EK Supremacy Review

[st0ned]

EDIT:

New V2 Bench rig with this block









Again don't forget to support me by liking my facebook It's so easy !

full review on .pdf

__________________________________________________ ____________________________________


Hey guys I finally had the time to join all the data together and compile the results.

https://dl.dropbox.com/u/20116032/Re...cos/Tables.jpg

https://dl.dropbox.com/u/20116032/Re...os/Results.jpg

https://dl.dropbox.com/u/20116032/Re...s/Results2.jpg


Full review for download here.

And please if you like what I did like my profile on facebook so I can show partners how many ppl I've interested in what I'm doing

[rge]

Thanks for doing the review st0ned! And for using IHS like most do

[scamps]

Deleted after getting additional info.

[Six]

scamps
Another test shows other results: http://www.hardwaremax.net/wasserkue...wall=0&start=3

[tiborrr]

Very nice work. I love how you logged the temperature with two thermistors and log the data with the camera. Very time consuming, yet still very precise. I hope you aquire digital data logging station soon!
As for the heatload - fixed Prime FFTs help achieving more constant load results although as long as you use the same method every time with every block it's less of importance. Cut the best and the worst result of each mount.

@Six: In every showdown (may be CPU or GPU full-cover) water blocks from German manufacturers are in top spots, is this coincidence? It's German review site though :P

Another test from Rigmods, not as accurate but gives a good indication:
http://rigmods.com/wp/blog/ek-suprem...-block-review/

[Six]

Its no coincidence, they make good products

[zigzag]

Its no coincidence, they make good products

Just a shame about the looks ^^

[BeepBeep2]

Thanks for your efforts!
But: You should really improve your testing (and monitoring) methods. At least use something like Aquaero to measure ambient and water temps, fix Prime @10 k (or 8 k) to get a constant load on CPU cores and delete clearly wrong data due to very bad mounts.
Average core temps due to different load stats of Prime are good for nothing.

I hope you are joking.

[Six]

He is not joking. When there is a variation of 4 degree in this test then he realy should improve the testequipment.
The variation schould have a maximum of 1 degree for serveral mountings.

Also i agree with scamps that it is very important to fix Prime by 8k to get a constant load.
If not there will be differences from about 1 degree because prime95 only.

[st0ned]

Thanks for your efforts!
But: You should really improve your testing (and monitoring) methods. At least use something like Aquaero to measure ambient and water temps, fix Prime @10 k (or 8 k) to get a constant load on CPU cores and delete clearly wrong data due to very bad mounts.
Average core temps due to different load stats of Prime are good for nothing.

Prime load is fixed at 1024k so it's a constant load As for the other I got fairly consistent results I think. But why not improve

He is not joking. When there is a variation of 4 degree in this test then he realy should improve the testequipment.
The variation schould have a maximum of 1 degree for serveral mountings.

Also i agree with scamps that it is very important to fix Prime by 8k to get a constant load.
If not there will be differences from about 1 degree because prime95 only.

As I sad before prime was fixed. Not the 8k that you guys talk about but what I felt was the best at the time I started testing ( aka 1024k ). As for the 4ºC deviation, I coment on that saying that I messed up with that mount and that I didn't count it towards the final results. Maximum deviation from average was around .4ºC what I find really good beeing that the CPU temperature reads only have a 1ºC accuracy.


EDIT:

Oh and I forgot to say. To validate my method I did 4 "mounts" with the same mount, I mean I did 4 different loggings with the same mount for 2 days at different ambient temps. They all came at a .05ºC from each other, so I would say that's my method range atm 1/20th of deggree

[rge]

@st0ned, Core temps are rounded to 1C, but relative accuracy is ~0.1C according to companies that make core temp DTS. And you just proved that to yourself by logging same mount within .1C, averaging core temps over time eliminates the rounding. I do same mount test on mine as well for same reason.

@tiborrr, I just looked at that test you linked to, no ambient or water temp logging at all. Just used room temp ambients. And they got 3C difference at idle and 3C difference at load . At idle cpu is going to be a few watts, 10 watts at most, and load closer to 100W, so those two EK blocks would test within 0.3C at idle, if 3C difference were accurate at load. In other words 3C difference at idle, means 3C error which equals the difference at load. Hence that test is useless, as is any test that does not log at least water temps. Since you made the comment about the german site which does use testing equipment, why is it that these review sites that dont use any testing equipment at all, and clearly and obliviously show large errors... why is it that the error defies probability by consistently favoring the product they are reviewing at the time?

[st0ned]

As for idle temps I would have to agree I didn't care to log mine ( I did log air in/out , I just don't show it because I find it redundant ) but they all look around the same for the same water temp.

[scamps]

...
@Six: In every showdown (may be CPU or GPU full-cover) water blocks from German manufacturers are in top spots, is this coincidence? It's German review site though :P
...

It is a austrian review site. Perhaps not for you but for the Austrians this is a veeeery big difference. So what is this trash comment good for?
Six himself knows the best that he had no better results for the Mips Ice Force due to it´s "nationality".

This "national reasoning" here and elsewhere really sucks! I want to see proper reviews from where- and whoever! Deleted after getting additional info.
And I really would like to see the new EK-Superblock (but with old design please ) performing even 2 K better than the Supreme HF with jetplate #6. In that case noone would find a reason to buy Phobya´s or Alphacool´s new blocks (= I do not like german Aquatuning ^^).

[st0ned]

Well if you can point out where you did not like my methods feel free to do so. But by your previous coments both things you pointed out were wrong... I don't feel that having an aquaero ( an overpriced piece of hardware... ) is a key to do a proper review. Yeah it could have saved me a lot of work, I had to film all the sensors and then log them on excel but saying this is more inaccurate then having an aquaero... I can't agree. In the end I think you didn't even bother to read the review or my previous answer but still you come here and say, on two different posts, bad things about my methods...

As for phobya and alphacool I send them both an email and they didn't even bother to answer so... Nothing I can do there. AquaComputers sad they weren't interest. Though in part I've to agree with tiborr german blocks show up in their reviews much better then the other blocks whereas when they show up at martin's or skinnee or whoever they don't show up as good as they were suposed to be.

[Martinm210]

Great work!...keep it up!!

[scamps]

Deleted after getting additional info.

[st0ned]

Well I'll try to answer all of your questions.

- Yeah I just noticed that, I could swear I wrote that but I see now that I did not, you're right

- Well for the core differences between average and highest I can't comment on that, just now. I'll test it furthermore with 8k as you sad and I'll come up with some numbers

- Well my review is based on the delta the block can push between Core Avg and Water Avg, I don't care about max temps, I just showed them because there are ppl that care about it. But as I sad above I'll dig in to it.

- Uhm as I sad about I base my data on the average numbers and at that point there's a clear difference. As it was when I tried direct die with the Supreme pushing around a 30.5 Delt and with the Supremacy was a clear tendency to stay above 32ºC dT whatever was the injection plate I use. All this to say that there's improvement margin for my methods though their results are consistent.

- Sure I'll edit my post with the Supreme data so you can give it a go

- Well IB is a fairly cold chip. Well higher temps don't mean less testing error, what I mean is if I would be using a "default" IB with stock IHS TIM temps would be 20ºC higher but that would be a "DC" temp value due to the lack of thermal conductivity from the default TIM. But you are correct if I could get the Watts dissipated by the CPU higher it would give me more accurate data. I wish I could do that without pushing more volts ( a continuous LinX would be a dream... ), I might look in too 4.9Ghz with 1.42v testing.

- As for water sensor I used two thermistors zip-tied to a copper pipe with mx2 between the thermistors and the pipe, them some tape around them. This are placed at like 10cm from the CPU inlet in a clear and well ventilated area so I don't think they'll be getting heat from other sources.

EDIT:

Ther you go

Supreme table



Sensor ( I'm sorry I just now noticed one of the pics is totaly out of focus... )

[paulbagz]

Though in part I've to agree with tiborr german blocks show up in their reviews much better then the other blocks whereas when they show up at martin's or skinnee or whoever they don't show up as good as they were suposed to be.

Are you saying they imply nationalistic bias in their reviews?!...

-PB

[tiborrr]

It is a austrian review site. Perhaps not for you but for the Austrians this is a veeeery big difference. So what is this trash comment good for?
Six himself knows the best that he had no better results for the Mips Ice Force due to it´s "nationality".

This "national reasoning" here and elsewhere really sucks! I want to see proper reviews from where- and whoever! And this one is no proper review because of faulty methods. Over and out. Sorry st0ned.
And I really would like to see the new EK-Superblock (but with old design please ) performing even 2 K better than the Supreme HF with jetplate #6. In that case noone would find a reason to buy Phobya´s or Alphacool´s new blocks (= I do not like german Aquatuning ^^).

I apologize for being misunderstood. I have had no intentions of insulting anyone, I just stated what I've seen being obvious at first glance. I also want to apologize for marking HardwareMax a german review site, I stand corrected now.

All things aside, the results of this review are pretty much what I have got in my internal testing (I have a better average ~ 1.8°C, but then again I was testing 32nm CPU at higher voltage). I still believe this review has better testing methology (Prime load aside) than the one posted on HardwareMax, but that is just my opinion. I also do not like flow rate tests of water blocks that are presented as total system flow. These type of tests tell nothing about the hydraulic performance of the water block and subsequently show much smaller difference between each block than they really are. You see, each test has it's own fault in the testing methodology.

As for the high difference in mount #2, check the imprint (page 5), must have used uneven force while initially tightening the block.

@rge: That's what I said - it's more of a preview, but gives good indication as to where to put the block. As for the idle temperatures I think it's best to contact Dave himself over at his website.

[rge]

@ st0ned: You could just leave it a 4.8ghz and use 1.42v, so if your cpu requires a little more vcore down the road, you dont have to change testing parameters when testing more blocks, that is extremely annoying if that happens.

On mine, if using prime 27.7 newest version with avx (and 64 bit like you stated.. just wasnt sure which version using), it draws same power as linx newest on mine, if testing prime with 8, 10, or 12 small ffts. But like Scamps said, you will get more power draw and more constant load picking one of the smaller ffts, though dont know if that will change results much, but may make data cleaner.

But hope you keep at the testing, need more willing to do it, especially those that can take constructive criticism, and try it out.

And life would be alot easier for you with digital logging to an excel file, something like crystalfonz 633 and ~8 digital dallas 1 wire with USB connection and power connection (everything total is $105) and that you can put directly in water, though have to make inserts or use t-lines and drop probe all way down into flow. You will get burned out with camera logging and manual averaging.

Also given cpu/ihs differences, waterblock production differences milling channels/bow, different tims (thin + low contact resistance vs thick and low bulk resistance), I would expect some differences in testing. Be interesting to have 2 testers switch blocks with each other and retest.

[st0ned]

Are you saying they imply nationalistic bias in their reviews?!...

-PB

1st I wasn't talking about someone/somesite in specific. What I meant ( and I think tiborr also noticed it ) is that there are, in fact, some difference between german blocks tested in German speaking countries and those same blocks tested abroad ( on the US for exemple ). I had specificly in mind the results that I saw around about the AC Kryos on german written sites against those I saw from skinee ( comparing it with the Supreme HF from EK ).

Keep in mind this is something I noticed and it's based upon facts that can be seen on the web, I'm not trying to dent someones image or something like that. Also based on this I tryed to grab a kryos, so I could have my own truth ( if you know what I mean... ) but sadly AC wasn't interested. I was also interested in phobya's new block because if the numbers around are, in fact, accurate their performance is unbelievable and when factoring it's ( low ) price it would make it a killer block.

@ st0ned: You could just leave it a 4.8ghz and use 1.42v, so if your cpu requires a little more vcore down the road, you dont have to change testing parameters when testing more blocks, that is extremely annoying if that happens.

On mine, if using prime 27.7 newest version with avx (and 64 bit like you stated.. just wasnt sure which version using), it draws same power as linx newest on mine, if testing prime with 8, 10, or 12 small ffts. But like Scamps said, you will get more power draw and more constant load picking one of the smaller ffts, though dont know if that will change results much, but may make data cleaner.

But hope you keep at the testing, need more willing to do it, especially those that can take constructive criticism, and try it out.

And life would be alot easier for you with digital logging to an excel file, something like crystalfonz 633 and ~8 digital dallas 1 wire with USB connection and power connection (everything total is $105) and that you can put directly in water, though have to make inserts or use t-lines and drop probe all way down into flow. You will get burned out with camera logging and manual averaging.

Also given cpu/ihs differences, waterblock production differences milling channels/bow, different tims (thin + low contact resistance vs thick and low bulk resistance), I would expect some differences in testing. Be interesting to have 2 testers switch blocks with each other and retest.

Well if I bumped the voltage I would have to retest all the blocks anyhow... I still have to think about it.

As for logging I can't spend 100$ = 100€ ( here ) now, but I already have in mind what I want to buy for logging. Well what I'm doing is getting 1 point at each 30s ( that's around 24 points ) then I use mathlab to fit a 6th degree polynomial function to it. I might be not ideal but as I stated before it showed up being fairly consistent.

[rge]

Shame they wouldnt send you the phobya, with 68 0.2mm fins/channels and the novel jet plate, I would be interested to see other tests with it as well.

[st0ned]

That's not my problem, what I don't like from brands is to be ignored... I value a negative answer a lot more then a no-answer. I still have the DT5Noz to test and the DT sniper on the way too, after that if no one else moves forward I've to find another way to get my hands on some other blocks.

[paulbagz]

1st I wasn't talking about someone/somesite in specific. What I meant ( and I think tiborr also noticed it ) is that there are, in fact, some difference between german blocks tested in German speaking countries and those same blocks tested abroad ( on the US for exemple ). I had specificly in mind the results that I saw around about the AC Kryos on german written sites against those I saw from skinee ( comparing it with the Supreme HF from EK ).

Keep in mind this is something I noticed and it's based upon facts that can be seen on the web, I'm not trying to dent someones image or something like that. Also based on this I tryed to grab a kryos, so I could have my own truth ( if you know what I mean... ) but sadly AC wasn't interested. I was also interested in phobya's new block because if the numbers around are, in fact, accurate their performance is unbelievable and when factoring it's ( low ) price it would make it a killer block.

All good and I know what you mean No harm intended just wasn't sure what you were implying

Thank you for your hard work as well, its good to see EK are still producing good blocks, even if their current "style" is rustling jimmies around the internet.

As people have noted before, maybe there are difference between cherry picked "testing" blocks and the stuff that you can get @ retail.

I wouldn't be suprised if certain companies did this as its forums like XSF etc that are such a great medium for spreading information and collating opinions like a word of mouth convention. Plus its reviewers like Skinee/Martin that people use as their reasoning for making certain purchases as well as they are good with their independant testing and so forth.

-PB

[st0ned]

Thanks for the kind words

Btw I tested the Supremacy with 8k prime, it landed almost on my average mount lol