Afternoon all Just wondered who else on here has had a bash at overclocking a Haswell chip? Mainly wondering what volts you're using and what frequency you're getting with that? Just starting having a fiddle around with my setup and I seem to have got 4.3GHz stable using 1.15 volts. Getting temps of around 66°c using a H100i and quiet fans. Over to you! Cheers George
It varies a massive amount. My first 4770K needed 1.3v just to get 3.9GHz stable, so I returned that to Amazon. The second needed 1.3v for around 4.4GHz. I've got a second system with a 4670K, and that needs 1.3v for 4.6GHz. Overall it seems that up to 4.2GHz is normally ok around 1.25v or lower, while some chips can't feasibly do over 4.3GHz and others can get to 4.8GHz.
Did you crash? Some people have chips that can get 4.7GHz stable at 1.19v, others can't get to 4.3GHz at 1.35v. It's a terrible chip for overclocking compared to SB/IB - it's too unpredictable.
Yeah - bloody typical! It survived about 2½ hours of prime at 4.3GHz with 1.17v but then crashed again so nearly stable but not quite yet. I'll dial it back to 4.2GHz for now and have a bit of a game and then continue fiddling with it later I think.
I heard direct from 8Pack that with Haswell he binned huge amounts of chips to get good ones. The silicon lottery is quite poo. I am tempted to pick up a binned chip. As I have a contact but they are fussy about selling them
My 4770K needs 1.3v for 4.5GHz and hits 90c under a SB-E Extreme cooler, under Prime95 without AVX. With AVX, it throttles almost instantly. It's ridiculous.
I've given up with Prime95 testing. I try the Intel tuning tool stress test for 5 mins, and then play a few games - no crashes and I'll keep it. Since Haswell gets so hot so quickly on artificial stress tests I don't think it's very suited to them.
I'm at 4.3 on stock volts - I can't go any higher voltage because of the cooler I'm currently using (bear in mind I don't want noise ) so once I get a cooler that's a little more performant I think I'll be able to crank this thing high However I haven't actually tried pushing higher - I might when I get some time
I was watching this and reading this and it effectively says that something like Prime95 runs the processor harder than any normal application, to the point where you shouldn't use their Adaptive Mode for the voltage as when the processor notices you're running something with AVX instructions it'll ramp up the voltage by ~0.1v Interesting, if confusing in places, reading!
My thoughts exactly - it's an interesting concept but I'd always be paranoid that it was going to ramp up the voltages without me noticing. I suppose if it was only to ever do it when you were running programs using AVX instructions then that'd be fine but I'd worry it was doing it other times too!
The days when prime is a go to for stress testing ended with ivy really. Best thing is to do what you normally use a pc for. Most users will never max the CPU the way prime does so its not a very relivent test. If all you do is game most games won't even show a 30-40% load on a modern intel CPU. I run a quick intel burn test make sure temps don't hit 100c instantly then play a few games check temp if its sub 80c then I'm pretty happy. Don't overclock my other rig as stability is more critical.
I'd prefer a high load realistic test. For example, I use Linux as well and so a full kernel compilation using all cores is quite a good test. Another one would be a handbrake encode of a DVD or something. Varied testing is important though. I've found that I can be 100% stable in Prime for 30 mins, and yet still crash in a game a few minutes in. That's why I prefer a varied amount of shorter realistic tests.
That's why I like & repect the CPC benchmark suite actually as it's doing stuff we do regularly (other than gaming)
No no no PRIME is a great stress test because its above and beyond what you normally do and as such leave a amount of security in you OC. Better yet use F@H and boost the teams rating while you burn in. OC should be rock solid or just bench testing everything els it PFFF! ;P
Gaming is an excellent test though, the more CPU intensive ones tend to go up and down very quickly in CPU usage a lot, which is important to test in addition to flat out 100% usage.
Fair point. The thing I always wonder is why an overclock will fail after 3 hours if it's stable after half an hour - it can't be a temperature thing as the temperature has stabilised after 20 or 30 minutes, so what causes it to happen...
Varying calculations, it's not the same test over and over again. I think that Prime95 is an excellent test for systems that need 100% guaranteed stability running applications that will stress the machine, or for testing the stability of a CPU at stock clocks. For gaming and general use it's excessive, as it'll push the CPU harder than you'll ever experience in normal apps.
when I OC'd my C2D I ran Prime, F@H and had it set up with down clocking when not needed back in the FSB OC days . then I lost the settings and just let my P5K Delux auto OC, which it did like a champ. THe one time it died was when there was a Giant dust monster in my cooler. But I defeated it with an air compressor and a vacuum cleaner. That must have been a sight to see!
