Swiftech Apogee HD Review and Comparison to other Top Blocks

[Martinm210]

Interesting discussion, particularly in regards to test methods.

I have always been frustrated by the rather inconsistent mounting results I've been able to achive despite taking every effort to be careful.

I have also previously tried to reduce mounting mechanism variables by utilizing one set of hardware (Same backplate/springs/thumbnuts).

I've also been criticized (Manufacturers included) for NOT using the OEM mounting hardware with the thought that not doing so is not testing the whole package mounting system. I guess I tend to agree with that now, but it doesn't help reduce variables any. Unfortunately as already mentioned, many or most mounting systems do not fix the mounting pressure. I also agree that "Mechanical Contact" has become what separates the very minor differences in blocks and the mounting mechanism is perhaps more important than the block internals.

So should we be testing blocks or mounting mechanisms, and how can you test mounting mechanisms if there is no real way to define what the users will do when systems do not fix mounting pressure?

Finally comes the other variable I don't think any reviewer will capture and that is sample variance and IHS shape differences as previously noted. I have tried testing multiple block samples and have found differences. Unless you test 10 blocks over 10 mounts over 10 different processors, you're still probably only good to .5C.

It's not perfect, just one tested condition with one block with one processor giving an approximate result. IMHO, the only way to see any definitive trends is to have multiple testers over multiple methods and average the results.

Unfortunately, testing in itself is really boring after you've done a few dozen blocks. In my Q6600 testing I only squeezed out a whopping 4 degree difference over 15 blocks, and now less than 3 degrees over the 7 or so 2600K tests I've run. So...overall, we are seeing less than 7 degrees difference over 22 blocks designed over several years.

Then...you might take a step back and ask the question, is 7 degrees really going to net an increase in overclock? Probably not. That's not to say blocks shouldn't seek these tiny advancements, but performance alone is IMHO a pretty small part of the package to consider anymore. The mounting mechanism is one of the top in my book, then there are things like cost, looks, quality control, features, customer service, packaging, accessories, etc...most of which are fairly subjective and hard to pin down in a review.

[Utnorris]

Gabe, I've been speaking English my entire life and still suck at it apparently. I think "gimmick" might not have been the best choice in verbiage. From a standpoint of flexibility, I can see how this would potentially be a great option for a loop (i.e. split off into a parallel loop sort of setup) - not trying to bash design at all here. From a purely performance standpoint (which is what this entire topic is about - how well the new block performs), my personal guess is that dual inlet/dual outlet won't give much change to temps. In other words, multiple inlets won't change how heat transfers from the block into the fluid. This may also just be me showing a complete lack of understanding here.

That make more sense? Of course, I do stand to be proven a total moron here. It certainly won't be the first time. I may be a computer geek, but I will never claim to be so much as even competent in the theories of thermodynamics.

You do realize it is not a dual inlet/dual outlet design, correct? It's one inlet and three outlets so that you can split the loop between the chipset blocks and gpu blocks which don't necessarily benefit from high flow and still allow the cpu block to receive the full flow which benefits more from flow. The design does make sense when using either their new rad since it has multiple inlets. Of course, if you have a rad that has multiple inlets/outlets then it should work also.

@Gabe - I like the look, not so much the white, but the overall design to me is better than the chrome plated hold down bracket that is on my XT model. This is just personal preference and some will like it and some will not. What I am sure 99% will agree on is that we appreciate Swiftech's hard work and innovation. For the last year or so you guys have thought out of the box and brought us our first PMW pump, a new design in rads and a new design in cpu blocks, not just the standard repackaging that some companies have been doing lately. Granted, there are a few other companies out there with newly designed rads like the Aquacompter ones or new res/pump combo's like Koolance or Danger Den, but you guys seem to be hitting more than one area, which I like. Don't take what some are saying to heart, we as a community appreciate all of Swiftech's efforts and your involvement in the community. All you have to look at is sales over the last 5 years and you can tell Swiftech is doing it right 99% of the time. Thank you again for the innovative products.

@Gabe - One last thing, any chance you will have other colors besides black and white for the tops, i.e. green or blue or red?

[Utnorris]

Interesting discussion, particularly in regards to test methods.

I have always been frustrated by the rather inconsistent mounting results I've been able to achive despite taking every effort to be careful.

I have also previously tried to reduce mounting mechanism variables by utilizing one set of hardware (Same backplate/springs/thumbnuts).

I've also been criticized (Manufacturers included) for NOT using the OEM mounting hardware with the thought that not doing so is not testing the whole package mounting system. I guess I tend to agree with that now, but it doesn't help reduce variables any. Unfortunately as already mentioned, many or most mounting systems do not fix the mounting pressure. I also agree that "Mechanical Contact" has become what separates the very minor differences in blocks and the mounting mechanism is perhaps more important than the block internals.

So should we be testing blocks or mounting mechanisms, and how can you test mounting mechanisms if there is no real way to define what the users will do when systems do not fix mounting pressure?

Finally comes the other variable I don't think any reviewer will capture and that is sample variance and IHS shape differences as previously noted. I have tried testing multiple block samples and have found differences. Unless you test 10 blocks over 10 mounts over 10 different processors, you're still probably only good to .5C.

It's not perfect, just one tested condition with one block with one processor giving an approximate result. Don't like it, go home and do your own testing...

Interesting thought and well put. This is one of my biggest complaints with CPU blocks is that there has been very little improvement in the area of mounting. Sure, we have some tweaks, but it's pretty much the same old 4 bolts and 4 screws holding the blocks down which can vary the pressure from one corner to the other. Even the models that tighten all the way down can still vary some. I am not sure how you would do it, but a hold down mechanism like what the Thermalright Venomous uses with the screw in the middle to apply pressure evenly might be something that would be easier to use and get more even pressure since it applies pressure in the center. Not sure people would like the look of a screw in the middle of the block, but you never know till you see it.

[jayhall0315]

Put up an illustration of the pressure clamp we made. Busy at the moment, but will explain more in a bit.

More data later today.

[gabe]

Gabe, I've been speaking English my entire life and still suck at it apparently. I think "gimmick" might not have been the best choice in verbiage. From a standpoint of flexibility, I can see how this would potentially be a great option for a loop (i.e. split off into a parallel loop sort of setup) - not trying to bash design at all here. From a purely performance standpoint (which is what this entire topic is about - how well the new block performs), my personal guess is that dual inlet/dual outlet won't give much change to temps. In other words, multiple inlets won't change how heat transfers from the block into the fluid. This may also just be me showing a complete lack of understanding here.

That make more sense? Of course, I do stand to be proven a total moron here. It certainly won't be the first time. I may be a computer geek, but I will never claim to be so much as even competent in the theories of thermodynamics.

From a pure performance standpoint, there are two cases:

CPU-only config: these features won't matter.
Multiple-device environment: using the features will substantially improve flow-rate, and improve overall loop performance. This point cannot be argued because it is based on known laws of fluid dynamics.

Not to say that you couldn't accomplish the same performance gains using "Ys" - We just claim that it makes things easier and cleaner.

This is one more tool to faciliate the life of eXtreme users. That is all we claim. The fact that we explain it at length might make it sound like we are making a big deal out of it. But this is not the case. We completely understand (and knew this from the get go) that this feature touches only a limited number of users (which btw probably represents a majority of users at XS and other extreme forums); yet 95% of these users either set up their system in series using dual pumps, or split their system in two independant loops. This innovation is Stephen's attempt at showing a smarter way to do things.

[gabe]

Other colors: Been working hard on this, but hit a brick wall. Were' not giving up. It just takes time. Meanwhile, we'll come out with some limited editions which we hope will satisfy ppl in search of custom looks. Stay tuned.

[stephenswiftech]

Gabe, I've been speaking English my entire life and still suck at it apparently. I think "gimmick" might not have been the best choice in verbiage. From a standpoint of flexibility, I can see how this would potentially be a great option for a loop (i.e. split off into a parallel loop sort of setup) - not trying to bash design at all here. From a purely performance standpoint (which is what this entire topic is about - how well the new block performs), my personal guess is that dual inlet/dual outlet won't give much change to temps. In other words, multiple inlets won't change how heat transfers from the block into the fluid. This may also just be me showing a complete lack of understanding here.

That make more sense? Of course, I do stand to be proven a total moron here. It certainly won't be the first time. I may be a computer geek, but I will never claim to be so much as even competent in the theories of thermodynamics.

the HD is not a 2 in 2 out design at all. more like 1 in and 3 out in the regular configuration. it's the fact that you parallelize "what's after the HD" that improves the flow in the HD itself, because the pressure drop of the parallel setup that's after the HD would be much less restrictive than stacking whatever blocks in series.

[stephenswiftech]

Yeah my understanding is its more for flexibility of loop creation than for cpu block performance. Whether they can do this without losing block performance will be the impressive part . I believe Gabe has already said elsewhere there is no performance difference between running one outlet and 3 outlets in a cpu block only loop. What im interested in is if theres any need to balance restriction between sub loops to maintain optimimum waterflow through the cpu block (ie would we see differences in temps between cores or even a general increase in some situations).

You could spend a month testing this block alone now with all the combinations it can give.

remember that in the standard configuration (1in 3out) the HD is always in series with the pump/rad. so no matter what the loop is it always gets 100% of the flow pushed by the pump. Like I said above, if you have 3 other blocks (other than cpu), if you put them in series your pressure drop will be much higher than of them in parallel. From the cpu stand point you will want to do this parallel configuration after the CPU to maximize the flow rate in the HD.

re. balancing: balancing is only a question if you were in a configuration of a low-low restriction chipset block in parallel of a high pressure drop GPU block for example. then the question becomes: wouldn't that kill the GPU block performance. the answer is likely no in most instances as long as the flow in the gpu block is greater than ~.25gpm in our tests. I am preparing another test where I'll drop the flow in the gpu down to .1gpm to see what happens. But all in all, if you take an average chipset block and an average full cover gpu block, chances are you will have more than .5gpm in the gpu which is good enough for today's GPU and today's gpu blocks.

[PiLsY]

I was thinking more along the lines of whether it could affect flow direction across the base plate, and whether that would even matter if it did happen .

I did my own little test with full cover gpu blocks back when I had dual 260's. They had no issues at all right down to a roughly calculated 0.25gpm per card after which temps started going up quite quick. Temp difference between approx 1gpm and 0.25gpm was around 6c, but still only just over half the temps the cards hit on air. That was a very low flow loop with an apogee xt, 2 x ex fc260gtx and a chipset block all with a jingway pump (9v to the pump gave around 0.5gpm at the outlet).

What would be the effect of having a second pump in a sub loop off the cpu block? Would this just increase flow in the whole loop or would there be a gradient effect in the sub loop its on due to the parallel flow?

I hope Jay realises how much testing he's going to get asked to do in this thread!

[ryan92084]

Other colors: Been working hard on this, but hit a brick wall. Were' not giving up. It just takes time. Meanwhile, we'll come out with some limited editions which we hope will satisfy ppl in search of custom looks. Stay tuned.

will you be selling the white top by itself so we can experiment with dying at least?

i wonder which vendors' (if any) mounting system actually obtains 240N of even pressure

[gabe]

will you be selling the white top by itself so we can experiment with dying at least?

i wonder which vendors' (if any) mounting system actually obtains 240N of even pressure

You can dye Nylon, but this is POM. AFAIK it is dyed during manufacturing process, not after. doesn't mean we can't sell the housing sperately, in fact we do intend to.

[ryan92084]

delrin (a type of POM) is dyable its just a pain and you have to worry about warping due to heat. There are plenty of hobbyist threads floating around (mosty yoyo or RC based) http://www.rccrawler.com/forum/paint...ng-delrin.html has some pics of guys that used RIT

[stephenswiftech]

I was thinking more along the lines of whether it could affect flow direction across the base plate, and whether that would even matter if it did happen .

I did my own little test with full cover gpu blocks back when I had dual 260's. They had no issues at all right down to a roughly calculated 0.25gpm per card after which temps started going up quite quick. Temp difference between approx 1gpm and 0.25gpm was around 6c, but still only just over half the temps the cards hit on air. That was a very low flow loop with an apogee xt, 2 x ex fc260gtx and a chipset block all with a jingway pump (9v to the pump gave around 0.5gpm at the outlet).

What would be the effect of having a second pump in a sub loop off the cpu block? Would this just increase flow in the whole loop or would there be a gradient effect in the sub loop its on due to the parallel flow?

I hope Jay realises how much testing he's going to get asked to do in this thread!

the 3 outlets are connected with a channel which is a highway for water compared of going sideways through pin channels.

[stephenswiftech]

Interesting discussion, particularly in regards to test methods.

I have always been frustrated by the rather inconsistent mounting results I've been able to achive despite taking every effort to be careful.

I have also previously tried to reduce mounting mechanism variables by utilizing one set of hardware (Same backplate/springs/thumbnuts).

I've also been criticized (Manufacturers included) for NOT using the OEM mounting hardware with the thought that not doing so is not testing the whole package mounting system. I guess I tend to agree with that now, but it doesn't help reduce variables any. Unfortunately as already mentioned, many or most mounting systems do not fix the mounting pressure. I also agree that "Mechanical Contact" has become what separates the very minor differences in blocks and the mounting mechanism is perhaps more important than the block internals.

So should we be testing blocks or mounting mechanisms, and how can you test mounting mechanisms if there is no real way to define what the users will do when systems do not fix mounting pressure?

Finally comes the other variable I don't think any reviewer will capture and that is sample variance and IHS shape differences as previously noted. I have tried testing multiple block samples and have found differences. Unless you test 10 blocks over 10 mounts over 10 different processors, you're still probably only good to .5C.

It's not perfect, just one tested condition with one block with one processor giving an approximate result. IMHO, the only way to see any definitive trends is to have multiple testers over multiple methods and average the results.

Unfortunately, testing in itself is really boring after you've done a few dozen blocks. In my Q6600 testing I only squeezed out a whopping 4 degree difference over 15 blocks, and now less than 3 degrees over the 7 or so 2600K tests I've run. So...overall, we are seeing less than 7 degrees difference over 22 blocks designed over several years.

Then...you might take a step back and ask the question, is 7 degrees really going to net an increase in overclock? Probably not. That's not to say blocks shouldn't seek these tiny advancements, but performance alone is IMHO a pretty small part of the package to consider anymore. The mounting mechanism is one of the top in my book, then there are things like cost, looks, quality control, features, customer service, packaging, accessories, etc...most of which are fairly subjective and hard to pin down in a review.

I think from the reviewers/users stand point, testing the whole package is the only true way to go. Unless your readers are all willing to take out the original mounting system to replace it with custom screws/springs or whatever, what is the point of showing what the results are with a more pressure? IMO the more things you do that deviate from what the typical users do the more you risk to confuse them with the results. So I like the seeing blocks tested as is, as a true end-user review.

On my side I usually do both, but in a typical design workflow I use a single and custom made retention mechanism that I use for all the blocks I tested. That includes the numerous prototypes we build that may or may not be compatible with our standard thumbscrew+spring. It just makes my workflow more efficient. I usually validate all results with a couple final tests with original mounting system. Again, if I was just comparing blocks I would just do what you are now doing.


The mounting system is a big part of the block I couldn't agree more, especially today as TIM and mechanical contact are pretty much what it's all about... This is exactly why we came up with this thumbscrew+spring+e-ring / back plate package with Apogee GTZ: repeatable, super fast and looks better than other mechanism for most people. This was more than 3 years ago and it did take a long time for others to catch up with it... but the fact they eventually did, illustrates very well the importance of the retention mechanism... I really miss the "fluid dynamics" discussions from 6-7 years ago and truth is there is very little chance for those discussions to pop up on these boards anymore. There is still a lot of pressure drop / flow rate discussions but as soon as people understand they are grossly overrated (other than for CPU) there won't be much of hydraulics discussions either.

[jayhall0315]

Part of our tests Stephen are with the blocks OEM thumb screws and back plate. I will put that up tonight

I simply did the pressure plate, because I got tired of one attachment mechanism giving 180 N and another 260 N.

As to fluid dynamics, I have mentioned this in three or four detailed posts, but not trying to be arrogant, it requires some mathematical background and knowledge of Navier-Stokes, which most dont have. They then get small details mixed up and go off on the wrong track. With all the blocks I am testing now, there is not much variance, they are all the same flat planed copper finned arrays with minor differences. So, I havent brought it up. It takes too much from the day anyway to defend in post after post, arguing with someone who doesnt know how a Reynold's calculation is done, or what a partial derivative or pde is. Doesnt mean I think they are stupid, just that like so many North Americans, their math/physics background is sorely lacking.

I will get the data up when I stop bye the lab later tonight.

[jayhall0315]

Data is up for the average of five mounts using the original manufacturer's OEM mounting thumbscrews and backplates, for the top three blocks.

Will add flow data tomorrow, but very similar to what Martinm already found.

[PiLsY]

Thanks for answering Stephen .

Regards mounting mechanism - I wonder how many people are like me and are re-using their own custom mount with blocks? I've got a frankenstein mount made from various heatsinks and waterblocks that I use with everything (except the XT obviously). Basically a Titan backplate with mount posts, 2 ek spacers per post instead of springs and some zalman thumbscrews. I just tighten down to the limit of the thread each time, very consistent pressure.

As you said mounting mechanisms arent good - infact Id go as far as to say that they all suck in terms of repeatable mounts except swiftechs . Its such a simple thing to get right as well with a bit of thought. I've often wondered why everyone still uses springs though when a hardmount is so easy provided you limit thread depth (and therefore how far it can be tightened down).

[rge]

Regards mounting mechanism - I wonder how many people are like me and are re-using their own custom mount with blocks? I've got a frankenstein mount made from various heatsinks and waterblocks that I use with everything

From pics I have seen, more than just you and I. I am using Swiftech screws (only screws that dont suck), EK waterblock (wanted all black wb), ? HK nylon spacers/screws that happened to fit Swiftechs, ? whose backplate pic here.

Only time I used OEM mounts was testing (Swiftech's aside, though only issue I have with theirs is I would like a "soft" stopping point, then be able to exceed if I desire).

Regarding what to test with, most OEM mechanisms you can vary pressure on, so theoretically hit the correct mounting pressure, it is just you likely dont know when you get there. Testing the blocks at same mounting pressure, imo, is best way to minimize testing error to see the capability of block. Testing with OEM mounts is useful if you cant tighten a particular OEM to the suggested pressure and to inform the user how easy/difficult it is to get OEM to correct mounting pressure. Both ways have merit, but granted not practical to always test both. (on the mount I will use, I can slowly walk pressure down and retest, though granted most people arent buying testing equipment for such).

[stephenswiftech]

Keep in mind that from the reviewer stand point, showing results that people can and will repeat is very important. If you test differently than what the average Joe does, then make sure you are very clear about what you do so that people don't read the results in the wrong way.

[Church]

+1 to testing with standard mounting sys. It's what majority of users will see. Imho 90% of users will use stock mounting system unaltered, so for roundup/comparison to be usable for those 9/10 of community one should test that way.
Though i wonder what to do if some vendor's stock mounting system allows to mount with more pressure then vendor has written in instructions vs blocks where mounting system has more or less consistent and fixed maximum pressure. I suspect a bit unfair advantage for former compared to later given above info about how much mounting pressure matters for top block relative performance. Hmm, in such case probably one should test at both scenarios, one with mounts following instructions to the letter, and one with more pressure.
I don't think that same said about stock mounting sys should be attributed to stock TIM in block's set though. I suspect at least half of LC crowd using whatever TIM they find best, replacing it is also cheap enough, so imho best would be to test all blocks using same TIM (as you probably did?). I thought of suggesting indigo xtreme, as giving very consistent results, but then thought that LM/IX might be not so common, and it also lessens a bit real-world importance of mount pressure, so probably some easy to apply paste that you are used enough to apply for more consistency & average of several mounts should be closer to what users might wish to know.

[rge]

That is what I was getting at. Swiftech and newer EK's mounting system have defined stopping points, and if all blocks did, it would be a no-brainer to test all with their original mounting systems. But what happens with waterblocks with variable pressure mounting mechanisms and no clear instructions on how to achieve optimum pressure (EK's old mounting system for one), and one reviewer has pressure in 250N-280N range in testing and another from 150N-180N in testing. And what are the odds that a given user will duplicate eithers pressure. The best of both worlds would be to use original mount, and somehow measure mounting pressure.

[meanmoe]

Part of our tests Stephen are with the blocks OEM thumb screws and back plate. I will put that up tonight

I simply did the pressure plate, because I got tired of one attachment mechanism giving 180 N and another 260 N.

As to fluid dynamics, I have mentioned this in three or four detailed posts, but not trying to be arrogant, it requires some mathematical background and knowledge of Navier-Stokes, which most dont have. They then get small details mixed up and go off on the wrong track. With all the blocks I am testing now, there is not much variance, they are all the same flat planed copper finned arrays with minor differences. So, I havent brought it up. It takes too much from the day anyway to defend in post after post, arguing with someone who doesnt know how a Reynold's calculation is done, or what a partial derivative or pde is. Doesnt mean I think they are stupid, just that like so many North Americans, their math/physics background is sorely lacking.

I will get the data up when I stop bye the lab later tonight.

Lol, those poorly educated north americans... as if anyone, with few exceptions, outside of a grad program+ in aero/mech/computational engineering should ever stare too long at Navier-Stokes. Most people that do CFD don't build their own solvers even. (read: attempt at not being arrogant failed - or maybe I just under estimate the community )

On another note, I dig the approach and the discussion here. And I agree with Stephen's emphasis on the "test as its used" philosophy. Test it all though - of course

[jayhall0315]

Went surfing today, so I will post the flow results tomorrow.

MeanMoe - If it makes you feel any better, I have taught thermodynamics II and engineering mechanics to undergrads before, and even I have a tough time getting my head around the intricacies of N-S. It is after all a Millennium Problem. If you can solve the accurate 3D description at some non-trivial time T, you have won yourself one million dollars (seriously). And maybe even a Fields Medal.

And, yes, .... take 100 average folks off a typical American street and compare them to 100 average Koreans and they would be massacred in math/physics knowledge.

[jayhall0315]

Flow and restriction results up. The Apogee HD is better than the XT Rev 2 when it comes to restriction, but far from the Raystorm.

I may look into the Apogee HD multi port options later this week if time permits. Of course, one engineering nerd is never the final word, and I invite Martinm and others to redo my tests and add more input.

I think most of the questions have been answered too, but if you have specific points you want addressed, let me know.

[bds71]

So measuring the various blocks heat release ability is actually fairly easy if you account for everything correctly including block attachment force. And that is why we need a lab made rig to correctly apply force to the blocks, to get exact results.

why not take force/pressure and TIM both out of the equation - erase them as a variable? i've seen this argument (how pressure affects measurability/consistency) before. it is definately a valid argument - and one that should be addressed. i also agree that it accounts for many of the inconsistencies seen through review sites. there is a much simpler solution though. i'm not trying to tout anyone's product here, but in my opinion (and the data from the many reviews i've read on the subject seem to agree) indigo xtreme solves this problem. i reference Skinnee's processor reviews and his TIM reviews. the EK Supreme HF block is VERY pressure sensative (Skinnee's review shows a 2.75C difference between a typical mount and IX) - just as the Koolance 370 is more flow sensative (probably due to its higher restriction). but, when using IX the temperatures remain constant REGARDLESS OF MOUNTING PRESSURE OR TIM. right off the bat you can eliminate both of those as variables. personally, i think that is huge, and wonder why more folks don't use it (aside from cost considerations). with IX you are measuring the quality of the block and its ability to transfer/dissipate heat - rather than how effective your TIM is at a specific (or non-specific) pressure.