy a bottom end i7 thats been binned to run at 7700k speeds for example.

 

[pantsaregood]

The 6800K continues to be a headache to overclock.

Cores are stable at 4.2 GHz/1.25V - they drop to 3.6 GHz/1.19V during AVX/AVX2 workloads. Attempting to overclock the cache at all results in crashes.

The above mentioned overclock has passed 24 hours of Prime95 1344K FFTs using SSE2, as well as 12 hours each of AVX and FMA3.

As soon as I threw RealBench at it, it crashed in only an hour. Looks like the IMC can't handle the DDR4-3200 it has been running for a few months. Going to drop it to 2666 and try again.

 

[ehume]

I have forgotten what IBT stands for, and can't find it by searching the net.

 

[RichUK]

I have forgotten what IBT stands for, and can't find it by searching the net.

Intel Burn Test

 

[superstition]

I have forgotten what IBT stands for, and can't find it by searching the net.

It uses Intel's Linpack math library, the main alternative to Prime95 for stress testing. LinX, which is also tricky to find with Google (especially since you get a lot of outdated versions) is an alternative GUI for Linpack. OCCT also includes Linpack.

 

[ehume]

Thanks.

 

[sm625]

Only care if my games are don't crash.

Yep. I will OC past the point where IBT would show a failure. I dont care because no game puts that kind of load on a cpu.