X99 Deluxe - 5930K - 32GB DDR4 3000

[Category 5]

I'd be absolutely floored if that killed something. Definitely clear the CMOS and try again.

Me too! But you never know. I've read in a dozen places P95 can kill Haswell CPUs. Makes no sense since some people may actually be building a machine to search for new primes, but I guess you can't use intel for that. lol I like Prime better than AIDA because AIDA pretty much freezes the system up. P95 lets you still use the machine while you test. If I can get stable though, then I can use my machine...

I'm back in Windows at 4.5GHz with 1.3V and 100MHz bus speed now. Ra, is running 2133 at 1.2V (which it's also rated for). AIDA64 going for 30 minutes or so.

 

[Woomack]

You can't really kill a CPU unless it's overheating and you keep it like that for too long. Then it will degrade faster but shouldn't instantly die. CPUs have overheating protection. Depends from CPU but usually near 100*C it will shut down. The latest CPU instructions may cause temps bump but if you have a good cooler then all should be fine.

 

[Category 5]

I'm testing with the side cover off and temps are much more reasonable. I have to rethink my airflow. I have my fans pulling in cold air from the top but there's just not enough exhaust to keep it flowing. I think I need to bring cold in from the 200mm fan in front and expel it out the top, through my radiator. Even though it will be heated air from the case I think my temps will be better thanks to the more consistent and constant airflow.

3 hours of AIDA64 at 4.5/1.3 now Temps around mid to upper 70s with occasional peak at 80-81. How long should AIDA go? 24ghours? What should be my next line of testing? Or is AIDA comprehensive enough to call it stable and move on to working up the RAM?

 

[EarthDog]

Front/sides = intake, top/rear = exhaust for good airFLOW.

 

[Woomack]

Just to add something to what ED already said. Hot air is going up so in tower build, exhaust is recommended to be on top for easier airflow. The same colder air is usually near the floor ( on which is usually standing pc ) and because of the most popular pc case build, intake is in the front/bottom.
There are of course exceptions if you have less standard case but if you are buying new case with already installed fans then you will see that simple rule regardless of brand.

If you have PSU on the bottom of the case then in most new designs it's recommended to install PSU with fan down so it can take cold air from outside the case and won't interrupt airflow inside the case.

 

[Category 5]

Just to add something to what ED already said. Hot air is going up so in tower build, exhaust is recommended to be on top for easier airflow. The same colder air is usually near the floor ( on which is usually standing pc ) and because of the most popular pc case build, intake is in the front/bottom.
There are of course exceptions if you have less standard case but if you are buying new case with already installed fans then you will see that simple rule regardless of brand.

If you have PSU on the bottom of the case then in most new designs it's recommended to install PSU with fan down so it can take cold air from outside the case and won't interrupt airflow inside the case.

Yeah. That philosophy makes sense. I did it the other way because i read a lot of accounts where pulling cold air through the radiator was better that pulling hot air from the case but if you have air coming in and not enough going out then it's a fail, as my configuration shows when the cover is on.

4.5 1.3V at DDR4 2133 (100FSB) ran AIDA for 27 hours until I got home tonight. Victory #1.

 

[Brando]

NICE! now if it bombs with ram in xmp mode you know the problem.

 

[Category 5]

Well, I was playing a little bit of tomb raider and did have the game unexpectedly quit. that was once in about an hour and a half. Still susoecting RAM (SA voltage?) because I know 32GB puts a strain on the IMC. Well, at least that was the case the last few OC's I did (970, 3370K)

What other stability tests are recommended now that P95 is the devil? Is IBT still good?

 

[Category 5]

BTW, cache multiplier is set to 24x min/max. Leave that at default or should it match RAM/CPU speed or something?

I see a lot of people test in AIDA only cpu and memory selected. I was testing with FPU and cache also enabled. What is the best/most reliable config for 24/7 stability test.

 

[EarthDog]

Leave it at auto for cache...

I run with FPU and Cache (everything) as well.

 

[Category 5]

Thanks Earthdog. You and the others are a super help here and it's MUCH appreciated. Wouldn't have gotten here without the advice.

So mine isn't set on auto though. It's set to 24 for both min/max. That the same thing? It's not a pull-down, just a field I can type in. I suppose I can type AUTO and see if that works too.

 

[EarthDog]

Not sure on your board... mine has auto and other values.

 

[Category 5]

Gotcha. So it's realbench that's causing restarts so teh good news is I have something to expose my weakness. the bad news is, nit sure where to look.

4.5, 1.3V, FSB 100 - RAM 1066 - SA V 1.02

Where would you go next? I raise CPU V by a notch and didn't get anywhere. 4.5 seems reasonable for this chip, but should I dial back the clock then?

 

[EarthDog]

I would stop looking for instability...

Do you 'real bench' for a living? If not, I would assume AIDA64 stress test, P95, or Intel XTU stress test for a couple of hours would suffice. I wouldn't try it out it much more than one of those three, and a game, or whatever you may do productivity wise the with PC.

You will find that everyone has an 'idea' of stable... be it this application or that one... 2 hours or 24... The key is to find something that works for you. I stick to AIDA64 and XTU and a game I play as I do not do anything else on my PC that would work it harder than those items. It works well for me!

 

[Woomack]

I'm not sure what can cause instability when you have memory @2133. At this frequency memory controller shouldn't need any additional voltage.
You can check it other way. Set everything at auto but enable XMP and check if it's stable in games. If it does then change CPU ratio and CPU voltage to higher. Everything else keep at auto.
Cache shouldn't need higher voltages at ~3300MHz.

 

[Category 5]

Yeah. Not looking for it but thought AIDA meant I was good, and then tomb raider crashed on me last night so i figured it was back to the drawing board. I did set the NB to AUTO and it clocked it at 3000 so that 2400 must have been set when I enabled XMP.

I'll keep at it. I'm close since I can do things in windows. Just crashing under stress. Maybe i'll dial it back to 4.4 or even 4.2 Both would still be respectable by most counts.

 

[EarthDog]

I run 4.250 at less than 1.2V for daily driving. I can bench the thing to over 4.750 GHz (5820K) and am rock solid at 4.5GHz. I am sure I could stabilize 4.75GHz+ if I chose...

 

[Woomack]

I run 4.250 at less than 1.2V for daily driving. I can bench the thing to over 4.750 GHz (5820K) and am rock solid at 4.5GHz. I am sure I could stabilize 4.75GHz+ if I chose...

I had similar settings and for longer tests and reviews I was always using ~4.2-4.25GHz at ~1.15-1.2V just because it's easy to stabilize on even worse chips and it's good for any comparison on air/water. My brother is running [email protected] with 32GB G.Skill@3200 ( it's XMP ) like 24/7. He got my better 5820K which can make ~4.8GHz on water but really there is no point to keep it so high for games or anything else.
Tbh I was never using X99 for games. When I think harder on that then I always had worse PC for games than for tests/overclocking.

 

[EarthDog]

The problem, for me, is that I need to keep our 'review platform' (Z170/6700K)... I suppose I could 'daily drive' another just like it, but, I wanted to keep the HEDT system for its 6 cores and 3D...


....... like I benchmark much these days...

 

[Woomack]

I keep X99 mainly for tests/reviews but I wasn't using it much lately. I actually switched to [email protected]/[email protected] for reviews and additional tests but had to make RMA couple of days ago. Even though X99 is faster and is simply higher platform then all is about Skylake nowadays. Most new memory kits are in dual channel and X99 can't really run at high memory clock ( above 3200 ) so all higher memory tests have to be performed on Z170.

I also barely benchmark anything lately. One day in Dec I was benching on cold and nothing since then.