IBT, Real Temp and the AVSIM S/H Guide for FSX

[Ted Striker 338]

It's worth mentioning that both Asus, and Intel do "not" recommend IBT for stress testing Ivy Bridge. It doesn't test all the CPU's parameters and tends to overheat

 

Aida 64 is recommended. Prime seems to be okay now with the latest version.

 

Martin, do you have any links for these recommendations? I am not challenging you, I would just like to read some official recommendations on this and other topics. All I have been able to find is opinions from other users.

Much Appreciated,

Ted

 

 

Paul, You can use a free trial version of Aida 64 for 30 days. The trial version does a stability test. I'm considering springing for the full version.

[Paul J 349]

yeah - I'm not challenging you either, Martin! I have a "Guide" to write, and any info I can beg, borrow or steal I'll go for it!! I use LinX, OCCt and Prime occasionally, but IBT is my personal fave.. :biggrin:

 

From what I understand from Martin's good post, it only applies to the IB: as I have the SB, and haven't noticed any difference in IBT's operation between the C2D, the 950, the 2600K and the 2700K - I don't think it applies to those: I could be wrong, though, as I wasn't looking for any problems with the tool, nor had I heard of any. I've not touched the IB, and so have never dug deep into it.

[coa1117 53]

Martin, do you have any links for these recommendations? I am not challenging you, I would just like to read some official recommendations on this and other topics. All I have been able to find is opinions from other users.

<snip>

Paul, You can use a free trial version of Aida 64 for 30 days. The trial version does a stability test. I'm considering springing for the full version.

 

See

, especially at 03:55.

[martin-w 3,254]

Regarding the recommendation to use Aida64, Martin: this costs U$40... and unless one is a serious overclocker, or does this as a hobby or profession - not too many cost-conscious FSX flyers are going to buy that in addition to all the other costs of building a pc.

 

If ASUS and Intel both do not recommend IBT - which, I believe uses LinX - what do they recommend besides Aida?

 

Also - that Prime95 "seems" ok, doesn't IBT "seem" ok, too? What does it do that will show this?

 

Hi Paul. The trial version of Aida 64 is free to use. For one month. Personally, when I overclocked Ivy Bridge, I used a combination of Aida, and Prime. I would say prime is fine now, as long as you use the latest version. When a stress testing utility states, "unknown CPU" caution is advised.

 

In regard to Aida, temps will be lower than in Prime.

 

Also - that Prime95 "seems" ok, doesn't IBT "seem" ok, too? What does it do that will show this?

 

I'm assuming the latest version of Prime is okay.... as it's the latest version, and thus probably utilises more of the CPU's instructions/features... My Choice and conclusion, not definitive.

 

Aida utilises iinstructions such as ABX, ADS, SSC. Thus Accounting for all of the complex logic on the chip.

 

IBT/LinX doesn't... unless there's a new version out that does.

 

The other point to consider is that LinX is vicious, and has always been known to very heavily stress the CPU and generate a LOT of heat. This isn't something that I personally would recommend for Ivy Bridge, a CPU that is known to have thermal issues at higher voltages.

[martin-w 3,254]

See

, especially at 03:55.

 

Thanks for tracking the video down. :smile:

 

Seems to me, that as Intel created LinX in the first place, to test "older" CPU architectures, that we should consider their words, if they do indeed frown upon LinX/IBT, for Ivy.

 

Could well be of course, that there's some kind of "arrangement" between the two companies to promote Aida. Who knows, but the video is there for all to consider.

 

@Paul Perhaps you could contact Intel and Asus and query this, if you require clarification for your guide?

[martin-w 3,254]

Just to mention, never use two utilities in the same time to measure a temperatures, both could show the wrong results. Use Core Temp or Real Temp for that.

 

BTW - I heard the same like Martin W. about Ivy Bridge and stress testing.

 

In my opinion, de-lidding is the must for Ivy Bridge, you should do this and use some good termal paste. Your temps are too high for that overclock

 

Regardng de-lidding. I would have to disagree with you there. Personally, I don't regard de-lidding as a "must". It's true that there are a handful of Ivy CPU's out there with shocking TIM application under the IHS. For those "defective" CPU's then yes, de-lidd, or send the defective CPU back to the supplier.

 

However, the average Ivy bridge can overclock up to 4.5-4.6 without de-lidding. For me, in regrd to my 3770K, I can get to 4.6. as soon as I hit 4.7 the temps shoot up dramatically, to unacceptable levels.

 

Now the question to ask, is how much performance is it possible to gain by de-lidding. And the answer, is 10% at the most. So what does that 10% give you, in the scenarios where you actually need it? At 10 FPS you would gain a mere ONE frame per second. At 20 frames per second, you would gain an almighty TWO frames per second. Hardly a massive jump in frame rate is it? Hardly significant.

 

So, logically, for me, I couldn't justify de-lidding. Especialy considering that I get high frame rate anyway.

 

I have no problem with others deciding to de-lidd though, if they so desire. I have been tempted myself, for the hell of it, for fun, out on interest, to see what I could achieve. But if I do in the future... it won't be for the minimal performance increase.

[lodestar 186]

Regardng de-lidding. I would have to disagree with you there. Personally, I don't regard de-lidding as a "must". It's true that there are a handful of Ivy CPU's out there with shocking TIM application under the IHS. For those "defective" CPU's then yes, de-lidd, or send the defective CPU back to the supplier.

 

Now the question to ask, is how much performance is it possible to gain by de-lidding. And the answer, is 10% at the most. So what does that 10% give you, in the scenarios where you actually need it? At 10 FPS you would gain a mere ONE frame per second. At 20 frames per second, you would gain an almighty TWO frames per second. Hardly a massive jump in frame rate is it? Hardly significant.

 

So, logically, for me, I couldn't justify de-lidding. Especialy considering that I get high frame rate anyway.

 

I have no problem with others deciding to de-lidd though, if they so desire. I have been tempted myself, for the hell of it, for fun, out on interest, to see what I could achieve. But if I do in the future... it won't be for the minimal performance increase.

 

Ok, i said that for ivy bridge de-lidding is the must in my opinion. That's becouse i always try to get maximum of my PC. I think that is very important attitude if you want max from fsx. For me 2fps in fsx are very important. I have high fps, but for example - taking off from Aerosoft EDDM in bad weather with NGX, your fps will go down for sure, and in this situation 25-26fps feels very different(smoothnest and stutters) then 23-24. Or 23-24 vs 21-22.

 

It's simple: de-lidding = higher overclock = beter fps(and overall performance)

 

I had my PC at 5GHz and i can tell you that you can see the difference for sure vs 4.8GHz(my current OC), and it's just a couple of fps difference, but fsx is much, much smoother. I will go for 5Ghz again after i change my cooler.

 

However, the average Ivy bridge can overclock up to 4.5-4.6 without de-lidding. For me, in regrd to my 3770K, I can get to 4.6. as soon as I hit 4.7 the temps shoot up dramatically, to unacceptable levels.

 

I would de-lidd anyway, it's better to be at 4.6GHz with temps at 50c then 80-90c

 

I think it's about what you want from fsx, you don't need to overclock if you don't use complex airliners for example. In fsx, you want get 200+fps with NGX or MAddog, ORBX, Aerosoft etc... You need 20 to 30fps - so 2fps means a lot.

 

So Martin, i'm not saying that you are wrong, i just tried to explain my point of view

[TechguyMaxC 516]

Martin: don't forget about performance gained from running high speed RAM which requires additional voltage for the CPU/IMC, headroom that lidded CPUs don't have. According to results in the FSXMark 11 thread from myself and others, an additional 10-12% gain in FPS has been observed based on increasing RAM speed only.

[Dirk98 354]

Guys, from my humble experience CPU de-lidding is a much much more pleasant way of tinkering than, say, lapping of the block and ihs. Does not take long, improves CPU temps dramatically, just be reasonably careful with the razor and your fingers, make sure to apply a good TIM (like Coollaboratory's Liquid Pro) between the chip and IHS and seat the lid properly on the chip before fastening it to the CPU socket. Not a big deal really (thanks again to TechguyMax and BimmerCop!)

 

Cheers,

Dirk.

[martin-w 3,254]

So Martin, i'm not saying that you are wrong, i just tried to explain my point of view

 

Yep, absolutely. This isn't about right or wrong, we are just sharing opinions and preferences. :smile:

 

 

I would de-lidd anyway, it's better to be at 4.6GHz with temps at 50c then 80-90c

 

I am at 50c... without delidding. About 55 actually. That's at 4.5. At 4.6 Ghz, I can't remember what the temp was, but it still wasn't high. The point about Ivy Bridge, is that you hit a thermal brick wall, where temps suddenly shoot up. Below that threshold temps are fine.

 

Martin: don't forget about performance gained from running high speed RAM which requires additional voltage for the CPU/IMC, headroom that lidded CPUs don't have. According to results in the FSXMark 11 thread from myself and others, an additional 10-12% gain in FPS has been observed based on increasing RAM speed only.

 

I do run reasonably high speed RAM, and have no temperature issues, at 4.6Ghz, without delidding.

 

I think it's about what you want from fsx, you don't need to overclock if you don't use complex airliners for example. In fsx, you want get 200+fps with NGX or MAddog, ORBX, Aerosoft etc... You need 20 to 30fps - so 2fps means a lot.

 

So Martin, I'mm not saying that you are wrong, i just tried to explain my point of view

 

Fair enough, each to his own as they say. But personally I couldn't get excited at a mere two frames per second, whether I'm at sub 30 FPS or not. But then I hate frame rate in the 20's anyway. I'm usually in the 40's in bad REX weather, parked at an FSX detailed airport, in the NGX. Simply because I don't run with hyper complex airport add-ons, or high sliders. Frame rate wins over eye candy everyday for me, not that my sim lacks eye candy, it looks great to me.

 

 

 

I run with GEX, REX, UTX.

[Ted Striker 338]

Thanks for posting the video link Martin. Sounds like a good reason to use Aida64 to me. It suprised me that he never stated what a safe CPU temperature was. Also he said not to go over 1.325 vcore and then they proceeded to go right to 1.4 v in their demonstration.

[MarioB 4]

OC is new to me but seems necessary with the fancier glass cockpits and add-ons. So I decided to take the plunge and learn... thanks to all for your input

 

Friday, I got a Corsair H80i, installed it properly and got 80-90 C @4.3 on the standard level of stress in IBT (around 1.2v core). Couldn't push it to High/Maximum stress at all. These are worse results than with the Coolmaster EVO 212 previously installed. I do not have the exact figures to share but Real Temp is s good 20-25C over the Corsair Link2 reading. These figures may result from using two monitoring systems at once as mentioned above.

 

Others on the web show amazing results of 4.5 GHz @1.25v with with temps in the 60C with a high air/water cooling system. I ran a quick test flight from CYBW (Calgary Springbank - where I work) to CYYC Calgary loading the F1 T182T, REX, UTX. The Corsair Link 2 temps indicated around 60-65 max temp.

 

Separately, I ran a quick Prime95 and observed lower temperatures - 75-80C. With all the data gathered and the high temps I have, something must be wrong either with my cooling installation or the CPU itself. I will leave this to my buddies at my local computer store. I will keep you posted.

 

I am itching to resume my IFR training...

[martin-w 3,254]

Thanks for posting the video link Martin. Sounds like a good reason to use Aida64 to me. It suprised me that he never stated what a safe CPU temperature was. Also he said not to go over 1.325 vcore and then they proceeded to go right to 1.4 v in their demonstration.

 

Actually it wasn't me that posted it, but it was the video i was talking about.

 

I wouldn't shy away from the latest version of Prime to be honest I used both Prime and Aida when i overclocked.

 

I'm not surprised he didn't say what a safe temp was to be honest. Even Intel don't specify, Intel just tell you what TJMax is and say anything below is within the CPU's operating range.You and I know of course that the hotter you run, the more degradation. Acceptable temp when overclocking is something we have to decide ourselves. For me when stress testing, I don't go higher than the high 70's. Maybe I would go to 80. But then it's a stress test, and you won't see anything like those temps in everyday use.

 

After running Prime/Aida etc, the best test of an overclock in my view is to run the software you usually run on your system, if all is fine, totally stable, no high temps after extended use, you're onto a winner.

 

Regrding voltage, he did exceed his own recomended max, buit that was purely for demonstration purposes. I'd agree with his voltage recomendation.

[martin-w 3,254]

OC is new to me but seems necessary with the fancier glass cockpits and add-ons. So I decided to take the plunge and learn... thanks to all for your input

 

Friday, I got a Corsair H80i, installed it properly and got 80-90 C @4.3 on the standard level of stress in IBT (around 1.2v core). Couldn't push it to High/Maximum stress at all. These are worse results than with the Coolmaster EVO 212 previously installed. I do not have the exact figures to share but Real Temp is s good 20-25C over the Corsair Link2 reading. These figures may result from using two monitoring systems at once as mentioned above.

 

Temp varies dependant on where the temperature is taken from. TCase, TJunction,or a probe on the MB in the region of the CPU socket. And remember all CPU's are different. Mine for example seems to require higher volts than average.

 

 

H80i isn't the best cooler around, the H100i beats it. For air cooling, I'd recommend the Noctua NH-D14, I can't praise this almighty cooler enough.

 

The high temps you experienced with IBT are no surprise. IBT is vicious, especially for Ivy Bridge. I would avoid IBT for Ivy Bridge, given the advice of Intel and Asus. Why subject a CPU that has known thermal issues at high overclocks to a stress test that's known to heat up a CPU dramatically. And if Asus are correct, and IBT doesn't test all of the CPU's parameters, then it's not an ideal test to use for Ivy Bridge.

 

I ran a quick test flight from CYBW (Calgary Springbank - where I work) to CYYC Calgary loading the F1 T182T, REX, UTX. The Corsair Link 2 temps indicated around 60-65 max temp.

 

Sounds about right for the H80i. At 4.3GHz. Don't quote me though, I don't have one. That depends where Corsair link is measuring the temp though. If it's not core temp, it's high.

 

Separately, I ran a quick Prime95 and observed lower temperatures - 75-80C. With all the data gathered and the high temps I have, something must be wrong either with my cooling installation or the CPU itself. I will leave this to my buddies at my local computer store. I will keep you posted.

 

 

Why do you regard that as wrong? Your at 4.3GHz, right? 75-80 is what expect for a H80i, under Prime95. That temperature would be lower with a better cooler.

 

I get temps about 75 on one of my cores, with my 3770k, at 4.5GHz. With my NH-D14.

 

P.S. Don't forget, if using RealTemp, check to make sure your CPU's TJMax is set correctly in RealTemp settings. Otherwise RealTemp may be inaccurate. Make sure it's the latest version of RealTemp as well, and that it correctly identifies your CPU.

[MarioB 4]

The high temps you experienced with IBT are no surprise. IBT is vicious, especially for Ivy Bridge. I would avoid IBT for Ivy Bridge

 

Thanks, just learned that today from this post. My knowledge of OC is newbie level, I just follow the suggestions in the AVSIM guide for FSX.

 

Why do you regard that as wrong? Your at 4.3GHz, right? 75-80 is what expect for a H80i, under Prime95. That temperature would be lower with a better cooler.

I get temps about 75 on one of my cores, with my 3770k, at 4.5GHz. With my NH-D14.

 

I considered my values wrong because:

1. the guide mentions an OC of 4.5 GHz with the stock cooler
   
2. I compared my values with others on the web who achieved much better results with regular air coolers like the CM EVO 212
   
3. It was my understanding that liquid cooling was better than air cooling
   

The values may not be wrong, but I expected better results from this H80i, that's all.

 

This is all new to me, so I seek other people's opinions, results and recommendations, hence this post.