[Review] Round 2: EK Supreme HF

[wez]

It's worth noting, however, that Supreme HF comes with hardware for all sockets out of the box so for some of us it is the cheaper option (I have an AM2 system atm, but can see myself moving to LGA1156 at some point).

Yea, AM2 mounting out of the box, and very low price compared to other blocks, at least in europe, that was what made me go with EK Using a supreme atm, and waiting on a HF.

But comes down to look and price imo, judging by those charts

[prava]

Here in Spain you can get the HF for 55€, which means that besides being the best block out there it's also the cheapest (regarding top tier cpu blocks, of course).

[affiliate13]

A copper top for the HF, I wonder if Eddy will nickel plate that...mmmm nice.

[HuffPCair]

Just ordered this block from sidewinders. Got the all black acetal/nickel and cant not wait to see it in person

[kpablo]

Uff... if Eddy decide to nickel plate the top...
(Hope he decide to do it )

[Eddy_EK]

Wait no more... or a day more...
Full copper and fulll nickel plated will be available very soon.

[Vapor]

I'd be careful jumping to the "obsolete" train. Differences equal to fractions of 1C can be more of a statistical error than anything and are hardly enough to call one product vastly superior to another product. At the end of the day it's the temperatures that matter and nobody here will see any difference between a block that runs 65C, 65.2C, and 65.5C. It's just silly. I would call the HF clear winner in the flow area, and basically tied with the rest of the popular blocks for temperature performance.

As other have said, I appreciate all the work that has been put into this, but I just find that there is a little too much "vastly superior" and "so much better" and "the best thing since sliced bread" going on when the reality is that you have three blocks that are basically tied for 1st place.

I see the three blocks like this...

Swiftech XT - budget conscious option, average aesthetics, high performance, very good mounting mechanism, average build quality
Supreme HF - mid-range price, aesthetically weak, high performance, weak mounting mechanism, good build quality
HeatKiller Rev3.0 CU - high-end price*, aesthetically very strong, high performance, average mounting mechanism, very good build quality

* if you throw the LT into the mix then I think things change drastically. Since the LT can be bought for ~$60 and it performs basically on par with the CU version, and therefore the XT and the HF the HK LT really becomes the overall winner for $/performance/looks/mounting/quality in the high-end segment.

Edit: Before I get rotten tomatoes thrown at me, this is my opinion only and you are free to constructively agree or disagree with me

When I talk about superiority, I don't think about people upgrading...I think about people who are choosing between the various blocks for a fresh build. Upgrading has to take into account value, which is very subjective (for some, $80 for 1C and a chance to rebuild their system again is worth it....for others, $80 is a huge pill to swallow for 1C and the hassle of rebuilding their system). Obviously I'm unable to assess a block from everyone's perspective--so I choose the most straightforward one: the perspective of someone who looking at the market offerings and trying to decide and maybe needs some help.

That said, in a fight between the HK3.0 and the Supreme HF (where only things that can be objectively measured are counted), it's really not a fight at all.

Thermal performance: Supreme HF wins (and ultimately wins big if you discount the rarely used silicone mod).
Restriction: Supreme HF wins.
Mounting flexibility: Supreme HF wins big (comes with a backplate and is compatible with every major socket....HK3.0 does not come with a backplate and new mounting kits and backplates [which add up to be expensive] must be bought each time you switch sockets, whether or not you upgrade your system or relegate the block to a secondary system).
Fitting compatibility: Supreme HF wins (works with large compressions out of the box, HK3.0 requires adapters that aren't cheap).
Board compatibility: Supreme HF has no known compatibility issues....HK3.0s have issues with Gigabyte boards (and any other board with caps tall enough and close enough to interfere with mounting).
Manufacturer support: I suppose both prefer you go through the reseller for RMA/etc., so a tie? Eddy does come on XS a lot though, so that's a big plus.

In terms of things that are psuedo-objective:
Mounting system use: Supreme HF gets a 2.5/10, HK3.0 gets a 2/10 (in my book). Both are horrible, piecemeal mounting kits, but the Supreme at least uses thumb-tightening parts for the final assembly steps.
Build quality: both have issues (all blocks do)...Supreme HF's structural screws can rust easily (but are non-wetted parts so while they shouldn't, they still can). HK3.0's base has machine marks, the head of the structural screws strip really easily, internal o-rings are hard to replace if they fall out (they're slightly oversized). I'm still flushing out copper shards from the tapping on the HK3.0 Cu here that's on loan from a forum member....he used it extensively before me, I used it for a full suite of a tests, and it's still 'shedding.'

In terms of objective things, the only place the HK3.0 is superior to the Supreme HF is in price and you have to get the LT with no backplate for that to be true (and at that point, it's only a few dollars between a Supreme HF and an HK3.0LT + backplate).

In terms of looks, the Supreme HF is a little stale in my book, but that's a totally subjective thing. It's not a bad look, just indistinguishable from its predecessor.

So again, it's Clottey vs. Pacquiao in my mind....not a knockout performance by the Supreme HF, but it's still a clear win overall.

[Humminn55]

*SNIP*

Well said, Vapor. (I hope not quoting all your text isn't a problem. )


Presently, I'm running a HK 3.0 LT. But I'm moving my build into a Silverstone TJ07 case next week, providing the interior painting goes as planned.

And since I'll have my loop completely disassembled, it gives me a chance to upgrade, so to speak. Already going to replace all tubing, so why not look at other parts of it, too?

I wouldn't look at the Supreme HF normally, given how close the HK is to the Supreme HF's performance, but given that my loop will be completely torn apart, may as well look into alternatives. And I can recoup some of the cost of the EK block with the sale of the HK block.

But, for one, I HATE HK's mounting system. "For this mount pressure, measure between the mount plate and motherboard....if it's this much, it's this pressure, etc." And the mounting screws are just pathetic.....you absolutely need a Hex wrench at all times to turn the mounting screws beyond loose hand tight.

Thank goodness I have calipers, otherwise I'd have to guess which is no way to mount a cpu block. Wish both EK and Watercool would look into licensing/copying/whatever Swiftech's mount system.

[prava]

I don't see the problem with EK current mounting mechanism. It's just the simple and reliable washer + nut + make as much pressure as you think it's needed. That's all, no matter whether you lap your cpu or block (or both) you can have as much pressure as you want to, unlike Swiftech's system that doesn't work on moded cpu's (as it's designed taking cpu height in mind and if you change that value then you don't get max pressure).

You see, before getting into WC I thought that because of so many people complaining about EK mounting system It had to be a total crap or PITA, which it's not. Just a plain and simple method: what really is a PITA is to mount a TRUE...the springs are so hard I was scared to make a mess to the mobo with my screwdriver if somehow something happend.

[rge]

for the nickel plated copper Supreme HF, to me nickel dramatically increases aesthetics, I will buy new one or top for that. I was leaning towards the nickel plated HK for new build, but having to dremel it to fit my board and 1.6C improvement with plate 1 on Supreme HF, made difference for me...+ something new to test myself, though thanks to vapor I will only try plate 1 vs my HK.

No doubt like dejahn said, without equipment, most wont see that small difference or notice if upgrading, but then again 1.6C is same gain from best to worst performing rads in many scenarios, so guess I can see both sides of that issue, and new builds definitely something to consider. But again that assumes others get same difference, and not one extreme of inevitable variance based on different rigs, mounting pressures, variability in block quality, etc.

Regarding mounts, I liked the Swiftech at first, but now one flaw I dont like is cant vary pressure, like Prava said. Even before lapping my cpu, I repeatedly got about .5C better temps with my HK by increasing pressure 1mm more than HK instructions, and given variability rig to rig, I would like to vary mounting pressure and test myself. I would prefer markings, not a hard stop. But large thumbscrews and ease of swiftechs is awesome.

And one other issue, I can put two non-rotating compression fittings on supreme HF and still have max performance.

[jayhall0315]

Someone cant do math correctly. I have read several reviews of the Apogee XT and three reviews of the EK Supreme HF (Vapor's, the German showdown with like 20 blocks and Pure OC) and all show the Apogee XT as being more restrictive than the Apogee GTZ. However, go to Swiftech's Apogee XT page and they show the XT as being less restrictive than the GTZ. With that dichotomy, clearly someone has no math skils or they are simply fudging the numbers. Here is what I get with the following loop:

Swiftech Micro-Res -> 8 inches of 1/2 ID Clearflex -> MCP 655 set at Power 5 -> 18 inches of 1/2 ID Clearflex -> Thermochill 120.2 -> 10 inches of 1/2 ID Clearflex -> CPU block under consideration on a i7 920 at 4.2@ 1.5 V under full load -> 10 inches " " back to Micro-Res

(I take the best of 5 mounts (out of 10) for each of the four blocks, all as close to 150 Newtons as I can guesstimate based on spring constants)

Apogee XT (stock config) - 58.17 deg C and 1.45 GPM
EK Supreme HF with Plate #4 - 58.81 C and 1.56 GPM
EK Supreme HF with Plate #1 - 58.58 C and 1.43 GPM
Apogee GTZ (stock config) - 60.34 C and 1.42 GPM

These results do not agree with some of the other stuff out there I know, but there it is. I dont honestly believe some of these folks are doing the reviews they claim to do. Basically, Swiftech saying that the XT is less restrictive than the GTZ is true. I am no Swiftech (or any other company's) fanboy, it is just that the Apogee XT performs better.

[Sadasius]

Your rad is only a 120.2. I think your kind of bound in getting rid of heat efficiently don't you think? What fans are you using and at what settings. People will get different results with different systems and that is very common. But if your thermally limited in your ability to get rid of the heat it will not show much difference between blocks at all if any under certain circumstances. A good way to test that is to show your water temps and compare that to your ambient temps. If your only a few degrees off then you know your system is running efficiently. However if you have a small rad and are running in 'silent mode' then I am sure there will be a much greater gap in your water temps and ambient and then it will not matter really what CPU block you use.

[Vapor]

Yeah, Gabe and I couldn't corroborate our restriction results. He even sent me his the Swiftech reference GTZ to test on my testbed....it matched the two GTZs I already had (one GTZ, one GTZ SE) with flowrate within 2%.

IIRC, he pretested my XT vs. the reference XT for flow and they matched....so there's something different about our loops causing the discrepancy (I use the same barbs for every block, maybe that's it?). Odd that your Supreme HF is so restrictive, wtf?

[dejanh]

Yeah, Gabe and I couldn't corroborate our restriction results. He even sent me his the Swiftech reference GTZ to test on my testbed....it matched the two GTZs I already had (one GTZ, one GTZ SE) with flowrate within 2%.

IIRC, he pretested my XT vs. the reference XT for flow and they matched....so there's something different about our loops causing the discrepancy (I use the same barbs for every block, maybe that's it?). Odd that your Supreme HF is so restrictive, wtf?

@jayhall0315, Vapor - Considering how close all of the flow values are I'd be inclined to say that something else is the cause of the restriction in his case. The difference between his HF, GTZ, and XT is insignificant but it also cannot be compared to your setup since the test bed is completely different. Pardon my ignorance here, but why does nobody test the restriction of these blocks without any other components in the loop, i.e., just go with block, pump, and flow-meter?

I realize that we all have other stuff in the loop as well, but if you want to see how much restriction a block really provides and always be able to compare it to another block then test it on its own, no?

[Vapor]

I only have pumps, block, flowmeters, and a res in my loop.

[dejanh]

I only have pumps, block, flowmeters, and a res in my loop.

So I take it you don't push the fluid through your radiators then? You actually isolate the flow to res > pump > block > flow meter > res? If that's the case then there is your answer for the difference. The rest of his setup must be providing the added restriction then.

Either way, this is why I like and dislike these reviews. Realistically they are good to get an idea of the expected performance but the variability of each of our setups can always result in different practical results. I am absolutely not saying that I do not see the values in these reviews, just that they are always contentious.

[Vapor]

I have two loops coming off my res.

Subloop1: res -> pump1 -> pump2 -> pump3 -> pump4 -> block -> flowmeter1 -> flowmeter2 -> res
Subloop2: res -> pump5 -> rads -> flowmeter3 -> res

This way the flowrate through my radiators is always constant, no matter the flowrate through the block. Overall, with 3 pumps 'off' and 2 flowmeters, my block-subloop restriction is fairly high, higher than a basic tubing + res + rad + block set.

Mathematically speaking, ancillary components should create a ceiling for flowrate, not have a floor for flowrate (i.e., the effect of the restriction of the block diminishes). Because HF-p4 is getting a higher flowrate, his (Jay's) secondary components are not creating a ceiling and his numbers say that GTZ ~= XT ~= HF-p1 in restriction. My guess is a fairly extreme variation in production quality may be at play here. Maybe the machining on his Supreme HF has resulted in shallow microchannels (relative to spec and/or mine). That would reduce surface area, decrease cross-section of the flowpaths, and increase the distance between the water and the IHS (all in all creating a noticeably negative effect thermally and increasing restriction).

[dejanh]

I have two loops coming off my res.

Subloop1: res -> pump1 -> pump2 -> pump3 -> pump4 -> block -> flowmeter1 -> flowmeter2 -> res
Subloop2: res -> pump5 -> rads -> flowmeter3 -> res

This way the flowrate through my radiators is always constant, no matter the flowrate through the block.

Mathematically speaking, ancillary components should create a ceiling for flowrate, not have a floor for flowrate. Because HF-p4 is getting a higher flowrate, his secondary components are not creating a ceiling and his numbers say that GTZ ~= XT ~= HF-p1 in restriction. My guess is a fairly extreme variation in production quality may be at play here.

Touché, his plate #4 flow is higher which would imply his Supreme HF being the most restrictive component in the loop.

[jayhall0315]

Sadasius - I am using two 120 mm Scythe S-Flex's at 12 V and 49 CFM. The Thermochill 120.2 can handle 150 watts without problems.

Dejanh - I used the loop I did because this is an average representation of what to expect for both temps and flow in a real installed and running setup. The flow meter idea is fine, but as long as same loop is used, and no variable changes but block, .... it is apples to apples.

Vapor - Bernoulli's hydrostatic equations for closed connected loops. I am not in an argumentative mood, just I have to disagree with all the pumps turned off and the side loop. The best testing setup you can acquire with the tools listed in your methodology section is similar to the setup this guy used at the systemsextreme website back in 2001 -2004 (cant remember name now but will get it for you). The XT, GTZ and HF are all machined fine and without problems (I checked their bases but pictures can be provided if needed). P4 is getting higher flow rates cause it has 7 channels cut in it. P1 is getting less flowrate cause it has 2 channels cut it.

[Vapor]

You can tell the difference between a channel that's <1.5mm deep and ~2.0mm deep without measuring? I'm suggesting that your Supreme HF doesn't have channels that are as deep as mine--a manufacturing variance (or flaw, depends on what spec is). Mine are 1.9-2.0mm deep, it's not very easy to measure with a lot of precision considering how narrow the channel is and considering the bottom surface of the channel is not flat. Slightly shallower channels would explain all our discrepancies in performance.

And as a matter of fact, I do know that the restriction of my two flowmeters + three 'off' pumps is higher than the restriction of a PA120.2 + a moderate amount of tubing (at least at all flowrates above .25GPM...below that I don't have any viable data). I also know, because I've measured it, that when flowrate through the block subloop (Subloop1) is anywhere between .3GPM and 4.1GPM that the flowrate through the radiators is maintained at 1.5GPM within 1% (measurement limitation)

[Devil_Dog]

Wait no more... or a day more...
Full copper and fulll nickel plated will be available very soon.

[Prozium]

Wait no more... or a day more...
Full copper and fulll nickel plated will be available very soon.

may I poke a prod a bit and ask just how soon? i plan to buy my WC set up soon (like with in 2-3 weeks) I would defiantly be interested. would that be WAY too soon to see them in American web shops like sidewinder?

[jayhall0315]

You can tell the difference between a channel that's <1.5mm deep and ~2.0mm deep without measuring? I'm suggesting that your Supreme HF doesn't have channels that are as deep as mine--a manufacturing variance (or flaw, depends on what spec is). Mine are 1.9-2.0mm deep, it's not very easy to measure with a lot of precision considering how narrow the channel is and considering the bottom surface of the channel is not flat. Slightly shallower channels would explain all our discrepancies in performance.

And as a matter of fact, I do know that the restriction of my two flowmeters + three 'off' pumps is higher than the restriction of a PA120.2 + a moderate amount of tubing (at least at all flowrates above .25GPM...below that I don't have any viable data). I also know, because I've measured it, that when flowrate through the block subloop (Subloop1) is anywhere between .3GPM and 4.1GPM that the flowrate through the radiators is maintained at 1.5GPM within 1% (measurement limitation)

No I did not use a micrometer on it Vapor, but I did look at the bases and no bent channels, depth irregularities, bent pins (on the XT) etc... Simply saying that I think my EK HF is an average sample and representative of other HF's and is not a statistical outlier

Your duel loop should only be 1% variance if your resevoir approaches a large size. At smaller sizes and with your extra pumps on, the fluid pressure going into and exerted on the walls of the resevoir will be higher (the smaller the resevoir the higher the fluid pressure) and this will definitely affect your second loop. The solution to this a second order ODE state space setup and is part of Bernoulli hydrostatics.

[JoeBar]

Wait no more... or a day more...
Full copper and fulll nickel plated will be available very soon.

Very nice additions...

[prava]

Someone cant do math correctly. I have read several reviews of the Apogee XT and three reviews of the EK Supreme HF (Vapor's, the German showdown with like 20 blocks and Pure OC) and all show the Apogee XT as being more restrictive than the Apogee GTZ. However, go to Swiftech's Apogee XT page and they show the XT as being less restrictive than the GTZ. With that dichotomy, clearly someone has no math skils or they are simply fudging the numbers. Here is what I get with the following loop:

Swiftech Micro-Res -> 8 inches of 1/2 ID Clearflex -> MCP 655 set at Power 5 -> 18 inches of 1/2 ID Clearflex -> Thermochill 120.2 -> 10 inches of 1/2 ID Clearflex -> CPU block under consideration on a i7 920 at 4.2@ 1.5 V under full load -> 10 inches " " back to Micro-Res

(I take the best of 5 mounts (out of 10) for each of the four blocks, all as close to 150 Newtons as I can guesstimate based on spring constants)

Apogee XT (stock config) - 58.17 deg C and 1.45 GPM
EK Supreme HF with Plate #4 - 58.81 C and 1.56 GPM
EK Supreme HF with Plate #1 - 58.58 C and 1.43 GPM
Apogee GTZ (stock config) - 60.34 C and 1.42 GPM

These results do not agree with some of the other stuff out there I know, but there it is. I dont honestly believe some of these folks are doing the reviews they claim to do. Basically, Swiftech saying that the XT is less restrictive than the GTZ is true. I am no Swiftech (or any other company's) fanboy, it is just that the Apogee XT performs better.

Mmmm, here it is. You are only accepting the best results instead of the average (Vapor eliminates the best and the worst results, accounting them as statistical deviation), which is not good for testing purposes. Also, your test-bed is not good enough, as its pretty much rad-restricted in all senses. Of course, what one has to realize when reading Vapor's review is that in a BEST CASE SCENARIO the difference between Supreme HF and other blocks is XX, but that is because he can put out of the equation all other factors: the dissipation on his loop is hugely oversized.

You see, in your test-bench you can pretty much say that all the blocks perform the same (all data is so close you can't be sure who is on top), which means that for the same conditions as yours all of them will be equally fine. But, take in mind that Vapor's aproach is from a perspective in which the limiting factor is only the cpu block and nothing else. Yes, it is not a real life test as I don't think nobody has two triple rads + 3 pumps on its loop, but this is the only way of testing subjectively something.