i5-4690k really concerning temps?

Toggle sidebar Toggle sidebar
Deus Gladiorum

Deus Gladiorum

Distinguished
Jun 29, 2013
1,434
0
19,960
Hey everyone, this weekend I OC'ed my i5-4690k to 4.5 GHz at 1.2 Volts and now I'm really concerned. My temperature in OCCT seems alright, only going to 73 absolute max after 2 hours of stress testing on it, but otherwise passed the test in terms of stability. However, in Intel Burn Test I was hitting 85 and but managed to complete 10 passes without fail on standard, and Prime95 almost immediately had me at 90 degrees. I was way too concerned to let that go, but I think one of my workers kept failing, but I was too concerned with that temp to pay attention. However, according to CPU-Z my CPU isn't throttling so honestly, should I be concerned?
 
Solution
CompuTronix
Deus Gladiorum,

Here's the problem:

" ... Core i 2nd, 3rd and 4th Generation CPU's have AVX (Advanced Vector Extension) instruction sets. Recent versions of Prime95 run AVX code on the Floating Point Unit (FPU) math coprocessor, which produces unrealistically extreme temperatures. The FPU test in the software utility AIDA64 shows the same results.

It's not necessary to run AVX code for thermal testing. Prime95 v26.6 produces temperatures on 3rd and 4th Generation processors more consistent with 2nd Generation, which also have AVX instructions, but do not suffer from thermal extremes due to having a soldered Integrated Heat Spreader and a 35% larger Die. ... "

That's from the Intel Temperature Guide -...
Sort by date Sort by votes
I already have a Hyper 212 EVO, and I think I seated it just fine since my idle temps are around 25 - 35. I chickened out, and became way too concerned about those load temps and the potential instability, so I just brought myself back to default settings. I have a Z97 Extreme4 mobo. Weirdly, when I was back at 4.5 GHz, even though my fan was at "standard" mode in my UEFI, it was working at 95% (nearly 1900 RPM) when it should have been working at 1400 RPM since the temps at idle were at most like 35. I thought that was very concerning so I just created a custom profile for my fan and that solved the issue just fine, but now I'm thinking why would it have done that in the first place at standard mode, when my overclocked idle temperatures almost the exact same as my stock clocked idle temps?

Furthermore, when I OC'ed to 4.5 GHz via the multiplier, my UEFI automatically set my CPU input voltage to 'Fixed' at 1.9 Volts and set my LLC to level 1. I think the LLC was fine, but that 1.9 Volts was weird.

Anyway, now that I'm back to stock clocks, there's a 20 degree delta for OCCT (~70 degrees OC, ~50 degrees Stock) and a massive 30 degree delta for Prime95 (~90 degrees OC, 60 degrees Stock)! This seems ridiculous, right? My cooler is clearly seated just fine and everything, so I'm wondering why my temps are shooting so high? I started OCing my i5 this weekend, but I thought I had gotten lucky when I was stable at 4.5 GHz, 1.2 Volts after 2 hours of OCCT. I guess my dirty overclock wasn't too lucky after all. This is something best saved for when I have a week's worth of time to devote to studying everything I can about this. But in the meantime, do you or anyone else know of any possible reasons my temps are so outrageous?
 
Upvote 0 Downvote


put less stress on the IMC, loosen timings of the RAM. maybe 4.4Ghz is the sweet spot, sometimes that .1ghz can make all the difference in terms of stability/temps.
 
Upvote 0 Downvote
Running into the same issues with my i5-4690k and my H80i cooler, it instantly hits 90-100 degrees in AIDA64 when running at 4.5 GHz about 1.28V, maybe just me and I need to do some more fine tuning but this is gonna be a lengthy process.
 
Upvote 0 Downvote
I don't think it's a memory controller issue. My RAM is Corsair Vengeance DDR3-1600 9-9-9-24, with my XMP set to its default profile and I haven't touched my BCLK or CPU strap, so that should be fine, right? Also I tried 4.4 GHz and that was still reaching 90 degrees at 1.2 Volts on Prime95. I was trying 4.3 GHz at 1.15 V at some point as well but only did it with OCCT for about 10 minutes but I can't remember what the temps were -- only that they were good enough that I wasn't freaking out about them. However, I don't believe I tested that 4.3 on Prime95. I'll go back to that at some point later and let you guys know, but can someone tell me what the Fixed CPU Input Voltage at 1.9 V is about? It seems concerning.

Also, ideally I think when I finally start doing the appropriate research on this, I'll go for a per-core overclock, so can someone tell me if this logic follows through? I think if I could get all 4 cores to operate at 4.3 GHz under load, I'd be pretty happy. In terms of gaming, any game that actually uses all 4 cores is probably optimized well enough to maintain 60 fps at 3.5 GHz, let alone 4.3. Then I think I'll try to get 2 cores to reach 4.6 GHz or maybe 4.7 if possible, since games like Borderlands 2 which don't use more than 2 cores really need those cores to be powerful. Then for the worst optimized games that only really use 1 core, i.e. Crysis 1 and Crysis Warhead, I'd try to get a single core as close to 5 GHz as possible without heat constraints. So does that make sense and does that seem feasible?
 
Upvote 0 Downvote




just set the ram to auto temporarily to see if the temps are any better, then we can rule out stress on the imc
 
Upvote 0 Downvote
Deus Gladiorum,

Here's the problem:

" ... Core i 2nd, 3rd and 4th Generation CPU's have AVX (Advanced Vector Extension) instruction sets. Recent versions of Prime95 run AVX code on the Floating Point Unit (FPU) math coprocessor, which produces unrealistically extreme temperatures. The FPU test in the software utility AIDA64 shows the same results.

It's not necessary to run AVX code for thermal testing. Prime95 v26.6 produces temperatures on 3rd and 4th Generation processors more consistent with 2nd Generation, which also have AVX instructions, but do not suffer from thermal extremes due to having a soldered Integrated Heat Spreader and a 35% larger Die. ... "

That's from the Intel Temperature Guide - http://www.tomshardware.com/forum/id-1808604/intel-temperature-guide.html You might want to give it a read.

You can download Prime95 version 26.6 from here: - http://windows-downloads-center.blogspot.com/2011/04/prime95-266.html

Try running Small FFT's ... you'll be happy with your temperatures.

CT :sol:
 
Upvote 0 Downvote
Solution


Very interesting! I'll download version 26.6 and see what it's like and I'll certainly give that guide a thorough read. Thanks a lot! However, I'm not so sure that's the source of the problem, or at least, not all of it. I had an FX-6300 before this which I overclocked immensely and ran a near-24 hour stress test on Prime95 at 4.5 GHz and 1.428 Volts. Though I don't remember the exact temps, I'm almost positive they never went over 60 -- in fact, going off memory and some old forum threads I've created here I'm almost positive they never even hit 60! Yet the piledriver microarchitecture supports AVX, and this was only a year ago or actually less since I stress tested and overclocked my FX-6300, well after version 27.7 came out in 2012. So why would there be such a massive discrepancy in temperatures? I'll still give that post a thorough read when I get a chance though, in case you answered that there, but any theories on that you could tell me about right now?
 
Upvote 0 Downvote
Damn you, Intel... and damn you equally, AMD, for not lighting a fire under Intel to provide better quality in their CPUs with regards to overclocking... But anyway, isn't using version 26.6 a little counterintuitive? The point of stress testing is to give a worst case scenario for your CPU, so if AVX isn't being used, you're not finding a worst case scenario and hence not really stress testing, right? Or am I missing something?
 
Upvote 0 Downvote
No, you're not missing anything.

"Stress" testing is for establishing stability, which is a broad and debated topic. What's the point in stress testing for lower overclock settings just to accommodate extreme processor features which you may never use?

"Thermal" testing is for establishing a Core temperature benchmark, which is a 10 minute test. There's a specific method to properly conduct a test. P95 v26.6 Small FFT's is used.

It's all explained in the temp guide.
 
Upvote 0 Downvote
I suppose. I read the whole guide and I understand the emphasis on separating stress and thermal testing, but even so I'd like to do everything possible to reduce overall temperatures short of spending more money or deliding. After all, even if the AVX instruction extensions are a ridiculous thermal testing method, they're still good for general stress testing, but I'm guessing that letting my CPU run at 90 degrees for 4 hours wouldn't be a great idea. Even if the max temp is 100, letting it hit 90 or even 80 for any duration of time is ridiculous and probably not recommend, right?
 
Upvote 0 Downvote
Have you tried 26.6 yet?

I would recommend running whatever software you use on your rig to test for stability including apps, games, utilities and benchmarks. If you don't run any apps that use AVX / FMA then there's no point in down-tuning your OC just to accommodate FPU stability testing with P95 28.5 Small FFT's or the FPU test in AIDA64.

The 4.7 GHz OC I'm running on my 4770K is 100% stable for everything I use, but it will blue-screen if I run the above tests. My average Core temperature at 22C ambient with 1.280 Vcore on P95 26.6 Small FFT's is just 76C with big air, but I'm also de-lidded and lapped.

CT :sol:
 
Upvote 0 Downvote
Yea, in 26.6 they run closer to 70, but I think the more pressing matter is that it's not stable at 4.5 at 1.2 Volts according to Prime95. I don't feel like bumping voltage and generating even more heat, so I think I'll push all 4 cores down to 4.3. However, I can't get the per-core overclock to work unless I actually disable the other 3 cores. I thought that the per core overclock was supposed to allow 1 core to get up to a certain OC (in this case 4.7) as long as the other cores weren't under load, but from what I was seeing, that wasn't happening at all. In fact, all 4 cores managed to sit at 3.5 even though under load 4 cores were supposed to sit at 4.3. Bleh. I need that full week to devote to this. Anyone know any really good Haswell and/or Devil's Canyon OC guides?
 
Upvote 0 Downvote


Here are some solid reads:
http://www.overclock.net/t/1401976/the-gigabyte-z87-haswell-overclocking-oc-guide
Scroll down on and look at OCing steps
http://www.overclock.net/t/1490324/the-intel-devils-canyon-owners-club
http://www.anandtech.com/show/8227/devils-canyon-review-intel-core-i7-4790k-and-i5-4690k/2

Currently running mine at 4.5 GHz at 1.31V with temps going into 70s under full load in P95 and v26.6 P95 FFT, similar results with OCCT, AIDA64, and POV Ray

 
Upvote 0 Downvote
I tried OC my 4690K as well using the OC profile AsRock Motherboard UEFI has (which I didnt even know exists O_O). OC'ed to 4.5 and stress tested with Prime95, after 2 tests my temp already hitting the low 90degree Celsius. I got scared so I backed down to 4.2 and temp circled at 40degree celsius on idle, I didn't notice any difference in speed so I went back to stock Intel turbo boost 3.9 to prolong my chip's life. I'm using Noctua NH-U9B
 
Upvote 0 Downvote
Not that I don't appreciate your help, CompuTronix, but I'm from the school of thought that when it comes to stress or temp testing, I want to see my PC under the absolute worst case scenario, so I stuck with the recent version of Prime95 for my temp test. At 4.3 GHz at 1.5 Volts, I performed a stress test on version 28.5 of Prime95 on the Small FFTs run for 20 minutes straight. None of my cores failed, and my max temps were as follows:
Core 0: 88
Core 1: 88
Core 2: 89
Core 3: 83
Package: 89

I can see what people mean about Intel's TIM under the heat spreader not being distributed evenly when I look at that Core 3 temp, which has always been consistently 6 - 7 degrees lower than any other core in every stress/temp test I've done. Anyway, that aside, how do these temps look? After 20 minutes, I never hit 90 which makes me feel confident for a 20 minute temperature test. But to the rest of you, does that seem alright for a recent P95 test?
 
Upvote 0 Downvote
1.5 Vcore? That can't be right. Perhaps you meant 1.15?

Without knowing your ambient, those temperatures are excellent. Also, temperature differences between cores may not be due to TIM; it can simply be due to the fact that Intel's specification for sensor accuracy is +/- 5%. The most likely reality is that to a some extent it's a combination of both.
 
Upvote 0 Downvote
Sorry, yeah I meant 1.15 ahaha. It'd both be extremely impressive and extremely worrisome if I was reaching only 89 degrees at 1.5 Volts without my CPU frying. As for my ambient, I forgot to post it in the last post but I recorded it. Putting a non-contact thermometer which measures room temperature on my front intake fan, the recorded temperature was 22.4 degrees Celsius. So for 89 degrees, that sounds pretty good, right?

In addition, I just did a 1 hour stress test on Blend through Prime95 version 28.5, and managed to pass. I'm planning to try for a longer one, at least a 6 hour one, but really it was just a preliminary to ensure I'm at least fairly stable.
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom