NV30- the fan will last how long, we must dust it how often?

I think "annoying" has a nice ring to it as well.

Fred-

And what observations might these be? There is a clear difference between concerns and observations. Concerns lead to buyer hesitation.. and buyer hesitation leads to the need to greater exceed expectations for consumers to "take the plunge".. so to speak.

An example might suffice- if you take product A and product B, and hypothetically price them similarly and have them both perform near equally, a consumer may pass over product B if it's 2x the size, massive, loud, noisy and unwieldy to install. Product B in this scenario would need to have advantages far above and beyond product A in order to ease any hesitation caused by the visual/perceived issues of size, volume, maintenance or similar.

And that's the real point here. I'd like to have an NV30 when it comes out- but I seriously dont know about three things- 1) will it crowd (or make contact/heat) the dual-channel PCI UW SCSI card that HAS to be in PCI slot #2 (slot 1 is shared with AGP so is empty anyways, slot 2 is the only "master" PCI slot without shared mainboard resources), 2) Will the sponge filter be efficient enough to prevent the fans from packing/failing over time (read my previous posts about my front intake fans which ARE filtered by the way) 3) How loud is this sucker going to be. I've already tossed a thermaltake P4 fan as the 6-9000 rpm thing sounded like an F16 under my desk even in the middle and low rpm settings.

 

Original Radeon and NV30...

I usually lurk, but I thought I should point this out.

With the original Radeon, ATI choose to include a fan on the heatsink even though they didn't need one. There were articles that mentioned this, though I could only find one still around:

http://www.onepc.net/articles/0001/page2.shtml

And the explanation ATI gave back then was that consumers wanted to see a fan on their cards because that reassured them that they were getting an advanced and powerful piece of technology.

Most of the people in this forum know better than that, but isn't it possible that Nvidia is hoping for a similiar response? We might look at the cooling solution from an engineering standpoint and think "that's excessive and reflects poor design" but an ordinary consumer might look at it and think "That must be one badass card."

And that's why ATI included the fan. A fan based solution is more expensive, louder, and less reliable than a passive heatsink; but ATI choose it anyway for marketing reasons.

Another possibility is that Nvidia really believes in the goal of creating a quiet cooling solution. There are alot of ignorant people here who apparently don't understand that radial fans (the design Nvidia is using) are not necessarily "dustbusters" and can in fact be extremely quiet. One of the most quiet hsf available is the Noise Control Silverado which uses two huge radial fans, but runs virtually silent. This is from Tom's Hardware:

http://www17.tomshardware.com/cpu/01q1/010306/cooler-10.html

Some people use huge heatsinks, ducts and large fans to overclock. Other people, like myself, use them to achieve low noise computers (On my Athlon I have a huge copper Zalman heatsink and 90mm fan running silently at 900 rpm.) It all depends on what rpm the fan is running at. Abit OTES runs at 7000 rpm, which is why it is loud. If you don't know what rpm Nvidia's fan is running at, then you have no possible way of knowing what they are trying to do. And instead of flaming them, you should maybe wait to see and hear the actual card first. Duh!

Since Nvidia has explicitly said they were trying to achieve a relatively low noise solution, we should at least recognize the possibility that this is what they are doing.

As for reliability, the board makers are the ones who will have to deal with the warranties and such, and I'm certain they'll demand a product that doesn't cost them a boatload in product returns. I can't imagine outside dust being that big an issue. It would have been trivial to design the duct to take its air from inside the case (like Abit does) so there is no reason to take it from the air from the outside unless they knew that wouldn't pose a problem.

At any rate, you need to actually test the card before passing judgement on it.

 

Re: Original Radeon and NV30...

Yes, the info on fan rpm is key! As I wrote before it's difficult to judge this solution without knowing that kind of air flow (and thus fan rpm) that is needed under load. My only reason to be a bit sceptical about this is that it looks like the air flow will be somewhat limited by the air intake unless you use a high rpm.

BTW: I have a Silverado on my Athlon XP and while it is very quiet and removes the heat well, I still need a good air flow in the case in order to avoid a massive heat built up inside.

Welcome to the forum! 8)

 

Ghandi:

We've already heard testimonials from people saying the NV30 reference fans are NOT particulary silent, and in fact quite darn noisy actually. Why else would it need a throttling function, hm?

As for your speculation the extreme cooler is a dildo of gigantic proportions to attract buyers, well that is simply laughable at best. Sorry, but I don't buy that at all. It's very far from providing a small 50mm fan for cosmetic reasons to providing one that takes up two card slots in width and sucks in air from the outside.

*G*

 

I just wonder what kind of design margin the NV30 cooling system has. Yes dust will accumulate on the heat transfer surfaces especially if in a smokers home. This would cause a larger thicker laminar flow over the heat sink decreasing the heat transfer coefficient. If the surface area is sufficient and the velocity of air is high enough even if the cooling capacity goes down it may still be sufficient to keep the core cool below critical tempertures for a given temperture enviroment. I have my doubts of the viability of this solution due to the fact it doesn't look like it is made to be serviceable as in cleaning or being a filtered system. It really does look ad-hoc to me. Here are some basic Heat transfer equations:

Qgpu = UA(Tsink - Tavg of air)

Qgpu - Is the heat transfer rate of the gpu, what ever units you want to use (watts, joules/sec, BTU's, Tons, whatever).

U - Overall heat transfer coefficient - basically how well does the setup transfers heat.

A - Surface area, increase the surface area you have more heat transfer capability obviously.

(Tsink - Tavg of air) - The difference in temperature between Heatsink and cooling medium. You can also go straight to the heat source the gpu and write this instead:

Qgpu = UA(Tgpu - Tavg of air)

But a good assumption that all heat transfer occurs through the heat sink allows the first equation to hold up. BIG NOTE: In order to have heat transfer you must have a difference in temperture. meaning the the higher the difference -> higher the heat transfer rate.

First trick in Nvidia's design was to take cooler air in from the outside meaning it gives you a highest delta T or difference in temperture without cooling the ambient air temperture. This could be neat in that you could easily route an air conditioning duct to the suction side of this card easily and get increase cooling .

Second trick was to enclosed the heat sinks with a high velocity fan creating less laminar flow. Laminar flow is the dead air clinging next to a surface which the molecules or atoms flow virtally parallel and not mixing with in this case the cooler air in the air flow. More turbulent the air flow less laminar flow will occur. More dust build up, tar buildup from smokers the more laminar flow occurs and this will decrease the U overall heat transfer coefficient.

From my experience in nuclear engineering and mechanical engineering this design will fail in the long term. Why?

No filtration of system and no apparrent way to do routine cleaning maintenance on it. Now back to my orginal question, what kind of design margin will this cooling system have will decide if this card will last 3 months, a year, years or eternity .

 

Well Said Noko. My faith in a major company like Nvidia or ATI leads me to believe and a agree with "Others" statements here that no matter what the opinions are of the appearance, the space it takes up or the noise it makes, it will work and work well for an acceptable amount of time. As for the reasons of why or when they went to this design is really a guessing game unless someone has qualified info from the company.
The question that keeps popping up into my mind is the issue of pricing.
ATI has had a nice run now as the top card available in its category.
Nvidia isnt on the shelves yet and we really dont have any finely determined idea when it will be.
Nvidia will surely have to charge a premium price in order to recoup its development investment.
ATI on the other hand has a huge headstart in this area.
Scenario: Nvidia comes out with a Card that sells for $425 US.
ATI drops the pricing on their offerings to $250 US.
Yes all this is only speculation and Im sure im on the radical side of the spectrum but this will be a factor for buyers in this category of cards as well, imho.

 

There are some holes in your analogies I think, atleast when I was considering the same things before I concluded differently.

My experience with nuclear and mechanical engineering would cause me to be concerned about those factors as well...but we're not talking about a nuclear reactor with crack propogation due to thermal stresses and neutron embrittlement being a major factor, and corrosion due to prolonged exposure to corrosive agents under a corrosive environemnt (high pressure, high temperature, high flux). We're just talking about heat transfer with the conditions much less stringent.

While the change in the heat transfer coefficient will be affected exactly as you say over time and likely by such factors as you state, there is no reason to suppose that the degree to which this will occur under those conditions need ever reach a point to precipitate failure. It is reasonable to believe the design could be realized such that such factors would not impact heat transfer signficantly before the expected mechanical failure of the fan (i.e., years). This is primarily because, unlike in a nuclear reactor, the occurence of laminar flow need not appreciably impact heat transfer...the primary impact of the buildup will be solely due to insulation not the increase in the laminar flow layer (I'd even say I think it would be more likely to prevent laminar flow and increase turbulent flow, such that the reduction in turbulent flow/enhanced heat transfer you imply with the phrase "increase laminar flow layer" would require a thickness of dust that I don't foresee being very likely...unlike corrosion, dust is not likely to be able to build up significantly without some poor design choices).

For one thing, we are talking about gas not liquid, and it is not occuring at extreme pressure such that laminar flow can result in characteristics of the phase changes of the heat transfer medium such that heat transfer is significantly disrupted For another, it is, as I said a much less stressful and tolerance critical situation so that such factors matter much less.

The most dramatic impact of dust build up will be solely (I think) due to insulation effect on the heat transfer surface, and the filtration only has to be succesful enough to prevent that from occuring in a short time period to address that concern. My own concerns are more for the mechanical failure of the fan than any significant impact of dust in such a fashion as air velocity would likely be sufficient to prevent a significant layer of build up (it is much easier for this device to be sufficiently "frictionless" to most dust than a channel wall in a nuclear reactor be resistant to corrosion). And actually I'm fully confident that a fan design resistant to mechanical failure from dust (for almost all dust environments atleast) is not unduly difficult.

As has been mentioned, laptop cooling being designed around similar principles goes a fair way to assuring that it is possible for this particular concern to be a non-issue.

 

I know we are not talking about fracture touhness either . I think you are missing a basic concept which alot of people miss that is that the difference in temperture between the heat sink and layers of air is the driving force for heat transfer, not the amount of air passing through .

Dust hinders air flow (increasing laminar flow) as well as to trap air (hey, increasing laminar flow again). In any case when the heat sink is clean the molecules/atoms hitting the heat sink are a lower average kinetic energy thus offereing a lower temperture for better heat transfer. When the heat sink get all dirty (hey my cpu heat sink in 3 months causes my cpu tempertureto go up about 5 degrees C) you are right, insulation, the average kinetic energy of the molecules next to the heatsink air side will be much higher. Then again what is your typical insulation except layers of air molecules/atoms clinging to the surface which will get thicker and thicker in this case as the NV30 cooler clogs up with nasties hindering heat transfer.

Air is a fluid system by the way and what you learned dealing with liquid systems will also pretty much apply to air systems. Even pump laws work out in ventilation and cooling systems . The heat transfer equation is universal and is correct for this application and no there is no neutron embrittlement.

Qgpu = UA(Tsink - Tavg of air)

When you change the air flow through any heat sink what factor or factors change?

My only design problem with the design is that it is unfiltered and doesn't look like it will be easy to clean. I don't know about the noise nor how long will the GF FX GPU will last with this settup. To many unknowns.

 

Well that is two people and I am sure many more have seen the effects of dust on heat sinks and fans. I think your case is more servere then mine but I have, actually my wife has cats and they like to hide or play behind my computer. Amazing how you can find cat hairs in the most inopportune places. I know what you mean having a gummy fan. The more gummed up it gets the more crap it accumulates. Now to have a blow dryer back there for the cats would be interesting. I am sure they will not mind the ventilation.

The more I look at Nvidia design the more I think it was a after thought to fix a problem.

 

I didn't propose that the amount of air passing through had to do with anything except the ability of the dust to build up. Please re-read if that isn't clear...?

Hmm? Are you trying to maintain that dust increases viscosity? I'd say with the concentration of dust we are talking about and the fact that we are dealing with a gas, this effect is insignificant...even the viscosity effect due to temperature increase should not be much of an issue for the fan likely to be used. Laminar flow occurs next to a smooth surface and depends on the stratification of flow speed in a linear and fixed fashion. Dust buildup does not seem likely to facilitate the increase of this, especially not in a gas which while a "fluid system" as you describe has significant variance in behavior to a liquid on the scale we are discussing.
I do think that enough dust build up will restrict flow to such a point that air flow (including the turbulent flow that most effectively facilitates heat transfer) could be restricted, but 1) that is not the same as saying laminar flow increases 2) I don't think is much of a concern as I address in my comments about dust buildup.

You are describing insulation effect, not laminar flow....that's why I say insulation effect has a significant impact and laminar flow does not seem like it would.

Well, it is the dust layer that will insulate. Air pockets will form in the layer of dust and those air molecules will fit your description, but again not because of laminar flow.

Yes, I do realize that, but I also mentioned phase change being a major concern with regard to laminar flow and when it occurs in a coolant channel.

Nuclear reactor: coolant at high pressure and in liquid form...coolant in a different phase (gaseous) is drastically less able to transfer heat, therefore it is important that laminar flow not occur to a degree such that the coolant flow changes phase under the temperature and pressure conditions next to the fuel plating (see my comment about temperature at intake and exhaust and why flow speed does matter there), therefore flow must imperatively be maintained to control the rise in temperature from travel along the coolant channel, therefore laminar flow "boundary thickness" is a critical factor, and "surface drag" that affect this are extremely important.

NV30 fan assembly: notably different to what is described above.

Really, you sure?

The amount of flow only changes, as far as temperatures, the temperature delta from intake to exhaust, which is a different equation that we aren't really concerned about in this discussion (though it can be related to some of what we discuss). Note that as I mentioned before my mention of air flow is to do with dust build up atleast in this application (small size, high tolerance atleast relative to a reactor)

I'm not convinced the filtration matters very much, other than to prevent things large enough to actually clog the passage to enter. For concerns of cleanliness of the heat transfer surfaces, a rudimentary filter, barely deserving of the name "filter", should suffice as long as airflow is high enough. Of course, there are substances that could enter the assembly and build up quite a layer, but I don't think that is a realistic scenario unless the presence of such would also affect the rest of the system in any case.

Again:

Laminar flow does not seem to me to be the large factor you describe, for the reasons I've stated.

The tolerance for performance degradation is not as small as you seem to be concerned about, namely because this is not a nuclear reactor.

Dust build-up and insulation effect could be a problem, but I don't think it will be due to the likely amount of airflow, and as already mentioned we seem to have an example of an application of this type of design that exhibits that it works without some of these issues.

 

That is by no means a statement that holds universally, or even "most" of the time. In fact, with an adequate heatsink the required delta T to transfer the specified amount of heat can be very small. Sometimes the UA term can vary more than the possible delta T can.

Perhaps you meant by "driving factor" that the delta T is "what makes heat flow" (duh )not "most important part of the equation?" If so, that just wasn't very clear.

Isn't that obvious?

I love how your equation just has a U in there... that U being the most difficult part of the equation to even define, let alone solve. U, or h as I've seen it referred to for convective problems, embodies information about fin efficiency and airflow (including turbulence, velocity, etc.).

So, the short answer is that U changes in response to a change in velocity (which also has an impact on turbulence, which also affects U). This change can be quite large depending on heatsink design... so large that the delta T goes from "frying an egg" to "cool to the touch."

 

I just want to put in a couple of things:

1. There's absolutely no reason to suspect that the GeForce FX will have any more of a dust problem than any other fan in the computer case.

2. There's also no reason to suspect that it will be any louder.

Just wait for some reviews. I see a lot of shots in the dark here that are just pointless.

 

My knowledge of thermodynamics is a little limited, but first thing

The heat transfer equation you used applies to a steady state sytem.

In this case, there is no equilibrium established. So it should read

dQ/dt = A k(t) * (Tw (t) - Tinfinity (t)) / d

d = conductive barrier thickness.
k(t) = thermal conductivity coefficient

The problem with the analogy, is that convection is not properly taken into account. AFAICS Newtons convenction law is what you used above.

The problem is that this is a forced convection system, so there are pressure terms driving the system (via the mechanical energy of the fan)

This changes the effective value of k(t) if you model it as a linear sum (convection + conduction) as well as the temperature gradient. Since the curl term in the local navier stokes air flow equations is nonvanishing.

From what I remember (now looking at only convection)

there is a delta term that acts sort of like the thickness parameter for conduction. eg delta (x) ~ (v * x/V(infinity)) .. This is combined into something known as a Nusselt number. The temperature gradient at the wall is a system of differential equations, which allow you to solve for the aforementioned Nusselt number. Minor problem is there is a factor of C, expressing a friction coefficient as a function of thereynolds number (which have to be solved for numerically)

And im getting myself into trouble here, b/c this can get arbitrarily long, and Im overthinking this and I might be mistaken.

The point is, the equations above are a vastly oversimplified system. It can get arbitrarily complicated.

 

Ahh....the Navier Stokes equation. I remember watching my professor derive that. Haven't used it yet, but I do remember seeing it derived. That's the funny thing about undergrad physics. A very significant part of it is about finding the equations, and very little goes into actually solving the more complex ones. I guess we just leave it up to the engineers to solve those pieces of the puzzle...

 

The fact the GeforceFX draws *directly* from outside the case is reason enough to "suspect" that dust/debris would be more of a problem.

As already noted on this thread, testimonials of folks at debut report it's not exactly quiet. Moreover, having a big hunk of hollowed plastic over the fan would most definately act like a resonator, which can clearly cause an increase in noise factor.

 

Chalnoth:

"I just want to put in a couple of things:"

Of course you do. Always ready to jump to Nvidia's defense!

"1. There's absolutely no reason to suspect that the GeForce FX will have any more of a dust problem than any other fan in the computer case."

Why would there be no reason? Because you want it to be that way/say it is so? Not good enough reasons. Air inside the case is already "filtered" to some degree by the rest of the computer. GFFX sucks "unfiltered" air from the outside. I much doubt there will be a filter on the intake as the fan most likely isn't strong enough to overcome the drag created by a filter dense enough to make any kind of difference (intake opening is very small). Besides, an uncleaned filter would be even worse than no filter at all as it would clog much faster than the fan and heatsink will...

Nvidia uses a radial fan. That means many densely packed small fins. The more fins, the more places for dust to settle. The denser the arrangement, the quicker it will clog up. Also, those long tubes of copper in the heatsink look ideal for dust to settle in.

"2. There's also no reason to suspect that it will be any louder."

Oh wait, totally ignoring all the people who say it is rather damn loud you mean? Totally ignoring the "silent running" speed throttle feature as well?

Now, Nvidia said they'd work on getting the fan quieter, but according to the information we have NOW, it does make a fair amount of noise. I see no reason to try to FUD this over.

Sharkfood:
Not only does the thing have a plastic hood/resonator thingy mounted on it, the fan actually seems to be mounted ON THE HOOD ITSELF...


*G*

 

Geeforcer,

giving the talent we have here I would be dissapointed if we did not see new jokes

 

Actually, it's perfectly acceptable to design for a steady state condition of maximum power draw (heat dissapation) under maximum load. Your GPU and CPU do reach equilibrium under any extended period of heavy usage. They better, or your chip will soon suffer a meltdown!

From what I could gather, the equation as originally posted assumed no conductive losses, and addressed only the convective problem from heatsink to fluid. Probably not a gross oversimiplification.

It's not that the convection was not properly "accounted" for, but rather that it was "hidden away" in the nice little U term (or h being more commonly used for convective coefficients).

The rest of your post wasn't out in right field... the problem can and does get very complicated very quickly. But, the distinction between natural and forced convection is a factor that only affects the final value of h. Prandtl, Nusselt, Reynold's numbers... all will come into play in calculating h.

Which is precisely what I was referring to a bit earlier when I said "I love how your equation just has a U in there... that U being the most difficult part of the equation to even define, let alone solve."