Apogee Test at Systemcooling

[nikhsub1]

Not good for Swiftech IMO. http://www.systemcooling.com/swiftech_apogee-01.html

[ExtraPickles]

Performance on a CPU with IHS seems OK, but that top sure is looking ridiculously thin.

-pickles

[nikhsub1]

Yeah, OK, but how do we explain it performing WORSE than a koolance block on the die sim?

[ExtraPickles]

That's a good question... In my own testing there's no difference in performance between the Apogee and Storm on an Opteron 170 with IHS on a single CPU loop only. After reading that review and seeing how thin the top is though I won't be putting the Apogee back in another system like I originally planned.

-pickles

[chinkgai]

guess people should wait for a rev.2

[Eddy_EK]

"Something awfully screwy is going around here!"

By that I mean that I don't understand, why such a big difference in simulating and real CPU testing. :confused:

[MaxxxRacer]

Looks like Lee is kinda selling out to Swiftech if you read parts of it. While he just about stated that the live cpu test was useless he went ahead and did it anyway. Kind of Silly. And he also defends the Apogee in his conclusion..


BUT his test data with the die sims confirms it.. Apogee is a product that should have NEVER seen the light of day... Swiftech just wasted 10-15k in tooling costs for that delrin top.

[snowwie]

robotech isn't selling out to swiftech, i think he would be offended by your saying that.

he is interpreting his results on his testbed, something i think he is much more qualified to do than you or nikhsub, but all the sudden you guys (or nikhsub at least) think this is your chance to jump up and cry "i told you so" and continue to claim that your suspicions of the new swiftech (TTV) setup are original and founded. billa (and he IS a definitive source of knowledge when it comes to WB testing, NO ONE has the experience he has) said to "scratch" his old data (1 cm^2 copper die), he found a nexxos xp performed poorly on his old simulator but on the new swiftech simulator it performs well. He believes the TTV better simulates a CPU. and I trust his judgement. he has no agenda to find one test setup "more convenient" than another...he just follows his knowledge and experience (and i'll repeat, which is more than anyone else).
I'm sorry to go on like that nikhsub, but your postings show a grave misunderstanding of robotech's review i think.

"but how do we explain it performing WORSE than a koolance block on the die sim?"
you don't. only robotech can explain such, as it's his data. You keep referring to this data like it's your own to interpret and understand.

[Fairydust]

I am not sure i like this Swiftech Testbed. If I understand correctly the diode is at the outer edge of the IHS? Coolers with a more uniform cooling will receive better results than coolers with core centered cooling performance. Unless the temperature is measured as close to the core as possible, it is quite useless.

[specialblend218]

I am not sure i like this Swiftech Testbed. If I understand correctly the diode is at the outer edge of the IHS? Coolers with a more uniform cooling will receive better results than coolers with core centered cooling performance. Unless the temperature is measured as close to the core as possible, it is quite useless.


That was the first I thought when they stuck the probe to the side of the IHS.

[aMp]

Another nonscientific method here: Why not see how far you can OC two or three difference chips (e.g. P4 570, FX-57, Opty 175, IHS on/off) using the same test rig but different blocks?

[MaxxxRacer]

What fairydust said.

snowwie: i understand you have a company to run here, but we all, including Intel, know that the TTV is a poor way to test CPU waterblocks. As nik as posted here, it says right in the Intel PDF that the TTV is merely for validation and should not be used for long term quantitative testing.

It is quite obvious that the die sim is the most accurate and repeatable test method for waterblocks. Dereks (pH) Athlon XP comes in a pretty close second, but as Derek will admit the Athlon XP's heat output will vary a bit during the test and from test to test. This is no fault of derek, but inherit in cpus.

[nikhsub1]

I'm sorry to go on like that nikhsub, but your postings show a grave misunderstanding of robotech's review i think.

Dood I think you are mistaking me for someone else. I understand just fine thx. :rolleyes:

[Bloody_Sorcerer]

So did I, and I feel my original conceptions are correct about this block; when BillA left, swiftech quickly took a turn for the worse.
I respect BillA's opinions on the TTV, but I disagree with using it and it alone in thermal tests. If we at least knew something about it, such as die size, I wouldn't mind as much.
What has me most concerned is not the performance, but the utter lack of any QA at all. .030 inch delrin is NOT acceptable; .030 inch copper or bronze isn't even acceptable. Copper shavings in blocks (hurried manufacturing process) is NOT acceptable. Is it just me or has swiftech taken a turn for the worse as of late?

[snowwie]

we all, including Intel, know that the TTV is a poor way to test CPU waterblocks....It is quite obvious that the die sim is the most accurate and repeatable test method for waterblocks.

if billa hadn't stopped posting here as a result of your censorship he would have issue with your statement right there. who is we? and how do YOU know? So Intel has tested the TTV for use as a vehicle in performance testing on waterblocks?

nikhsub, anyone who reads the review and then claims a koolance block, or even the mcw-6000 will provide better performance on today's processors than the apogee is misunderstanding robotech.

[Bloody_Sorcerer]

Intel claims in their docs that the TTV should NOT be used for reviewing purposes and should ONLY be used to see whether or not a HEATSINK would have a snowball's chance in hell of keeping a CPU cool enough to be stable or not. They then go on to say that HEATSINKs should also be VERIFIED with OTHER heatsources. NOWHERE do they say that it should even be considered for high-accuracy numbers, especially because we have utterly no idea about the TTV; tell us its die size and we might :banana::banana::banana::banana::banana: a bit less... might.

[snowwie]

Intel claims in their docs that the TTV should NOT be used for reviewing purposes and should ONLY be used to see whether or not a HEATSINK would have a snowball's chance in hell of keeping a CPU cool enough to be stable or not. They then go on to say that HEATSINKs should also be VERIFIED with OTHER heatsources. NOWHERE do they say that it should even be considered for high-accuracy numbers, especially because we have utterly no idea about the TTV; tell us its die size and we might :banana::banana::banana::banana::banana: a bit less... might.
do they? or is it possible that they only say that the TTV should only be used for heatsink validating purposes only and so haven't been qualified for testing? so what? this has been argued before already at procooling...where all of you have been present, including me, i read it too. so tell me something new or something i don't clearly understand and needs to be restated to me. I'll restate something: BillA has had experience with any kind of simulation for cpu testing that surpasses anyone else in watercooling. He believes the TTV is a good simulator (produces accurate results that can be understood by the tester) for waterblock testing. there are many kinds of simulators, all of which can produce good results. Understanding them and their relation to its purpose (thermal performance on modern processors) is the responsibility of the person who built the test bed. I don't think Robotech did a great job in his review. His explanation/conclusion seems to simply invalidate his testing and emphasize the CPU testing, which is a lot less than we would expect from him.

[Bloody_Sorcerer]

We all know where we stand on this, and no one is changing their opinions any time soon. Let's avoid yet another drawn out debate on this; we've had more than enough.

[tw33ter]

The only thing that concerns me about the apogee, is this (http://www.systemcooling.com/images/reviews/LiquidCooling/Swiftech_Apogee/image17big.jpg). It seems like there will be a few reports coming in of their apogee's breaking right there.

[AndrewZorn]

i feel like the kind of person who would break it just attaching the tubes, that is like mcdonalds thin

[Hu1kamania]

i feel like the kind of person who would break it just attaching the tubes, that is like mcdonalds thin

HaHa i hear that! :lol:

[nikhsub1]

The only thing that concerns me about the apogee, is this (http://www.systemcooling.com/images/reviews/LiquidCooling/Swiftech_Apogee/image17big.jpg). It seems like there will be a few reports coming in of their apogee's breaking right there.
IN all fairness to Swiftech, they have investigated and found no issue... this so far is from only one user IIRC.

[Hu1kamania]

IN all fairness to Swiftech, they have investigated and found no issue... this so far is from only one user IIRC.
Orkan mentioned his cracking in the same spot as the picture in the review.

[mcbarnet007]

Personally I think that the real life test results are about the same is because of bad IHS and core contact. If both water blocks effectively cool the IHS to close to ambient water temperature then the only variable is between the IHS and the Core. As shown in the review the Apogee performs really well for low heat density (large area). Therefore, cpu with IHS will see about the same temperature between Apogee and Storm.

However if you pop the IHS you will see a big difference between the two blocks on small die. For dual core opterons without IHS the result will be similar to the Large die similation.

From personal experience removing IHS (I've removed at least 5 IHS) will give you a 8 to 10 C drop in temperature. And this is only from a TDX. I am able to cool my opty 170 at 2.8ghz 1.43v to 30 C loaded. Ambient around 20 C. I use the #5 nozzel which has the highest pressure drop similar to the Storm.

So from the review my conclusion is if you are going to use IHS on your cpu, get the Apogee after they fix the molded top.

If you are going to remove IHS then by all means get the storm because it is very good at removing concentrated heat source.

I'll be installling the Storm on my opty 170 after finals and I'll report back my result.

[Plywood99]

It is quite obvious that the die sim is the most accurate and repeatable test method for waterblocks. Dereks (pH) Athlon XP comes in a pretty close second, but as Derek will admit the Athlon XP's heat output will vary a bit during the test and from test to test. This is no fault of derek, but inherit in cpus.


Agreed. Testing with ihs / tim joint is a mess. Bare die or die sim is best that can be done at the moment...

[lv_dicedealer]

Not bad for a first test/review...

The MAJOR thing that stands out to me is that the die sim data shows the Apogee performing one way, then the system test shows it performing another way....
Does anyone else get the impression that Swiftec in their own way are attempting to discredit BIllA's testing methodology (which is the basis for the majority of the reliable testbeds out there)?
I smell sour grapes and they aren't making chardonay, they are making whine......

[nikhsub1]

What does bill have to do with the testing?? He didnt supply the test data for the apogee.

[Gimmpy224]

god we are letting this get personal now, people are taking posts as someone attacking them now 0.o

these are called opinions for a reason 0.0


one thing I dont understand is why it matters what temp you get when testing, as long as the conditions can be repeated wouldnt the proportions be the same?
I guess I can kind of see how the diode being stuck to the side would maybe effect it.

[lv_dicedealer]

What does bill have to do with the testing?? He didnt supply the test data for the apogee.

You are correct that BillA didn't supply the data, however if I recall he was employed at Swiftec prior to the Apogee being released. This does not mean that he tested or even saw this design before leaving the company or for that fact that Swiftec tested it according to BIlla's methods.
He was also one of the early leaders in the watercooling arena who did start showing the general public the flaws which were involved in some of the testers methodolgies. His early tests and testbed became the basis for future tests and testbeds.
What I am trying to say (not very successfully) is that perhaps while at Swiftec BIlla had their R&D conducting testing one way, then when he left; someone there said let's test another way leading to those end results not agreeing at all with results derived from Billa's methods. All this leads to Swiftec trying to 'deprogram' the watercooling community in regards to testing methodology (hence the last test in systemcooling's article being new, as stated in the article).
Swiftec is basically saying 'Billa's tests are rubbish, here is how you should test and we have now got a part to show the world that this is the way it should be....'
I am in no way saying that anything is personal, nor am I saying that Billa's testing is rubbish, nor am I saying that Swiftec is right or wrong; just an idea.....

[Petra]

I'm curious as to whether Swiftech will continue to produce Cathar's Storm or not.... My guess would be that they won't and mainly for financial reasons. Let's just look at the situation for a moment...

-they stopped production of the MCW20
-they replaced the MCW50 with the MCW55
-iirc, it was mentioned in that review that they are discontinuing the MCW600x series
-they will likely discontinue the Storm
-all CPU blocks are replaced by the Apogee
-all blocks are now Delrin topped

The Storm is a relatively expensive block to produce due to the amount of machining time involved and its overall complexity (now, I'm not an expert but I'd tend to think that, at the very least, the Storm's very precise nozzle array would take a decent amount of machining time...even if using Delrin)... also, since it isn't Swiftech's design, they undoubtedly have some sort of licensing fee associated with its production. As a result of all of this, Swiftech's margin on the blocks likely wasn't good enough considering their limited production capacity and inability to fulfill their reseller's orders (let alone do so in a reasonable timeframe).

*EDIT* Bah! Sorry... I just noticed a difference which invalidates a lot of what I was saying below... the Apogee has a much higher pin density than the MCW55. :doh: :slap: (hey, what can you expect from a guy at 3:00am?)

So... What does Swiftech do to address this? Simple, really.... sell only one block with two different tops and a few minor alterations. After all, let's face it... the Apogee is an MCW55 with a few minor alterations: flat base vs. raised contact patch, different top, missing two notches on the sides, and missing that little groove in the 'diamond matrix'. Aside from those differences, the bases are the same part.
*please read above edit regarding quoted text block*

Anyway, I think what I'm getting at is that it appears that Swiftech has shifted from Performance & Quality to Low Production Cost & Volume. Is it just me, or does this not sound like a good change? :hm:

As a side note, I'm kinda interested to see actual test information as to how all three Swiftech blocks would compare when overclocking on a bare-die A64... I'd need an Apogee block to do that, though, since I already have the other two... hmm... perhaps later this month :idea:

...I really hate it when I shoot down part of my own conspiracy theory... even though there isn't really any "conspiracy" here... lol :rolleyes:

...maybe I should just go back to making watercooled cardboard box computers ;)

[Marci]

So Intel has tested the TTV for use as a vehicle in performance testing on waterblocks?

Yes they have. They also have all the correlative data to convert TTV results to produce predicted die results for every make and model of CPU from themselves and competitors, including heatflux variances. However, Intel don't and won't share that info. To do so benefits the competition too much. Hence you won't see any documented proof of such. TTV is fine to use provided we / the tester has all the info needed to produce the results we want. We / the tester don't. Simple as. TTV Documentation harps on about TTV-to-CPU correction / conversion formulae... but don't provide such formulae...

[Budwise]

i just ordered an Apex kit with the old Storm in it before they can no longer be found... Anyone ever order from voyeurmods? They seem ok but so far kinda slow...

[moonlightcheese]

i find it funny that a lot of you chide BillA for using the TTV but it's perfectly fine for Lee to use a die sim for conclusive results. it really is fortunate that he included actual CPU temps in this review and we can clearly see now that sims are an absolutely unacceptable way to test waterblocks.

it seems like the concensus of the senior members here on XS was perfectly on target. the Apogee performs only slightly worse than the Storm with the IHS included. my original theory was that the Apogee would perform better with the IHS removed, but it seems pretty obvious that it would not since the Apogee is geared more towards large SA heat removal and the Storm is heavily concentrated on cooling the die area alone. only testing will truly show this however.

it's true that the build quality of the Apogee is far less than acceptable. perhaps the Apogee can withstand hammering, but i'd hardly consider that conclusive testing. the problem is, with engineering like this, you are leaving too much room for error. it's very easy to have a small machining error and have a block pass QC (by the way, 40psi is a joke for testing) and fail when an end user receives it. i've seen it when i worked at Panasonic. leaving too much room for error, you are asking for it.

pH's tests will prove to be much more conclusive and i'm eagerly awaiting those results which will truly lay the topic of performance to rest, since we all know the quality is nowhere near that of the storm.

[mcbarnet007]

i find it funny that a lot of you chide BillA for using the TTV but it's perfectly fine for Lee to use a die sim for conclusive results. it really is fortunate that he included actual CPU temps in this review and we can clearly see now that sims are an absolutely unacceptable way to test waterblocks.

it seems like the concensus of the senior members here on XS was perfectly on target. the Apogee performs only slightly worse than the Storm with the IHS included. my original theory was that the Apogee would perform better with the IHS removed, but it seems pretty obvious that it would not since the Apogee is geared more towards large SA heat removal and the Storm is heavily concentrated on cooling the die area alone. only testing will truly show this however.

it's true that the build quality of the Apogee is far less than acceptable. perhaps the Apogee can withstand hammering, but i'd hardly consider that conclusive testing. the problem is, with engineering like this, you are leaving too much room for error. it's very easy to have a small machining error and have a block pass QC (by the way, 40psi is a joke for testing) and fail when an end user receives it. i've seen it when i worked at Panasonic. leaving too much room for error, you are asking for it.

pH's tests will prove to be much more conclusive and i'm eagerly awaiting those results which will truly lay the topic of performance to rest, since we all know the quality is nowhere near that of the storm.

The die sim is suppose to simulate bare core cpu without IHS. So if are you going to get a water block to cool a venice or a single core without the IHS then you better get the storm because Apogee sux with small area. For dual cores 2x 1mb cach cpu with IHS removed the storm will still come out ahead, just not by much.

[snowwie]

we can clearly see now that sims are an absolutely unacceptable way to test waterblocks.
????
c'mon now, as long as the reviewer understands the results that a simulator yields then a sim gives an infinitely better picture of waterblocks' performance than the "alternative" cpu testing....the cpu testing robotech did was only useful because he had sim testing to go with it. imagine he had just posted the review with only the "outside-the-IHS-temp" cpu testing and based his conclusions on that.

[MaxxxRacer]

Everyone seems to be forgetting something though. According to Robotechs tests the MCW6002 outperforms the Apogee in the large die tests at ANY flowrate...

Oh and yes.. the MCW6002 is more durable and the same price... So why in the name of anything would you want to get the apogee?? Because it looks pretty?? If so then I will sell you a picture of niki cox to cool your cpu with..

[Plywood99]

i find it funny that a lot of you chide BillA for using the TTV but it's perfectly fine for Lee to use a die sim for conclusive results. it really is fortunate that he included actual CPU temps in this review and we can clearly see now that sims are an absolutely unacceptable way to test waterblocks.

100% disagree. Die sim or bare die cpu is a much better way to test block performance. ihs / tim joint adds too many variables that can not be reliably accounted for. Why do you think that Apogee, Storm and 600x series scored so closely on the cpu test bench?

Robotech is currently testing the old Maze 3 or 4 on the cpu test bench. I think it will be a real eye opener...

[snowwie]

MR, you seems to be forgetting something though. The comparative data from robotech's die sim alone does not explicitly represent how performance will translate to any one wc setup.

http://www.hardforum.com/showthread.php?t=985553

he swapped out his mcw-6000 for an apogee and reports better performance.
maybe his setup is very unique, maybe he is misguided, but i doubt he is the only one

[Plywood99]

MR, you seems to be forgetting something though. The comparative data from robotech's die sim alone does not explicitly represent how performance will translate to any one wc setup.

http://www.hardforum.com/showthread.php?t=985553

he swapped out his mcw-6000 for an apogee and reports better performance.
maybe his setup is very unique, maybe he is misguided, but i doubt he is the only one

Or maybe he had a bad mount on the 6000. Or maybe it is wishful thinking...

[ExtraPickles]

Or maybe he had a bad mount on the 6000. Or maybe it is wishful thinking...Or just maybe the Apogee performed better for him. Swapping out the Apogee for a Storm (not a 6000) did nothing nada zip for my temps on an Opteron 170 with IHS. Granted the quality of the Apogee leaves a lot to be desired but the performance was there with my setup.

-pickles

[Plywood99]

Which is why I repeat what I said in post 38, "Die sim or bare die cpu is a much better way to test block performance. ihs / tim joint adds too many variables that can not be reliably accounted for."

[MaxxxRacer]

snowwie, I realize that Robo's results will not translate perfectly to every system on the planet, but they do show the general performance trends.

And to be frank, Robos smaller Die sim has a surface area of 196mm^2.. Guess how many square mm a X2 4800 is.. 199mm.. thats REALLY close if u ask me. Now Robos large die sim is 1024mm^2. That is 5.15 times larger than the X2 4800 core.. So unless AMD is releasing a 10 core cpu in the next few months the large die tests that robo did are not very useful.

Now you are going to say that they are useful becuase they represent how a cpu would perform with an IHS and that as we all know is just rubbish. In a die sim the entire die is heated relatively equally while in a cpu with an IHS the center of the IHS (where the cpu core is ) is transfering atleast 2/3 of the heat while the area outside of where the cpu core would lie transfers a considerably smaller amount of heat energy. In reality an IHS effects the waterblock testing merely by making the base of the waterblock thicker.

If you attached a 2mm thick plate (of copper) to the bottom of the Apogee, storm and MCW6002, and then placed it on a bare cpu (no IHS), you would see nearly the exact same results as if you just had the IHS on.

there is absolutley nothing special about the IHS. it is merely a hunk of nickel (i think its nickel) plated copper.



And now to get to the [H] member,

Anyone notice that he is using the onboard sensors to measure the temp.. And we all know the onboard sensors are good for nothing..

The good for nothings sensors are also displayed in robs testing as the cpu temp never varried (36C) in the stock speeed batch independent of the waterblock attached. And in the OC testing the temp only moved 1C and was in favor of the Storm. So going from that info the 6002 storm and apogee are equal across the board.. And we all know that isnt true.

So please do not EVER try to base ANY cooling performance data off of the onboard sensors. They are off by rediculous levels and only provide a vague estimate of the actual cpu temperature.

[chinkgai]

dude, how u gona sell a damn pic of niki? you email me that pic for FREE.

NOW.

[MaxxxRacer]

here you go :D

Niki Cox was here.

Picture removed to save keyboards around the world from puddles of drool.

[Bloody_Sorcerer]

anyone who DARES to trust sensor readings can be slapped with a trout. I have a screenshot of my video ram at -40C with aircooling. with an ambient temp around 25C. onboard sensors don't even have repeatability so don't give me that crap.
maxxx: yummy pictures ;)

[Fairydust]

OMG women like this exist? I have to seriously rethink my priorities.

[lv_dicedealer]

I looked at those pics for about 5 minutes, then realized there was a head attached... Dern, she's cute :slobber:

[MaxxxRacer]

lol @ you guys.. you need to get out more.

I shouldnt have posted that picture. I had a feeling that things would get off topic with that picture..

BUT now i know how to derail any thread.. just post that... and bam! FX60 to niki cox in 3 seconds flat.

[chinkgai]

*fap fap fap*

huh did you say something max?

:banana:

[moonlightcheese]

And to be frank, Robos smaller Die sim has a surface area of 196mm^2.. Guess how many square mm a X2 4800 is.. 199mm.. thats REALLY close if u ask me. Now Robos large die sim is 1024mm^2. That is 5.15 times larger than the X2 4800 core.. So unless AMD is releasing a 10 core cpu in the next few months the large die tests that robo did are not very useful.
i'm not arguing that point at all, they are very close to the same thing. i've already said this at least three times on several forums... a die sim is not a CPU. the results are NOT the same. i don't know many people who run die sims in their cases. in fact i don't know any. and to rebute.... the large die tests ARE useful. they are supposed to emulate an IHS.

yes those squares are very close in size... but the large die sim test (1024mm^2) is pretty damn close to the size of an IHS. so how do you explain the temps from the large sim test varying so much from actual CPU temps? well obviously it's because heat is not generated uniformly on an IHS as it is on a sim and therefore heat is dissipated more easily on an IHS as opposed to a sim.

by this same argument, temps on an actual die are not uniform either, as they are on a die sim. different parts of the die produce more stress and others produce less. this is why the sim tests vary so much from real world tests and why die sims are unacceptable as a source of conslusive results. this is why procooling is looking at testing with live CPUs with and without IHS as opposed to the quick and dirty die sim tests.

it's true that die sim tests are close to what you'd expect to real world results but, as you can see from the results in Lee's test here, "close" is not good enough. i'm an avid supporter of live tests and i'll continue to be one until someone produces a die sim that perfectly produces the exact same characteristics as a CPU die and produces the same results in testing. until that day, die sims are banned in my book.

[Gimmpy224]

psh I already got my 10 core cpu on backorder!

[MaxxxRacer]

Moonlight, I'm not disagreeing that a live cpu test, if ALL the poitns of error could be removed wouldnt be a great test, but that isnt really feasable. So die sims will be the more repeatable and accurate of the two tests.


so how do you explain the temps from the large sim test varying so much from actual CPU temps? well obviously it's because heat is not generated uniformly on an IHS as it is on a sim and therefore heat is dissipated more easily on an IHS as opposed to a sim.

You are entirely incorrect. you say heat is more reasily disipated from the IHS rather than the die sim.. This is NOT correct. Comparing the large die sim which heats up the entire die (IHS) rather uniformily, to an IHS which has a hotspot in the middle and is rather cool near the edges is a horrible comparison. Even though the IHS is large the heat is still only coming from the center, where as the heat source for the die sim is burried a bit lower so the heat transfers itself more uniformily over the entire surface of the die sim, which as any chemistry, physics, or engineering major will tell you will facilitate better cooling.

It really boils down to this. Heated surface area. Large die sim has a LARGE heated surface area, and a cpu with an IHS has a SMALL heated surface area.

To really display this you can put ur finger on the IHS of your cpu when its on and it will feel cool. infact the IHS On my opteron at 2.9ghz is icy cool even after hours of full load.

[moonlightcheese]

You are entirely incorrect. you say heat is more reasily disipated from the IHS rather than the die sim.. This is NOT correct. Comparing the large die sim which heats up the entire die (IHS) rather uniformily, to an IHS which has a hotspot in the middle and is rather cool near the edges is a horrible comparison. Even though the IHS is large the heat is still only coming from the center, where as the heat source for the die sim is burried a bit lower so the heat transfers itself more uniformily over the entire surface of the die sim, which as any chemistry, physics, or engineering major will tell you will facilitate better cooling.
this is exactly what i was trying to say. i think you just misread what i typed. i think the point of that large sim was to compare to the live cpu test in Lee's review. and i'm agreeing with you entirely on the fact that they are not the same whatsoever. this is why i'm saying sims are not satisfactory in testing. i don't care how easy it is to repeat results with a die sim, i don't run die sims in my computer and i don't expect my temps to resemble anything like a die sim and neither should anyone else.

and you'll have to explain to me... why you say heat is more easily dissipated from a die sim when the Apogee performs several C better on an IHS than on a die sim...

from the testing, the storm and MCW6002 show no change in dT when going from large sim to IHS but the apogee performs marginally better. so how is that statement you just made justified?

EDIT: woops. looked at that test data again and it looks like the storm performs worse on IHS where the apogee improves. but it still begs the question.

[Mr. Tinker]

After looking at the charts in the review, I would like to send My congratulations to Cathar, for what it's worth. You win.

[MaxxxRacer]

OOO i thought you were saying the large die test and IHS testing were comparable. Sorry for the confusion.

Well here is the thing about the IHS testing. As lee pointed out, measuring the side of the IHS is just short of useless as a block that cools the edges of the IHS will perform better than a block that cools the cpu core and not the IHS edges. Now you will notice something about what i said. Storm was designed to concentrate cooling on core area while apogee is designed to cool the entire IHS more evenly.

so obviously when measuring the temp on the side of the IHS the apogee will improve compared to the small die sim (equiv to 2x1mb cache X2).

the MCW is inbetween the two extremes of the apogee and storm. It centers its cooling in the center but uses the extra pins all over the block to "secondary cooling". so thats why it seems to not change.

but again you MUST understand that mesuring the side of the IHS temp gives only a weak correlation to the actual core temp. Looking at a curve of the IHS temp compared to actual core temp with different waterblocks will not give you consistent results. meaning that one block with a low IHS temp might have a high core temp and vise versa.


you also mentioned how the cpu core is not heating up perfectly even and that it has hotspots. While this is absolutely correct, it greatly varries from cpu to cpu. IE northwood, venice, toledo, preschott, conroe, dothan, yohan. All of their hotspots are in different places. So the logic of using an actual core is not very good as the tester may use a preshott but you have a toledo. guess what, the block will act noteably different on those two cores.

So a die sim is a happy medium between all of those cores.


and you'll have to explain to me... why you say heat is more easily dissipated from a die sim when the Apogee performs several C better on an IHS than on a die sim...

I hope with the above statements I explained your question.

[moonlightcheese]

Well here is the thing about the IHS testing. As lee pointed out, measuring the side of the IHS is just short of useless as a block that cools the edges of the IHS will perform better than a block that cools the cpu core and not the IHS edges. Now you will notice something about what i said. Storm was designed to concentrate cooling on core area while apogee is designed to cool the entire IHS more evenly.
Ah you are right on the money there. i can't believe i didn't consider that... :slap: so with that info it looks like the large die sim was almost right on (as far as it can be anyhow). i just noticed that he was using that Barrant thermometer on the side of the IHS. i thought he was using mobo sensors...

thanks d00d.

[MaxxxRacer]

np.

in actuality the best way to test temps with is to slap an IHS ontop of the small core die sim. I will contact Lee and offer to donate an IHS from my 3000 venice.

Then once u have it on there, drill a small hole in the side of the IHS so that lee can stick his thermal probe in there without any issues. do that and we will have MUCH more realistic and usable data for IHS testing.

[moonlightcheese]

np.

in actuality the best way to test temps with is to slap an IHS ontop of the small core die sim. I will contact Lee and offer to donate an IHS from my 3000 venice.

Then once u have it on there, drill a small hole in the side of the IHS so that lee can stick his thermal probe in there without any issues. do that and we will have MUCH more realistic and usable data for IHS testing.
PERFECT!

good suggestion dude. that would be much more accurate than the large 32mm test. naked core testing would be nice to see along with that as well (live test) just for the sake of comparison.

[nikhsub1]

Guys, you need to understand that the AMD IHS 'numbs' the cooling affect of almost all blocks. This is the reason why in Lee's CPU test all 3 blocks are nearly identical (he did read diode of CPU for this test IIRC AND the probe on the IHS). What does this mean to you? Well, If you run with an IHS, anything from a Spiral to a Storm will yeild you similar results as far as temp monitoring goes. Say what then? A Spiral as good as a Storm? Well, for all of you that argue the only valid testing is WITH an IHS, then your answer is YES. Go get a Maze 3 or 4, it will perform just the same as an Apogee or a Storm. Now we move to a different aspect, the max OC. IMO the best chance to acheive the highest overclock would be with the WB that performs best on a DIE SIM. Afterall, you want to cool the core, the core is under the IHS and a more focused block 'should' allow higher OC's than blocks like the Maze 4 or the Apogee. The IHS in testing introduces an uncontrollable variable which, even that variable is variable. The IHS > TIM will be totally different from CPU to CPU, why mess with that in testing?

ricecrisipi on procooling made a good analogy: "It's like racing on a track with potholes or a bad surface. Just because a car can handle the potholes better doesn't mean it's a faster car. It's just a a car that run well with tracks with potholes."

[Cathar]

Reading around on various forums I see a lot of rhetoric revolving around "real world" vs "artificial" performance. I also see some incorrect assumptions and parallels being drawn. Please allow me to explain the reasons why die simulators are better.

First, let's start with the issues with "real world" CPU's, and just why when we stick different blocks on a CPU, and then we see them all read the same. As we all know the temperatures that get reported by the CPU are being read from thermal probes often located in the coolest portion of the CPU die. A document from Intel details this fact: here (http://developer.intel.com/update/departments/initech/it04021.pdf) (sadly the document has been removed - rehosted here (http://www.employees.org/~slf/it04021.pdf)). While I have no link to show AMD is doing the same thing, if one does enough testing we can observe a similar pattern of behavior. The location of the diode means that often what it reports is totally "numb" to what's really going on in the hottest portions of the CPU. It works much like this. If you have a long rod of metal, and you heat it at one end with a flame, and then run cold water over it in the middle, and then measure the temperature at the other end, how representative is that of how hot the flame enveloped end of the rod is getting? This is precisely what occurs when you stick effective water-cooling on top of CPU's. The CPU may be getting hot, and waterblock are pretty good at removing that heat, so by the time the heat gets to where the user readable thermal probe is located, you're seeing a "numbed" picture of what's going on in the hottest sections.

Incidentally, as stated in that paper, Intel have a second TCC probe that control critical thermal shutdown of the CPU if it gets too hot, and that probe is located in the hottest section of the CPU, but the users can never see what it reports. However, we can get an idea of what's going on. Intel states that the TCC probe is specifically calibrated set to thermally shut-down the CPU when it reads 135C. XBitLabs conducted an experiment, here (http://www.xbitlabs.com/articles/cpu/print/p4-temp.html), which demonstrates the differences between what the user sees via the cool thermal probe, vs what the calibrated TCC probe is doing by slowing/shutting down the CPU at calibrated internal temperatures. This experiment by XBitLabs demonstrates the vast differences between what CPU's report to the user, vs what is actually going on.

Okay, so now that we understand why different waterblocks can report the same temperature on the same CPU, now let's understand why we want to use die sims, and more specifically, bare-die sim (no IHS involved).

When developing waterblock designs, the designer focuses on extracting the maximum possible thermal transfer rate between the metal of the waterblock, and the water flowing through the block. This transfer rate is commonly referred to as the "effective thermal convection transfer efficiency", and is denoted by the letter h in many engineering texts. Its units are W/m²K, or watts of energy, per unit area and degree kelvin (celcius) rise.

By using die sims of known size, and known even thermal output spread and applied to the base of the waterblock, and by calculating the temperature rise of that die sim we can arrive at a fairly confident approximative value for h. The higher the value of h, the more effective the waterblock is at transferring heat into the water flowing through it. When it comes to designing waterblocks for pure performance only, h is about the ONLY thing that matters. Get h up high, then tweak the base-plate thickness of the waterblock to suit your value of h, and you have your high performance waterblock. This is a simplistic description of the design process, but in a nutshell, that's what a waterblock designer/engineer does.

So how does a high h value help? It tells us how well that the waterblock will transfer heat into the water, regardless of what's making that heat, and whether or not an IHS is involved or not.

Where IHS's come into their own little world of confusion is this. The waterblock may be doing a fantastic job of transferring heat into the water, but the IHS may not be making even contact between the CPU die and/or the waterblock. How many times have we all seen people with very unflat IHS's? Lots. Almost every single CPU has a non-flat IHS because it is a mechanical joint. Variations in the glue that holds the IHS against the CPU packaging, the manufacturing process and the way that the IHS is formed, and cools, thereby causing warping, all adds up to a piece of metal that sits between your CPU and your waterblock that isn't likely to be contacting both evenly.

This is where testing with IHS's, even on die-sims, becomes an issue. Because the IHS can never be guaranteed to be sitting evenly, it will then be applying heat load to the waterblock in an uneven fashion. As you can imagine, trying to determine what a waterblock's h value is, when there's no guarantee that the surface area that the heat is being applied to is being spread evenly is a total nightmare. This is the sort of stuff that causes people to think that they have a fantastic performing waterblock, especially if they measure only the IHS temperature, and not what's going on at the CPU die level. If you're measuring what's going on at the CPU die level by using a real CPU's on-die diode, then read above. It's also telling you next to nothing.

The thing here is that a good waterblock will maximise the convectional transfer co-efficient where the heat is at its strongest. Then it doesn't really matter whether or not people use an IHS, if the waterblock can soak up the heat well, then even though the IHS may be warped you can be sure that your CPU is still be kept cool. Ideally we don't want the IHS there because it can flex, cooling only the edges of your CPU, or cooling only the center and not the edges, and then not transfer the heat evenly to the waterblock either. What is the best waterblock for gnarly IHS's is the subject of another debate, but I can tell you now that IHS's introduce such variability that they can make one block look crap, and another look good. Put the blocks onto a different CPU and the stories can change. That's not consistent enough around which to judge or even design waterblocks with. That's just rolling the dice.

Designers use bare-die's because it is good science, and because it provides the clearest and most succinct indication of how well the waterblock design is performing and how well it is transferring heat into the water. The use of IHS's, even in testing, and the reliance upon "real world" testing with "numb" thermal probes has heavily clouded this very important point, and that's purely because there are no guarantees with IHS's. They're random. You can form any conclusion you like with IHS's, because every tester will likely come up with a different conclusion on a different CPU. Does that serve the public?

In the end, it's up to the individual to understand the difference between marketing and good science. I hope that some of the above provides further information for people to chew on when considering how relevent "real world" and "IHS bound" tests are.

[ExtraPickles]

Thanks for taking the time to explain that, it cleared it up for me completely.

-pickles

[mad mikee]

:clap: :toast: :clap: :toast: :clap: :toast:

That is why I just bought a backup storm G4 b4 they go away.
Goes w/ my other G4 (wifeputer) and mine (G5) :D

And why I remove my IHS AND lap the CPU. All these go together to give me the best cooling for ME :D I try to remove as many problems as I can that Da Boss just mentioned (and the convex AMD CPU, which is never mentioned :shrug: ). I am lucky enough to be able to do these things.

For those others who cannot for whatever reason, just remember Cathar is like a sportcar designer: fine engineering of principles that he has spent years confirming by experience. None of the products that he has had a hand in designing or implementing have been anything other than EXCEPTIONAL. You can spend less money, and get Okay performance. But for the best, so far no one has beaten Cathar to the punch for that which he does :fact: (not to mention he can explain the relevant points SO well :D )

Now back to our regularly scheduled debate on the finer points of IHS (imaginary heat spreader) simulator testing....

[MaxxxRacer]

mike I read your post about your cpu lapping.

The first thing I though of was homebrew telescope manufacturing, and sure enough the next post u linked you mentioned you used to be into making ur own lenses.. In any case, If i ever do remove my IHS again (prolly wont) I will have to try that.

[chinkgai]

:worship: :worship: :worship:

[Budwise]

i just ordered the Apex kit from Voyeurmods and its supposed to be the non Ultra one. I called em up after ordering and they said it was being drop shipped from Swiftech. Im gonna be pissed if it gets here an i see an Apogee inside rather than a Storm. I told her on the phone that i ordered it because it is advertised as the Extreme Edition with the Storm not the Ultra. She said if thats what it said then thats what should come. Im skeptical...

[MaxxxRacer]

Thanks for the post Stew.

I do have an idea though.

Using the same method that telescope lenses are made (and what mad mike did) lap the die sim with the bottom of an IHS (the part that will be touching the IHS). thus the two surfaces will have the exact same concavity (be it convex of concave) and make excelent contact. Next is to lap the top of the IHS with as flat of a piece of glass as possible. The best way would be to use one of the granite blocks used for measuring surface flatness.

Next drill a hole in the side of the IHS so that a thermal probe can be attached to the core of the die sim.

Attach the thermal probe. to the die sim and then attach the IHS to the core of the die sim with some high quality thermal paste.

I believe that if these steps are taken that a good comparison can be made with how an IHS performs with a given waterblock compared to the NON IHS die sim.

While I realize that the IHS merely 'numbs' the waterblocks cooling (and I stated this in not so many words), it would be interesting and useful data to see how the numbness effects each waterblock.

I have a prediction that different waterblocks will be effected differently by presence of the IHS. In mathematical terms, if a cooling effeciency equation was made for these waterblocks with a numbness coeffecient (for the IHS), the coeffecient would be different for each block, all things being equal.




With the method I proposed above the problem of the temp of the outer edge of the IHS being cooler would not effect the test results as the temp is still taken from the core, and NOT the side of the IHS.

[MaxxxRacer]

yes but the waterblock will act a little to marginally different with the IHS. As i stated before the IHS is like adding a thicker base to the waterblock, and as we all (or atleast some of us) know, a thicker base does indeed effect waterblocks differently.

[gabe]

Reading around on various forums I see a lot of rhetoric revolving around "real world" vs "artificial" performance. I also see some incorrect assumptions and parallels being drawn. Please allow me to explain the reasons why die simulators are better.

[QUOTE=Cathar]Reading around on various forums I see a lot of rhetoric revolving around "real world" vs "artificial" performance. I also see some incorrect assumptions and parallels being drawn. Please allow me to explain the reasons why die simulators are better...


Well Stew, allow me to explain why I disagree with you.

First off, die simulators are not "better". They are simply another tool. I would also argue that good science surrounds itself with as many tools as possible to achieve a goal. In the case of the Apogee, the goal was a feature set consisting in performance (thermal and hydraulic), user convenience, high manufacturability, and obviously cost.

We use Intel TTV's because they are an industry standard, and the absolutely closest thing to a real CPU that we can put our hands on.

Stew, this simple truth cannot be denied: TTV's are specifically designed by the largest Industry player to emulate the behavior of their microprocessors for thermal design and testing purposes. Rejecting such tool in the course of the design process would be entirely foolish. Entirely basing the design on such tool would be equally foolish. Our design proved itself not on the TTV, not on the die-sim, and not on a CPU. It proved itself on the CFD. Then, and only then, we verified the data on test benches.

On the other hand, reporting the data generated by this tool, which I say again is an industry standard, is far more legitimate IMO than reporting date generated on a home-made die. We have been reporting TTV generated data for close to 2 years now and no one has ever complained until now.

Gabe

[MaxxxRacer]

Gabe,

CFD?

EDIT: Nvm.. google is your friend.

While fluid dynamics software is all well and good, one cannot rely on it to be the end all for thermal transfer.

IF the design of the Apogee was indeed designed in this manner, Lee (and hopefully Derek, if he ever does testing again) proved that the software isnt always right. As computer junkies we woud all like to think the computer is always right, but its only as good as the equations we put into it.

A queston for you Gabe. One of the earlier posts in this thread states that Intel says the TTV is ok for repeat long term testing as long as the proper adjustment calculations are done to the results. They dont state what those are for whatever reason, but does Swiftech have those vital equations? If not, according to Intel Swiftechs TTV testing is just about useless.

[Cathar]

Gabe, all I know is this: Attempting to quantify waterblock performance using a "numb" thermal die and/or a variable laden interface is nigh on impossible. Let me know if you don't agree with that.

If you don't agree with that, then we have a fundamental difference of opinion on waterblock performance assessment methodology.

BTW, I mentioned nothing about the TTV above. I mentioned nothing of Swiftech's testing above. I was referring to Robotech's review (the subject of this thread), and various people calling for testing on "real world" testbeds.

As for the TTV data for 2 years now. To date I still have not seen a single independent testbed ratify the apparant closeness of the MCW6000 vs the Storm (Procooling's being the closest, but on a very small CPU die which heavily favors thicker base-plates). Until the Apogee came along, Storm and MCW6000 were the only two public pieces of TTV derived CPU test data available. Of course no one complained though. Swiftech stated that the MCW6000 was a little bit worse than the Storm, independent test data agreed, the scale of which was in disagreement, but the ranking was not, so most people let it slide. With the Apogee though, now we have an anomoly that no one can ratify. Of course there's going to be complaints now, and not before.



this simple truth cannot be denied: TTV's are specifically designed by the largest Industry player to emulate the behavior of their microprocessors for thermal design and testing purposes.


I don't care who designed it, especially if it is being used for purposes for which it cannot possibly hope to perform as. The fact remains that it is a testbed that introduces unaccountable and unquantified thermal interface variations, and these are inherent in its very design. The TTV is great for what Intel say it is designed for, bulk, gross, high level pass/fail validation of thermal mechanisms that must be subsequently ratified in a full case-bound test setup. It is a validation tool. It is not a tool in which anyone can accurately assess the convectional thermal transfer characteristics of a waterblock's internals. One does not use a hammer to do the job of a torque wrench, not even if the hammer came from the heavens hand delivered by angels.

I am sorry that we have such a fundamental disagreement about the correct way to assess waterblock thermal performance.

[Marci]

They dont state what those are for whatever reason

Been digging further (still doing so) - datasheets per CPU. eg: Xeon 800Bus Datasheet gives a TTV correction factor of 0.2 iirc, but no accounting for die hotspots... Intel's own tech resource reports however reveal ultimate annoyance of IHS :- TIM1 (joint of Die to IHS) accounts for upto 40% of the total package thermal resistance. Variance of TIM1 could therefore create potentially WILDLY differing results... how good is Intel's Indium Solder process? Certainly better than AMDs process (suspected odd implementation of Enerdyne Indigo?)... I digress... one discussion across several forums... o/t.

[Marci]

General Observations on the issue...

You physically cannot remove the heatspreader from a P4 now. It is indium soldered to the die. There are reasons for the use of IHS. Easier to qualify a heatsink for the CPU's thermal design basically. Without IHS, aircooling solutions would be noisier and more difficult to implement and more expensive. Only way to allow current CPUs to be cooled cheaply en-masse for the general public is to incorporate a heatspreader (acc to Intel docs). IHS is a necessity for the aircooling world. For watercoolers they're a curse. All points more in the direction of AMD being more suitable for watercoolers across the globe, purely because you can still remove their IHS. If that rings true in sales numbers (which it does - far more overclockers and watercoolers buy AMD than Intel IME - might be diff for other companies / countries), then if anything we should be attempting to acquire an AMD TTV as it would be more applicable to the scene. But that would bring along it's own set of issues akin to the Intel TTV issues.

Either way, that leaves you unable to define a testrig (either CPU or sim based) that gives a true representation of both platforms. 2x testbeds must exist. One IHS'd, one not. Two sets of figures produced and stated and marketed.

Problem. As long as companies such as Dell etc are churning out predominantly Intel CPUs, any company with their business brains in the right place will characterize everything for Intel products. It makes commercial sense to plummet R&D into qualifying on an Intel platform. Their marketshare is much larger, and therefore, so becomes your customer base. If you can acquire TTV, then this becomes an easy target. Screw AMD - not where the final big money is.

Ultimately, if TIM1 can potentially be a large variable (which it is) then watercoolers would be better off from a performance point of view to simply lose it from the equation. TIM1 joint then becomes die to block rather than die to IHS. This joint can and has been characterized and accounted for with diesims (depending on your choice of TIM and application). Only issue with diesims is inability to replicate genuine hotspots whilst being able to monitor those hotspots accurately. But ultimately things become a lot more controllable and definable without the IHS and current TIM1 joint in the equation.

With IHS in place, we are not guaging performance of a waterblock. We are guaging performance of TIM1 joint, IHS, TIM2 joint, and the waterblock. 2 points of direct variability within that - TIM1 and TIM2, and an indirect variability - flatness/application of IHS, varying in a 4th dimension with (very) different methods in place from each manufacturer. Unless those 2x (TIM1 and IHS - TIM2 has already been done) variables can be measured and accounted for it's impossible to characterize a waterblock for all CPUs in this scenario.

With baredie or diesim, we're guaging performance of TIM1 joint and waterblock. One variable. One fixed scale for comparative performance. More accurate solely due to reduced number of variables.

Only reason IHS exists is for the aircooling community. NOT to protect from core crushing etc. That's just a secondary benefit. You watercool for performance. Even if for silence, you still need the performance. Logic dictates, remove the IHS based on manufacturer's own comments in tech-docs. Watercooling gives you the ability to do so and doing so should result in significant performance increases based on what's stated in those tech-docs (or that's how I read them anyways)

Produce TTV figures for Intel owners (as long as IHS is non-removeable & correction factors are applied and STATED [as in which has been applied, or the entire list of correction factors provided alongside bareTTV data if non applied], it's a m00t point), Produce die/diesim figures for AMD owners. AMD Owners can potentially benefit more from watercooling than Intel (as long as AMD's IHS remains removeable)

Bare TTV data means nothing without appropriate correction factors for the enduser to translate into something meaningful to them. TTV was only developed to provide useful data for aircooling solutions. Figures will merely show watercooling as "better" but nothing truly meaningful

Reading Material:
http://www.enerdynesolutions.com/learn_thermalinterfaces.html
http://www.intel.com/technology/itj/2005/volume09issue04/art05_materialstech/p04_thermalinterface.htm
http://www.intel.com/technology/itj/2005/volume09issue04/art03_nanoandmicro/vol09_art03.pdf
http://www.apialliance.com/pdf/Archive_semiwest_05/AMD_Touzelbaev.pdf

[moonlightcheese]

i still don't agree that die sims are the best way to test actual CPU temps. however, given that the thermal diode on the CPU doesn't give temp info for all areas of the CPU then the die sims are really about all we have (unfortunately). i'd be satisfied with using die sims to test performance on a naked core but there is still a complaint from my tent...

some 80% of users use the IHS as they are scared to remove it, they are lazy or they just don't care. they are giong to be sorely dissapointed with the performance on some of these blocks and how much their results vary with the IHS still on. with this in mind, there MUST be a way to test waterblocks with some type of IHS simulation. it has to happen. i realise that the IHS can vary quite wildly due to concavity (my storm block went from producing 31C temps to 38C temps just from a remount before i removed the IHS) but it must be tested somehow.

i don't care how you do it and i don't know how you're going to do it and get repeatable results everytime but it has to be done.

[Marci]

Or wait til AMD Indium-Solder their IHS on, then the whole issue goes out the window *shrug*

[moonlightcheese]

nice info Marci, you guys are gods... ^_^

so then that still goes to show that you MUST have a test that somehow simulates IHS.

[nikhsub1]

i still don't agree that die sims are the best way to test actual CPU temps.
Umm, you arent testing for CPU temps, you are testing to see the efficiency of the WB. NO WB WILL BE EFFICIENT WITH AN IHS PERIOD.


some 80% of users use the IHS as they are scared to remove it, they are lazy or they just don't care. they are giong to be sorely dissapointed with the performance on some of these blocks and how much their results vary with the IHS still on.
Then those users should use any block that fits their budget, Maze 3 whatever it does not matter. Until AMD solders the IHS to the core (even then perhaps a bit iffy) then use anything, you will likely get the same results, otherwise remove the IHS altogether.

[orkan]

OMG...

I just unpacked my Storm. Comparing these two blocks side by side is like comparing a diesel Rabbit... to a Z06 corvette.

If the apogee is the future... I always did like history. The storm is going in my loop.

[nikhsub1]

OMG...

I just unpacked my Storm. Comparing these two blocks side by side is like comparing a diesel Rabbit... to a Z06 corvette.

If the apogee is the future... I always did like history. The storm is going in my loop.
Hehe, dare I say 'told ya so'. Enjoy it, it is an awesome block, a work of art :D

[gabe]

While fluid dynamics software is all well and good, one cannot rely on it to be the end all for thermal transfer. .

read my post: we used a series of tools.


IF the design of the Apogee was indeed designed in this manner, Lee (and hopefully Derek, if he ever does testing again) proved that the software isnt always right. As computer junkies we woud all like to think the computer is always right, but its only as good as the equations we put into it.

What is this IF all about? I just posted that we did. Design sequence consists in theoratical data first (CFD), then validation (thermal tools).



A queston for you Gabe. One of the earlier posts in this thread states that Intel says the TTV is ok for repeat long term testing as long as the proper adjustment calculations are done to the results. They dont state what those are for whatever reason, but does Swiftech have those vital equations? If not, according to Intel Swiftechs TTV testing is just about useless.

I am not at liberty do discuss any of the above. Please indulge me in assuming that we use our tools as professionals: the way they are suppoosed to be.

Gabe

[gabe]

Gabe, all I know is this: Attempting to quantify waterblock performance using a "numb" thermal die and/or a variable laden interface is nigh on impossible. Let me know if you don't agree with that.

If you don't agree with that, then we have a fundamental difference of opinion on waterblock performance assessment methodology.

As I said, we designed with CFD, then validated using three different test benches: TTV, Die Sim, and Real CPU - All the data converged with respect to expected performance of the Apogee.


but the ranking was not, so most people let it slide. With the Apogee though, now we have an anomoly that no one can ratify. Of course there's going to be complaints now, and not before.

Now you called it. It's all about ranking. There is no anomaly Stew. The consensus that is developping is clear: some blocks will perform better in some applications than others. Of that I am certain. We are betting on the future of dual core processors.



I don't care who designed it, especially if it is being used for purposes for which it cannot possibly hope to perform as.

What purposes? Design or Reporting ? I think I addressed both points in my post. We agree as to design. We disagree as to Reporting. Again, its all about ranking. And again, my point is that we published all three data sets. Emphasizing TTV results in the presentation of the data sets is a more legitimate approach from an industry standpoint IMO.



I am sorry that we have such a fundamental disagreement about the correct way to assess waterblock thermal performance.

I will reformulate: You say "it's my way or the highway", and I will reiterate: from a design standpoint we do not disagree, but from a Reporting standpoint, we certainly do. We do not place ourselves ABOVE industry standards. We abide by them. If a better tool becomes available to Report water-block performance that complies with Industry Standards, then we will use it. This discussion reminds me of what happened in the bench testing of graphics cards. Several tools are now necessary to evaluate graphics performance. The same now applies to water-blocks.

[Cathar]

As I said, we designed with CFD, then validated using three different test benches: TTV, Die Sim, and Real CPU - All the data converged with respect to expected performance of the Apogee.


Unsure how one could ever expect to to see convergence on testbeds (TTV & IHS capped CPU) in which there are unknown and unquantified variances. This is the very basis of the point I'm trying to make. It's not much different to expecting to see convergence when spinning a roulette wheel. Sure, you know that something between 00 and 36 is going to come up...



Now you called it. It's all about ranking. There is no anomaly Stew. The consensus that is developping is clear: some blocks will perform better in some applications than others. Of that I am certain. We are betting on the future of dual core processors.


You must be reading different independent tests to I then. Both the MCW6000 and the Storm are seen to exceed the Apogee. That is not an anomaly?

Future of dual-core processors? Unsure where you're going with this comment. The smaller of the AMD X2 CPU dies is 147mm^2, no bigger than an old single-core Intel P4 Northwood die, and the largest AMD X2 die is 199mm^2, or about the same size as Lee's "small" test die. If you're referring to dual-core CPU's suddenly getting much larger than that, then Lee's "enormous die" (at 1024mm^2) isn't exactly saying much there either for the Apogee in comparison to the MCW6000. The Intel dual-core CPU's are ~206mm^2, also about the same size as Lee's small die.

It is not economically feasible for CPU makers to produce dies of silicon significantly larger than 250mm^2. Doesn't matter if it's single core, dual or quad core. As processes shrink, CPU makers will continue to squeeze more cores into the same sorts of die areas as historically used by single-core CPU's.




What purposes? Design or Reporting ? I think I addressed both points in my post. We agree as to design. We disagree as to Reporting. Again, its all about ranking. And again, my point is that we published all three data sets. Emphasizing TTV results in the presentation of the data sets is a more legitimate approach from an industry standpoint IMO.


Design or Reporting. Both are at issue here. Primarily Design though. Can't design a waterblock and be truly certain that you're making forwards progress on a testbed with unquantified variables. Again that's the fundamental basis of my point. As to the act of Reporting values that a variable laden testbed is providing, we've all seen the sorts of fallout that occurs on various forums when independent testers do that...



I will reformulate: You say "it's my way or the highway", and I will reiterate: from a design standpoint we do not disagree, but from a Reporting standpoint, we certainly do.

No, see above. This is why I think you're misunderstanding where I'm coming from. In the past you've dismissed this as all being down to a matter of my professional pride based upon reported results. It is not. It is about my perception that some people are arguing for the dependence upon variable laden testbeds for "proof" that we're making forwards progress, and I see that as bad science, and I will always call it when I see it. Have been doing so for four years now. Doesn't matter who does it. In the scientific community it's called peer review. Right now I see a very concerning trend emerging that bodes ill for any hope of forwards progress. Not just from me. Not just for the Storm. For waterblock development as a whole.

I am sorry that the Storm seems to have clouded people's perceptions about the strength of my message. This has very little at all to do with the Storm. I would be saying the same even if the Storm was not involved.



We do not place ourselves ABOVE industry standards. We abide by them. If a better tool becomes available to Report water-block performance that complies with Industry Standards, then we will use it. This discussion reminds me of what happened in the bench testing of graphics cards. Several tools are now necessary to evaluate graphics performance. The same now applies to water-blocks.

I was unaware that Intel had actually produced an industry standard testbed specific to waterblock assessment. Again, hammer and torque wrench anaology. Just because Intel made an industry standard hammer, doesn't mean that it's appropriate to use to do the job of a torque wrench. As the saying goes, when all you have is a hammer, everything looks like a nail. This saying describes this situation perfectly.

[orkan]

As the saying goes, when all you have is a hammer, everything looks like a nail. This saying describes this situation perfectly.

Can't get any simpler than that really.

[nikhsub1]

Guys, you need to understand that the AMD IHS 'numbs' the cooling affect of almost all blocks. This is the reason why in Lee's CPU test all 3 blocks are nearly identical (he did read diode of CPU for this test IIRC AND the probe on the IHS). What does this mean to you? Well, If you run with an IHS, anything from a Spiral to a Storm will yeild you similar results as far as temp monitoring goes. Say what then? A Spiral as good as a Storm? Well, for all of you that argue the only valid testing is WITH an IHS, then your answer is YES. Go get a Maze 3 or 4, it will perform just the same as an Apogee or a Storm.
http://www.systemcooling.com/forums/index.php?s=&showtopic=3903&view=findpost&p=33173

[Cathar]

http://www.systemcooling.com/forums/index.php?s=&showtopic=3903&view=findpost&p=33173

Yep. Illustrates precisely why using IHS bound testbeds for "performance assessment" is a total joke, unless we choose to believe that the last 5 years of waterblock development has all been for nawt, even though testbeds that remove variability and numbness show perfectly well what's going on.

[situman]

I am not sure whether I am OT or not, but since this thread is related to the systemcooling article...Now Gabe, "Robotech" along with some users have found copper shavings in their blocks after they opened them. How and when will this potentially pump damaging flaw be remedied? If we don't want dust in our loop, we certainly don't want chunks of sharp copper "flying" around in there. If any watercooling hardware is damaged due to the copper shavings how will swiftech handle this? I am asking only cause you seem to be working for Swiftech.

[moonlightcheese]

I am not sure whether I am OT or not, but since this thread is related to the systemcooling article...Now Gabe, "Robotech" along with some users have found copper shavings in their blocks after they opened them. How and when will this potentially pump damaging flaw be remedied? If we don't want dust in our loop, we certainly don't want chunks of sharp copper "flying" around in there. If any watercooling hardware is damaged due to the copper shavings how will swiftech handle this? I am asking only cause you seem to be working for Swiftech.
check out the thread in the link at the end of the systemcooling.com review for the Apogee and Lee posted some comments from swiftech regarding QA and extra testing (which were insufficient and barbaric if you as me...)

[MaxxxRacer]

ROFL at that link nik. that REALLY shows you how useless onboard sensors are and how entirely useless taking temps from the IHS are.


You must be reading different independent tests to I then. Both the MCW6000 and the Storm are seen to exceed the Apogee. That is not an anomaly?

Future of dual-core processors? Unsure where you're going with this comment. The smaller of the AMD X2 CPU dies is 147mm^2, no bigger than an old single-core Intel P4 Northwood die, and the largest AMD X2 die is 199mm^2, or about the same size as Lee's "small" test die. If you're referring to dual-core CPU's suddenly getting much larger than that, then Lee's "enormous die" (at 1024mm^2) isn't exactly saying much there either for the Apogee in comparison to the MCW6000. The Intel dual-core CPU's are ~206mm^2, also about the same size as Lee's small die.

It is not economically feasible for CPU makers to produce dies of silicon significantly larger than 250mm^2. Doesn't matter if it's single core, dual or quad core. As processes shrink, CPU makers will continue to squeeze more cores into the same sorts of die areas as historically used by single-core CPU's.

Gabe, after reading your comment that you guys are looking ahead to dual cores, the response that I quoted above from cathar was EXACTLY what went through my mind. The core will NEVER get above 250mm^2 becuase as Stew said, its just too expensive. When Intel and AMD go to four cores it will be on 65nm wich is 30% smaller than the same core today in 90nm flavor. And in 2-3 years Intel will be on 45nm which is 1/2 the size of todays 90nm chips. So as Stew pointed out your comment on the future of dual core is totaly null and void.


As we all have figured out by now, the IHS really puts a cramp on getting good test results out of a waterblock. BUT I believe I have a solution. Rather simple I think.

To fix this problem someone merely needs to test with a die simulator with a real IHS on the die sim and without the IHS. Use the EXACT same equipment, just with IHS and without. Everything else being equal. using the dT (water in to die temp i suppose) divide the IHS temp with the non IHS temp. This will give us a IHS co-effecient. The higher the co-effecient, the WORSE the block performs with an IHS compared to its non-IHS performance. this number would be useful in letting the consumer know which block to choose from if they plan to run IHS or not and if they plan to ever remove the IHS after using a block with the IHS.

[da-key]

I am for better testing methods. Using copper shims under a IHS to simulate core position seems a good idea.

I still prefer my old maze 3 to everything else out there in my chiller loop.

I doubt there will be some magical waterblock that scores so much better than the rest anytime soon. I've tried most of em and the RBX kinda outshines the crowd with the right pump on a room temp setup.

[lv_dicedealer]

Why would you want to test with an IHS in the system? It is an uneeded variable that will only serve to screw up the results. You only have to look at the systemcooling test using the actual cpu which showed all the blocks very closely grouped. Now, before you all raise heck about the onboard temp sensor used in that test, remember it is the same one used for all the blocks; therefore its temps for block 'a' can be compared to block 'b' ONLY if they were both tested on that rig.

The last few years have been very exciting in the watercooling world. Since the original Maze1 from DD there has been a steady progression of technology employed in waterblock design, all the way up until the Apogee. Here is where for some reason a major company has taken a huge step backwards, and for what reason? Did it have to do with the business deals regarding blocks that were designed outside of the company? Did it have to do with a former employee going to work for a competitor? Why did Swiftec dust off the old 'diamond-pin' design and present it to the world as the new leader in design, unseating the previous holder, a block also produced by them (although not designed in house)? Oh the things that only Gabe knows the answers to.....


PS... Give me any two items and lab full of equipment and I can design tests which will make each item appear the winner.... Think about it ;)

[orkan]

PS... Give me any two items and lab full of equipment and I can design tests which will make each item appear the winner.... Think about it ;)

Exactly.

... and if you stand to profit from one item being the victor over another... it shall be so! ;)

Really the core of the matter if you ask me.

[Bloody_Sorcerer]

Now, before you all raise heck about the onboard temp sensor used in that test, remember it is the same one used for all the blocks; therefore its temps for block 'a' can be compared to block 'b' ONLY if they were both tested on that rig.
I beg to differ; I have personally observed stupidly abnormal anomalies in ondie sensors; I changed the RAM cooler on my video card and suddenly had -40C RAM temps with air cooling.

[MaxxxRacer]

much agreed bloody. Onboard temp sensors are nothing short of useless.