HSF tests are a load of bullhooey

[Angryviking]

Sorry for the title.

I started out running a Globalwin CAK38. Great temps, horrible noise; so I tried a Volcano 6Cu (4500rpm version) Low noise, ok temps but I wanted them a little lower, in preparation for summer temps.

So I started researching HSF that were readily available in my city. I was really taken by the Thermal Integration Dr.Thermal V77. Good performance, quiet fan and a wicked easy clip system. (Remember I had a Globalwin before !)

I asked around and I read literally 100's of reviews and opinions about this HSF. I couldn't get a straight answer on anything!

c/w, clip strength, temps...

Everyone said it would be the same or worse than the Volcano 6Cu (BTW is a great HSF for the $$$)

Well, sick of it and just plain bored; I went out and got a V77 today...

This HSF is great ! Super easy install, quiet, great cooling !

It idles where the CAK38 idled at 1200Mhz, except now I'm at 1467 !
:burn:

Idle 33 loaded 38

All this with a supposed c/w of .70 (CAK38 is .56 and Volcano .61)

Bullhooey...

:beer:

[rogerdugans]

I have found, in dealing with my hsfs and temps, that there are NO guarantees in air cooling systems: a sink/fan combo that is supposed to be junk may work quite well in one of my rigs- I had an Orb that worked absolutely great! Only reason I changed it was all the stuff I read about fan blades breaking. It took me some work to get that system cool with a more mainstream sink.

The reviews seem to be a guideline, not a guarantee; sort of saying " you are more likely to succeed with this one than with that one." I always reserve the right to make my own judgement.

[Angryviking]

I agree with that.

I also think, that until a universal measurement of C/W is developed, it should be dropped completely.

[rogerdugans]

I understand your point, I think, but I don't think it is possible to come up with a performance indicator that will be right in all cases, all the time. There are so many variables that can change!

I think c/w is probably a fairly accurate measurement of a heatsinks performance under a given set of conditions....

Its just that my conditions vary; I frequently try new fan locations etc, and it is amazing what moving just one fan can do!

The most important thing to me is that you did find a cooling solution that worked.

[Angryviking]

Thanks Roger

You are exactly right about all the factors in testing. It's incredible the amount of factors that determine the numbers. I guess my main gripe was; despite what everyone said and the tests...I ended up with a great cooling solution! :eek:

I would also say that what works great for your rig, may not be so great for mine and vice versa. Hence postings like "the CAK38 sux0rs" etc etc... :(

You won't know unless you try...

[Godfodda]

Thanks for that note viking. I ordered one of these last night off newegg simply because it "looked" like it would work. Glad to see someone has some good experience with it.

[Angryviking]

Glad to be of service Fodda :burn:

Do tell your numbers/experience once you have that puppy installed...

(which is sweet BTW, that lever action to lock it down is PURE genius!)

[Godfodda]

I'll post, but it'll be a couple of weeks. The goods will be here next week while I'm stuck working in Kentucky. :)

I didn't get a good look at the clip mechanism, but easy is good. My favorite for usefulness (though I've not used many) is on a generic Gladiator on my 1100. It has a little extension where a stock P3 HS clip has only a hole. The extension is just right for a flat blade screwdriver and the clip holds the HSF assembly very firm to the core.

Anyway, I'm gonna try it on my first AMD (1.33) for starters. That oughta be a nice chore for it. :)

[ButcherUK]

Another thing that always gets mistreated is noise. People say loud or quiet without any quantitive info at all. I've heard people call YS-TECH 120s silent, and one of them is louder than the entire noise from my system (inc. the scream of my 15krpm disk!)

[Angryviking]

I hear ya !

It all depends where they take the reading (in case or not) and how far away the measurement is taken from the fan.

:o

[rogerdugans]

True: my Delta is absolutely silent when I'm on this computer.

Since the delta is on a pc in the basement!

Many people just seem to get used to it, or something, but not me: I hooked that 7000 rpm fan up and my first thought was "what broke?"! And yet I have seen others that say they are used to it and LIKE it because they know the pc is ok if the delta is running!

Further proof that while research and seeking others opinions are very valuable tools, we each have to decide which compromises we want to make, and find out what works best in our situation.

[BillA]

for more words on C/Ws you might be interested in this thread (http://www.tekforums.co.uk/posts.php?threadId=3670)

my conclusion was/is that C/Ws are ONLY comparable when done by the same "reviewer" on the same rig

C/Ws from different sources mean nothing at all
bullhooey indeed

be cool

[Angryviking]

I wholeheartedly agree BillA :beer:

Thanks for the input!

[flounder43]

If I put a really high cfm fan on a cheap, extruded aluminum, stock Coolermaster, I get some pretty good performance. The fan really does help...

[William]

Its because Insocket thermistors vary from motherboard to motherboard and aren't accurate at all. C/W is the best way, but only if the C/W is taken correctly. AMD says to drill a hole through the base of the heatsink to a few mm over the core and attach a thermocouple there to get an accurate measurement of the core temp. Another method that is easier and almost as accurate is to nudge a thermocouple and attach it to the side of the core. The Thermoengine heatsinks and the spinoffs of it aren't fabulous, though they aren't awful either. They blow a lot of air into the socket artificially cooling the thermistor and skewing temps. This is what made the Thermoengine look lightyears better than anything out there when it performed about the same if a little bit better than everything else. The Glaciator I generally has temps reported that are higher than they actually are. Temp software is pretty pitiful too. AsusProbe tends to read 5-10C too high, so that can be added into the thermistor error. I don't trust any numbers from a review unless they use a CPU with an Internal Diode, use an External thermocouple attached to the side of the core, or attach it over the heatsink. OC.com uses the bore method, as wel as uses a P3 from time to time. There are numerous articles about how bad in socket thermistors are, I don't know any off hand though. So yes, MOST heatsink tests are a load of bullhooey. There are some that test correctly like OC.com and gideontech who at least have been using a nudged thermistor(at least, they were). I know there are others, but I don't know their names.

[Angryviking]

Quote
"Thermoengine heatsinks and the spinoffs... blow a lot of air into the socket artificially cooling the thermistor and skewing temps"

This I have heard about before; but how exactly can the airflow get to the thermistor under my CPU? I have a hard time believing it in my situation.
:eh?:

[MikeTimbers]

Dansdata (http://www.dansdata.com/coolercomp.htm) has a simulated cpu test rig and has tested shedloads of the damn things. Since all these were tested the same way, the C/W comparisons are valid.

Worth a look, IMO.

[Angryviking]

As much as I like DansData..that is the page that prompted me to start this post! :mad:

As controlled as his tests are, his c/w are way off on several high-performance HSFs. If I followed his c/w and MY test conditions remained the same (which they did..) I SHOULD have gotten the following results:

CAK38 - best cooler
Volcano 6Cu - Mid level cooler
Dr.Thermal V77 - basic level cooler

:eh?:

When in fact my findings WERE - v77, CAK and then Volcano.

So lemme get this straight Dan- the worst cooler in my lineup for c/w, did the best for me? How the hell does that help the average joe when deciding on a HSF based on your tests ???

Bullhooey... :beer:

[littlewoy1]

most CPU temps quoted around the internet and in reviews are wrong. They usually use the mobo thermistor which i have found to be incredibly inaccurate. On my Abit board it reads 10 oC lower than the actual temp! I use a probe mounted right next to the cpu core to get readings, and for nearly everything ive reviewed the temps are in the 40's or 50's. As I see it, a thermometer is never gonna read above the temperature, so the highest temp you can get is the most accurate one. I'd beware of reviews that dont tell you how they get temps.

Pete

[Angryviking]

I shouldn't say this, but I almost don't care what the temps are...

I mean seriously- HSF tests are run on everything from factory speed 700s, OCed XPs and BBQ starters! Temps are gonna be all over the place and as discussed there are already too many factors to consider! All I want to know is:

Which one cools better than the rest and how noisy is the fan?

Simple. I wish...



:eh?:

[MikeTimbers]

Originally posted by Angryviking
As much as I like DansData..that is the page that prompted me to start this post! :mad:

As controlled as his tests are, his c/w are way off on several high-performance HSFs. If I followed his c/w and MY test conditions remained the same (which they did..) I SHOULD have gotten the following results:

CAK38 - best cooler
Volcano 6Cu - Mid level cooler
Dr.Thermal V77 - basic level cooler

:eh?:

When in fact my findings WERE - v77, CAK and then Volcano.

So lemme get this straight Dan- the worst cooler in my lineup for c/w, did the best for me? How the hell does that help the average joe when deciding on a HSF based on your tests ???

Bullhooey... :beer:

If I remember rightly, the V77 is the one that looks like the ThermoEngine?

There seems to be a lot of evidence all over the net that the ThermoEngine design which allows air to flow onto the chip itself (although not the core, obviously) seems to affect the registered temperatures as per the in-socket thermistor. The temperature you are recording is from there and is not the core temperature itself. One of the points about a professionally recorded c/w is that you can use it to calculate how inaccurate your recorded temperatures are. The site for Arctic Silver goes into great detail about how to work this out and then you can amend MBM5 to show the true figures.

If dansdata's figures are different from yours perhaps that suggests that your figures are wrong, not his?

[Angryviking]

My statement still stands:

I don't care about the figures...just tell me what works. :burn:

My figures can't be depended on...that's a given. Whether my thermistor is 5, 10 or more degrees off is irrelevant..the V77 does better than the CAK and Volcano.

I also don't believe that my thermistor is being cooled by my thermoengine-lookalike; though I suppose more air is reaching the chipface- would that not inadvertently cool the core too? Even to a small degree that could be seen with better results- that's what we want right?

Dansdata figures are also coming up questionable against sites like Frostytech and HardOCP...2 against 1 wins? :confused:

I really don't know anymore... :eh?:

[MikeTimbers]

But you've already told us what works! For you, it's the V77. If you're unhappy with that, I would suggest an Alpha 8045 or a Swiftech MCX462.

My point about Dansdata and not wishing to belabour an already done-to-death thread is that esting with a thermistor inside the heater is more accurate than any other method. If you're going to go by the WWW heatsink tests, the SK6 or the Swiftech/Alphas are the way to go.

[Angryviking]

Trust me, I am happy :beer: but confused :confused:

No matter- I appear to be running cool, stable and quiet.

Thank you for everyone's input; this was fun.


:cool:

[William]

C/W is the ONLY way to translate temps across multiple chips. You multiple the C/W times the wattage of the CPU to get how many degrees above ambient the core temp will be. Raw temps don't mean a whole lot. I bet a Blorb does awesome on a 486. The Top Heatsinks are the Sk-6, Glactiator(I or II), Alpha 8045, and the Swifty 462.



The Socket does not come all the way down to the motherboard, it sits slightly off. Plus it also cools the ceramic, which cools the air in the socket, cooling hte thermistor. [H]ardOCP "solved" this by testing with masking tape on the CPU, but thats a bunch of crap as well. Check OC.coms ratings, they are very accurate.