Haswell Overclock results and temp

[NGXfanatic 85]

I think I have found my final config:

 

Core/Cache: 4.4/4.0

CPU Core Voltage Override: 1.290

CPU Cache Voltage Override: 1.290

CPU System Agent Voltage Override (VCCSA): 0.280

 

Per Idahosurge's advice, I ran OCCT cpu AND linpack tests, with 74 and 82 degrees max temp respectively.  VCCSA max voltage maxed out at 1.184, well within limits.

Any lower than 1.290 vcore is just not possible, and adjusting either cpu cache or system agent voltage overrides will also not work.

 

Going to start running FSX finally!  I think I may run some other testing programs, perhaps Prime95 or Intel burn test, just to be sure I don't run into issues with FSX, but I think I can finally put this to bed and see how FSX, as configured for my old 930 I7 setup, responds to the new engine under the hood.  

 

Thanks to everyone for your patience and sound advice, I really do appreciate it.

[martin-w 3,235]

I question the value of using artificial benchmark routines to establish stability unless the real world application puts the same stress as the the artificial benchmark routine will.

 

Wise words Noel.

 

This has always been my philosophy. But that doesn't mean something like Prime isn't useful. Like you, sensible use of Prime I favour.

 

Why do we build our rigs? We build them to run our software, the software we use daily... we don't build them to run CPU melting stress tests all day.

 

A system that is stable when running the software we built the system for is all we need.

 

The alternative is a little crazy. If we are stable in Linpack, what next, what if a new stress test comes along that our systems fail in? What then, do we regard our systems as unstable and tweak some more, add more voltage? And what if the new stress test is superseded by an even more demanding one? What then, do we continue the crazy scenario?

 

Of course not, all we need is a system that is stable day to day, with the software we run. 

[idahosurge 2]

For the record the way I defined to test an OC and how to lower voltages and retest is the exact same method used by NickN in his Haswell OC guide, that is where I got it.  Also the tests I said to use are the exact same tests used by Nick in his Haswell OC guide.

 

Regarding stability testing with OCCT as defined by Nick, once you pass these tests you have established that your OC is stable in these tests.  These tests only get you to a 99% to a 100% stable OC.  In his guide Nick says that even after you pass these tests it is still possible to have problems (BSOD, lockups, crashes, etc.) in FSX, most people do not, but some do.  If that happens after you pass these tests then you need to bump vcore by 0.01 and run FSX and keep doing this until FSX is stable.  You also might have to bump CPU cache by 0.01.

 

Regarding stability testing, for Haswell you only need to use the tests outlined by Nick, for Haswell there are no other tests you need to do and anything else that comes out in the future will not need to be used.  No other real world applications that most of us use will stress an OC like the two OCCT tests and other than FSX I know of no program that will BSOD, crash or lockup your OC once you pass the two OCCT tests.  You may run into problems with FSX only because of the amount of 3D rendering required by FSX.

 

If you want to test for 3D rendering you can use 3DMark06, this is also a great benchmark to see how well FSX will run on you PC.  If interested you can read up on using 3DMark06 over at SimForums.  Also you can use the free version, nothing to buy.

 

If you do not want to stress test an OC and wait and see if you may or may not have problems is a choice each person makes, but I will stress test my OC, trim for FSX as required and then I am done.  Any problems in the future I can be sure are not due to an unstable OC, but something else and when I have a problem I am not running around trying to fix an OC when the problem is elsewhere or I am not trying to figure out why a program will lockup when all the time it is an unstable OC.

[Boomer 1,619]

Maybe a thread derail, dunno, but I am moving from i7/920 o/c to 3.9 to a haswell that a plan to o/c.

 

For FSX specifically, all things being equal, what kind of performance or FPS increase could I anticipate?

[Noel 4,499]

Of course not, all we need is a system that is stable day to day, with the software we run. 

 

I decided to increase back to 4.4 from 4.3 however at the same vCore as at 4.3:  1.29v.  What do you know, no instability, two long flights completed and all was well.  

 

As an experiment to demonstrate just how poor an approximation Prime95 is as a surrogate for running FSX I compared temps in Prime95 w/ temps in FSX, using HT enabled or HT disabled.  This also demonstrates why HT enabled does not increase temps appreciably in FSX.  In Prime95 at 4.4Ghz (or 4.3) core temp shoots to 65C within a few seconds (HT on or off), at which point I shut it down as I'm not too keen on stressing my CPU just to see how much it can endure.   In FSX temps rarely get to 60C after sustained use, and that's HT on or off.   This is proof positive of just how much more stress Prime95 puts on the CPU than FSX does, hence this approach to establish 'stability' is really a poor one, primarily leading to overstressing one's CPU unnecessarily. 

 

NickN and I had an interesting interaction in here many moons ago which points out the perspective he delivers advice his unquestioned authoritative advice from.  He railed on me for sending 40F air at a low flow rate into my PC case from my A/C unit as previously described, admonishing this dangerous practice on the grounds that ambient temps must remain very constant else bad things will most assuredly occur.  He referenced 'mission critical' l server or workstation environments that must be kept very environmentally controlled.  I've done this for over 10y and have had no problems.  Suggesting people put their FSX boxes thru this sort of stability-determining routine comes from the same perspective:  yes, if we are playing 'Global Thermonuclear War' on the WOPR, why then sure I can support that.  But...FSX?  I will always argue this approach does more harm than good!

[Boomer 1,619]

NickN, I am afraid, is not the last word on these affairs.  He is simply wrong on some issues or his OPINION open to dispute.

[NGXfanatic 85]

Maybe a thread derail, dunno, but I am moving from i7/920 o/c to 3.9 to a haswell that a plan to o/c.

 

For FSX specifically, all things being equal, what kind of performance or FPS increase could I anticipate?

I've seen a massive increase in performance, from the 4.2ghz o/c I7 930 to my Haswell at 4.4ghz.  This is not a scientifc test, but I did a run a quick benchmark:

EGLL at runway 9R, 12:00 noon with the clear skies FSX weather theme, 100% AI (a mix of UT2 and WOAI)

 

930 I7/480GTX : 9.5 fps

4770k/780GTX: 15.3 fps

 

So I saw a 60% increase in that quick test alone.  This should give you an idea of what to expect.  I have tested some saved scenarios that I used to run on the old rig, and using the old systems FSX settings, where I used to see mid to low teens I now experience mid to high 20's, often at 30fps locked limit! 

There is no question that the Haswell is a worthwhile upgrade to the first gen I7s, not so sure for the SB/IB. 

[martin-w 3,235]

The Corsair H110 is a liquid cooler isn't it?

 

I think it's a bit easier to achieve higher clock speeds with liquid cooling as opposed to air cooling like the OP is using.

 

That depends on which coolers you're talking about.

 

The H110 is a great performer and quiet. Don't right off air cooling though, The NH-D14 does a great job and very quiet. Many of these cooling reviews don't consider the fans, namely that some of the AIO coolers have very high RPM, very noisy fans at full tilt. To truly compare, you need to compare coolers while using the same fans.

[NGXfanatic 85]

The Thermalright Archon SB-E x2 I now use, surprise surprise, because NickN has it, is a monster!   Not big enough to force you to remove the heatsinks from RAM modules, but it has two fans, one on each side.  My previous heatsink was a bit smaller with just one fan.   One thing I have noticed about the fans are that they have a nice little hum to them, kinda like the sound of the PMDG 777!  And I notice it spooling up when the loads rise in FSX.  Not sure if it's a nuisance or not, but I will say the chassis fans are much more quieter in my case, so that may explain why I'm noticing it. 

[martin-w 3,235]

Nice coolers the Thermalrights, I've always liked them.

 

It doesn't beat the D14 though does it? Quiet a bit warmer I think? Not surprising given it's a single tower.

 

The Silver Arrow equals the D14, almost. The extreme beats it, but it does so because of the powerful noisy fans. So not as a result of cooler efficiency...it cheats in other words.

[NGXfanatic 85]

I haven't seen any reviews or data, but I presume the D14 beats the Thermalright hands down.   I just had a look at for the first time, now I know why it blocks tall RAM heatsinks.  I thought my new heatsink was huge, that one is a monstrosity!

[TheFamilyMan 280]

This chart from xbitlabs may prove be useful:

http://www.xbitlabs.com/picture/?src=/images/coolers/thermalright-silver-arrow-sb-e/zchart_table_big.png

 

I'm 100% satisfied with my Silver Arrow SB-E (not extreme) with its PWM fans, which is a feature that the NH-D14 fans suprisingly lacks.

[Zikous82 5]

I have i5 4760k oc'ed from 3,4 GHz to 4,4 GHz just by one click in BIOS - I have ASRock MB and there are some safe presets. Temperature of all cores doesn't go over 70°C (I'm using Noctua NH-U9B SE2). Working very vell. FSX is nice and smooth even with tons of addons (PMDG 737 + FTXG + REX + AS2012), sliders set to max. Test flight around KJFK during heavy tunderstorm - 25 and more FPS.

[PigsInSpace 21]

This chart from xbitlabs may prove be useful:

http://www.xbitlabs.com/picture/?src=/images/coolers/thermalright-silver-arrow-sb-e/zchart_table_big.png

 

I'm 100% satisfied with my Silver Arrow SB-E (not extreme) with its PWM fans, which is a feature that the NH-D14 fans suprisingly lacks.

 

Put three fans running at just 750rpm on the D14 and then see what happens.  :blink:

 

PWN fans on my D14 here (150mm to boot):

 

[TheFamilyMan 280]

Excellent air cooler setup!  I would have gone with the NH-D14 if its stock fans were PWM for their non-2011 socket model.