AMD R9 390X Coming With Cooler Master Liquid Cooler + Estimated Performance

[Khvarrioiren]

True, but power consumption must be higher than TDP, so it gives a hand-wavy indication of power consumption. Energy lost as heat, too, is considered wasted, so TDP has a direct relation to efficiency.

Go ahead and explain my 280x's sub-200 watt power consumption, seeing how the power consumption must always exceed the TDP.

[othertomperson]

Go ahead and explain my 280x's sub-200 watt power consumption, seeing how the power consumption must always exceed the TDP.

 

http://en.wikipedia.org/wiki/Conservation_of_energy

 

Technically TDP is the dissipation from the cooler needed, though, to keep the temperature sufficiently low, not the actual total heat energy produced. Otherwise (pretending a 100% effective cooler could exist) your card's temperature would never increase beyond ambient.

[Khvarrioiren]

http://en.wikipedia.org/wiki/Conservation_of_energy

Technically TDP is the dissipation from the cooler needed, though, to keep the temperature sufficiently low, not the actual total heat energy produced. Otherwise (pretending a 100% effective cooler could exist) your card's temperature would never increase beyond ambient.

So with my rig currently drawing 500 watts from the wall with overclocks that far exceed the stock clocks when really pushed with 1200 watts worth of heat dissipation topping at 48c on both the CPU and GPU defy law of thermodynamics in not one, but two ways?

[othertomperson]

So with my rig currently drawing 500 watts from the wall with overclocks that far exceed the stock clocks when really pushed with 1200 watts worth of heat dissipation topping at 48c on both the CPU and GPU defy law of thermodynamics in not one, but two ways?

 

What are you actually arguing? The heat output from your GPU cannot be more than the power it takes from your power supply. The Thermal Design Power is the dissipation needed to keep the GPU running at what AMD deem to be a safe temperature.

[patrickjp93]

 

• 120 mm x 1 fan Xtreme form factor dual-core radiator
• 158mm x 133mm x 54mm (L x W x H)
• 16 FPI 25 Micron Copper Fins
• Now optimized for sub-800 rpm ultra-stealth fans
• Supercruise optimizations for scalable performance with higher speed fans
• 15% more tubing area in the same Black Ice® GTX™ 120 form factor

That was taken directly from Hardware Labs website on the Black Ice GTX 120.  You really believe that the 25 micron thin copper fins are hollow and water actually flows through them?  Are you really that daft?!

 

The following picture is a cutaway of the Alphacool Nexxos, clearly illustrating where the water pipes are.  In the lower left picture, you can even see that some of the fin sections aren't even attached to the pipes. 

 

This is the EK Coolstream.  Again, the same exact structural design as the above two radiators that you mentioned.  The liquid travels vertically through the pipes, and not through the fins in any way, shape, or form.

 

Again, you have absolutely no idea what you are talking about. Stop stating misinformation as gospel.

 

Those pipes split into the fins immediately after that block. The vertical ridges are there only for structure. Cut one of them clean in half right at the midsection and you'll see I'm right.

[patrickjp93]

What are you actually arguing? The heat output from your GPU cannot be more than the power it takes from your power supply. The Thermal Design Power is the dissipation needed to keep the GPU running at what AMD deem to be a safe temperature.

Yes it can. Power dissipated by a resistor is = current^2 * resistance. What you're misunderstanding here is that TDP is a measure of instantaneous heat presence, not power usage. One line of electricity can go through many resistors and generate a lot of heat summing to more than the power which drove the current through. That heat comes not only from the electricity but from the physical matter of the circuitry itself, just as accelerating any body will force it to lose mass over time and why accelerating an object with mass to the speed of light isn't possible.

 

This is exactly why overvolting shortens the lifespan of any non-superconducting circuit.

[decripple]

The dual 8 pin power has me skeptical about it being single GPU that would only be necessary with a TDP of over 300 watts. Which is ridiculous for a single GPU reference card IMO. If they can keep it relatively cool and quiet then I am all for it especially if that memory bandwidth is true.

[patrickjp93]

The dual 8 pin power has me skeptical about it being single GPU that would only be necessary with a TDP of over 300 watts. Which is ridiculous for a single GPU reference card IMO. If they can keep it relatively cool and quiet then I am all for it especially if that memory bandwidth is true.

Overclocking headroom.

[Khvarrioiren]

What are you actually arguing? The heat output from your GPU cannot be more than the power it takes from your power supply. The Thermal Design Power is the dissipation needed to keep the GPU running at what AMD deem to be a safe temperature.

 

There's still the point of you saying that the power consumption must exceed the TDP. What were you actually going at with that statement?

[patrickjp93]

There's still the point of you saying that the power consumption must exceed the TDP. What were you actually going at with that statement?

1) he's wrong about TDP HAVING to exceed power consumption (see Nvidia's 980 power consumption vs. TDP)

2) Usually it does nowadays, but AMD has very leaky silicon, making the TDP and power consumption very close.

[Khvarrioiren]

Those pipes split into the fins immediately after that block. The vertical ridges are there only for structure. Cut one of them clean in half right at the midsection and you'll see I'm right.

 

They don't. I have several snapped fins on my Alphacool rads, and lo behold, no liquid leaking out of there. Because,you know, they go through the pipes. If the liquid went through the fins themselves, they would rupture immediately.

[Memphaestus]

 

Those pipes split into the fins immediately after that block. The vertical ridges are there only for structure. Cut one of them clean in half right at the midsection and you'll see I'm right.

 

Look at this photo again (click on it) paying particularly close attention to the top left quadrant.

Notice that the cut is made in the middle of the radiator, with zero evidence of water flowing through the fins (just that they were destroyed by the saw). 

 

Perhaps instead of me scouring the web for ANY pc radiator with this miraculous new metallurgy technology with hollow 25 micron thin fins, how about you show me proof of your statement.  You would think that this being a revolutionary technology, vastly different than every other radiator in existence, that these manufacturers would post on their website somewhere that the water flows through the fins.  There is absolutely no mention of this on any of those websites.  

 

I have done a significant amount of research on watercooling for my own rig as well.  Why is it that no watercooling guide has ever mentioned radiators like this, especially if it is truly better performance like you claim.

 

Additionally, why is there no Heat Exchanger Patent showcasing this design? The reason being that it isn't possible at that thickness of metal with present day metallurgy technology.

 

Please link one credible source illustrating water flowing through the fins of any of those 3 radiators.  

[Sakkura]

TDP is thermal design power, or the cooling power required by the thermal dissipation system to keep the chip from thermal throttling.

 

Nvidia does not at all underestimate its TDP. Consumers make the mistake of thinking TDP = power consumption, something totally incorrect. Think of TDP as the sum of each (current^2 * resistance) quantity of every path in the GPU chip circuit at each moment. Current is a function of the amount of power being drawn in and the potentials (voltages) needed across each circuit path). If the resistance of the paths is low enough, even if AMD's chip draws less power, Nvidia's can have a lower TDP due to less heat being discharged per moment by the sum of the quantities I gave earlier. Also be aware this is far oversimplified and doesn't take leakage current or resistance scaling with heat of the circuits into account, but this is beyond what I could reasonably display here, and eventually you'd get lost without a thorough background in electrophysics.

 

That's just ridiculous. At steady state, heat dissipation is identical to power draw. If an R9 290 draws 210W and a GTX 970 draws 175W, they are dissipating 210W and 175W respectively. It's that simple.

 

What you're arguing, that some power disappears, violates conservation of energy, a fundamental concept in physics.

[Sakkura]

You clearly don't know what TDP is because you quoted the part that stated that that was the power consumption of the card, not the heat output. They are not the same thing. Just because they are both measured in Watts (that is, Energy in Joules per second) does not make heat out and electricity in the same thing!

 

So you also don't believe in physics. The heat output comes directly from the power consumption, at steady state those wattages are going to be identical. All the power drawn by a graphics card is converted to heat (a very tiny fraction is first converted to fan noise etc. but that quickly ends up as heat too).

[tnlich]

Did u just say "only 300w tdp part" ?!

Please give me Nvidia GPU that ON ITS OWN (not full system consumption) eats 300w.

Vram thingy is nice is agree. I just need to see the performance in games or other applications (if it does support other ones).

It clearly destroys 980 but when 390x is out Nvidia will have 1080 coming out and 390x can be forgotten. All depends on pricing really.

 

 

 Yeaaaaaa .. Your right on that about the 390X but the 395 X2 like the 295 X2 was made to leapfrog a generation or2 of Nvidia. Also about the Water cooling you was touching on earlier. I emagine that AMD is stepping there foot into that department not ONLY because of the heat factor and such on the card. But they are sidestepping the (FORGIVE ME) Aftermarket water cooling business like Frozen CPU. Have any of you guys priced what it would take to actually have a liquid cooled video card and what it takes to money wize to have this option. 800-1500 American is a bargan to what you get into and spend on the REAL water cooling solutions.

[DIV1D3]

The dual 8 pin power has me skeptical about it being single GPU that would only be necessary with a TDP of over 300 watts. Which is ridiculous for a single GPU reference card IMO. If they can keep it relatively cool and quiet then I am all for it especially if that memory bandwidth is true.

Well there are a few 290Xs with two 6+2 Pin connectors, and the 390X I think will consume even more power (sadly), but definitely kick some ass. It'll be up against the *speculated* GTX 980 Ti, and the 390 should equal the 980 in performance. 

[othertomperson]

 

Yes it can. Power dissipated by a resistor is = current^2 * resistance. What you're misunderstanding here is that TDP is a measure of instantaneous heat presence, not power usage. One line of electricity can go through many resistors and generate a lot of heat summing to more than the power which drove the current through. That heat comes not only from the electricity but from the physical matter of the circuitry itself, just as accelerating any body will force it to lose mass over time and why accelerating an object with mass to the speed of light isn't possible.

 

This is exactly why overvolting shortens the lifespan of any non-superconducting circuit.

 

Over-volting shortens the lifespan of a processor because the increased energy increases the probability of quantum tunnelling across the classically forbidden potential well, ie an insulator. If this probability is high enough then for all intents and purposes your semiconductor just becomes a conductor -- it no longer functions.

 

Accelerating a body doesn't cause it to lose mass... overzealous science teachers say that accelerating a body increases its mass, but they're not entirely correct. The relativistic mass increases, but that's a slightly different quantity.

 

Yes P=I²R=dE/dt=VI. You cannot have more heat energy leave a system than went in. What you are describing is perpetual motion.

 

So you also don't believe in physics. The heat output comes directly from the power consumption, at steady state those wattages are going to be identical. All the power drawn by a graphics card is converted to heat (a very tiny fraction is first converted to fan noise etc. but that quickly ends up as heat too).

 

Assuming that the GPU instantaneously converts all of the electrical energy it receives into heat and nothing else. Remember, Power is the rate of energy transfer. While the net energy transfer must be in equilibrium, it's not a given that the GPU is in steady state. In fact it cannot be because the GPU must convert to other forms of energy before ultimately ending up as heat. You're also assuming that all of the useful energy transfers occurring throughout the card do also completely become heat. I'm not sure that's a given, which is why I said that heat out must be less than the power in. It might be, but that only changes the statement to include "less than or equal to".

[KeltonDSMer]

That's just ridiculous. At steady state, heat dissipation is identical to power draw. If an R9 290 draws 210W and a GTX 970 draws 175W, they are dissipating 210W and 175W respectively. It's that simple.

 

What you're arguing, that some power disappears, violates conservation of energy, a fundamental concept in physics.

 

I was typing pretty much exactly this as your post popped up. Sure power consumption might not be equal to TDP in some real world scenarios, but power consumption always equals heat output. That's like physics 101: conservation of energy. A graphics card is 0% efficient in the classical sense. It does not have a "useful" work output. All of its electrical energy consumption is converted to heat. A graphics card drawing 200W is putting out 200W of thermal energy (well let's say 198W because of a couple ~1W fan motors).

[mr moose]

I think most of you guys should leave the physics to the guys that actually understand it.

 

Total Energy in must = Total energy out,    TDP is only a portion of energy out, there for it must be less than total energy in.

[KeltonDSMer]

TDP ratings (when discussing GPUs) are useless as far as I am concerned and only confuse consumers. TDP ratings for coolers are even more worthless. It boils my blood when I hear an AIBP or tower cooler maker talking up how many watts their new design can mediate. A cooler (heat exchanger) system has to be described by (energy/time)/deltaT or power/deltaT. If the "ambient" fluid (air) is the same temp as the chip there is zero energy transfer, making this TDP number worthless without the extra variable. It's the same as describing how fast a CPU is by clockspeed alone, doesn't work that way. Put a "300w" cooler up in space and it won't do anything, just like a stock Intel/AMD cooler works great if the tower is in a freezer.
 

 

TDP is not a set value of energy or power in this system, it is a made up number that doesn't mean much. TDP is not its own kind of energy, it is not relevant in an energy balance equation. It should be somewhat representative of power consumption under normal operating conditions, but we all know this isn't always the case (especially if a component is operating outside of factory specs, ie: OC/overvolt/undervolt, this number can be thrown out the window).

 

What is important is that energy consumption (actual power draw) is always equal to heat output for this kind of system. 100% of the electrical energy consumed is converted to heat. Graphics cards are 0% efficient, they do not have a "useful" work output (light from a flashlight or mechanical energy from a motor for example).

 

Of course it is fair to compare GPUs by looking at FPS/watt, which is often referred to as "efficiency". This is a valid metric, but keep in mind this is not actually a measurement of efficiency. Last time I checked "frames" are not a unit of energy.

 

OT: I am really excited to see what HBM brings to the table, can't wait for the 3xx series release!

[patrickjp93]

That's just ridiculous. At steady state, heat dissipation is identical to power draw. If an R9 290 draws 210W and a GTX 970 draws 175W, they are dissipating 210W and 175W respectively. It's that simple.

What you're arguing, that some power disappears, violates conservation of energy, a fundamental concept in physics.

Not at all. The entire reason silicon chips die is due to loss of mass over time. Depending on the resistance of the circuit you can easily generate more heat than the amount of power you dump into it. Go ahead and yank out your resistor circuit problem sets from your high school/college/university days. The extra energy comes from the circuit itself. There is no violation of the conservation of energy. That said, if you want the fully detailed version I'm going to have to get into quantum mechanics and multivariable calculus, something this website is not at all a good medium for.

[patrickjp93]

I think most of you guys should leave the physics to the guys that actually understand it.

Total Energy in must = Total energy out, TDP is only a portion of energy out, there for it must be less than total energy in.

Again, no! You are all forgetting the other source of energy in: the mass of the circuit. Any time you push electricity through a non-superconducting circuit, you shave off extremely small amounts of mass. It's just entropy at work. That's why TDP can exceed power usage quite easily. This is entry-level laws of thermodynamics and consequences of resistance in a circuit. This is second semester physics at any decent college/university in the US if you didn't take AP Physics C in high school.

[patrickjp93]

Over-volting shortens the lifespan of a processor because the increased energy increases the probability of quantum tunnelling across the classically forbidden potential well, ie an insulator. If this probability is high enough then for all intents and purposes your semiconductor just becomes a conductor -- it no longer functions.

Accelerating a body doesn't cause it to lose mass... overzealous science teachers say that accelerating a body increases its mass, but they're not entirely correct. The relativistic mass increases, but that's a slightly different quantity.

Yes P=I²R=dE/dt=VI. You cannot have more heat energy leave a system than went in. What you are describing is perpetual motion.

Assuming that the GPU instantaneously converts all of the electrical energy it receives into heat and nothing else. Remember, Power is the rate of energy transfer. While the net energy transfer must be in equilibrium, it's not a given that the GPU is in steady state. In fact it cannot be because the GPU must convert to other forms of energy before ultimately ending up as heat. You're also assuming that all of the useful energy transfers occurring throughout the card do also completely become heat. I'm not sure that's a given, which is why I said that heat out must be less than the power in. It might be, but that only changes the statement to include "less than or equal to".

Accelerating an object increases its momentum, but it also shaves mass in the tiniest of quantities. Also, I'm not violating the laws of conservation of energy. The circuit itself is losing mass over time, and a single atom has a ton of energy to give. Take out any circuit problem set from your first year in physics. You'll find the total power dissipated by the circuit can be greater than the electrical power pumped through it depending on the resistors you put in the way.

[othertomperson]

Accelerating an object increases its momentum, but it also shaves mass in the tiniest of quantities. Also, I'm not violating the laws of conservation of energy. The circuit itself is losing mass over time, and a single atom has a ton of energy to give. Take out any circuit problem set from your first year in physics. You'll find the total power dissipated by the circuit can be greater than the electrical power pumped through it depending on the resistors you put in the way.

 

p=mv, decreasing an object's mass would decrease its momentum. Furthermore, reducing an object's mass would make it easier to accelerate, making it easier to increase the velocity of a massive object the closer it approached c. The usual (mis)conception is that accelerating a massive object increases its mass. In reality mass is invariant but relativistic mass is a quantity which undergoes Lorentz transformations.

 

 

TDP ratings (when discussing GPUs) are useless as far as I am concerned and only confuse consumers. TDP ratings for coolers are even more worthless. It boils my blood when I hear an AIBP or tower cooler maker talking up how many watts their new design can mediate. A cooler (heat exchanger) system has to be described by (energy/time)/deltaT or power/deltaT. If the "ambient" fluid (air) is the same temp as the chip there is zero energy transfer, making this TDP number worthless without the extra variable. It's the same as describing how fast a CPU is by clockspeed alone, doesn't work that way. Put a "300w" cooler up in space and it won't do anything, just like a stock Intel/AMD cooler works great if the tower is in a freezer.

 

And there was me thinking that measuring a cooler's TDP at a specified temperature, such as 20-25 degrees C to match what is usually taken as "room temperature" would be the only sensible way of doing it, and certainly not in a vacuum. Still, your radiator would actually be a radiator in that case, losing heat only through BBR

[Notional]

Ugh, so many people trying to lecture me about TDP, and getting it wrong at that.

 

Either way, the entire point I was making, is that you cannot criticize AMD's powerusage or TDP, without taking it's performance into account. A 390x having 80% of a Titan Z' TDP is not a bad thing, if the perfomance is at least 80% of the Titan Z.

 

It's all speculation atm anyways, so let's get back to it, when this card is actually reviewed.