In the beginning I ran 4.5 GHz for a while, and after going to 4.7 I got 10-15% more

 

Magical thinking I'm afraid Arjen. Oh you may have thought you 'got 10-15% more', but it was something else beyond the 200Mhz that brought this about: you changed something else, you didn't do an apples-apples comparison, etc.

 

 

Magical thinking I'm afraid Arjen. Oh you may have thought you 'got 10-15% more', but it was something else beyond the 200Mhz that brought this about: you changed something else, you didn't do an apples-apples comparison, etc.

 

Why do you make conclusions when you have no idea what you're talking about? I mean, I know you are a smart person regarding PC's, but you do not own a Sandy Bridge system and therefore you cannot comment.

 

I have just tested it again with the following settings:

 

BP=0, HMF=1, FFTF=0.33, TML=4096, AM=14, FPS unlimited, 1920x1080x32, anisotropic, AA on, global texture resolution=maximum, DX10 off, lens flare on, light bloom off, advanced animations on, aircraft shadows off, aircraft landing lights illuminate the ground=on, LOD_RADIUS=4.5, mesh complexity max, mesh resolution max, texture resolution max, water HIGH 2x, autogen max, land detail textures on, ground scenery shadows off, weather maximum, cloud draw distance 100 nm, all AI traffic including road and boat traffic off.

 

Scenario:

 

PMDG 737-700NGX KLM on Aerosoft Schiphol EHAM runway 18C, 12:00 zulu, summertime, clear skies.

 

Other add-ons used: REX, GEX, UTX, FSUIPC. OpusFSX was not active at this time.

 

4.5 GHz: 33 FPS

 

4.7 GHz: 37 FPS

 

4.8 GHz (1.45V): 38-39 FPS.

 

Perhaps the higher clockspeeds allow for an even bigger performance increase?

 

I will now try the difference between 4.0, 4.1, 4.2 and 4.3 GHz and report back.

Why do you make conclusions when you have no idea what you're talking about? I mean, I know you are a smart person regarding PC's, but you do not own a Sandy Bridge system and therefore you cannot comment.

 

Sure I can comment. You don't have to consider my comments which is your choice of course. Perhaps you didn't let the machine boot completely for your first test, or something else. Perhaps you have some competing processes that were happening w/ test 2 but not test 1. I'm not convinced you, Hasse, Word Not Allowed, Jim, or anyone else can dial up clockspeed by 4% and see '10-15%' increase in performance, especially since your video card is so much more than you can use for the scenario you described. Just be careful w/ the overvolting issue unless you're ok w/ having to buy a replacement. I don't think the model of processor has anything to do w/ the argument either. You are describing a non-linear performance change which *could* be explained by some odd timing issues which I've previously conjectured was in the differential diagnosis of this stated claim from others. But if that's true, you *may* then find 4.8, 4.9Ghz WORSENS performance scaling: increase clockspeed by 4%, get 1% increased performance.

Some test results:

 

1.6 GHz: 4-6 FPS, cockpit stays white and refuses to load

2.8 GHz: 15 FPS

3.3 GHz: 21 FPS

4.0 GHz: 26 FPS

4.1 GHz: 27-28 FPS

4.2 GHz: 29 FPS

4.3 GHz: 31 FPS

4.5 GHz: 33-34 FPS

4.7 GHz: 36-37 FPS

4.8 GHz: 38-39 FPS

 

I tested the 4.5 vs 4.7 GHz two times no, with the same results.

 

So it seems that in the low GHz there is a big difference in FPS between every 100 MHz, in higher MHz this difference is relatively smaller, 1-2 FPS, but still there seems to be a relatively big spike in FPS increase in the 4.5 - 4.7 GHz area. I don't know why exactly, but these are the results I'm getting. Maybe it's because a certain file, effect, add-on and something similar suddenly starts to run much better at 4.7 which causes the big increase.

 

Anyways, sometimes I'm getting nowhere on these forums with my actual questions, topic always get highjacked because some people don't agree or can't stand the test results others are getting.

 

1.6 GHz: 4-6 FPS, cockpit stays white and refuses to load

2.8 GHz: 15 FPS

3.3 GHz: 21 FPS

4.0 GHz: 26 FPS

4.1 GHz: 27-28 FPS

4.2 GHz: 29 FPS

4.3 GHz: 31 FPS

4.5 GHz: 33-34 FPS

4.7 GHz: 36-37 FPS

4.8 GHz: 38-39 FPS

 

Hi Arjen,

 

You seem to have confirmed exactly what occured in my scenario while overclocking.

 

For quite a while I ran my 2700K @ 4.8GHZ and I was mostly happy. The only time I got irratated was when I would be flying into an JFK, LAX, etc with heavy rain and the NGX. - My frames typically fell into the mid 20s and things got a bit choppy.

In desperation, I pushed my overclock a bit higher to 4.9GHZ. After some testing, I found a tangible increase in smoothness and performance. I found framerates were about 2-4FPS higher in very demanding scenarios and smoothness was at a much higher level.

 

I'd like to do some tests soon and further investigate the effects of clockspeed and FSX FPS.

 

 

Hi Arjen,

 

You seem to have confirmed exactly what occured in my scenario while overclocking.

 

For quite a while I ran my 2700K @ 4.8GHZ and I was mostly happy. The only time I got irratated was when I would be flying into an JFK, LAX, etc with heavy rain and the NGX. - My frames typically fell into the mid 20s and things got a bit choppy.

In desperation, I pushed my overclock a bit higher to 4.9GHZ. After some testing, I found a tangible increase in smoothness and performance. I found framerates were about 2-4FPS higher in very demanding scenarios and smoothness was at a much higher level.

 

I'd like to do some tests soon and further investigate the effects of clockspeed and FSX FPS.

 

Thanks for confirming Ben. You're lucky to get it to 4.9 GHz on the 1.4V boundary. I can only get it to 4.7 but 4.9 is going to be possible at 1.48V or so. But I don't want it to die in a matter of months.

Arjen,

 

Here is your report findings formatted to show change in performance relative to change in clock speed:

 

Clockspeed Change in Clockspeed Frame Rate Change in Performance Performance Change Relative to Clockspeed Change 2.8 15 3.3 17.9% 21 40.0% 124% 4 21.2% 26 23.8% 12% 4.1 2.5% 27 3.8% 54% 4.2 2.4% 29 7.4% 204% 4.3 2.4% 31 6.9% 190% 4.5 4.7% 33 6.5% 39% 4.7 4.4% 36 9.1% 105% 4.8 2.1% 38 5.6% 161%

 

 

What conclusions can we draw in general and with regard the highly non-linear, seemingly random distribution of these results, provided you did a maximally controlled test?

• General trend: frame rate always increases more than the change in clockspeed change. That's good to know and is not compatible w/ my prior experience doing controlled tests.
  
• There are 'unknown factors' that may be relative to individual systems, that cause the general trend to be highly variable & completely non-linear & random, which serves to instruct the would-be overclocker to be sure the added system stress is truly worth it.
  
  • Timing issues in the individual system. I think there is merit in this guess.
    
  • Failure to appreciate the test conditions were not truly comparable. I'm not accusing anyone of not trying their hardest to duplicate conditions, however there may be more in the background that is coming into play unbeknownst to the experimenter. I don't know about you all, but if there is anything that can be said about FSX it is that sometimes the thing flies fabulously in all respects, and other times, not so much, despite the seemingly same conditions, and it's hard to pinpoint exactly why. This is why I take all claims of 'smoothness', 'performance' etc w/ a large salt crystal.
     
     
    

Arjen,

 

Here is your report findings formatted to show change in performance relative to change in clock speed:

 

Clockspeed Change in Clockspeed Frame Rate Change in Performance Performance Change Relative to Clockspeed Change 2.8 15 3.3 17.9% 21 40.0% 124% 4 21.2% 26 23.8% 12% 4.1 2.5% 27 3.8% 54% 4.2 2.4% 29 7.4% 204% 4.3 2.4% 31 6.9% 190% 4.5 4.7% 33 6.5% 39% 4.7 4.4% 36 9.1% 105% 4.8 2.1% 38 5.6% 161%

 

 

What conclusions can we draw in general and with regard the highly non-linear, seemingly random distribution of these results, provided you did a maximally controlled test?

• General trend: frame rate always increases more than the change in clockspeed change. That's good to know and is not compatible w/ my prior experience doing controlled tests.
  
• There are 'unknown factors' that may be relative to individual systems, that cause the general trend to be highly variable & completely non-linear & random, which serves to instruct the would-be overclocker to be sure the added system stress is truly worth it.
  
  • Timing issues in the individual system. I think there is merit in this guess.
    
  • Failure to appreciate the test conditions were not truly comparable. I'm not accusing anyone of not trying their hardest to duplicate conditions, however there may be more in the background that is coming into play unbeknownst to the experimenter. I don't know about you all, but if there is anything that can be said about FSX it is that sometimes the thing flies fabulously in all respects, and other times, not so much, despite the seemingly same conditions, and it's hard to pinpoint exactly why. This is why I take all claims of 'smoothness', 'performance' etc w/ a large salt crystal.
     
     
    

 

Noel,I do understand what you're saying and I can imagine you thinking that my test results doesn't make sense. And normally, they wouldn't make sense if you run a stable application where performance would increase linearly. But FSX is different, it's a strange program. And there are multiple factors and parts in FSX that need different "minimum system requirements" to unleash performance. Just to give an example: in my test scenario the cockpit of the NGX stays white and refuses to load when running on 1.6 GHz. Slowly bumping up the frequency to 2.0 GHz doesn't change anything, not even the framerate which just stays at 4 FPS. But when I tried 2.1 GHz, the cockpit suddenly loaded up just fine and FPS got unleashed to 8 (doubled). This makes me come to the conclusion that certain parts of FSX need a minimum of hardware power to run properly. Perhaps, at this 4.5 GHz boundary, something similar happened to something in FSX. Perhaps it was a scenery or add-on, or something in the cockpit of the NGX that got held back by the 4.5 GHz and needed 4.7 GHz to run properly which caused the big FPS increase.

 

And I don't think a test scenario is THAT important. I use a saved flight for testing and don't simply touch my fsx.CFG in between. I know my PC really well and I know my performance has always been stable, I know what applications are automatically loaded, and you are not going to convince me that on the next boot of changing something, suddenly all kinds of random DLL's or applications got loaded which caused framerates to change...

...you are not going to convince me that on the next boot of changing something, suddenly all kinds of random DLL's or applications got loaded which caused framerates to change...

 

What you are calling 'strange' I am calling 'unknown factors' that make it difficult to draw solid conclusions, other than higher the clock the higher the overall performance--maybe. I say maybe because one can have higher frame rates (what your testing evaluated) w/o higher smoothness for example, especially if one is on the bleeding edge of instability. I think the biggest culprit is in the domain of timing issues: timing between all factors unique to each system involved, which I think is why you call FSX strange--I think we can agree on that. My initial reply had to do w/ your initial question, 'Is it safe...' Given the idiosyncratic nature of scaling, I'd say best stay on the safe side unless you want to increase your odds of buying a replacement as I had to do.

@Noel

 

Do you remember the test I did with my 3770k, in the delidding thread?

 

The overclock scaled exactly. For a given percentage overclock, I got exactly the same percentage increase in frame rate.

 

Personally I don't see why Sandybridge should be different to Ivy Bridge in this regard.

 

My advice to the OP is the same as before. If you have to ask, "is it safe" to overclock further, then surely it makes more sense to try a non risky alternative first... lower one of your sliders slightly.

 

I notice the OP, in his posts above, runs with max Autogen. Why? Knock it back a notch or two. It will still look great.

As for Arjen's original question... "Is it safe to increase the overclcok by 100-200 MHz?"

 

We can't tell you. Only you can answer that question, because only you can discover what the resulting temperature is when you try it.

The overclock scaled exactly. For a given percentage overclock, I got exactly the same percentage increase in frame rate.

 

Yes, I do remember. This discussion has prompted me to do exactly the same thing: I set up the sim exactly as Arjen had--same everything, except where I did the test since I don't have that 3rd party terminal. Instead, I went to KSEA 16R. Mind you, I'm on a Core 2 Quad. Used the NGX, however the UA livery.

 

I set up my machine at default 3.0Ghz, DRAM at 1333: 21.5 (21-22) fps. Not bad for the old box eh? As I say, all other sliders same same as w/ Arjen.

 

I then increased to my default o'c: 3.71Ghz, DRAM 1641 (I think that's where the FSB puts it): 26.5 (26-27) fps: exactly down to the nano decimal, 1:1 scaling. Apparently, C2Q and IB are predictable, but SB is wildly random in scaling? There is one conspicuous common denominator: the folks who describe these results happen to be the same folks who are very focused on overclocking as a means of achieving flight sim Nirvana. Yes, sure I'm being a bit sarcastic--but I hope you can see why w/ the IB and C2Q tests just described. I gave a nod to timing issues as a possible explanation, now I'm doubting this even. Arjen, Ben, Hasse: what do you make of your results, versus what you just read here? What explanations are there?

1.6 GHz: 4-6 FPS, cockpit stays white and refuses to load

2.8 GHz: 15 FPS

3.3 GHz: 21 FPS

4.0 GHz: 26 FPS

4.1 GHz: 27-28 FPS

4.2 GHz: 29 FPS

4.3 GHz: 31 FPS

4.5 GHz: 33-34 FPS

4.7 GHz: 36-37 FPS

4.8 GHz: 38-39 FPS

 

 

Arjen,

 

I created a chart based upon the values provided. I used the exponential average function to vizualize the curve of performance.... Very interesting results IMO

 

 

There is one conspicuous common denominator: the folks who describe these results happen to be the same folks who are very focused on overclocking as a means of achieving flight sim Nirvana.

 

Confirmation bias.

Confirmation bias.

 

How about your provide numbers and tests for your own 'beliefs'?