IHS or not?

[Brians256]

For more information, see this thread (http://forums.procooling.com/vbb/showthread.php?t=12475) on the Apogee review done by Lee "Robotech".

What is most interesting is that we now have some information about how badly the IHS affects temperatures! Testing with the ancient Maze3 on an Athlon64 chip with an IHS shows that it performs nearly as well as a modern Storm WB! That is, within 1C. See this (http://forums.procooling.com/vbb/showpost.php?p=151092&postcount=93) post by Lee. Quite informative!

Should all watercooling afficianados remove their IHS? What does this mean for future testing? Is it necessary if no one wants to remove their IHS? After all, if even ancient designs perform adequately on IHS equipped CPUs, then new designs have to be awefully bad to even register a difference in on-die temperatures.

[Etacovda]

depends on your definition of 'performs well'. We DONT know how well they perform, because the temp monitoring on such cpu's is !#@$ing useless.

Overclocking test is needed, imo. Screw the temps, get to the end result - we're beating around the bush with temperatures (and will be for some time), and the end result is the overclock. Jump straight to the conclusion, no need for the plot twist...

[jaydee]

depends on your definition of 'performs well'. We DONT know how well they perform, because the temp monitoring on such cpu's is !#@$ing useless.

Overclocking test is needed, imo. Screw the temps, get to the end result - we're beating around the bush with temperatures (and will be for some time), and the end result is the overclock. Jump straight to the conclusion, no need for the plot twist...
What would overclocks tell us? All CPU's overclock different. One might OC 500mhz with water and another of the same make might OC 1000mhz on air.

[Brians256]

nikhsub1 believes that even though temps are not noticably different, the "focused cooling" of storm designs will always help overclocking. Maybe so?

Orkan (on another forum) notes that he sees a 100MHz increase in OC on his CPU difference between Apogee and Storm despite seeing zero change in temperature.

[Etacovda]

What would overclocks tell us? All CPU's overclock different. One might OC 500mhz with water and another of the same make might OC 1000mhz on air.

I dont mean using multiple cpu's; just one. But, in saying that, if the IHS is not contacting the core properly, it'll be different depending on the block design.

http://forums.procooling.com/vbb/showpost.php?p=151114&postcount=104

[bigben2k]

Assume that all temp measurements are wrong. What do you have left?

Is overclockeability a measurement? A Storm's cooling is very localized, Apogee's is not.

Too many variables; there's nothing here.

[Etacovda]

And, pray tell, what is the end PURPOSE of a waterblock ben?

Damn, theres getting interested in the science of waterblocks, then theres COMPLETELY IGNORING THEIR END TASK.

"Is overclockability a measurement? A Storm's cooling is very localized, Apogee's is not." So? NOONE with a shred of a brain cares what their reported cpu temperature is. They care what their overclock is, thats what a waterblocks end purpose is. Its just that, until now, it was/is considered not scientific enough for testing purposes. I think thats out the window with the 'throw a tc on the side of an IHS edge' testing method thats doing the rounds now. If we're going to get this off-hand about how blocks should be tested, i think its perfectly valid that the only actual end varible that matters (the overclock) should be tested, even if it is imperfectly (because NOONE can call these IHS tests 'perfect'). Obviously this test would need to be done with a waterchiller etc, its not just a straight madshrimps window open, drug addled moronic l3d review - there would have to be constants involved...

I never assumed all temp measurements were wrong - i said the cpu tests that Lee did were CLEARLY not showing us anything, as was shown by the maze3 test.

spending too much time planning your test bed and not enough actually thinking about the end task of the DUT?

Im all for testing, and getting it RIGHT - but lets not forget the end task of the damn DUT, because otherwise it wouldnt exist. As i also said in the post i linked, it wont be a be-all and end-all, its a nice side discussion whilst other things get done.
Id also really like to see the mushroom shaped, single piece heat die based on a dual core and an IHS solution - but that wont be done for a while, so why not do the overclocking tests?

Another side topic (im good at this) - if this was a german block tested this way, say, a year ago - theres no way in HELL anyone would accept these results. This whole thing just seems so... un-procooling like.

Im looking forward to what pH's results are, actually.

[BillA]

"What is most interesting is that we now have some information about how badly the IHS affects temperatures! Testing with the ancient Maze3 on an Athlon64 chip with an IHS shows that it performs nearly as well as a modern Storm WB! That is, within 1C. See this post by Lee. Quite informative!"
??
shown was the effect of Lee's technique, to which you have nothing to compare
sorry, enough only for me to say more and better is needed
by what soaring leap of the imagination was it concluded that temps in the center are the same as the edge ? or that the extent of compression would be the same ?

anyone recall this image ? the 'range' at the edge is 0, 100% compression if you would
http://thermal-management-testing.com/wbTCholes.jpg

it is simple and correct to observe that temps are reduced by removing an IHS, and it is generally held/true that lower temps will boost the OC (sometimes ?)
for enthusiasts whose goal is OCing it would seem AMD products are mandated if removing the IHS intended

no controversy

[clocker]

depends on your definition of 'performs well'. We DONT know how well they perform, because the temp monitoring on such cpu's is !#@$ing useless.

Overclocking test is needed, imo. Screw the temps, get to the end result - we're beating around the bush with temperatures (and will be for some time), and the end result is the overclock. Jump straight to the conclusion, no need for the plot twist...
Why isn't cooling efficiency a valid end result even if divorced from overclockability?
You imply that a drop in operating temp is not worth anything unless I also spend extra time trying to wring a few MHz more out of my PC.

[Etacovda]

increased overclock stability would indicate that the chip is being cooled better in the areas it needs it, yeah? If it gets cooler in the diode area and makes no difference to the clocking of the chip, im going to go out on a limb and suggest that the slightly lower temperature will have no effect on the chips lifespan (if thats what you're after)...

If you're just talking about general cooling efficiency - then for what? so you can say that your cpus diode is colder, even though theres no tangible benefits for it being so?

Watercooling is generally for 3 reasons

Overclocking (obviously chip stability comes into this)
Silence (again, chip stability - means you can have it slightly warmer and be slightly quieter)
Bling (completely irrelevant to anything else)

cooling efficiency comes directly under how far it will overclock... If its more stable (ie will clock higher) with a different waterblock, even though it shows the same temp, id call it more efficient... what is efficiency?
#

1. The ratio of the effective or useful output to the total input in any system.

Id underline useful here, but hey, you get the point :p

[clocker]

If you're just talking about general cooling efficiency - then for what? so you can say that your cpus diode is colder, even though theres no tangible benefits for it being so?
Maybe just the warm, fuzzy feeling I get from optimising the cooling system?

BTW, what is the tangible effect of an extra 100MHz on a machine that's running close to say, 3GHz already?

[koslov]

Right now, as the only desirable overclocking chips are AMD anyway, you should be testing AMD chips with the IHS removed. This is your audience. Unless you want to cater to those who are either too timid to butcher their overpriced CPUs or those doing this for bling, you should be doing whatever the pro-level enthusiast is doing. If, right now, this means ditching the IHS on AMD cpus, do it. If in the future, Intel is the cpu de rigueur, and IHS isn't going anywhere, then test that way.

[Etacovda]

clocker: personally, im in the same boat as you - i dont care about that last 100mhz, but if you're using the machine for something intensive (ie, rendering, which i actually do use my pc for), increasing the pc's speed by 5% can mean a lot when you're doing a render/animation that will take 2 days... plus, in a dualcore, that benefit is two fold.

However, i do care about silence - so a better waterblock could theoretically let me lower my fan speed (probably pretty significantly, really) for the same overclock; allowing for the same productivity at a lower noise level.

Also, a watercooling system is plain extravagance anyway - theres no justification in it, may as well get the most benefits out of it that you can....

Chip stability is the important thing. The temperature of the diode doesnt neccessarily represent that stability.

[jaydee]

CPU's overclocking ability can change over time aswell. Probably more so than the TIM joint issue. Also using the same CPU still tells us nothing more than others tests as the results are only good for that CPU.

Also I refuse to remove the IHS even on my $60 Sempron. I get what I get with what I got. This from a person using the stock AMD all aluminum heat sink though with no need to change it. :) My system runs great as is.

[bigben2k]

Etacovda: the end purpose is not forgotten, but the testing method may or may not give the right picture.

Eg; Lee (Robotech) test on a 32mm by 32 mm die: what did you learn?

The associated CPU test: what did YOU learn?


I'm taking a few days off; back Tuesday.

[clocker]

Right now, as the only desirable overclocking chips are AMD anyway, you should be testing AMD chips with the IHS removed. This is your audience.
You really think the majority of folks here have removed their IHS?
Unless you want to cater to those who are either to timid to butcher their overpriced CPUs...
Hard to discern your gist here..."too timid to butcher"?
"Overpriced CPUs"?
OK, I'll own up...indeed I do feel trepidation when contemplating the possible demise of my $350 CPU, not to mention the hit it takes on the resale market.
... or those doing this for bling, you should be doing whatever the pro-level enthusiast is doing. If, right now, this means ditching the IHS on AMD cpus, do it. If in the future, Intel is the cpu de rigueur, and IHS isn't going anywhere, then test that way.
What the hey is a "pro-level" enthusiast and how can I sign up for a salaried postion?
How many such creatures are there anyway?
At what level interest/involvement/investment does one qualify?

[koslov]

What's pro-cooling? Obviously, except for a handful of people here, we're all cooling amateurs.

OK, so pro-level is a silly, indefinable term. I'm talking about skilled coolers who will do pretty much everything they can within their homes/garages to realize a significant gain in temps. And using a hobby knife to remove an IHS definitely is within their capacity.

Monetary investment has nothing to do with it.

[jaydee]

Salaried position? lol. I have lost about $7,000 in my cooling adventures just to end up being satisfied with a stock $5.00 cooler. :)

[clocker]

Salaried position? lol. I have lost about $7,000 in my cooling adventures just to end up being satisfied with a stock $5.00 cooler. :)
An old cooling proverb say:
"The road to cooling nirvana is paved with broken Visa cards."

At the very least you could have a very interesting garage sale...

[Brians256]

The point to all this furor is to find out WHAT we need to investigate and HOW we need to present it. Or even if we need to investigate!

Temps of the on-CPU diode are one metric. Another is overclockability. Add both of those with/and w/out IHS. Then, there's die-sim tests and finally simulations. Finally, we can just test overclockability.

All of them have merit, it seems, and we can't single out a particular number and say "This is the ranking". The closest seems to be CPU diode and die-sim temps, but overclocking really gives an effectiveness number that trumps the other numbers. However, without testing a large sample, that overclockability is hard to trust.

Blah.

I think pH is on the right track, but I'm still worried about making sure the numbers are important to you guys.

[BillA]

all in the pot with a vigorous mix, produces goulash
until the issues are separated, and considered separately, the present mess will continue

suggestion:
consider die sim and CPU based testing apart
- consider bare die apart (AMD only)
- consider IHS clad CPUs apart (Intel only ?)

I note that the IHS temp was not included as a metric
this thread is an extension of the avoidance of an intellectual discussion of IHS characteristics

agreed, Blah.

I lost the opportunity to influence pH's track, but do hope he brings some rigor to the discussion

[jaydee]

An old cooling proverb say:
"The road to cooling nirvana is paved with broken Visa cards."

At the very least you could have a very interesting garage sale...
Well all I have to show for that $7,000 is a little CNC mill I bought for $1,800 and about $1,500 worth of test equipment. The rest is metal shavings, scrap copper and aluminum (in the form of water blocks and die sims) and wore out tooling.

As for Bill's comments I agree.

What we need to investigate is which methods gives our readers the results they can use the most. The other methods we can use for our own interests I guess.

I don't see overclocking as an option as the CPU's are not all equal. The results could not be compared to anything but that CPU which = useless to our readers.

I don't see testing without the IHS as an option because I refuse to believe most readers here will decapitate their CPU's. Myself included. So that = useless to our readers.

I don't see die sim testing as useful to our readers because die sims do not react like CPU's do. So those results are useless to our readers.

Testing with the IHS with the system as stock as possible seem to only way I would think about attempting it. Especially being many coolers mounting systems are designed with the IHS's extra height in mind. The problem with this is temp monitoring and the IHS's TIM joint not only the quality of it but repeated mounting over time (forgetting all smaller variable's of course). Intels soldered on IHS might help that but still need a decent temp monitoring method. Grooving the IHS is an option but not many can get that done or are willing to risk it.


That is my opinion at this point anyway. It is all :hammer: at this point though.

[pHaestus]

I lost the opportunity to influence pH's track, but do hope he brings some rigor to the discussion

LOL

"This is incomprehensible crap that should never be posted" I think sums your comments up nicely. Or did you mean with my test system?

I'm HAPPY to do IHS testing with a grooved CPU and a diode reader to monitor effects over time. But I can't make that be on my main test bench for the simple fact that I'll destroy that CPU if I am remounting wbs on it several times per day.

[koslov]

OC numbers don't really make sense when you consider all the variables involved. You might find that a nice hsf will outperform many a good wb simply because you're cooling the fets more. We really don't even know all the reasons why a chip will clock higher under certain conditions than under others. For the ones that we know of, there's already cpu temp, voltage, voltage stability, fet temp... even some weird phenomenon that could probably be ignored (but who knows?) such as electromigration, alpha particles, and cosmic rays. AND the relation between the factors is not simple.

I'm not saying it can't be done, but you must construct a system where at the very least, OCs vary predictably with CPU temps.

jaydee, not sure exactly what you mean by "react." Results from a single CPU are no more indicative of the performance on other CPU designs than are results from a die sim. Maybe worse. Even for dies with similar geometry, every design has different hotspots.

Here's a suggestion: poll your readers (not just the forum nerds) to see what they are doing. What CPU, willing to remove IHS, current cooling, etc? But keep it short and easy! You want to get as large a sample as possible.

I guess a prerequisite to this is to *have* readers or even articles to read for that matter ;). So once you get an article you're certain will receive fairly wide circulation, *make* them take the survey before the last page or something along those lines. Joe's smart, I'm sure he can figure something clever out. And I hear he doesn't mind extra coding at all! Then you'll know exactly who you're readers are, and you can go from there trying to determine how best to serve them. JoeC recognized this fundamental issue long ago on oc.com (http://overclockers.com/articles1271/index02.asp).

[jaydee]

jaydee, not sure exactly what you mean by "react." Results from a single CPU are no more indicative of the performance on other CPU designs than are results from a die sim. Maybe worse. Even for dies with similar geometry, every design has different hotspots.


Pretty simple. A die sim is a large chunk of copper (or whatever), a CPU is a small chunk of silicone with transistors and what not in it. They do not react the same to power input or cooling effect from the block.

That is also why lasers and conduction are useless IMO and why a gave up on that other thread. Taking base plate temps is stupid IMO. You want to know how well the block cools the die not how well it cools itself.

Anyway I am done with testing mumbo jumbo for a while.

[BillA]

dishonest participation ?
an honest oversight ?

from another forum, a post bt eva2000:
"G4 Storm works perfectly everytime on my LGA775 rig just AMD64 get's all weird on me i.e. http://www.xtremesystems.org/forums/...hlight=mcw6002 "

must now admit to being totally disgusted with the non-intellectual non-technical 'assessments' of the IHS and its impact on testing and performance

too bad I cannot come to an accomodiation with Cathar,
I will never comment on a heat die again if Cathar will stick to factual data about the 775 Tim1 joint

how about it Stew ?

[pHaestus]

My only comments on the IHS are from my own experience: CPUs pop off with the cooler when they're attached, and this will rip wires off my modded processors. If that's non-intellectual then so be it.

You know there was at least one other Swiftech Storm review on an IHS-equipped CPUs that also showed it perform rather poorly, wasn't there?

[BillA]

boy I can't write for shit, sorry pH

IHS
775
Tim1 variability
period, full stop

re CPU pins; yes, we agree (same issue on P4 TTVs)
so ?
no pins on a 775, and the IHS itself is retained by the socket 'clamp' (impossible to separate CPU from socket w/o releasing clamp)

I have no issue with anyone testing with a bare AMD CPU, does this not represent the enthauast/procooling community's typical manner of use ?
testing as it will be used, excellent stuff

my issue is not testing as no one has done that of which I am speaking, it is with the ongoing bias in the discussion of testing with an IHS (thread title)
I have suggested consideration of a single Tim1 configuration; while a pissing contest then occured over a different Tim1 configuration (which had been excluded from the onset)

[pHaestus]

My personal problem with the discussion is my inability to properly test the AMD TIM1 joint. I would LOVE to take a 939 and groove the IHS and wire it for diode readings and then just test with it. But I am not at all convinced I could use it for long enough to get any useful data before it broke.

Do you have access to someone that'd machine an IHS as per spec Bill? I can do the diode soldering and I actually have a spare 939 mobo here...

[BillA]

jeez, I sure could not recommend grooving an AMD IHS w/o a program yielding something useful
??
what would be the positive result ?

I know the shop that grooved some before, zero assumption of responsibility, weeks to get to, not cheap (think $150 ? ea)
suggest holding off a bit to see if IHS clad CPU test data may be found acceptable by the forums (lol)
do not want to be sending 1 at a time

EDIT
yes, the 939 Tim1 WILL debond - there is no question in my mind
I would guess that you will observe the number of sink removals 'possible' before Tim1 decouples

[pHaestus]

Well I was thinking that would kind of be the point. Collect CPU case and CPU die temps while cycling system and then while remounting regularly and just watch the joint degrade. Or do we all concede it happens and don't care about the magnitude of the effect?

[BillA]

I have never contested the described degradation of the AMD Tim1 joint, but any test data of that joint cannot be extrapolated to the 775 Tim1 joint as they are different - and very much so

you may not see a gradual decline, also possible is an abrupt shift in temps due to the IHS being (eventually) pulled loose
and after a single test, . . . .

[koslov]

BillA: skip this.

Heh.

Preparing a CPU for testing with IHS: Remove IHS, clean away all TIM joint material. Take clean IHS and fill with thermal paste with thermal properties equivalent to AMD TIM. Fill past normal IHS enclosed volume and take care not to leave any air pockets. Place paste-filled IHS on clean CPU and press firmly. Wipe away excess paste from IHS edges. Seal IHS edges to package with epoxy.

Now you've got a CPU with IHS that should be able to handle any number of remounts. True, the properties will be different, but I don't see why you couldn't get a good match. It's up to the tester to find the right paste, which may not exist. Also, the IHS flex will probably change.

[BillA]

koslov
why not run that 'bare die' poll ?

[koslov]

Poll:
http://forums.procooling.com/vbb/showthread.php?t=12531