OC and CPU cooling

[westman 257]

Hi

 

I want to clarify some myths about CPU cooling.

I have OC for more than 10 years on high level, evrything from Air to LN2.

Even feed the water block from the garden house tube 4bar 5-8 degrees C.

 

1. Stage1 Stock cooler , no comments.

 

2. Stage2 AfterMarket air cooling, depending on Qualty the gain is good.

 

3. Stage3 AfterMarket "Fake water cooling" gain 0-200mhz over Air cooler depending on Quality.

But in my world the "best bung for bucks"

.

4. Stage4 ProWather cooling, You gain 100mhz if your lucky,limiting factor the ambient temp or

room temp normal 20-25C you get no headroom for much more OC almost the same temp on

the IHS as Stage2 and 3.

As a see it the price of a good Pro system is high to what you get "waste of money".

The only thing you can do for more mhz is to add a chiller and tune it to 10C wather

temp if you not want to have condensation, with good insulation you tune for subzero

-10 to -20c or add a TEC -15C.

 

5. Stage5 SubZero cooling now we talking, with a SS ( Single stage Vapor phase change cooling)

we gain at -40C load + 300 to 500mhz than wather cooling.

Drawbacks Insulation, it works for 24/7 if you do it right

 

In 2004 to 2006 i run Fs9 on Prometia -30C 24/7 on a AMD64 3200+ @3.1 thats was a huge diff

from my watercooling system.

 

I building a SS for 24/7 usage to the 3770K , the suction line and evap is orderd.

The rest Compressor is ready only fill with gas and tune to -40C load.

 

6. Stage6 Xtreeme nothing for normal OC, 2 or 3 StageCascade and LN2 cooling -90 to -120C load.

 

Here is a Pic when I and my great friend Kinc have a 3DM session on a big lanparty,

it was 2005 on the golden AMD days we have mine FX57 @3.9 -95C and 2xGTX7800 Sli -80C

we brake 3 or 4 3DM WR that day.

 

http://dl.dropbox.com/u/54475361/casx3.gif

[HLJames 2,568]

Very good!

and I remember the golden AMD days.

 

HLJAMES

[Dirk98 354]

Hi

 

I want to clarify some myths about CPU cooling.

I have OC for more than 10 years on high level, evrything from Air to LN2.

Even feed the water block from the garden house tube 4bar 5-8 degrees C.

 

1. Stage1 Stock cooler , no comments.

 

http://dl.dropbox.co...75361/casx3.gif

 

Very informative post, westman. I opt for Stage 4 method.

 

Thanks,

Dirk.

[Noel 4,531]

After this detailed overview my conclusion is that even at Stage 5, the gains are in the domain of irrelevant when it comes to FSX users. I have never seen 100% linear scaling w/ CPU over clocks personally. If you just look at raw frame rate, which I will agree is only one facet of 'performance', my experience has shown for every 10% increase in raw clock speed you might see 50-70% increase in frame rate. Given this, when you go thru all of this hoopla to get 500Mhz increase using Stage 5, which no one I hear of is using other than the truly dedicated, you are looking at let's say 4.5Ghz on excellent air or Stage 3 'fake water cooling', +500Mhz= 5.0Ghz, or a 9% increase in clock speed. 9% x 70% = 6%. Now ride that NGX into JFK w/ everything hard right, and you go from a frame rate of 24 to 25.44. All of that, for a whopping 1.56 fps 'improvement'. Ok, it's a hobby and a fun one for many no doubt, but that's really about all you get. I can see why westman calls Stage 3 the best bang for buck. Plus, you hopefully get something a little quieter and quite easy to install. One possible easy method to improve some on Stage 3 might be to move it into the domain of an evaporative cooler since subjecting the radiator on an H100 to a water source to improve heat transfer and take advantage of the huge change in kCals it takes to move water from liquid to vapor phase--a mini swamp cooler--very inexpensive to operate.

[ZachLW 37]

After this detailed overview my conclusion is that even at Stage 5, the gains are in the domain of irrelevant when it comes to FSX users. I have never seen 100% linear scaling w/ CPU over clocks personally. If you just look at raw frame rate, which I will agree is only one facet of 'performance', my experience has shown for every 10% increase in raw clock speed you might see 50-70% increase in frame rate. Given this, when you go thru all of this hoopla to get 500Mhz increase using Stage 5, which no one I hear of is using other than the truly dedicated, you are looking at let's say 4.5Ghz on excellent air or Stage 3 'fake water cooling', +500Mhz= 5.0Ghz, or a 9% increase in clock speed. 9% x 70% = 6%. Now ride that NGX into JFK w/ everything hard right, and you go from a frame rate of 24 to 25.44. All of that, for a whopping 1.56 fps 'improvement'. Ok, it's a hobby and a fun one for many no doubt, but that's really about all you get. I can see why westman calls Stage 3 the best bang for buck. Plus, you hopefully get something a little quieter and quite easy to install. One possible easy method to improve some on Stage 3 might be to move it into the domain of an evaporative cooler since subjecting the radiator on an H100 to a water source to improve heat transfer and take advantage of the huge change in kCals it takes to move water from liquid to vapor phase--a mini swamp cooler--very inexpensive to operate.

 

Nail on the hear, IMO. When this rig and earlier my 2600k were new, I was aiming for 5.0GHz no questions asked. After playing around for a while, the above was my experience. More volts, more heat, negligible performance gain.

 

I only overclock for the tangible results. ie I'm no benchmarker!

[Noel 4,531]

If you just look at raw frame rate, which I will agree is only one facet of 'performance', my experience has shown for every 10% increase in raw clock speed you might see 50-70% increase in frame rate.

 

Oops, meant to say 5-7% increase in frame rate for every 10% bump up in clock speed.

[westman 257]

After this detailed overview my conclusion is that even at Stage 5, the gains are in the domain of irrelevant when it comes to FSX users. I have never seen 100% linear scaling w/ CPU over clocks personally

 

Do you ref to your PC specs Q9650?

I dont know but could it be 35fps avg in FSMark11 ?

 

3770K scale about 3% 0.1ghz with same hardware mems at same speed.

 

I have done tests with both 2700k and 3770k with dif cooling systems from ,

Air , H100 , Pro Watercooling , Pro with chiller and Phase Change.

 

Today a run Phase Change 24/7 the only downside is the time i take to start the PC aprox 3min.

 

It runs @5.2ghz 1.34v , to run that with Air or a unlapped H100 is @4.6ghz 1.34v some needs less or more voltage.

 

FSMark11 @ 4.6ghz is around 48fps avg and @5.2ghz 56 fps avg. its 8fps more

 

If we look at FSMark11 it have a span from 25 - 55 fps thats a huge dif , and if you dont

see a diff when flying FSX wy uppgrade.

 

Hasse

[Noel 4,531]

FSMark11 @ 4.6ghz is around 48fps avg and @5.2ghz 56 fps avg. its 8fps more

Hasse

 

If these results are correct, you are saying at 4.6Ghz you are seeing 10.4 fps/Ghz, whereas at 5.2Ghz you are seeing 10.7 fps/Ghz. So, the frame rate per Ghz goes up. What explanation is there for this, which is suggesting greater than proportional scaling? This seems very counterintuitive, but perhaps could be explained by better GPU performance.

[westman 257]

This seems very counterintuitive, but perhaps could be explained by better GPU performance.

 

If you look at my runs at 5040x1050 @4.6ghz a had the same 39fps botth at base clock 1110 and 660 on

GPU but @5.14 a have 45fps at 1110 and 43fps at 660.

The CPU scale the same 8.4fps ghz with GPU @660 and 8.75fps ghz @1110 gpu clock.

 

Have not tested to downclock the GPU to see the diff at 1680x1050 a not normally run the CPU @4.6 and GPU @660.

 

My 24/7 base is CPU @5.2 GPU @1110 then a marginal for CPU @5.4 with raised voltage and @1250 for GPU

 

Cheers Hasse

[CharliesP51 0]

Thank you very much as i had been considering some OC on my i5 but see now it might be fun but way past the point of diminishing returns. An informative post and I am very inclined to heed your advice and appreciate your post highlighting your OC experiences.

Edited August 21, 2012 by CharliesP51

[ZachLW 37]

Thank you very much as i had been considering some OC on my i5 but see now it might be fun but way past the point of diminishing returns. An informative post and I am very inclined to heed your advice and appreciate your post highlighting your OC experiences.

 

I think you missed the point. Overclocking from stock can give you quite the performance boost in FSX. In my case, it's somewhere around 20-25%, with lowest FPS being considerably higher. There is, however, a barrier at a certain point. Around 4.5GHz for most Sandy Bridge and later processors. At this point, frames per 100MHz gained are negligible, where voltage and therefore heat per the same 100MHz is considerable.

 

At any rate, I'm pretty sure overclocking your Gateway isn't an option.

 

PS. New(ish) forum rule for signatures with PC specs. Suggest you take a look.

[Ben Cap 89]

At this point, frames per 100MHz gained are negligible

 

I can't remember where the post is, but the performance increases are linear with clockspeed increases.

[SAAB340 28]

I can't remember where the post is, but the performance increases are linear with clockspeed increases.

For the FPS to increase linearly with CPU speed the workload needs to be fully CPU limited.

That means that at that moment we're not limited by the GPU or PCIe bus.

We also need to have the "other treads" (often incorrectly called fibers) offloaded from the main thread.

[ZachLW 37]

I can't remember where the post is, but the performance increases are linear with clockspeed increases.

 

I doubt the gain is linear. Highly.

 

*edited for inappropriate reference

[Noel 4,531]

What is the difference in decibel level w/ an H100 over a comparable air cooling solution? 25% lower w/ the H100? 10%? I would consider lower decibels.