Cooler recommendation

[wilhelmsson]

Hi,

I plan to oc my e8400 to about 3.3ghz and want it to be able to run at full load at both cores for a couple of days (doing statistical simulations that take awhile)

With the stockfan I get temps measured by PC probe at about 62C while running only one core at max @3.3ghz so I realize that I must change the cooler. The three most important things are in descending order

1) Keeping reasonable temps during full load @3.3ghz for both cores for several days on continuous full load

2) Not too loud

3) Nice if it can be installed without having to remove the MB from the case

Relevant specs: CPU Intel e8400, Case Antec 900, MB Asus P5N-T Deluxe

 

[iluvgillgill]

if you really not aiming for more speed then 62C is TOTALLY SAFE.

but if you do have to get a new cooler then get the OCZ V2 thats very good for the requirement of yours.

 

[wilhelmsson]

if you really not aiming for more speed then 62C is TOTALLY SAFE.

but if you do have to get a new cooler then get the OCZ V2 thats very good for the requirement of yours.

62C was with one core at full load and the other idle, I want to be able to run both cores at full so I'll check out your recommendation, thanks!

 

[iluvgillgill]

use prime 95 25.6 to do the stress test.remember small ffts though.

 

[bobbknight]

Look here you are after an after market cooler for a modest OC.
What you want is to select from the range of coolers with the best cooling specs and cooling ability. Thermal Resistance measured in ℃/W is the cooling spec you want to look for.

1) For your modest OC you really don't need more than the Arctic Cooling Freezer 7 Pro. If you catch Newegg on the right day and time it can be had for $25 shipped, the price always fluctuates, seemingly with out reason. At the time I type this Newegg is $40 for this item.
Thermal Resistance 0.17°C/Watt

2) For more ambitious OC attempts and to get even more cooling. The Xigmatek HDT-S1284EE offers even better cooling ability. With this cooler you do want to use the extra mounting bracket. Right now this option is $60.
Thermal Resistance ‧ 0.14 ℃/W

The thing is you can spend a lot of money and get poorer cooling.
Thermal Resistance is the number you need to look for, if you can find it they are hiding something from you, (poor performance).

 

[wilhelmsson]

Ok thanks to both of you, of course I would not mind oc'ing more than to 3.3 I just thought that was resonable since the system needs to be very stable.

I live in Sweden and it seems no retailer is getting the ocz v2 until the end of October so I'll have to go with something else. I'll look into your suggestions bobbknight.

 

[Dopekitten]

You need a new cooler if its 62C with only 1 core under load (and with PC probe)

I don't believe that PC probe uses the correct TJ max for 45nm (should be 100C)

Download this: http://www.techpowerup.com/downloads/SysInfo/Real_Temp/

and

this: http://files.extremeoverclocking.com/file.php?f=103

Then run both, and run prime95 small ffts for a min or two, then tell us your load temps (my guess, based on what you told me, is probably above 70 C, which is bad)

 

[wilhelmsson]

You need a new cooler if its 62C with only 1 core under load (and with PC probe)

I don't believe that PC probe uses the correct TJ max for 45nm (should be 100C)

Download this: http://www.techpowerup.com/downloads/SysInfo/Real_Temp/

and

this: http://files.extremeoverclocking.com/file.php?f=103

Then run both, and run prime95 small ffts for a min or two, then tell us your load temps (my guess, based on what you told me, is probably above 70 C, which is bad)

I get 66C from the stresstest measured with real temp @3.0ghz and 83C @3.3ghz so, yes, it seems I need a new cooler to run it @3.3

 

[dagger]

http://www.frostytech.com/top5heatsinks.cfm

 

[geofelt]

Almost any oem tower type cooler will do a better job than the intel stock cooler, and do so with less noise.

Plan on removing the motherboard from the case in order to install the cooler. If your cooler has a back plate, then the motherboard HAS to be removed.
If your cooler has push pins, then you SOHULD remove the motherboard so that you can look at the back side of the motherboard to be certain that all the pushpins are completely through and locked. If those tricky push pins are not secured properly, your cooler will wiggle and be only marginally effective.

 

[Dopekitten]

Yea, 83 C is far to high, get the Xigmatek S1283, that should be all you would need (get the bolt-through kit also, the pushpins don't give enough pressure)

 

[modtech]

I can vouch for the sunbeam core contact freezer and the zerotherm FZ120 as I have experience with them on multiple systems. However the previously mentioned Xigmatek S1283 will also perform well as long as you "remember to forget about push pins".

 

[Superhal]

i made a little buyer's guide:
http://forums.amd.com/forum/messageview.cfm?catid=11&threadid=100223&enterthread=y

 

[Zorg]

If you can find it in Sweden and you have the space, the clear winner in my book for price and performance is the HDT-S1283 along with the Crossbow I7751.

Don't make the mistake that so many have made, remove the mobo and use a backplate. The push pin mounting system that mimics the stock cooler is complete crap and absolutely will give you trouble sooner or later.

If you get the S1283, then spread a thin film of TIM over the whole CPU IHS. The heat pipes don't allow for good spreading of the TIM by contact.

 

[wilhelmsson]

Thanks for the info everyone, I'll post back later when I have decided what to get.

 

[wilhelmsson]

I was lazy and went with the Cooling Freezer 7 Pro which I installed without removing the mobo. Maybe the pushpins will cause problems in the future, but hopefully not.

I'm happy with the result at least, 58C @3.3ghz running prime95 small ffts , temp measured with Real Temp 2.70.

Also completed the simulation results this morning, total runtime 73h which had been 103h without OC'ing. (The runtime just scales linearly with the number of simulation runs so it is easy to calculate the total run time for the program by just running it for a little while).

Thanks for all the suggestions people.

 

[Superhal]

oh i just noticed something in your original post. Pc Probe only measures tjunctionmax. get CoreTemp and speedfan. Use CoreTemp to find tcasemax in speedfan, use pc probe to match up tjunctionmax in speedfan. if speedfan lists 2 cpu's, one is the graphics card. get ntune, nvmonitor (i dunno the ATI equivalent) or gpu-z and figure out which cpu temp is the gpu.

the antec cases have nearly legendary cooling. if your tjunctionmax is 62C under load, that means your cpu is maybe 53C, which is good. You still need aftermarket cooling to overclock though.

this article will help you diagnose if you have a heating issue and how to use various sensor reading software:
http://forums.amd.com/forum/messageview.cfm?catid=11&threadid=98342&enterthread=y

also, in that article, there's links to read about all the tXmax stuff we're talking about.

note: i wrote that article for amd chips, i understand that intel chips have temperature quirks like cores that don't run at the same temperature.

 

[modtech]

You've got it backwards. PCProbe does not measure tjunctionmax or anything else related to tjunction.

 

[Superhal]

You've got it backwards. PCProbe does not measure tjunctionmax or anything else related to tjunction.

I think you intel guys have it turned around. Tjunctionmax is the temperature from the thermostat under the chip, installed in the motherboard. TCaseMax is the diode inside the chip. when you guys talk about "calibrating" tjunctionmax, you guys are actually lowering the accurate measure to an approximation of the internal cpu temp.

From: http://www.overclockers.com/index.php?option=com_content&view=article&id=3765&catid=58:software&Itemid=4264

1. TCaseMax: This value stands for the maximum temperature at the dead-center at the top of the processor core or the IHS which will ensure 100% stability and safety.

2. Tjunction (or TjunctionMax): This value stands for the maximum temperature at the junction between the processor die and the PCB it sits on, this is usually much higher than the TCaseMax value.

According to the definition above, I know that PcProbe measures tjunctionmax, and Core Temp measures TCaseMax, and speedfan measures both, but labels them confusingly.

 

[modtech]

Core temp and realtemp are measuring distance to TJmax and then calculating the delta from a TJmax value that is assumed (although this number has been recently released by intel).

 

[genzeleam]

Xigmatek HDT-S1283 with retention bracket. You won't be sorry.