Some OCing questions

[elezraita]

So, recently I got a 4790k to replace a 4670k for uns in my msi z87 gd65 gaming mobo. I wanted the extra threads for video rendering and the like, and I'm not affluent enough to afford x99 at present. Also, my 4670k was NOT a good overclocker. Before you freak out and tell me that it was a pointless upgrade for the money, let's just say that I made a deal with a friend that subtantially mitigated the cost. And, yes, he chip is definitely supported by the mobo.

 

Anyway, I'm working on dialing in a decent overclock for the 4790k. Basically, I started by putting locking the cache ratio at stock (mult=40), with the ring voltage set to auto. I disabled turbo boost and EIST, set my crucial ballistix 1866 MHz ram to 1333 MHz at 1.5V, I set vccin=1.800V, and I set vcore=1.25 V in override mode with the core multiplier=46. I then booted into windows and stress tested it using Intel Extreme Tuning Utility for 10 hours with no problem. I then set vcore=1.24V and stressed it in IXTU for about 6 hours before I decided that the XTU stress test wasn't really doing a good job looking for stability and it was a waste of time to dial in the core multiplier and voltage this way.

 

I guess that that's where I am now. I've read that the IntelBurnTest is pretty good and will find obvious errors and instability relatively quickly, but it isn't fullproof without other tests (which can be said of any test, I know). I have Prime95, but I've read multiple claims that Haswell doesn't play nice with Prime95, at least not the latest version. However, there are those who seem to live and die by Prime95. Then there are people who love Aida64 and x264. The list goes on. Everybody and their aunt's dog have an opinion, and I'm willing to listen to them. I'm also willing to spend the time to be sure of stability and that I've got the best possible overclock, but I'm wondering if there is a shorter way to dial it in and then test substantially. What do people think?

 

You probably want to know what temperatures I'm getting. I'm using a Nepton 280L AIO and my idle temps are typically in the mid 20's, but that might just be that the ambient temps are pretty low. At 1.25V the highest any core ever got was 63C in xtu. At 1.24V, all the cores were in the mid to high 50s. Both of these were at 4.6 GHz.

 

The other thing that is worrying me is something that happened I was just testing the 4790k at stock speeds right after I put it in. My friend wanted to see the xmp (1866 MHz) in action, so I enabled it, but I changed nothing else. We started watching a youtube video and got a bsod about halfway through. I don't remember what the code was. I need to go look, I know. I then went in an disabled the xmp and was able to watch the video and do a few other things at stock before I turned off the computer. Later, it was bugging me, so I enabled the OC genie just to see what it would give me. In typical conservative fashion, it gave me 4.5GHz on all cores at 1.22V and enabled the xmp. I then ran xtu for a few minutes with no crashes. Granted, that doesn't really tell me much, but I'm confused as to whether or not the xmp caused the bsod at stock cpu speeds and voltage. I feel like the the memory controller should have no problem with 1866 MHz, especially at stock, and my 4670k had no problem with it. Well, I guess I can't really say that, as I never had it enabled with the 4670k at stock settings. So, is it likely that the stock voltage on the CPU wasn't enough to handle the xmp? That seems unlikely to me, based on my admittedly limited experience. I haven't tried to reproduce the error, but I definitely haven't seen it at higher vcore. Has anyone seen this type of thing before?

 

Sorry for the long post.

[LogicalDrm]

First question. I use combination of OCCT, IBT and XTU at the moment for stress testing. Prime95 is good for non-Haswell CPU's from Intel and all AMD chips. Personally I think it as something from old ages of overclocking. I mean you need really long stress periods (24-48h) to see if system is stable. With OCCT it can just be 3h, and then another 4-6h with XTU or Aida. Much more affortable for someone with only one PC and it hanging right next to your bed.

 

Second question. I'm not that experienced in overclocking that I would be able to explain this in depth. But what I've read, overclocking RAM is prone to make CPU overclocks unstable. That what happened. XMP is basicly overclocking like you would with software wizard with CPU.

[elezraita]

Thanks for responding. I might give OCCT a shot. I'm a bit confused as to the functional difference between OCCT and IBT; as far as I can tell, they both use linpak. Are they equally effective, or do they test different instruction sets somehow? It looks like OCCT has other cpu/memory tests. Is there something else I'm missing?

 

As far as the second question, yeah, that's kind of what I thought. I know that the xmp is considered overclocking, but I was just surprised that enabling it would make the stock settings unstable. I guess I am a bit worried that that kind of instability at stock will cause problems at higher clocks, even with higher vcore. I guess that's why we stress test...

[LogicalDrm]

Thanks for responding. I might give OCCT a shot. I'm a bit confused as to the functional difference between OCCT and IBT; as far as I can tell, they both use linpak. Are they equally effective, or do they test different instruction sets somehow? It looks like OCCT has other cpu/memory tests. Is there something else I'm missing?

 

 

I don't know that much of technical background. But when I was testing how this chip OC's, I got it stable with IBT but at the same time it crashed with OCCT. Thats why I currently use IBT for only temp checking.

[elezraita]

Okay, so last night I finally got it stable for 3h and 30min with occt without the xmp. I'm currently running the the same thing with the xmp. If that passes, it's probably stable, but I'll still run x264 for a few hours to simulate encoding, I think. If this doesn't pass, I'll be somewhat disappointed with my RAM.

[elezraita]

It was stable in occt for 3.5 hrs again at 4.6GHz and 1.26 V fixed, but it crashed while rendering a video in adobe media encoder. That's really annoying.