Pro/Forums - Apogee from Swiftech...

Pro/Forums (http://forums.procooling.com/vbb/index.php)

- General Liquid/Water Cooling Discussion (http://forums.procooling.com/vbb/forumdisplay.php?f=9)

- - Apogee from Swiftech... (http://forums.procooling.com/vbb/showthread.php?t=12376)

Bill > I didn't say to take only die T° at a certain point (copper isn't silicon, processor get numerous hot spots, flux has a very complicated shape and never "flat" like a bare die, thermal probe in processor die could be near an edge or elsewhere). If I had a my die done, I'll take several T° as well the IHS internal T° to get a global view, but IHS center T° above core is normally enough to characterize a system thermal resistance (with a groove on it like Intel TTV or with a tiny hole to fit TC/RTD just above center die at 0.2-0.3mm from IHS surface with my version), die T° reacts as IHS center do. Absolute T° in copper die won't be never the same as a processor so we can't say if I got 2°C difference with my die between X and Y, I'll have 2°C better on my processor between X and Y, this is not true. Die T° is just to complete data and see if relations are linear between 2 different coolers, we could have a better view of how an central impingement will affect IHS T° and core T° in comparison of a simple WB for example, flux spreading, etc. We can imagine 5 thermal probes in IHS, one for center and one in each quadrant, why not..

round and round we go.

Let me ask this simply:
How does varying levels of heat, change the ability of a waterblock to remove said heat?

So from one cpu to the next, there are varying levels of heat being transferred through the IHS. So what. Every single person that buys these blocks knows that someone else with the exact same setup, may have cooler, or higher temps based on the fact their CPU's may put out less or more heat.

The heat being trasferred to the block by the cpu being higher or lower does NOT affect the ability of the waterblock to remove that heat.

The amount of talking and lack of doing astonishes me. Ok, so you say you can't re-mount a waterblock to the same cpu because the IHS is changing its contact every time you do. The law of greater numbers will take over when you average out, and give you solid numbers either way. (solid enough for me and 90% of the community anyway)

You have 2 waterblocks. You want to compare them. You say if you use the same processor/comptuter setup to test them, the IHS will be different every time you re-mount. So mount it five times, alternating the block you use. ... so the variations will be as close as you can get. Then do the same thing on FOUR other computers.

I get the distinct feeling that everyone just keeps talking, making things more complex and more complex until no one knows at all what anyone else is talking about... but they sure did talk alot.

Bottom line: - as a consumer
Who is going to figure out which block performs better?
How are they going to figure it out?

Not that it would really matter... as demonstrated by this thread. No matter who did the testing or how they did it... half of you would believe them, and the other half would not.

Round and round we go. Someone pass the pepsi.

LMAO @ Orkan. It seems as if this whole discussion is upsetting you Orkan. This is how perhaps progress can be made in the world of testing WB's. Or not. Good dialogue for sure. Let me explain again, to you; the IHS issue at least with AMD and older northwood Intel CPU's. The IHS is affixed to the PCB of the chip, not to the core itself. You slide a blade round the IHS on an amd and whammo! the IHS will come off. This was also the case with the NW's of which I've owned many and have seen DRASTIC irregularities in temps between like CPU's with IHS on, however, pop it off and each CPU is well within what one would consider a 'normal' margin, ie they thermally behave very similar at similar speed and voltage. How could this possibly be?

Quote:

Originally Posted by nikhsub1

FWIW, I tried to get a soldered/epoxy'd ihs off of an Intel cpu, to no success. I stuck in in a 400 degree oven for an hour, froze it overnight, soaked it in googone, nothing worked. Believe it or not, the CPU still runs to this day.

:eek:

The part I don't get

There's a few things that confuse me here. It's going to take me a while to get to the relevant point, but I will get there, don't worry.

For a closed system, power in *must* equal power out. That's a fact we know for sure.

Next, given a set flow-rate, and a set airflow through a rad, it's easy to characterize a radiator's performance. We know that too.

Given that knowledge, we can characterize the heat added to a system by the pump, friction, turbulence, etc. That heat can also be characterized by insulating the pump, and measuring it's power dissipation (closed system again here).

Would it not be easy to measure power into a system by measuring the temperature change across the barbs of the radiator, the ambient air temperature, and the coolant flow-rate?

Given that measurement of coolant temps is easy (used everywhere in industrial process control) and so is flow rate (if you want to pay for it, and have the space), how hard would it be to characterize full die->coolant performance--which is all we should care about anyway? I know the on-die CPU diode is not the best measure out there, but it's better than nothing, which is where we are now.

If a worthy tester was willing to grove the IHS of an actual CPU, the thermal output of the system can be characterized off of the radiator. Given the CPU diode and the IHS measurement, the quality of the IHS contact, and the thermal resistance could both be characterized for any given WB.

Others have said it, I'll say it too. The temperature of the IHS doesn't mean a thing to me. I care how well the whole system works together.

We need wholistic testing because--really--who cares if the IHS is a few degrees colder if the CPU keeps overheating and crashing?

Quote:

where is the DIE TEMP to be measured, and how ? (I missed this in your post)

Quote:

For those struggling to picture what I mean, imagine something like a 13x13mm (or whatever sized) copper heat die with a "mushroom" IHS lid on it, but all one solid piece. In fact the word "mushroom" pretty much exactly describes the sort of one-piece thing I'm talking about. Heat energy is applied at the mushroom stalk, the temperature of the die is taken from the center of the stalk just below where the IHS "head" starts to fan out. Heck, groove the IHS head and stick a TC in there too, ala Intel TTV style, and then we can validate that the relationship between IHS surface temp and stem temp is consistent.

Of course as said this has no correlation to CPU die temp, or some such correlation would have to be developed (if of interest).

Quote:

do you know what the TTV has that no die will ever have ?
repeatability, lesser absolute performance but a greatly reduced range

How is a TTV more repeatable than a heat die?

Quote:

Originally Posted by unregistered

but how is anyone to produce comprable results for multiple blocks over time without using the same "old" chip......

Cathar
everything can be considered variable, we accept that
it is the degree of variation that is in question
I would not use an AMD CPU because of what I have heard about the IHS, I have no personal experience however
I have not heard of IHS related problems with current Intel products, nor in asking do I hear of them
so I do NOT assume AMD and Intel have the SAME CPU IHS problems

the TTVs use a TC so resolution is limited to 0.1°C, cross ckg TTVs indicated no differences that I could measure
are they all the same; forever ? - no, nothing is

Cathar, I have no devotion to TTVs, I will probably never have one; but they are a repeatable heat source
if the TTV numbers are wrong, they need a correction factor (as Intel provides for specific CPUs)
if the Swiftech numbers are wrong, they need an (appropriate) correction factor

I repeat, the TTV is a fine tool; the data's interpretation may be/is the problem
baby and bathwater issue
the TTV is only a tool, and 'produces' test results; a TTV does no cooling
as Intel says, confirm the results (Swiftech seems to have done so on a CPU)
-> a conflict between TTV and CPU results ?
the CPU every time, what else ? (the TTV is not a cooler)

get real Cathar, why would I "tear down" a tester for something they cannot get ?

I see the latest posts and now we have people believing that IHS and silicon temps are independent ?
ah, progress

Quote:

Originally Posted by Roscal

We see well the limits of testing here. Swiftech made several new dozen of measurements on different systems with Apogee and Storm, data are valid (some will be made on a true processor very soon to get more data) and finally, they got opposite results according to the testbeds. What to conclude then? Yes it's a good block but is it the best? Not ! What best mean? on which system? With which others parts? Etc. Swiftech want principally to reduce costs manufacturing without reduce performances too much, they suceeded a priori and they could release good blocks at lower price, so more people could enjoy in WC because it's cheaper. It's not a Storm bis, apparence is almost the same because of their process. It's not a crime to keep a design to be their new design reference, or it's like to said "all radiators use the same design", it's the same ridiculous sentence. We need to have a global view of their decision, not only to be stubborn on performances which don't make any interesting differences on most real configurations in typical applications anyway (mounting, paste, load, etc.), some people forget this point...Swiftech is honnest and they say clearly that core geometry will affect efficiency between their 2 top notch WBs, no mystery here. Global and universal statements can't be made on a such complex field like thermal management for processors due to diversity and a lot of people here forget this point too !

Some are playing with dies which are, per definition, far from actual reality and non linear systems like true CPU with an IHS. Robotech will show soon that he got better results with Storm than Apogee on his large naked die. Does it mean Storm is always better? Absolutly not, it's better on his specific testbed, not in all the real cases and perhap's Apogee is always the best in real applications (with multiple processor geometries), who's know. Argue that Storm is the best of the best is true BS, and the same could be applied to any WB obviously, not especially to Storm. It depends of the context and if you aren't be able to understand or extrapolate what differences could be provided between different heated systems, it is because you don't know anything to thermal management, no less no more.... It makes me laugh when I see people on varied forums who are very surprised that Storm could be outperformed a bit by another WB in particular circumstances. No one made an giant & perfect WB testing with all WB worldwide to give an universal answer I believe, so why they need to impose a biased point of view? There's so much mind restrictions that it's impossible to be better for these people, false statement again. It's always the game of who gets the bigger...You look at american forums, you'll see Storm is sainted, you look at european forums, you'll see that Alphacool or others european manufacturer are sainted. Who's right? Let people choose what they want...

Sure, a heated die provides great accuracy if well done, but it ends here and correlations with reality are a real pain and not so easy as some people think. It's easy to see that comparisons between data provided by several people using a die are completly disparate generally ! Who's right? Who's false? None of them again... It depends, that's the universal answer to give you. An IHS has a role in heat flux spreading and almost all actual processors use one, you can't neglect it at all. It's not because a dozen of people pop off the IHS that it's an universal case ! The TTV Swiftech use, in parallel of the BillA's die, is surely nearer from reality than a naked die. IHS acts like a spreader (hot spots are filtered and flux on the IHS top appears to be more constant and a bit less important in terms of power density) and WBs are probably less sensitive to the core dimensions below... If you change the die area or contact, you change its behaviour too because flux are not the same (correlation are complex) and we have a good demonstration here. Is it bad? No. Personnally, I would prefer a controled testbed on real instrumented parts (mobo, processor, etc.) with multiple mounting to ensure consistency like Ph, or I, did. When after months you'll try again to measure again a WB to see if you were correct, and you get same data @+/- 0.1°C you could be happy of that, no need perfs @0.01°C in a world where designs variations are great (I don't say that precision is useless for all measurements, eh). One thing is sure, the more data you'll get from various persons, the more interesting it will be, true data are essential, assumptions or extrapolations remain extrapolations. Swiftech did that using several die systems to get more data, no complaints about that, they don't invent anything.

One key is the downscaling to achieve better convection efficiency thanks to mini/micro structure for example (next generation of Apogee will down scale probably with tinier pins), Storm and others american WB generally are enormous for nothing, 50-70% of their weight/volume is useless (it costs) and we can achieve same results and better integration with WB like my EDM protos if you know them (40x40x10mm only and I could reduce them by a factor 2 if one day I continue this project) or simple WB like MP05 or similar. But external design is a major problem because who want to have a simple small rectangular WB without bling or harmonious curves ? If these esthetics considerations were abandonned by people to keep only efficiency and simplicity, costs and manufacturing processes will be simpler, because Storm needs a lot of milling work for example. Jet impingement is a good technic (the second one with microstructures) but not an ultimate solution and principally because pumps are limited. Some tricks could be made to enhance thermal transfer and Cathar tried some, but you are always dependant of the jet speed and pressure drop related... Remember too that actual WBs are near hitting a effectiveness wall because it become more and more difficult to have a better performer, convection isn't the major factor in thermal transfer now... Apogee has some advantages in comparison of Storm for many circuits (less pressure drop, no clogging, etc.), why don't let it a chance? Not all people are restricted to get an hypothetical 0.5°C less @100W you know...

A lot of time I would say that because it hurts me when I read some headstrong people and their false perceptions. Open your mind a little, it's not a bad thing...

worth quoting as there is a lot of info here

great stuff Stew, my wife was most impressed (ok me too @230k)

Quote:

Originally Posted by unregistered

great stuff Stew, my wife was most impressed (ok me too @230k)

That wasn't me on that bike (that was a GSX-R1000, mine is an R1), but it's the same sort of idea, except I don't pull standing wheelies.

sounds nice shame about the mic :( 14 thow rpm?

Quote:

Originally Posted by flatline

sounds nice shame about the mic :( 14 thow rpm?

Indeed. ~175hp or so from a 999cc naturally aspirated in-line 4cyl engine.

The upcoming 2006 R6's rev out to 17500rpm, as witnessed on this flash/animation sound-byte of a dyno run.. 127hp & 600cc.

Sorry for the short interlude while I dream of things more pleasant than this thread.

i prefer the sound of the 1st the other on rolling road sounds a little 2 cleen not a husky roar like the 1st thow this may be cos its not under full load (air drag rider grip on road and so on) or may just be desine

I can relate. :) my gixxer is about to start crying... as winter is setting in up here in the midwest... and she must be parked till spring. :(

To clarify for the popcorn munchers... (I'll play the "roadside commentator", as Cathar zooms by ;) )

While testing, a raw core-side temperature measurement will be taken.

One of the points brought up is that the TIM joint under the AMD and Intel IHS may vary in quality, over time. The result is that it becomes more difficult to apply an offset (to the raw core-side measurement), and predict what actual temperature an AMD CPU would report. It also makes using an actual AMD / Intel processor as a testing platform a bit more complicated.

Cathar's objection, as I understand it, is not that the TTV's TIM joint under the IHS may be variable, but that it is not checked for a variation. Bill's reply; it's just a (repeatable) heat source.

Orkan; we test at fixed power levels, usually on the high end (70 to 100 W). Cooling solutions perform less at different power levels. Accuracy and repeatability are prime; we're shooting high here.

Annirak; secondary losses exist, and they are very hard to quantify. Intel has parameters for the testing environment.

Otherwise the issue of having a die simulator whose size is different than an actual processor, remains. We did however agree on 10mm by 10mm and 14mm by 14 (in another thread) as it covers most processors. We can revisit yearly, that's fine, but must start somewhere (I'm so far behind!!!).

I'm all for the proposed mushroom cap; seems fine, just because the Intel TTV is hard to get (?). Can we agree on 1.4mm as IHS thickness? Can we get together on one mushroom cap? Can we still have a TIM joint of some sort? (Raw copper stock in 36mm by 36mm is not obvious, but 36mm by 1.4mm bar stock is easy) (let the popcorn flow!)

Bill's point remains: -> how is the die face to be maintained ?
Can we do nickel plating?

I am excited about the thermocouple solution, but fear the cost of EDM'ing anything to .3mm .

5 TCs in the IHS? Wouldn't be more reliable to have 5 IHS with one TC in each quadrant, then center? Either way, it would be for additional study, no direct relation to cooler testing, right?

Me looks down at his endcan-less bike... not goin' nowhere at the mo *snif*

*reaches for the MiG* Can't be having a winter without the 2 stroke... s'just not right...

Quote:

Originally Posted by unregistered

no it is improper to characterize all CPU/IHS TIM joints as being intrinsically variable to a huge extent
that some may degrade so over time and use has nothing to do with testing (don't use old shit)
that some may be inconsistantally mfgd is another separate issue (don't buy cheap shit)

For what it's worth.
http://www.xtremesystems.org/forums/...p/t-59068.html
Quote:
jess1313
08-16-2004, 03:40 PM
I had removed My IHS, but I couldnt get good contact with it off.
After this I put it back on with AS5 & gain 150mhz clock with 22c lower on idle.
This sounds like alot but I can clock way higher with less voltage now.
Mine was not making good contact with the IHS, there was a big bubble in the center of the factory clue.
Well worth it. I am on air BTW.

wow still the same thing again about the ttv?

Am I hearing from Bill that using the Intel CPU would be preferable to an AMD one for my test bed? Because the limitations of the intel platform are pretty large ones: 1C resolution, complete reliance on the motherboard temp monitoring, and no access to the "hot" diode only the cooler diagnostic one.

I would think ditching the IHS (or popping it and reaffixing it somehow?) on an AMD processor would still be preferable to the above problems?

which is what Marci was saying re AMD
- strange that he would put the IHS back on, interesting #s

anyone seeing such with 775 pkgs ? (yea, how would one know ?)

Quote:

Originally Posted by pHaestus

I would think ditching the IHS (or popping it and reaffixing it somehow?) on an AMD processor would still be preferable to the above problems?

I don't see how that will reflect a real-world situation... as we have established most people will keep their IHS on.

Well what about that then. Think one could reaffix an IHS acceptably well to make it worth putzing with? An extra TIM joint doesn't make me happy in the nether regions.

The discussion is good and it will probably prompt me to test a block w/ IHS and then pop it off for retesting.

Orkan no offense but most people don't solder wires onto their CPUs to read temperatures off the diode to 0.1C or spend 20 hours remounting and adjusting flow rates to test out a waterblock either.

If I do it how "most people" do it then we won't get any useful information.

I know you said you don't need to have a Ph.D to use logic, but education DOES help one logically design experiments (at least in my experience).

Don't think I'll do something off the wall that doesn't compare to anything that "regular people" will do, but sometimes things have to be done a bit different from "stock" to get useful info out of the system.

don't really know pH
from an individual perspective all I need is a source, I can figure out what the #s mean
from a community perspective a common platform is needed

TTVs, good or bad, are out due to nonavailability

a heat die is needed, how much similitude is to be sought ?
AND HOW WILL THE FACE BE MAINTAINED ?
this topic cannot progress (with me) w/o addressing this issue
I had offered to lap dies while with Swiftech, this is now out - what next ?
I can hand lap because I have a Flatscope (optical flat reflex viewer), do it 'till its right
what does everyone else do ?

let us NOT hear foolishness about 'good enough', this is the heart of the whole system
not flat = variable and certainly lower results

EDIT
a good (sensitive) system will begin to show degradation after 20 mounts or so on a 10x10mm soft copper die

I guess it really doesn't matter if you pop the top or not... so long as the IHS does not create a wider heat source than a cpu die.

Seems logical that it would, however. If you completely dismiss it, then no one that runs an IHS will truely know how this block is meant to perform on their system.

Quote:

Originally Posted by pHaestus

Actually, this was quite common in the northwood days, people would pop the IHS, remove the crap TIM from intel and put Ceramique or AS-5 and would have quite good results... usually only 1-2C hotter than with no IHS. Some folks would cut out the socket from an old mobo, insert the CPU to the socket then use some sort of epoxy to seal the IHS back down to the PCB. They would then put massive weight (15-20 pounds maybe) on the CPU until the epoxy cured.

I believe 'blank' IHSs can be sourced, just need a TIM joint procedure
how is the IHS edge to be supported, this is the load bearing element
the edges cannot float, eh ?

if an IHS is to be used, the CPU package must be replicated adound the heat die
(or develop a correlation, but we've been here already)

Here is a thread of a guy putting an IHS onto a Dothan... http://www.xtremesystems.org/forums/...ad.php?t=68378

{SNIP}

Bin that idea...

@Bill re edges - black silicon sealant as used in both Intel and AMD by the looksa things... how much downforce does press exhert and for how long is now the question, to be able to replicate a "true" mounting...