Strange 6700k OC results (HT on/off)

[honanhal 1,420]

A few months back I posted about temps higher than I'd like when running P3D -- usually mid-to-high 70s C, but with spikes to up to 84C under certain conditions. At the time, I decided to just live with it. I was using a mobo-standard 4.6 GHz profile for my 6700k, with HT on. Voltages set to auto in bios. Noctua NH-D15S for air cooling. I figured the auto voltages were probably overdoing things a bit, which accounted for the heat, although I did notice that P3D ran about 10C hotter than even other demanding programs. Even RealBench topped out at about 79C running a stress test. Oh well, I thought.

After putting my system in a new case (Fractal R6) and with a new PSU recently, I again got the itch to tinker. My pre-tinkering temps were not significantly different from before (no huge surprise, since the cooler was the same). I played around a bit with manually setting voltages, but quickly discovered that the auto rules seemed to be pretty effectively maximizing the stability/voltage curve already. I was about ready to give up for a second time, when I decided to try HT off. I knew that would give me about 10C headroom. But I was surprised by what I found next.

As expected, RealBench ran a lot cooler at 4.6 with HT off. I then tried bumping the OC up to 4.7 just for fun, to see how things would shake out. No big surprise in terms of RealBench: it now still topped out at 79C, but running 100 MHz faster. OK. Probably not worth keeping HT off if that's the difference. On to P3D.

Imagine my surprise when I found that P3D now ran more than 10C cooler than it had before, even though the single-core speed had increased from 4.6 to 4.7 GHz. I never in my wildest fever-dreams expected to both solve the temperature problem and bump up my OC, but here I am. In a P3D test I just ran, it never even broke 70C in a scenario where it had previously gone above 80C.

Is this expected? What could account for this? I don't think I'll miss HT at all!

(Don't get me wrong, I'm thrilled. I'm just curious what could possibly be going on here.)

James

[martin-w 3,234]

Unless I'm misunderstanding, which wouldn't't be a surprise as I'm on my phone...

You increased OC by 100 MHz and are running with HT off.

4.6 to 4.7 hasn't impacted your temp much and HT off has reduced temp by 10-12 degrees. So not much of a mystery really.

 

Sorry if I'm misunderstanding.

[zmak 331]

Does your new case have extra fans built in or are they larger diameter? My case has 4 large diameter fans and my OC of 4.6ghz is on a $50 cool master cooler with temps never higher than high 60"s

[honanhal 1,420]

3 hours ago, martin-w said:

4.6 to 4.7 hasn't impacted your temp much and HT off has reduced temp by 10-12 degrees. So not much of a mystery really.

 

Sorry if I'm misunderstanding.

Except that my temperatures in RealBench stress testing have not gone down by 10-12 degrees. They are the same as they were at 4.6 with HT on. It’s only P3D with this decrease.

[honanhal 1,420]

2 hours ago, zmak said:

Does your new case have extra fans built in or are they larger diameter? 

Yes (larger diameter). But before I switched HT off I was getting nearly the exact same temperatures as with my old case when P3D was under load.

[Iceman2 110]

How’s the performance in P3D, FPS and stutters. Have you seen an improvement. I read here somewhere that the later version of P3D run better with HT on.

IM

[westman 257]

Not sure i understand you

You oc 100mhz ht off,  in  realbench you  see not temp drop  as in P3d. 

Do you run with manual vcore or auto 

First i Think not realbench run with avx as P3d, avx with ht on produce  lot off heat

Realbench normally you get a increase in heat with 100mhz oc and a sligtly decrease with ht off , with same vcore but if you have auto the auto can increase the vcore 

The 6700k have no avx downclock setting as the later  cpus from 7700k.

Think Its normalt, if i understand it right 

 

Edited May 7, 2018 by westman

[honanhal 1,420]

Performance seems much smoother with HT off, actually, although there may be slightly increased blurries. I’d figure the smoothness is mostly simply from the higher clock speed, but it really seems like a large qualitative difference, especially with the FSL A320, which was always kind of jerky on my system. More testing needed.

I’m using auto voltage, so it’s possible that the voltages just haven’t changed much from 4.6. But again, why would RealBench temps be higher in that case?

You’re right that the lack of AVX offset for the 6700k is a major reason I was doing this at all. HT on at 4.6 worked great for everything else on my computer, but was just running too hot in P3D. Maybe there was some way to moderate it with an AM setting (I’ve stuck with the default) but all the discussions of AM I’ve seen here have gone significantly over my head...

James

 

[martin-w 3,234]

4 hours ago, honanhal said:

Except that my temperatures in RealBench stress testing have not gone down by 10-12 degrees. They are the same as they were at 4.6 with HT on. It’s only P3D with this decrease.

 

Was there a variation in ambient temp between 4.6 and 4.7 testing.

HT off drops temps by 10 degrees, so P3D is behaving normally.

 

As for RealBench I wouldn't take much notice, as it depends how its stressing the CPU. If you haven't seen a temp drop then there's a variable you aren't accounting for.

Edited May 7, 2018 by martin-w

[honanhal 1,420]

No meaningful ambient temp difference between 4.6 and 4.7. Absolute maximum difference of 1C, and I’d guess less.

It’s very possible I’m failing to account for a variable, Martin, which is exactly what I’m trying to figure out. That said, I think westman is on to something with the AVX and HT interplay. If AVX with HT off produces much less heat, that could account for the P3D temp difference even with most temps across the board being comparable between the two clock speeds.

Maybe I didn’t explain what I saw as the real puzzle here very well:

4.6 w/HT on: P3D peaks 5C hotter than RealBench stress test

4.7 w/HT off: RealBench peaks 10C hotter than P3D

It’s that delta of 15C (or even the 10C delta between the two programs at 4.7) that surprised me, not the fact that HT off lowered temps.

James

[martin-w 3,234]

Quote

 

deleted

Edited May 7, 2018 by martin-w

[honanhal 1,420]

4 minutes ago, martin-w said:

 

CPU cooler fan set to PWM? So the higher temp compensated for by faster fan RPM? When testing, fan RPM fixed is the best way.

 

 

You’re right, PWM fan. Possibly explains the RealBench temps, but in P3D the fan never even gets to full speed at 4.7 with HT off (as opposed to 4.6 with HT on where it is at 100% about half the time under load). So something very different is still happening in P3D. It’s definitely not that the fan is running more — just the opposite is true.

James

[martin-w 3,234]

 

Quote

 

4.6 w/HT on: P3D peaks 5C hotter than RealBench stress test

 

4.7 w/HT off: RealBench peaks 10C hotter than P3D

 

 

 

It's just down to the way RealBench functions, how it's coded I would say. We could try and fathom out precisely why, but in the end it's just how it's designed, how it stresses the CPU. I don't think it's reasonable to expect different applications to react the same way, thermally, to HT being off or on.

[honanhal 1,420]

Yeah, that's fair. Like I said, at the end of the day it's a good result whether or not I can fully explain it!

After more testing, I've found that HT off actually does seem to make my sim significantly smoother. Sadly, I found that the initial Aerofly-like levels of smooth-as-butter flight were partly due to the fact that autogen was barely loading at all, but switching to locked frames at 60 and an FFTF setting of 0.05 solved the autogen loading while preserving close to consistent 60 FPS even with PMDG aircraft. That's not bad at all.

Given the fact that I needed to switch to locked/bump up FFTF to get autogen loading normally, it's clearly true that HT on does produce some benefit in P3D in terms of terrain/autogen loading. HOWEVER, even aside from the unacceptable levels of heat that was generating (and with a 6700k, I don't have the option of using an AVX voltage offset), the higher clockspeed with HT off seems to have easily compensated for that benefit.

Bottom line: performance/smoothness is as good if not better at 4.7/HT Off as it was at 4.6/HT On, but with temps that are 10C lower. It's wonderful to be able to just fly without fretting about temperatures and having my computer fans sound like they're getting ready for liftoff!

James

[martin-w 3,234]

Quote

performance/smoothness is as good if not better at 4.7/HT Off

 

Changing track somewhat... my 8700K isn't a particular good overclocker (although it may simply be an immature BIOS issue). With HT on, quite high voltage required for 5 GHz. Think it was around 1.435 if I recall correctly. However... switch off HT and 5 GHz is a piece of cake, voltage pretty low and nice and cool. Given how so many titles make little use of HT, seems a way better approach.

Having said that, I'm not overclocking at all at the moment. Don't sim anymore, and games are all about the graphics card.