Overclocking On Air: 10 LGA 1156-Compatible Performance Coolers

Page 4 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.

rambo117

Distinguished
Jun 25, 2008
1,157
0
19,290
4
[citation][nom]Crashman[/nom]Normally if you're going thermoelectric you need water cooling to keep the "hot side" of the TEC from overheating.[/citation]
Well, wouldn't that eliminate the purpose of going thermoelectric?
 

Crashman

Polypheme
Former Staff
[citation][nom]rambo117[/nom]Well, wouldn't that eliminate the purpose of going thermoelectric?[/citation]
A TEC element moves heat from one side to the other. Therefore it has a cold side and a hot side. If you want the cold side to remove a bunch of heat, you need to remove a bunch of heat from the hot side. To do that you either need a much larger than normal sink, or a water cooler.

What, you thought the heat just...vanished? TEC's aren't 100% efficient, they actually make more heat as they're moving heat from one side to the other. There are TEC-equipped air coolers, but they can't remove much heat because the sink isn't big enough for that purpose.

It's explained in more detail here:

http://www.tomshardware.com/reviews/vigors-monsoon-ii-tec-cpu-cooler,1565.html
 

rambo117

Distinguished
Jun 25, 2008
1,157
0
19,290
4
[citation][nom]Crashman[/nom]A TEC element moves heat from one side to the other. Therefore it has a cold side and a hot side. If you want the cold side to remove a bunch of heat, you need to remove a bunch of heat from the hot side. To do that you either need a much larger than normal sink, or a water cooler.What, you thought the heat just...vanished? TEC's aren't 100% efficient, they actually make more heat as they're moving heat from one side to the other. There are TEC-equipped air coolers, but they can't remove much heat because the sink isn't big enough for that purpose.It's explained in more detail here:http://www.tomshardware.com/review [...] ,1565.html[/citation]
Thank you very much for the info, kinda changed my mind about going thermoelectric; used to think it was the best of both worlds. I'll give that review a read tonight, how would a TEC stand against one of these more modern HSF solutions? Or even an H50? had my eye on that one for awhile until I saw the Sunbeam core contact freezer pretty much destroy it..
 

BigStack

Distinguished
Jan 25, 2010
32
0
18,530
0
The one thing I didn't notice is if they said whether the system was stabile at the overclocked speed with all the coolers. A subsidiary question is how cool does the CPU need to be in order to be stabile. I'd like to be able to overclock an i7 860 to 4GHz with a CM Hyper 212. I think it would do the job, but from this article, I'm not sure if they're saying it would or not.
 

Crashman

Polypheme
Former Staff
[citation][nom]BigStack[/nom]The one thing I didn't notice is if they said whether the system was stabile at the overclocked speed with all the coolers. A subsidiary question is how cool does the CPU need to be in order to be stabile. I'd like to be able to overclock an i7 860 to 4GHz with a CM Hyper 212. I think it would do the job, but from this article, I'm not sure if they're saying it would or not.[/citation]

Eight threads of 64-bit Prime 95 were used to reach peak temperature, so stability wasn't a problem. The only problem was keeping the smaller Artic Cooling unit cool, the CPU is stable up to its 100C limit.
 

BigStack

Distinguished
Jan 25, 2010
32
0
18,530
0
So really, unless yo want to push an overclock to the highest possible limit, any of the units, with the exception of the Artic Cooling cooler would do.
 

rambo117

Distinguished
Jun 25, 2008
1,157
0
19,290
4
[citation][nom]BigStack[/nom]So really, unless yo want to push an overclock to the highest possible limit, any of the units, with the exception of the Artic Cooling cooler would do.[/citation]
Pretty much. It was neat seeing that little Thermaltake Frio keep up with the more expensive solutions.
 
G

Guest

Guest
TacomaSailor
I am still confused by the ambient/above ambient measures. My i5 is running at 3.4GHz - ambient is 24C and CoreTemp shows 82C with HyperPI 32M. Is my number that is comparable to your chart 58C above ambient?

I am running a stock Intel fan @100% in an open case - Seems odd that my stock cooler performs about the same as many of the coolers you tested. What am I doing wrong?
 

rambo117

Distinguished
Jun 25, 2008
1,157
0
19,290
4
[citation][nom]TacomaSailor[/nom]TacomaSailorI am still confused by the ambient/above ambient measures. My i5 is running at 3.4GHz - ambient is 24C and CoreTemp shows 82C with HyperPI 32M. Is my number that is comparable to your chart 58C above ambient?I am running a stock Intel fan @100% in an open case - Seems odd that my stock cooler performs about the same as many of the coolers you tested. What am I doing wrong?[/citation]
Well, it could be the fact that your CPU doesnt have 8 threads being stressed and not nearly as OCed as the 920 theyre using.
 

Crashman

Polypheme
Former Staff
[citation][nom]TacomaSailor[/nom]TacomaSailorI am still confused by the ambient/above ambient measures. My i5 is running at 3.4GHz - ambient is 24C and CoreTemp shows 82C with HyperPI 32M. Is my number that is comparable to your chart 58C above ambient?I am running a stock Intel fan @100% in an open case - Seems odd that my stock cooler performs about the same as many of the coolers you tested. What am I doing wrong?[/citation]

First of all, you're running 3.4 GHz. Heat increases by a huge amount going from there to 3.8 GHz, and then by another huge amount going to 4.0 GHz at 1.388V. Second, your processor doesn't do hyperthreading and, as rambo117 said, eight threads have a higher CPU load than four. Your system is in no way comparable to the one that was tested.
 

tacomasailor

Distinguished
Jan 30, 2010
3
0
18,510
0
I guess I am very confused about the purpose of HyperThreads/virtual processors.

My understanding of i7/HT is that it minimizes the TIME (and therefore possibly make more efficient use of previously unused clock cycles) required to switch threads but it cannot magically generate more clock cycles than are available in the real core. If every one of those REAL clock cycles is already being used (isn't that the point of a CPU computationally bound program - it NEVER enters a wait state and therefore uses every available clock cycle – it loops within tight bounds and never needs to fetch instructions past the closest cache) - then another virtual processor cannot use or generate more cycles and therefore more heat.

Clock Cycles/Instruction Cycles per unit of time are what generate heat (given a constant V) and once all internal processors are 100% busy they cannot get any hotter. How can Switching thru 8 virtual threads (implemented on 4 real cores) use more cycles than are available in hardware? Once all four real cores are 100% busy; how can more VIRTUAL threads make them more busy (use more cycles/unit time) and generate more heat?

An HT system with 8 threads could make a four real core processor run hotter IF the programs running on the 4 real cores were not properly written and did enter a state where an HT switch could use clock cycles that would be otherwise unused. But, my impression is that most stress programs are now very tightly coded and fully exploit all the clock cycles in an i7 real core.

Please explain what I do not understand.
 

tacomasailor

Distinguished
Jan 30, 2010
3
0
18,510
0
My i5 is OC'd 29% above stock (171x20) and the test i7 CPU was OC'd 36% above stock so my OC is within 14% OC of the test machine. Are you saying an i5 does not need aftermarket cooling when OC’d because it is slower than an i7? If that is the case there has never been a need for aftermarket cooling in the past because everything was slower than an i7.

Or – are you saying the stock i5 cooler is the same as the stock i7 cooler and therefore the i5 can OC more without additional cooling?

Let me ask the initial question another way.

If my OC'd i5 never gets above 82C on a stock cooler - is that too hot? I thought it was but it appears none of the tested coolers makes much of a difference compared to my 56C above ambient at my 28% OC.

I thought it was important for processor longevity to keep the max temp at TjMax (99 for an i5/i7 ??) minus 20C. So if ambient is 20C the desirable upper limit for temp delta/ambient would be 59C (99-20-20)and my stock cooler keeps the delta over ambient at 56.

Only three of the tested coolers achieved that value (20C below TjMax) yet no one seems concerned or even noted the fact. Is it now acceptable to run an i5/i7 at 85C (absolute) for long periods?

 

Crashman

Polypheme
Former Staff
[citation][nom]tacomasailor[/nom]If my OC'd i5 never gets above 82C on a stock cooler - is that too hot? I thought it was but it appears none of the tested coolers makes much of a difference compared to my 56C above ambient at my 28% OC.I thought it was important for processor longevity to keep the max temp at TjMax (99 for an i5/i7 ??) minus 20C. So if ambient is 20C the desirable upper limit for temp delta/ambient would be 59C (99-20-20)and my stock cooler keeps the delta over ambient at 56. Only three of the tested coolers achieved that value (20C below TjMax) yet no one seems concerned or even noted the fact. Is it now acceptable to run an i5/i7 at 85C (absolute) for long periods?[/citation]

I'm saying that you're still making a bunch of assertions that, because you read my previous response, you already know are untrue. Like I said, you can't compare any 3.4 GHz overclock to any 4.0 GHz overclock. I'll further quantify that it takes around 3x the cooling to get similar temperatures at the higher frequency.

Furthermore, I don't know where you get the TJMax-20C rule from, but generally speaking the lower the temparature the longer your CPU will last. You could probably run your particular settings at less than 60C on most of the coolers in this article, but I know that you're still going to say something about these coolers not being better because your lower overclock runs cooler.

I wouldn't want to run 85C for super-long periods, but most poeple don't run 100% CPU load for long periods.
 
G

Guest

Guest
What type of fan did you use in VRM temperature test? The temperatures are pretty low. :) I'd like to replace my one in Arctic Cooling freezing PRo 7 rev2 with that one you were using. OR this is due to 11 temperature in the lab?
 

Crashman

Polypheme
Former Staff
[citation][nom]WT[/nom]What type of fan did you use in VRM temperature test? The temperatures are pretty low. I'd like to replace my one in Arctic Cooling freezing PRo 7 rev2 with that one you were using. OR this is due to 11 temperature in the lab?[/citation]

Haha, you can't, it doesn't fit and was simply set on the board in front of the cooler. It's the only cooler that didn't allow the fan to be attached!
 

tacomasailor

Distinguished
Jan 30, 2010
3
0
18,510
0
Crashman –I am not sure where your 3X number for heat dissipation (going from 3.4 to 4.0 GHz) came from but it is impossible. At maximum power consumption (stock V) the i7 temperature is only 56% above it’s idle temperature.

Looking at Intel Thermal Specifications for the i7 (http://download.intel.com/design/processor/datashts/320834.pdf) - table 6.1. You can see the increase in temperature is LINEAR with increase in power consumption (clock cycles). The formula is TEMP = 43.2 + 0.19 * P(WATTS). The temperature increases 1.9C for each 10-watt increase in power used.
In the extreme case - going from 10 watts to 100 watts (a 10 fold increase) only increases the heat by a factor of 38% (45C to 62C). To reach your 3X increase in heat would require 458 watts used by the i7. I don’t think that is possible.

You will also find some very interesting data at the following site (http://www.guru3d.com/article/core-i5-750-core-i7-860-870-processor-review-test/10) where they OC’d an i7-870 from the stock 2.93 GHz to 4.130 GHz. The total increase in power at 100% load went from 162 watts to 295 watts. That means the OC accounted for a 133 watt increase in power consumption at 100% load. Or a 82% increase (133/162) in power/heat to dissipate. The Intel formula then predicts about a 26 degree rise in processor temperature above the 45C idle temp. Guru3D measured 71C at max OC load- exactly what the formula predicts. Seems pretty much in line with experience!

This is science/math not repeated rumors.

 

Crashman

Polypheme
Former Staff
[citation][nom]tacomasailor[/nom]Crashman –I am not sure where your 3X number for heat dissipation (going from 3.4 to 4.0 GHz) came from but it is impossible.[/citation]

Impossible? I'm only telling you what I see, it's up to you to figure out what you want to believe. System draws 160W at full CPU load at YOUR settings and 300W at 4.00 GHz. Take away what the other parts are consuming (around 50W +10% for power supply losses) and you're left with a CPU power increase from 100W at 3.4 GHz to 220W at 4.00 GHz. 2.2x might not be 3x, but that observation-based estimate is looking far more realistic than your assumptive math.

But you earlier mentioned stuff about "my overclock is 29%, that one is 36%" which means absolutely zero. You have the same core, 4.00 GHz is 4.00 GHz. Regardless of the speed Intel stamped on the box, 4-threads at 4.00 GHz is going to produce around the same amount of heat when the same core voltage is used. Eight threads will always consume a little more power and produce more heat, though your processor doesn't support eight threads.

So, no matter how you try to justify your comparison, it doesn't work.
 

GrandePrairie

Distinguished
Jan 14, 2010
7
0
18,510
0
I have an Arctic Cooling Freezer 7 Pro Rev 2, here are my results:

Ambient Temp: 16-17C

Idle @ 3.4GHZ: 22-26C
Prime95 8 hours @ 3.4GHZ: 52-55C

Specs
Case:CM HAF932
CPU: i5 750
GPU: EVGA 7950 GT
RAM: G.Skill Rip Jaws 4GBRL
Mobo: Gigabyte UD4P
 

Crashman

Polypheme
Former Staff
[citation][nom]GrandePrairie[/nom]I have an Arctic Cooling Freezer 7 Pro Rev 2, here are my results:Ambient Temp: 16-17CIdle @ 3.4GHZ: 22-26CPrime95 8 hours @ 3.4GHZ: 52-55CSpecsCase:CM HAF932CPU: i5 750GPU: EVGA 7950 GTRAM: G.Skill Rip Jaws 4GBRL Mobo: Gigabyte UD4P[/citation]

Yes, it's an awesome cooling value for running mild overclocks.
 

elasticman

Distinguished
Oct 1, 2009
46
0
18,530
0
i got a question

CPU: C2D E6300 1.86ghz.
Mobo: Gigabyte GA-965P-DS3 Rev2
RAM: TRANSCEND 8 gigz of DDR2 - 800 RAM
GPU: nVIDIA GeForce 8800 GTS
case: COMPUCASE CI-6a19

have been using the box cooler from intel which was giving me VERY high temps at idle - in the range of 50-52 with ambient temp as low as 17 degrees.

have bought the MUGEN 2 Rev B + the arctic cooling MX-3 thermal compound.

im using 2 types of software to monitor temps and this is what i get with them both :

Real Temp 3.4 - @idle 37/38 (each core)

Core temp 0.99.5 - @idle 32/33

i dont understand how come people here with an i5-750 -3.4ghz @idle get temps as low as 22-26 with an Arctic Cooling Freezer 7 Pro Rev 2 (which from my understanding is not as good as the MUGEN 2,correct me if im wrong) and my temps are still not as low...

can anybody help me out understanding this?

thanx
 
Status
Not open for further replies.

ASK THE COMMUNITY

TRENDING THREADS