Upgrade stock CPU fan?

jeffk

I have an AMD Phenom II X4 955 and the fan gets very noisy under normal usage (around 40% CPU usage). I was looking on Komplett, but am curious is it worth spending money on a new one?




Thanks for any help!

Deano12345

They'll run MUCH cooler and usually quite a bit quieter than stock coolers. Even if your not overclocking, id say its probably the best 30-40 quid you can spend on your computer.

5uspect

A good heat pipe based cooler is definitely worth considering if you have enough space in your case.

Here's a good resource:
http://www.frostytech.com/

jeffk

https://picasaweb.google.com/jeffking82/DropBox?authkey=Gv1sRgCPe92Iv73N2GTw#5592785050247227234

Thats the pc itself, if anyone can recommend one please. Im leaning towards buying on Komplett, the one below popped out as its a brand ive heard of before and seem something similar recommended on the net.

http://www.komplett.ie/Komplett/product/ZKB_01COM/13_COOLERS/03_CPU/productdetails/10021177/Cooler_Master_Hyper_TX_3_Processor_cooler_%28/RR_910_HTX3_GP/default.aspx

Deano12345

5uspect wrote: »

A good heat pipe based cooler is definitely worth considering if you have enough space in your case.

Here's a good resource:
http://www.frostytech.com/

Frostytech has the least consistent testing methology EVER (using different performing fans on different coolers on a hot plate last time I checked) so they shouldnt ever be trusted for giving comparisons between coolers.

http://www.hardwarecanucks.com/
http://www.legitreviews.com
http://www.guru3d.com

Those 3 sites will have reviewed nearly every CPU cooler on the market, and their testing methods are sound.

Edit : @jeffk : The TX3 is okay, its probably more suited to an application where you have less space. I'd go for its bigger brother, the Hyper 212+

http://www.komplett.ie/Komplett/product/ZKB_01COM/13_COOLERS/03_CPU/productdetails/10021605/Cooler_Master_Hyper_212_Plus_Processor_cooler_/RR_B10_212P_GP/default.aspx

jeffk

Great stuff thanks, once it keeps the fan level down its easily worth a few euros extra.

Im lookin on utube now, is it as messy as it looks to install?

Deano12345

jeffk wrote: »

Great stuff thanks, once it keeps the fan level down its easily worth a few euros extra.

Im lookin on utube now, is it as messy as it looks to install?

Plus you can change the fan if its still too loud for you !

From looking at it myself, its no tougher to install than any other air cooler, follow the instructions and you'll be fine, and sure you can always come back to boards if you need help

jeffk

Ah im not mega fussy, it’s just the fan at the minute sounds like a hoover when I’m barely doing anything on the PC. Also I haven’t over clocked it, well no more than what seems to be the default in the bios.



That’s true; you cant beat boards for help! Ill order this now and fingers crossed ill have it soon and no more noisy fan


Thanks for the help/info

5uspect

The best measure of a heat sink performance is its thermal resistance; the chip to ambient temperature difference divided by the thermal load in Watts. This is best achieved in a highly controlled environment with as many variables removed as possible. A PC case is not the best place to do this, even if the final place for such a product is exactly there. Of course ambient air currents are important but these are secondary is most cases.

On die thermal sensors are notoriously inaccurate and should not be used for characterising heat sinks. Add to this a set up where you have no idea what thermal load the CPU is under. I doubt anyone wants to drill a hole for a micro thermocouple in a heat spreader.

The frostytech setup is similar to Intel's own thermal test setup which uses highly accurate embedded thermocouples and a thermal block. I've used a similar insulated copper block rig to test CPU and GPU cooler designs (currently licensed to AMD). I've also tested them in working PCs and laptops under load too but more for demonstration purposes..

It's a highly idealised test that eliminates as many unknowns as possible and I would consider that much more consistent test.

Of course a problem occurs when you look at the actual thermal footprint of a CPU. A heat block will be extremely isothermal. A multi core chip will have distributed hot spots that the heat spreader probably won't fully dissipate. I've seen some IBM work on thermal imaging, here is an overview.

5uspect

jeffk wrote: »

Ah im not mega fussy, it’s just the fan at the minute sounds like a hoover when I’m barely doing anything on the PC. Also I haven’t over clocked it, well no more than what seems to be the default in the bios.



That’s true; you cant beat boards for help! Ill order this now and fingers crossed ill have it soon and no more noisy fan


Thanks for the help/info

Also have you checked the thermal interface material has been correctly applied? Are you're core temps higher than expected?

jeffk

5uspect wrote: »

Also have you checked the thermal interface material has been correctly applied? Are you're core temps higher than expected?

I have taken the fan off and put on some formula5 silver thermal compound and also made sure the CPU/fan and wires where in ok. Ive changed the fan level in the bios and also changed the cool and quiet to on and off to test that.

Most of the time I use 40% of the CPU and the fan goes mad and does an impression of a hoover.

5uspect

And the temperatures? Is it getting hot?

jeffk

Motherboard
ASRock 880G Extreme3 (CPUSocket) 50 °C
Hard Drives
488GB SAMSUNG SAMSUNG HD502HJ ATA Device (SATA) 33 °C

CPU has no temperature in it in speccy

When I ran speed fan(it doesn't support my motherboard) the CPU was at 50 °C

Deano12345

5uspect wrote: »

The best measure of a heat sink performance is its thermal resistance; the chip to ambient temperature difference divided by the thermal load in Watts. This is best achieved in a highly controlled environment with as many variables removed as possible. A PC case is not the best place to do this, even if the final place for such a product is exactly there. Of course ambient air currents are important but these are secondary is most cases.

On die thermal sensors are notoriously inaccurate and should not be used for characterising heat sinks. Add to this a set up where you have no idea what thermal load the CPU is under. I doubt anyone wants to drill a hole for a micro thermocouple in a heat spreader.

The frostytech setup is similar to Intel's own thermal test setup which uses highly accurate embedded thermocouples and a thermal block. I've used a similar insulated copper block rig to test CPU and GPU cooler designs (currently licensed to AMD). I've also tested them in working PCs and laptops under load too but more for demonstration purposes..

It's a highly idealised test that eliminates as many unknowns as possible and I would consider that much more consistent test.

Of course a problem occurs when you look at the actual thermal footprint of a CPU. A heat block will be extremely isothermal. A multi core chip will have distributed hot spots that the heat spreader probably won't fully dissipate. I've seen some IBM work on thermal imaging, here is an overview.

Problem being there, is that most chips very rarely run to their TDP under heavy daily usage (rendering etc). And whats the point in testing temps in an environment that the average user isn't going to encounter ( they will have something like a tech bench or case)

Also, it doesnt change the fact that FT dont use a control fan, which can have a huge effect on performance. They also only take into account the roughness or imperfections the base of a cooler, not any that are on a chip. Its all too scientific, and not enough real world.

Best Example of this is that they rate the Noctua C14 above the D14. No other review on the web does that because the D14 is the better cooler in terms of performance

Edit : @JeffK : Use CoreTemp

jeffk

https://picasaweb.google.com/jeffking82/Downloads?authkey=Gv1sRgCJ2ps927--ithQE#5604054271113834594

Deano12345

jeffk wrote: »

https://picasaweb.google.com/jeffking82/Downloads?authkey=Gv1sRgCJ2ps927--ithQE#5604054271113834594

Thats WAYYYY to hot, even for a stock cooler, what way did you apply the thermal paste ?

jeffk

Deano12345 wrote: »

Thats WAYYYY to hot, even for a stock cooler, what way did you apply the thermal paste ?

The normal way, on the fan and then onto the CPU. I checked it again and cleaned any excess off.

jeffk

It at 53 and ive a lot running, ive read 50 is the normal temperature?

5uspect

Deano12345 wrote: »

Problem being there, is that most chips very rarely run to their TDP under heavy daily usage (rendering etc). And whats the point in testing temps in an environment that the average user isn't going to encounter ( they will have something like a tech bench or case)

Also, it doesnt change the fact that FT dont use a control fan, which can have a huge effect on performance. They also only take into account the roughness or imperfections the base of a cooler, not any that are on a chip. Its all too scientific, and not enough real world.

Best Example of this is that they rate the Noctua C14 above the D14. No other review on the web does that because the D14 is the better cooler in terms of performance

Edit : @JeffK : Use CoreTemp

LOL at the bold bit!

The reason you test on a bench with a big temp difference is that most coolers are linear and you get more a accurate result that way since you minimise your instrumentation uncertainties. (We also tested with various degrees of blockage and crossflows to simulate real world conditions)

Since they use a standard block with know roughness that is simply a constant in their test. Obviously a CPU's heat spreader will have an unknown roughness but that doesn't matter when you can doing relative testing between different heat sinks on a standard test facility.

Of course fans have a huge impact, our GPU coolers had noticeable performance differences when small aerodynamic changes were made to the fan design, though our aerodynamics were much more unsteady than your typical laminar flow fin arrangements.

Deano12345

5uspect wrote: »

LOL at the bold bit!

The reason you test on a bench with a big temp difference is that most coolers are linear and you get more a accurate result that way since you minimise your instrumentation uncertainties. (We also tested with various degrees of blockage and crossflows to simulate real world conditions)

Since they use a standard block with know roughness that is simply a constant in their test. Obviously a CPU's heat spreader will have an unknown roughness but that doesn't matter when you can doing relative testing between different heat sinks on a standard test facility.

Of course fans have a huge impact, our GPU coolers had noticeable performance differences when small aerodynamic changes were made to the fan design, though our aerodynamics were much more unsteady than your typical laminar flow fin arrangements.

I still don't think you can beat testing in a case, especially since it doesnt involve modifying our very nice chips. Oh, just wondering, who's we ?

jeffk

I took the cpu out to check and now the pc boots but their is no display.

The cpu is in right and so are the wires. Anyone have any ideas?

5uspect

"We?" The research institute in the University where I was a postdoc.

From an engineering perspective testing in a case is a nightmare. You can't trust the CPU sensors for anything within a few degrees, the ambient air flow differs from case to case and the load is unknown. It's just not repeatable.

Deano12345

5uspect wrote: »

"We?" The research institute in the University where I was a postdoc.

From an engineering perspective testing in a case is a nightmare. You can't trust the CPU sensors for anything within a few degrees, the ambient air flow differs from case to case and the load is unknown. It's just not repeatable.

You'd cringe at the thought of trying to guess the core temp of an unlocked Phenom or Athlon then, since you've only got the CPU temp to go by and have to work back the ways

@Jeff, sounds like a grounding issue to me, check all your video card power connectors and video connections too ?

jeffk

Deano theres sd or d5 coming up on the red led lights on the motherboard if that makes sense?

Ill check the connections now also!

Deano12345

jeffk wrote: »

Deano theres sd or d5 coming up on the red led lights on the motherboard if that makes sense?

Ill check the connections now also!

D5 is signifying the BIOS is being prepared to load from RAM. That board has a Clear CMOS button, press that, maybe a CMOS reset is needed

jeffk

I tried that and take the battery out, also hold the power button turn on the psu, then turn on the pc itself

C.D.

As a reference, I have a stock 965 in an Antec P183 which hits 48C under max load (Prime 95). I have a Scythe Mugen 2 on it and the case would have low air flow with one intake and 2 exhaust 120mm fans, all low RPM.