RBX, How good is it?

[Rokk1972]

Checkout Danger Den's site, there is a new review linked there.

[Soja]

Hrm, we definitely need a better translation. I can barely make out the context using InterTran. Couldn't grasp their testing method either.

[nikhsub1]

If I'm not mistaken, the RBX by their chart was 2C better than the Maze 4... I think that says it all as the Maze 4 is a horrible performer.

[ILikeMy240sx]

Originally posted by nikhsub1
If I'm not mistaken, the RBX by their chart was 2C better than the Maze 4... I think that says it all as the Maze 4 is a horrible performer.

only 2C better than Maze4??? I expected alot more

[Soja]

More interesting is how the WW is 1c better than the Maze4 :p Anyone seen the insides of that Magnum block?

[ILikeMy240sx]

Originally posted by Soja
More interesting is how the WW is 1c better than the Maze4 :p Anyone seen the insides of that Magnum block?

wtf????? :rolleyes:

[Cathar]

Originally posted by ILikeMy240sx


only 2C better than Maze4??? I expected alot more

I personally regard most tests done by the Germans as invalid for the "English" speaking markets. This is not so much because they suck at testing, but because of the test conditions.

The test-bed in question is just a single mount of a water-block. There is no accounting for mounting variations.

The major issue here though is the flow rate and test setup used by many of these testers. The WCP testbed consist of an Eheim 1046 (5LPM, 1.2m pressure head rated pump). An Eheim 1046 is lucky to push 4LPM in wide-open mode with its fittings attached. The pump is then pushing through about 2m of 8mm tubing, and attached to a radiator that consists of 24 elbows of continuous copper 8mm tubing (just by eyeballing the pictures of the radiator). Before a waterblock is even attached, we're probably looking at well below the 2LPM mark in terms of actual flow rates.

Now the issue with performance for lower flow rates is that it is beneficial to have a thicker base-plate than a thin base-plate because the copper conduction "costs less" than the reduction in convectional cooling benefit through having the lower flow rates, and so it's better to have more copper to allow the heat to spread to a larger area to be cooled. As the flow rates are increased, the copper starts to get in the way and hinder the potential cooling performance.

This is perhaps best illustrated by Bill Adam's flow vs C/W graphs:

http://thermal-management-testing.com/summar4.gif

Notice how at below 2LPM that the MCW5002 and the White Water are pretty close? This is the thin base of the White Water coming into effect here dragging its performance back. The White Water was designed to operate in >4LPM environments and its base-plate was designed around that to achieve better performance as one pushes past the 4LPM mark. Now an Eheim 1048 attached to a heater-core and a White Water will achieve ~5LPM, so we're not talking about requiring large pumps here - just a regular smallish pump that most people would use.

Notice also how from somewhat humble beginnings that the White Water stretches its lead dramatically and it still continuing to show good improvement towards the end of the curve while the other blocks that started near it are almost flat-lining. This the the difference in base-plate thicknesses showing through here.

The RBX has a thicker base-plate than the White Water. The Cascade is even thinner than the White Water but you can still see the admirable showing that it puts in.

As one starts to boost the flow rates into "normal" levels, rather than the artificially low ~1.5LPM flow rates that the WCP test-bed uses, the performance picture will change quite dramatically.

As a side note, the Cascade tested there was a first rev block that had been modified by the buyer prior to sending it in for review, without my knowledge. A number of streamlining improvements have been made to the block since then, and I certainly would have appreciated an unmodified block being reviewed.

[Edit - I should add that I'm referring to the review of the RBX, Cascade and WW over here: http://www.watercoolplanet.de/index.php?open=4&show=1

[ILikeMy240sx]

thank you so much Cathar for that wonderful englihtment... I knew I shouldve thought about flow rate bias... I remember seeing this other time when cascade was second... They completely crippled the flowrate of cascade.

Now I know... thanx

[ILikeMy240sx]

Subquestion...

You mentioned that when low flowrate is given, thicker copper is beneficial... but what is the factor(s) that separates the block at low flowrate besides the thickness of the base?

In the graphs I see 3 groups at low flow rate. The top group being 5002 and WW

[zip22]

how thick is the base of the WW? i thought one of the main features of the RBX was that it was so thin in the center. on the wesite it says at the base of the "heat voids" it is .025".

[Cathar]

Originally posted by ILikeMy240sx
You mentioned that when low flowrate is given, thicker copper is beneficial... but what is the factor(s) that separates the block at low flowrate besides the thickness of the base?


I guess you can just lump it under "waterblock design efficiency". Each block is different in how it utilises the water flow to transfer the heat of the CPU into the water. This is highly dependent upon the block's design. The ones that you see at the upper side of the graph will generally have poor design characteristics for getting the heat into the water, such as largish expanses of copper without any turbulence or added surface area where it's most needed to facilitate better transfer of heat into the water.

When I think of waterblock design, I tend to think of it like this:

"How can I ensure that as much as possible of all the water-molecules that flow through the block can come into contact with the copper as close as possible to the CPU"

I guess that about sums up my waterblock design philosphy in a single sentence, and if you look at some designs where the water just rushes about but a large majority of it never gets close to where the heat is, then that will go a long way to helping to understand why some blocks perform worse than others.

[Cathar]

Originally posted by zip22
how thick is the base of the WW? i thought one of the main features of the RBX was that it was so thin in the center. on the wesite it says at the base of the "heat voids" it is .025".

The documentation on the RBX says that it has two base-plate thicknesses really, 0.025" and 0.05". On average you're probably looking at close to 1mm (0.04") for the RBX. The White Water is pretty close in thickness to the thinner sections of the RBX.

Really would like to see JoeC (Overclockers) do a comparison of the RBX and the White Water as his test-bed is closest to what people are doing today.

[Sneakytermite]

Originally posted by Cathar


"How can I ensure that as much as possible of all the water-molecules that flow through the block can come into contact with the copper as close as possible to the CPU"

Sweet idea. I don't think anyone can come up with a better design model than that. That's why he's Cathar.

If this is the case, then the RBX probably can't be as good as the WW or Cascade. The area covering the cpu contact point doesn't seem to create that much turbulence. Its slits are surprisingly at the two ends, far from the center. This is just my two cents.

[Ven0m]

Originally posted by Cathar


I guess you can just lump it under "waterblock design efficiency". Each block is different in how it utilises the water flow to transfer the heat of the CPU into the water. This is highly dependent upon the block's design. The ones that you see at the upper side of the graph will generally have poor design characteristics for getting the heat into the water, such as largish expanses of copper without any turbulence or added surface area where it's most needed to facilitate better transfer of heat into the water.

When I think of waterblock design, I tend to think of it like this:

"How can I ensure that as much as possible of all the water-molecules that flow through the block can come into contact with the copper as close as possible to the CPU"

I guess that about sums up my waterblock design philosphy in a single sentence, and if you look at some designs where the water just rushes about but a large majority of it never gets close to where the heat is, then that will go a long way to helping to understand why some blocks perform worse than others.

^^ That's worth a sticky (no kiddin)

[Cathar]

Originally posted by Sneakytermite

If this is the case, then the RBX probably can't be as good as the WW or Cascade. The area covering the cpu contact point doesn't seem to create that much turbulence. Its slits are surprisingly at the two ends, far from the center. This is just my two cents.

Hmmm, from an eyeball perspective, the RBX and the White Water would be fairly close in performance, but as always with caveats. The White Water has a higher fin density, but the RBX has added turbulence features. The White Water has a thinner base-plate on average that enjoys middling-higher flow rates, but the RBX has a slightly thicker base-plate (on average) which would favor marginally lower flow rates.

The two designs have their merits as to their respective approaches. As to which would perform better, I feel that is best left to an independent properly calibrated test-bed to decide, unlike many of the reviews that are about which are showing margins of error in the 3-4C range making it impossible to distinguish anything.

JoeC offers perhaps the best test-bed for that, but I'd also like to see him test at flow rates higher than the ~3.5LPM that he does test at, since that's still a little low IMO as even an Eheim 1048 with a decent heater-core will see flow rates quite a deal higher. Ideally JoeC should be testing with at least 4 flow rate points to give an idea of what people can expect across a range of pumping scenarios.

[Since87]

Originally posted by Cathar
The White Water has a higher fin density, but the RBX has added turbulence features.

It's not at all clear to me that the "turbulence features" will be beneficial. They look like places for water to get somewhat 'stuck' to me.

I'm definitely looking forward to seeing some quality testing results.

[Ven0m]

For me RBX base looks like it shouldn't make water stuck, maybe wit h these wider spreaders.

Adding good Cascade-style sprayers should be benecicial, tho...

[Sneakytermite]

The center of the RBX doesn't seem to be able to create that much turbulence since its holes are too small and may be neglectible.

[zip22]

what? could you elaborate on that sneakytermite

i would think that any irregularities in the surface would cause more area for the water to hit, even the slightly rough texture of the RBX.

[LPS]

Actually, i ordered the RBX also..
like last wednesday.
I'm waiting for it to setup my watercooling system
I wanna know the result also.

[#Rotor]

Cathar coined it so elegantly there.... :)

another thread (http://forums.procooling.com/vbb/showthread.php?s=&threadid=6786&highlight=molecule) that might enlighten people, when wanting to figure out why not all blocks are the same...

the zig-zag's in the RBX would have really worked well, if it only had 2 inlets. the fact that the entering water, comes in from above the zigzag area, greatly nullifies any gain to be had from it. [M2C]

[Sneakytermite]

Originally posted by zip22
what? could you elaborate on that sneakytermite

i would think that any irregularities in the surface would cause more area for the water to hit, even the slightly rough texture of the RBX.

It's true that there are much irregularities in the center of the RBX, but they are so small, which I think could be negligible in transferring the heat into the water. Their slits on both sides of the block don't seem like it would help much since the cpu heat will not disspate that far from the center as Cathar said. The slits just add more resistance to the water flow.

The WW slits seem to be able to create a controlled flow from the center to both sides of the block, which have better heat transfer to the water. It's hard for me to explain. I hope it makes sense.

[ZachM]

I was just looking at the Danger Den homepage and they said there will be a review from here at Overclockers up soon. Hopefully that will clear some things up and get some more consistent tests with less variables.

[Cathar]

Originally posted by ZachM
I was just looking at the Danger Den homepage and they said there will be a review from here at Overclockers up soon. Hopefully that will clear some things up and get some more consistent tests with less variables.

Well, it'll clear some things up if he tests the White Water as well.

[GTA]

Well, a test has just been posted, shows that its about as good as the Swifty, but needs higher flow.

Congrats on the testing oc.com, you're rantling through the tests very quickly, and the results seem to be what we would expect, and the accuracy seems to good, at least judging by the setup.

Thumbs up to you :)

[ILikeMy240sx]

FINALLY!!!! all this drama and finally some concrete evidence. He used 1260 and it was still that restrictive... Seems like RBX and 5002 will be competing a lot. Personally, I think swifty is better because of the option of pelt route if one chooses to do so and also, easier to mount, higher built quality

[nikhsub1]

Hmm, according to BillA's tests, the WW outperforms the 5002 by 3C at 100W. I guess we'll have to wait for JoeC to test the WW.

[Cathar]

Originally posted by nikhsub1
Hmm, according to BillA's tests, the WW outperforms the 5002 by 3C at 100W. I guess we'll have to wait for JoeC to test the WW.

JoeC's results to date have showed a fairly decent correlation to BillA's data in terms of block ranking, with some small variations which is somewhat natural given that most blocks don't always come perfectly lapped and some from a manufacturer come out as better/worse than others from the same manufacturer.

BillA tested the White Water, being one from myself, and his results can be seen here (http://thermal-management-testing.com/summar3.gif). Of note is the relation to the MCW5002. Now the DTek WW is an exact design copy of the one that Bill tested, but as we know not all of the earlier ones were as flat as the ones that I shipped. DTek says that this has been fixed with people already reporting as receiving better finished blocks, so basically as long as DTek has fixed the base flatness issue one would predict that the DTek WW results will correlate with the WW results from the block that I was shipping.

[ILikeMy240sx]

Ok just making sure, JoeC's CW is different than BillA because of the test setting he uses but in terms of how blocks perform relative to each other is similar. right? - just making sure im understanding Cathars POV above

[Les56]

Originally posted by ILikeMy240sx
Ok just making sure, JoeC's CW is different than BillA because of the test setting he uses but in terms of how blocks perform relative to each other is similar. right? ............

Too early to tell:
http://www.jr001b4751.pwp.blueyonder.co.uk/JoeC.jpg

[Cathar]

Originally posted by ILikeMy240sx
Ok just making sure, JoeC's CW is different than BillA because of the test setting he uses but in terms of how blocks perform relative to each other is similar. right? - just making sure im understanding Cathars POV above

Yes, block ranking should not change given the same sort of flow rate. JoeC does use a larger test die, which means a lower thermal resistance as thermal resistance is proportional to the inverse of the die surface area. This is why JoeC's numerical results are typically about 0.7x that of BillA's. There will also be some variation because JoeC's larger die size may mean that the internal "cooling furniture" to a block may affect things slightly depending on its proximity to the test die shape, but we're talking fairly minimal differences. Maybe enough to effectuate a block ranking change between the two testbeds for blocks that are very, very close together in ranking, but should not be enough to cause large deviations in relative measured performance.

Of course, there's still variations in block base flatness between blocks from the same manufacturer that will also cause variations. Bill tested the PolarFlo that did not have the "SuperFinish", whereas JoeC's test block did have the better finish. The SlitEdge blocks may also have had some variation here, but may also be affected somewhat by "block furniture" effects on the different die size as well.

This is why it's so hard to compare results between test-beds.

What we all need is a set of highly accurate test-beds. All will produce slightly different sets of results. The relative rankings, so long as the margins of error are low enough, will present a decent picture.

[nikhsub1]

Originally posted by ILikeMy240sx
Ok just making sure, JoeC's CW is different than BillA because of the test setting he uses but in terms of how blocks perform relative to each other is similar. right? - just making sure im understanding Cathars POV above
Well it is not that simple really. The problem with JoeC's testing is that if you look HERE (http://overclockers.com/articles373/wbsum.asp), you will see FREE FLOW at 1GPM and then BLOCK FLOW which VARIES for each block. It is very hard to discern differences when the RBX has .82GPM running through it and the SlitEdge has .96GPM running through it. To get a real concise picture, you need to do as BillA; that is to get an actual flowrate of say 1GPM through EVERY BLOCK. This is not an easy task, as you see JoeC ran the 1260 pump with the RBX and only achieved .82GPM through the block. To be fair, the RBX should outperform the 5002 by a bit IF you gave the RBX the same amount of flow, in JoeC's tests that would be .92GPM.

[Cathar]

What JoeC does is take a pump and artificially restricts it back to a known 1GPM flow rate. He then attaches a block and measures the final flow rate.

This is an attempt to give an indication of how the block will perform given any particular pump because if the block is restrictive, then the flow rate will always be lower with any given pump than it would be with a less restrictive block on the same pump.

So JoeC is trying to say that just looking at a fixed flow rate is not much use since most people don't have a fixed flow rate, they have a pump, and what the final flow rate will be is depending upon how restrictive the block is, so measuring at whatever flow rate comes out the other end (0.82GPM for the RBX, and 0.96GPM for the SlitEdge) is meant to be a fair comparison in terms of what will happen in real-world setups.

I'm not wholly convinced though, since even an Eheim 1048 with most any block and a heater-core will push way more than 1GPM (more like 1.5GPM for a low restriction block + heater-core).

I think that Joe's starting at an artificially low flow rate, and it perhaps would be better if he just used the Eheim 1048 in an unrestricted fashion and left it at that. At least then the flow rates would be up around what people would see in their setups.

Alternately, JoeC needs to work on a flow vs pressure drop (http://thermal-management-testing.com/summar5.gif) chart like BillA does. Right now he's doing something that a bit of a mish-mash between the two.

[nikhsub1]

Cathar I hear what you are saying, however, I believe in order to compare apples to apples, every block tested should have an IDENTICAL flowrate. I would love to see each block have 1GPM of flow so we can see which block performs better when all is equal. The RBX may pick up a .01 C/W (doubtful but maybe) if 1GPM were flowed through it as opposed to .82GPM.

[Ven0m]

Personally I preffer 2 graphs:
restriction
C/W / flow rate

This way we can see how blocks perform on specified flow rates and what flow rates (and efficiency) we can expect.

Of course c/w / flow graphs for blocks are not very likely to cross each other, but this way we can better see how that block works. That's why I like BillA's way of presenting cooling devices - there's everything what we may want.

[BillA]

I believe the PolarFLO I tested is different that the current production model,
the current one being presumably improved

be cool

[will_maltby]

Personally I prefer Joe C's flowrate method since, it is more 'fair' in a real world sense.

By taking BillA's method (no offence Bill :) ) of ensuring the same flowrate including the block, you are varying the input pressure and flow between blocks. Ultimately you are putting in more power to restrictive blocks to help them out.

Take an analogy with air heatsinks;

It's like measuring the output air flow of heatsink/fan combos and then using a higher powered fan for a restrictive heatsink and a lower powered fan for an 'easy' heatsink to make it even. You'd be using a more powerful fan for the restrictive heatsink and that definately would NOT be a fair test!

I don't see how waterblock testing is different...


Anyway, hopefully Joe'll test the WW pretty soon 'cos I wanna buy either a WW or RBX for xmas :)

[Cathar]

Originally posted by will_maltby
Personally I prefer Joe C's flowrate method since, it is more 'fair' in a real world sense.

By taking BillA's method (no offence Bill :) ) of ensuring the same flowrate including the block, you are varying the input pressure and flow between blocks. Ultimately you are putting in more power to restrictive blocks to help them out.


I guess then that you missed BillA's C/W vs pressure (pumping power) charts that he publishes with every block that he reviews.

Here's an amalgum chart of all the blocks he's tested:

http://thermal-management-testing.com/summar5.gif

What I like about this chart is that it helps to dispel the myth that high pressure drops need a powerful pump to work well. The White Water is a relatively high pressure drop block, yet at any particular pumping power (pressure drop) we can see that it's ahead of the pack.

This exactly the sort of data that JoeC needs to be putting out to give an indication of just how much "pump" one does need to get good performance out of a block.

[will_maltby]

Yeah I knew he goes all out and gives loads of flowrate comparisons. If you've got the time, then yeah, that's certainly the way to do it.

Joe's method allows for a faster review and a nice single graph containing all blocks. I reckon in a few months he's gunna have a highly respectable collection of results there, after which he'll no doubt be snapped up by some big name cooling company and someone else will have to start their own collection. It's a conspiracy I tell ya!



Anyway, I just like the way Joe keeps it real. He certainly knows what time it is, yes indeed he does! :cool:

[Ven0m]

The more data, the more you can read and guess less.
Most of developers prefer scientific way of presenting data - similiar to BillA's graph's. You can see how block works, how good it is, how good are techniques used, what are their good and bad sides. With a bit of calculations, you can check how block will perfrom in your rig, how to make it work better and so on...

On other hand Joe's data is easy to read. Many common users will find these results better as they need less time to get knowledge and check "which block is better", when they can get lost with BillA's charts.

[Since87]

Originally posted by Ven0m
On other hand Joe's data is easy to read. Many common users will find these results better as they need less time to get knowledge and check "which block is better", when they can get lost with BillA's charts.

Of course, users are more likely to make wrong choices WRT to which block will actually perform better in their system if all they consider is the single C/W value in Joe's ranking table.

[nikhsub1]

Originally posted by Since87


Of course, users are more likely to make wrong choices WRT to which block will actually perform better in their system if all they consider is the single C/W value in Joe's ranking table.
Very true Since87 and a very good point.

[GTA]

I agree that Bill's data is more usefull to people who've been watercolling for years, but I also agree that if you're looking for a waterblock to buy, then it could be slightly confusing, and a newcomer to watercooling would prefer a single figure.

If you're looking for a single figure, then Joe's method is as good as it gets, in my view.