new swiftech block?

[BalefireX]

A couple comments (unfortunately, I'm not flaming anyone, so if that dissapoints you, you can skip this post)

I'm pleased to see that someone (and in retrospect, I don't think there was any other player who could make this shift) has tried to push watercooling towards the workstation/server market. Price is a major factor here - most processors can be had with free retail fans, but everyone here knows their downsides - high processor temperature and lots of noise. A drop from ~50 to ~40 USD is big here; while it isnt much if you are buying one processor, if you have 50 dual processor workstations, its suddenly a $1000 difference (scale up to your heart's content). Of course, the MCP600 is still $85 which cancels out most of that advantage .

It seems me that the lifespan of traditional aircooling is reaching an end - especially when you are starting to see 4x89w Opteron processors in a case. Which of course makes me wonder, why are people with server processors being "punished" with a $2.51 premium per processor? I can't see production costs being any higher for an Opteron or Xeon mounting solution. I would think if any sector needed lower costs to push acceptance, it would be the commercial/server market - we know enthusiasts will pay $80+ for a quality waterblock.

Kick me in the face if I'm reading these graphs wrong (its late here) but if the 6000 outperforms the 5002 by nearly 2C at all flow rates, wouldn't that suggest we're looking at performance comparable to RBX/WW at flow rates under 2gpm?

[AngryAlpaca]

Kick me in the face if I'm reading these graphs wrong (its late here) but if the 6000 outperforms the 5002 by nearly 2C at all flow rates, wouldn't that suggest we're looking at performance comparable to RBX/WW at flow rates under 2gpm?

I believe it's pretty close. Look at the 1GPM range, and compare it to BillA's WW test. I think it's around 0.01C/W higher. It's pretty close, admittedly.

Bill: Can you at least tell us if we matter?

[BillA]

"Bill: Can you at least tell us if we matter?"

hugely, my job depends on YOU !!!
be more specific

does WCing matter ?
to us, sure
in the 'grander scheme', don't see no water there (IM not HO)

EDIT: balefirex
deltaTs are not C/Ws, conversion based on heat load, not exactly straightforward

[Cathar]

Originally Posted by AngryAlpaca

I believe it's pretty close. Look at the 1GPM range, and compare it to BillA's WW test. I think it's around 0.01C/W higher. It's pretty close, admittedly.

Bill: Can you at least tell us if we matter?

The flow-rate resistances are essentially equal between a WW and the 6000.

Fairly easy to get 1.5-2gpm through a White Water when in a loop by itself.

Looking at the graphs, one would say 0.01-0.015C/W difference. Pretty close, and definitely very price/performance competitive.

[#Rotor]

liquid-cooling a server..........
only reasonable reasons for such endeavor is
A: reliability
B: reliability
and
C: reliability....

that is it.... cost is irrelevant. for most server applications one hour of down time will nullify any savings made by buying cheap hardware.

centralized liquid cooling on large clusters is not new...(Cray) it is actually a very good and feasible way of doing it reliably.

[Cathar]

Originally Posted by unregistered

in the 'grander scheme', don't see no water there (IM not HO)

Have to agree. I don't ever see WC going main-stream. Far too many hassles for the OEM group, and the OEM people is what truly defines "main-stream". Only in totally sealed user-non-servicable scenarios (eg. water-cooled laptops from Hitachi). Cannot get the cost down low enough. Cannot get MTBF up high enough without a fair amount of expense.

CPU makers will have to stop increasing heat soon, because the OEM's will push back very hard. Already we see Intel radically re-thinking its Prescott strategy and making internal upheavals to more strongly promote the Pentium-M style architecture and method of doing things.

Also, environmentalist lobby groups won't be happy with 150W juice-suckers sitting on every desk. I don't foresee a world where 150W CPU's can ever be mainstream, and the cooling limitations are not the only reasons why.

[snowwie]

Originally Posted by unregistered

snowwie, why should we be concerned about who YOU know (of) ?

I wasn't talking about me

I know who #rotor is, I joined this site about a year ago, lurked approx. 6 mo before that

or do just like using me as the example of the lay person?
heh

[bigben2k]

Originally Posted by joemac

How can that be? Using smaller barbs slows down the flow. How can that be a good thing?

It reduces the flow rate, but increases the flow speed...

[joemac]

Originally Posted by bigben2k

It reduces the flow rate, but increases the flow speed...

Huh

I guess I will let Bill respond to that.

[BalefireX]

Originally Posted by Cathar

I don't foresee a world where 150W CPU's can ever be mainstream, and the cooling limitations are not the only reasons why.

Sounds an awful lot like "640K is enough for anyone"

Looking at the big picture, the electrical usage over the last century has increased at an incredible rate - 50 years ago the combined electical usage for the US was around 200 billion kilowatthours... now its closer to 4 trillion. Moore's law keeps being extended by some new discovery, but one of the downsides of adding more and more transistors is more and more heat. Shifting to smaller processes has helped that, but eventually air cooling as we know it is going to have to go. Whether it is replaced by nano-lightning bolts, or synthetic diamond heatspreaders, or by phase change systems, or by fully sealed water cooling systems is yet to be seen.

#rotor raises a good point with the reliability issue - actually, its my major concern with water cooling for workstations and small servers - yes, when implemented correctly in one large solution (like Cray) it works well. But when you have lots of servers or workstations, the chances of a pump failure or leak increase to near certainty, even with competent installation. If a HSF dies on a modern CPU, the system will shut down or warn you of overheating and usually there will be no damage and minimum downtime. If a pump fails, you have to drain the system, replace the pump, fill and bleed, and then start it up again - takes a good bit longer than an aircooling replacement. If you get a leak... well, quite possibly, goodbye expensive server and lots of downtime.

[AngryAlpaca]

Why would Bill respond to that? Sure, he termed it oddly, but it's still accurate. Here's a hint: look up speed. I believe they teach the scientific definition in Grade 10 (stupid education.) I thought that manufacturers would be forced to cool down, but I didn't want to appear as a brainwashed Ed Stroligo reader, and I didn't know if everyone else agreed with me. Bill: Do enthusiasts make up the majority of your sales, or is that too specific? Can anyone give me any indication of how big the WCing market in NA is? joemac: Are you affiliated with www.aquajoe.com? If so, I'm really not impressed. Balefire: Look at the physical limitations. Look at the work towards cooling down. They're both pretty great (meaning large or immense/ we used it in the perjorative sense!)

Edit: On the issue of server reliability, wouldn't an SP-97 with a tornado be more reliable? Fans seem much less likely to fail than a pump and fans together.

[joemac]

Originally Posted by BalefireX

Sounds an awful lot like "640K is enough for anyone"

.

Either you are showing your age or you have been reading computer history books
One can point out many, many cases in the past where the expert say enough no more can come from this or it can’t be done.Yeah I remember the sick look I got the first time I told a computer “Expert” that I was using a fan to aid in CPU cooling. Very good point BalefireX.

[Cathar]

Originally Posted by BalefireX

Sounds an awful lot like "640K is enough for anyone"

No, it's more to do with the cost of electricity and the environmental impact. Nothing at all to do with some nob not having any foresight.

What I'm saying is that there are more pressures at work here than merely "How can we cool 150+W effectively?".

[AngryAlpaca]

Are you sure? People and OEM's don't want to deal with the hassle of watercooling, nor the noise of extreme aircooling. We'll soon have things that make http://www.thermaltake.com/coolers/c...24tower112.htm look small, and people will stop taking the high heat crap. Also, will an ordinary person pay more for their heatsink than their processor? Doubtful at best.

[joemac]

Originally Posted by AngryAlpaca

Are you sure? People and OEM's don't want to deal with the hassle of watercooling, nor the noise of extreme aircooling. We'll soon have things that make http://www.thermaltake.com/coolers/c...24tower112.htm look small, and people will stop taking the high heat crap. Also, will an ordinary person pay more for their heatsink than their processor? Doubtful at best.

For what its worth you are 100% correct on this (at least for the major OEM and at this point in time). I have visited some and they cringe at the thought of water-cooling one even made the comment “Unless intel tells us to water cool we wont”

[AngryAlpaca]

That's a small OEM. I'm pretty sure Intel is Dell's bitch. Dell can switch to AMD in a moment's notice, and only have a slight drop in sales, but Intel can't switch, to let's say, Alienware, and survive.

[Cathar]

Originally Posted by AngryAlpaca

joemac: Are you affiliated with www.aquajoe.com? If so, I'm really not impressed.

Since he didn't answer you - yes - joemac one of the co-founders of aquajoe, hence the company name.

[joemac]

Originally Posted by Cathar

Since he didn't answer you - yes - joemac one of the co-founders of aquajoe, hence the company name.

oh.. I thought he was being facetious. I didn't think that needed clarifying
Yes I am

[bigben2k]

Originally Posted by joemac

Huh

I guess I will let Bill respond to that.

You should understand that...

Flow rate, i.e. gpm (gallons per minute)

Flow speed, i.e. fps (feet per second)

The higher flow speed creates the turbulence that increases the performance... bla bla bla... by reducing the boundary layer... sound familiar?

[Anemone]

If you think Dell rules Intel you have the relationship reversed. There are easily half a dozen companies who would practically die to have Dell's relationship with Intel, and if Dell were to screw up in the slightest they'd be first in line to help poor Intel. (haha)

Dell is more like the favored slave, allowed to speak up and say its mind as long as it doesn't go too far in doing so. Any significant deviance has a brief memo from Intel that is as close to a whipping as we need in the modern world. You want to know why Dell doesn't sell AMD? Because if they did Intel would take the "we made you and we can break you" stance.

Anyway the new design looks very interesting indeed, I won't claim the expertise of others here, but certainly there are a lot of 3/8" users out there, and the mention that a 1/2" is coming certainly fixes the pro concerns. And while I do realize that every C matters, I'd say that to some degree a few tenths of one difference of one block to the next is really splitting hairs, and could easily be compromised by the nature of the installation, other blockages, bends, even older equipment (not like any of you have old equipment still in a pc, hehe). So my thought is that this block performs in the same league as the best, pending pH's review, and yet costs less, has superb fit and finish, and to my eyes sets a reasonable mark in the industry. I don't wish to disparage the great designers and sellers present, but rather give some credit for what looks like a very nice product.

$.02

[AngryAlpaca]

You want to know why Dell doesn't sell AMD? Because if they did Intel would take the "we made you and we can break you" stance.

Why use AMD when Intel is fine? However, as soon as Intel isn't fine, Dell doesn't sell Intel. Dell isn't selling Prescotts. Why? Because Prescotts aren't good. Intel can't say, "Sell this!" The buyer ALWAYS has the advantage. However Dell knows that Intel sells, and delivers, based upon their history, but Intel also knows that if Dell stops selling Intels, their sales will drop 10's of millions per year. People go for the Dell name, not the Intel name. Hell, I doubt they KNOW the Intel name. Intel would probably rather sell to a lesser OEM, so they can charge more, except they wouldn't get as much money, because their sales would plummet.

[Groth]

So many posts, so little of relevance.

When can we get our hands it?

[deathBOB]

BillA, you mentioned it was designed to be used with the MCp600 pump, which is way out of my price range. How will it work with a lesser pump such as a HydorL30 or Mag3?

[BillA]

a pump just moves the coolant
slightly less powerful pump, slightly less performance
all a matter of degree, and how important those degrees are to the individual user

[AngryAlpaca]

Originally Posted by deathBOB

How will it work with a lesser pump such as a HydorL30 or Mag3?

You might want to check the specs of the Mag 3, if you think it's a lesser pump. Flow has diminishing returns, and pressure drop works the opposite way. High flow rates are probably not worth what it takes to get them, but that depends on your situation.