D-Tek+All Nozzle kit Vs EK Supreme+Option Accelerate Plate Tested (56k Warning!)

[headala]

With due respect, you didn't have to tell me that it is borked, I knew and know it is, now however, after seeing his imprint on the EK, my EK is most definitely FLAT. Eddy is already upset with me so I doubt he will send me another block... the WHOLE POINT OF MY COMMENTARY HERE is the fact that I knew something was wrong with the block and did NOT publish numbers on it. Now, would that be fair to anyone if I did? I think if ZoLKoRn were a 'mench' (it's a yiddish term, google it) he would pull this and request or attain a new V1 fuzion and retest it. Just to clarify here, I am NOT questioning the numbers that ZoLKoRn recorded, meaning I don't think he lied or fudged it, I am questioning (as should he) the integrity of the base of the Fuzion.

You don't see huge differences in performance between different blocks on an IHS unless one is flat, and the other bowed or stepped, its just a fact.

Scott, we all appreciate your experience and rigorous testing. It takes the right mix of perseverance and being freaking anal to get it done right. But, I am unclear. Do you mean that testers should keep trying blocks until they find one that works like they think it should? If that's what you mean, then I think that there is less integrity in that than if you just get a block, test it, and post the results.

I see what you're saying, and I agree that all of the results we see around the web must be looked at with a bit of sobriety and suspicion. But, I don't think that we should cherry-pick blocks to test by finding the one with the exact amount of bow, etc (and would the right amount of bow depend on the state of your IHS? Thicker/thinner solder, lapping, etc. could all affect that maybe?).

[Martinm210]

Yeah, I agree.

We shouldn't be afraid to publish results here regardless of testing methods used. I think we're all a little overly critical in the "Hobby", the truth is there are just way too many variables to completely isolate performance on a set test method.

The best thing you can do with variables and statistics is MORE NUMBER, MORE TESTING, and MORE TESTERS. While I really like the "5 mount method", I don't think people should hold back results for a single mount. If we had 5 people chip in and run 5 single mount tests, you've got essentially the same thing. CPU testing is extremely boring!! If I didn't have my crystalfontz to do the work for me, I wouldn't do it...

The only real solution to determining block performance specs would be for manufacturers coming together and agreeing on a c/w testing method and posting their results THEMSELVES! They would have to AGREE to a very defined method and do so for every processor type out there. But that'll never happen, so you get what you pay for and I have yet to see anyone get paid for testing....the best you can do is get some free parts out of the deal. But even with that you have to have a passion to doing work to help other people.

We don't have that, so we're stuck with helping each other out and the last thing we should do is discourage more testing.

My 2c...how about some more testing, I appreciate anyone's testing and always learn something from it so...keep it up!!..

[Kilyin]

I think there's something wrong with my Fuzion v1 too. I get a 27 degree celsius difference from idle to load on an E6850 with 1.5 vcore (no vdroop).

Is there a way to bow the block if it's flat? It already has the washer and a nozzle.

I have the exact same setup on a s939 system with a Storm instead of Fuzion and it has a 10-12 degree delta from idle to load on an Opteron 165 with 1.57 vcore.

[aspire.comptech]

1.5V vcore is pretty high for a core architecture chip.

[NaeKuh]

With due respect, you didn't have to tell me that it is borked, I knew and know it is,

It just you still kept doubting my comment about the EK. Ive been telling people from the start its a better block on quadcores.

So how many people actually own a RD-30 for a watercooling setup? Very little. Test results with these pumps might be interesting, but you're looking for a setup that'd apply to what most watercoolers would go with.

ZoLKoRn, I think your test results are interesting as they differ from what the general expectancy is. Good effort.

Quite a bit. And i know quite a bit who double up on pumps, and i also know quite a bit who own both setups. Me being one of them.

You can ask martin directly, everything thats being proved right now, ive been telling him.

Add more pressure on that block, and it will pull ahead of the d-tek by a larger margin.

Once you exceed the 22-23V on the RD-30, you'll lose performance due to the added heat dump by the pump. :\


ZolKoRn, dont get me wrong, i thank you for your test. You answered a question i had tho, how would the ek go against the v2 fusions. :T

[nikhsub1]

But, I am unclear. Do you mean that testers should keep trying blocks until they find one that works like they think it should? If that's what you mean, then I think that there is less integrity in that than if you just get a block, test it, and post the results.

No that is not what I mean. I'll try to explain better. I knew from various sources that the EK block performs well, close to or better than a STOCK Fuzion. Ok, with that bit of knowledge I mount my EK block and do a run. The numbers are BAD. Like, worse than any other block I've tested. OK, I think, hmm maybe it's the horrible mount hitting the MOBO mosfet sinks. Petra's sends me a 775 mount for the block. Put the 775 mount on to make sure the other mount was not the issue. Retest. Same thing, terrible. Remount. Same thing, awful. At this point, I call Alex asking him is he has any ideas. He says, "Have revalidated your testbed with a known block?". Hmm, obviously he is smarter than me, so I retest the test fuzion I've got and it checks out where it should. OK, what should I do? What does this tell me? This tells me the EK is BORKED and it can't be published.

By no means to I think testers should cherry pick anything but at the same time if you get a defect or a dud as I have with the EK block you ought to seek out WHY. It is MORE THAN OBVIOUS that the Fuzion he has is borked, the base is too flat. Furthermore, you cant test a block that has been in general use for 7 months vs. a brand new block from the Mfgr, which is the jist I got from Zolkorn, pehaps I'm wrong though.

Next, if you look at ZolKorn's FIRST roundup here: http://www.xtremesystems.org/forums/...hlight=roundup you will see that the EK and the Fuzion are VERY CLOSE, like almost identical. Now all of a sudden in this more recent test the EK is owning it by almost 10C? This isn't rocket science, and I agree with Martin but BAD numbers are worse than no numbers IMO.

[NaeKuh]

Furthermore, you cant test a block that has been in general use for 7 months vs. a brand new block from the Mfgr, which is the jist I got from Zolkorn, pehaps I'm wrong though.

Next, if you look at ZolKorn's FIRST roundup here: http://www.xtremesystems.org/forums/...hlight=roundup you will see that the EK and the Fuzion are VERY CLOSE, like almost identical. Now all of a sudden in this more recent test the EK is owning it by almost 10C? This isn't rocket science, and I agree with Martin but BAD numbers are worse than no numbers IMO.

actually i have to agree with this statement. I wasnt looking at the results very carefully, only the fact that the ek came ahead. :\

[Petra]

So how many people actually own a RD-30 for a watercooling setup? Very little. Test results with these pumps might be interesting, but you're looking for a setup that'd apply to what most watercoolers would go with.

From my point of view, it doesn't matter how many people do or don't use the Iwaki RD-series pumps. Having a powerful pump, like the RD-30, in a testbed is useful because in what I would consider a "proper" test, you're plotting performance (either in dT (CPU temp minus coolant in temp) or thermal resistance) versus coolant flowrate. Testing in this manner provides a much more complete picture of waterblock performance and allows for the direct comparison of waterblocks within specific flowrate ranges (this is where the power of the RD-30 comes in handy). In a good test, you're trying to control as many variables as possible so that you're testing just the waterblock and it's in this scenario that the rest of the hardware being used doesn't really matter so much because, if you've controlled your variables well, your results will be independent of the hardware used (referring to radiators, pumps, fans, ambient temperature, etc.).

As a side note, I'm with Scott on this one--he may seem harsh but he's usually right.

[Kilyin]

1.5V vcore is pretty high for a core architecture chip.

For 65nm? I don't think so. I see people here claiming to be running more than that through 45nm chips.

I currently use a TR backplate and screws/nuts. Thinking of trying the pro mount, but I'm not so sure it's going to make a drastic difference if any. So I guess it would be bad to put an o-ring on top of the washer in the Fuzion?

[nikhsub1]

As a side note, I'm with Scott on this one--he may seem harsh but he's usually right.

Man I really don't mean to sound harsh! I'm just trying to get folks to understand my logic here and to try to push the tester in the right direction. I know what PITA testing is and how much time he spent! Anyway, I think I've now said more than enough.

[Petra]

Man I really don't mean to sound harsh! I'm just trying to get folks to understand my logic here and to try to push the tester in the right direction.

Yeah, I know. It doesn't come off as being harsh to me, but that's because I know what you're going for... other people just seem to take it the wrong way sometimes. Some of Bill's past comments, on the other hand, I found a little abrasive

[NaeKuh]

Man I really don't mean to sound harsh! I'm just trying to get folks to understand my logic here and to try to push the tester in the right direction. I know what PITA testing is and how much time he spent! Anyway, I think I've now said more than enough.

yeah i get scared asking you questions sometimes.

You and Nol Both do this to me when i have a question regarding theory.

But i know you guys dont mean any hate to it. But it does sound harsh sometimes.

[Martinm210]

Yeah, the only problem with flow rate performance relations is as Cathar pointed out here that it ignores pressure drop.

If you compared two C/W curves of a restrictive block to one that's not restrictive it would "Appear" the more restrictive block is better. When it's entirely possible that the lower restriction block nets much higher flow rates and actually performs better.

In the end Cathar's solution was a dT curve relative to pumping power, but that still requires alot of work. I do like the idea, here is Cathar's suggestion on a good method of comparison:


But now we have multiple "Types" of processors, so you'd really have to do this for single core, dual core, quad core..etc...

Scott has brought to light the importance of multiple mounting and I can't thank him enough for setting the example there, it really helps tighten up the results.

I'm just afraid it all get way too complex and time consuming to also capture flow rate effects and I'm not entirely sure it would come across as something that easy to understand.

I think ideally a test would capture 5 mounts minimum and two pumping systems. Perhaps one "Average" pumping system like a D5/DDC etc, then one "High Performance" pumping system like an RD-30 or dual D5/DDC's etc.

The only non-commercial thermal test that incorporated flow rate effects was vkbms on his fuzion nozzle testing here.

http://www.nexthardware.com/focus/scheda/80_701.htm

I thought it was a great example. The only problem I'm having is measuring that level of accuracy. With 5 mounts I'm still at best only accurate to .2C, and it makes me wonder if part of that error is TAT itself or the internal diodes measureing core temperatures. He measured good gains clear down to the smallest nozzle, yet when I tried to measure gains on the 4.5mm nozzles vs stock on V2...I couldn't do it..

So either actual core temperatures can't be simulated using exterior probes, or the internals sensors just arn't that good..

I don't know. I think using an actual processor and heat signature from that processor is probably the best thing to do, but I'm really tempted to build a die simulator for another means to compare. I just don't know there could be a relationship made...?

In the end though, I think most people appreciate seeing a simple bar chart. I'm not sure it would be worth testing with two pump systems, there could be some switching around of performance rankings, or it could be insignificant. I like like the end result method of testing, pick a pump and go with it. If it's low restriction it automatically gains from the lower pressure drop and higher flow rates and at the same time it accounts for added heat dump by the pump for that added flow rate.

I went with a D5 myself because I figured most people were running a single pump like a D5 or DDC with top. Chances are the majority water coolers use a single pump and multiple block loops as well. We're probably in the single digit percentages for users running dual pumps or high performance pumps like the Iwalki's, although this is extreme systems, so there are users here.

Good discussion, since I'm switching to quad core...I'm considering tweaking my testing method as well..

[Xilikon]

Yeah, the only problem with flow rate performance relations is as Cathar pointed out here that it ignores pressure drop.

If you compared two C/W curves of a restrictive block to one that's not restrictive it would "Appear" the more restrictive block is better. When it's entirely possible that the lower restriction block nets much higher flow rates and actually performs better.

In the end Cathar's solution was a dT curve relative to pumping power, but that still requires alot of work.

Scott has brought to light the importance of multiple mounting and I can't thank him enough for setting the example there, it really helps tighten up the results.

I'm just afraid it all get way too complex and time consuming to also capture flow rate effects and I'm not entirely sure it would come across as something that easy to understand.

I think ideally a test would capture 5 mounts minimum and two pumping systems. Perhaps one "Average" pumping system like a D5/DDC etc, then one "High Performance" pumping system like an RD-30 or dual D5/DDC's etc.

The only non-commercial thermal test that incorporated flow rate effects was vkbms on his fuzion nozzle testing here.

http://www.nexthardware.com/focus/scheda/80_701.htm

I thought it was a great example. The only problem I'm having is measuring that level of accuracy. With 5 mounts I'm still at best only accurate to .2C, and it makes me wonder if part of that error is TAT itself or the internal diodes measureing core temperatures. He measured good gains clear down to the smallest nozzle, yet when I tried to measure gains on the 4.5mm nozzles vs stock on V2...I couldn't do it..

So either actual core temperatures can't be simulated using exterior probes, or the internals sensors just arn't that good..

I don't know. I think using and actual processor and heat signature from that processor is probably the best thing to do, but I'm really tempted to build a die simulator for another means to compare.

To complicate matters further, the thermal diode in the die isn't that accurate. Intel's way to make accurate temp reading is to mill a channel on the IHS then epoxy a thermal sensor in it then lap it flat. No one beside a professional tester will be able to comply to this method.

Even 2 testing setup might have variances so it's almost impossible to directly compare both results. The only way to reproduce a testing setup is to be able to get the exact same parts and use the same methods.

[Martinm210]

To complicate matters further, the thermal diode in the die isn't that accurate. Intel's way to make accurate temp reading is to mill a channel on the IHS then epoxy a thermal sensor in it then lap it flat. No one beside a professional tester will be able to comply to this method.

Even 2 testing setup might have variances so it's almost impossible to directly compare both results. The only way to reproduce a testing setup is to be able to get the exact same parts and use the same methods.

Yeah, I've seen that, but it also leaves you with performance at the center of the IHS and doesn't capture how well a block cools directly over the die locations. That's probably not a big deal for HSF testing because they generally cool over a large area well enough (big heat pipes and base masses), but water blocks are much more particular (Very thin bases and nozzles, etc). I think you would almost have to mill out a sensor directly over the die locations for that to represent well...

[ZoLKoRn]

Thank you everyone for your comment and some introduce to me

So from this tested, Yes! on my ideal in first time i think one mount it'll ok but after i try ans try... well i see some different from little to much same as Martin told

So i have plan for all new testing same as this time but i'll order new D-Tek V1 on new testing and i'll order V2 too and sure next testing will be multi mount same as Martin told because it'll good for results

btw. Thanks NaeKuh for kind word from your

Best Regards
ZoLKoRn

[nikhsub1]

Even 2 testing setup might have variances so it's almost impossible to directly compare both results. The only way to reproduce a testing setup is to be able to get the exact same parts and use the same methods.

Yeah its a pain, I wont even update my bios on my test bed lol. Even windoze updates scare the out of me

[philbrown23]

WOW!! Hey eddieEK how do you get these optional plates I cannot seem to find them?? I need a set bad, then I can switch out my Dtek