First power bill for my 6 GH/s rig

[DeathAndTaxes]

Again, if this statement were true, we'd all have to shut down our cars in the summer.

Can you please stop using the broken card analogy.  You may be right but your car analogy annoys the every loving shit out of me. 

Car operating temp is ~200F (90C).   The higher the heat differential the easier it is to move heat.  So car @ 200F and ambient at 100F = 100F difference.  You can accomplish same amount of cooling with a lot less airflow than trying to keep a GPU @ 60C/140F with ambient @ 100F = 40F difference.

Also car engines are watercooled for a reason.  It moves the heat into a radiator with massive surface area. 

So in summary
* the hotter the device is relative to ambient = easier to cool
* the larger the surface area is relative to ambient = easier to cool
* the higher the airflow = easier to cool

Just because a car can be kept "cool" (when cool = 100F over ambient temp) doesn't mean it applies to GPUs.

So please stop with the car analogy nonsense.  It doesn't help your case.

[1713520924]

1713520924

[sveetsnelda]

The analogy is far from "broken".

The car analogy wasn't meant to be overly technical or precise.  Judging by your posts, you obviously understand the basics of thermaldynamics.  My arguement was with someone who apparently does not.  If my argument was with someone who did, no broad analogy would have been needed.

That being said, everything in your post just explained exactly why my internal combustion engine analogy makes perfect sense.  He said something along the lines of "Video cards are hot.  When the air temperature warms up, enough heat can't possibly be removed".  My response was "Yeah?  Engines are hot too and we keep them temperature controlled in the summer just fine."

Yep.  Most modern automobile engines are water cooled for many reasons...  the biggest of which being that the engine is crammed into a compartment for aestetics and protection.  Noise and emissions are up there too.  I don't plan on putting my GPUs in a case and rolling them down the freeway anytime soon though.  I also don't burn hydrocarbons in my GPUs, so I don't need to worry about NOX levels.    Another reason they're water cooled is because it's easier to regulate the temperature to tighter tolerances than air cooling (using a thermostat).  If an internal combustion engine runs too cold, performance/efficiency suffers heavily.  There are plenty of modern air cooled internal combustion engines out there as well though.  I could give a pretty good dissertation on the subject because I study/build/repair/modify gasoline engines.  It's not very relevant to the discussion though.

Just because a car can be kept "cool" (when cool = 100F over ambient temp) doesn't mean it applies to GPUs.

So please stop with the car analogy nonsense.  It doesn't help your case.

Combustion engines produce substantial heat.  GPUs produce substantial heat.  Both require heat removal to prevent failure and both have very large temperature deltas compared to ambient temperature air.  Yeah, I see they are both completely unrelated.

[DeathAndTaxes]

Combustion engines produce substantial heat.  GPUs produce substantial heat.  Both require heat removal to prevent failure and both have very large temperature deltas compared to ambient temperature air.  Yeah, I see they are both completely unrelated.

A long winded nothing and you missed the only point.

Car's cooling system can effectively dump the heatload because a car can operate at 200F.  That is 100F OVER ambient under worst case scenario.   If a car's cooling system had to keep car's internal temp at <140F then a car (as built today) would be insufficient to operate when outside temp is 100F+.  If a car could operate effectively at 300F it would need a SMALLER cooling system (relative to heat load) to keep temps <300F when ambient is 100F.

Car = 200F operating temp = 100F Delta T to ambient*
GPU = <140F operating temp = <40F Delta T to ambient*

On edit:
* further and likely unecessary clarificaiton because sveetsnelda is either an idiot or a troll ...

Code:

ambient in this case would be 100F input air time. The input air time he indicates is sufficient to cool his farm.  Yes it may also be cooler parts of the year but your idiotic trolling aside it wasn't your claim that you could cool the GPU with outside air when airtime if 0F your claim was you could cool them when outside airtime was 100F.

The point of the 100F vs 40F is that given 100F AMBIENT AIR TEMP it is much easier to cool something (easier as in surface area and airflow requires) when you only need to keep it at 100F OVER AMBIENT.  It is much harder (as in amount of surface area and airflow required) to keep operating temps at only 40F OVER AMBIENT.  Yes sveetsnelda I know what over ambient means.  Maybe you forgot but your claims was you could cool GPU with 100F input air.  Not sure why it is sooooooooooooooo complicated but if your are cooling something and the T-In is 100F then that is your ambient temperature.  You can't cool anything below ambient without phase change so that becomes the baseline.   A perfect cooling system could keep card temp at Ambient.  Such a system would be massive though so the Delta over ambient determines the amount of cooling (in terms of surface area, and airflow) necessary.  40 Delta over ambient is harder to achieve than a Delta of 100F over ambient for the same heat load.

The fact that a car can operate (at 100F over ambient) doesn't prove anything.  A nuclear reactor can continue to operate at 600F over ambient that doesn't mean a water cooled rig that keeps a GPU at 600F over ambient will be effective.  Why do reactors run at 600F?  Because the cooling system can be smaller/cheaper/more efficient than if it needed to operate at 500F, or 200F (car), or 140F (GPU).

40F Delta T is not much to work with.  The smaller the Delta T the more volume necessary to transfer the same amount of thermal energy and the same surface area.  If you can keep an entire farm stable 24/7 with 100F input air temps well you are better than most.  My guess is you 100F temps are with low humidity? Still your stupid car analogy is just that ... stupid.

[Sannyasi]

I'm glad you don't design data centers for a living.  You would kill a lot of hardware with your lack of knowledge.

lawl

but more on topic = Invest slowly and build your farm over time. Huge bitcoin mining investments usually fail (Going off past posts I've found here). Slowly working your way up helps you learn alot of things you'd never consider otherwise, plus if you work up over time it's much easier to deal with anything that might go wrong.

[sveetsnelda]

 

A GPU can not operate at 100F over ambient.

In the winter, my inlet air is around 0F (or lower).  My GPUs are regulated by fans to around 160F (71C).  I guess mine are magical because that's a 160 degree difference and I have a few cards that are 3 years old and still doing GPGPU computation 24/7.  I'll just assume that you are trying to say something different here or you are setting some specific value for "ambient" in your head...

So using car as an analogy is as dubious as using a pressurized water reactor as an analogy.

You are completely overcomplicating the argument.  The original statement/arguement was that a GPU cannot be cooled with 90-100F air.  I said that the statement/arguement was wrong and silly because a GPU most certainly can if you provide it with a constant supply of fresh air.  You're arguing my side and apparently not realizing it.

GPU also don't work well at 345C either.

I never said or implied that they do.

If a car needed to operate at no more than 40F over ambient then cooling system would need to be radically changed.

Agreed.  At only 40F over ambient, the fuel would never combust.    The cooling system could just be stripped out entirely.     Jokes aside, I'll assume you're saying "coolant temperature at only 40F over ambient" because you were apparently quoting coolant temperature before when you used the 200F figure.  If you wanted a vehicle's coolant to run at only 40F above ambient on the freeway, just remove the thermostat.    No radical redesign necessary.  Don't say that I didn't warn you that it'll only burn 5 percent of its fuel though. 

GPU need to operate at no more than 140F (worst case scenario that is only 40F over ambient) thus car is a dubious analogy.

Err...  if that's the case, then everyone who fires up Crysis 2, Furmark, OCCT, or Furmark without manually controlling their fan speed will trash their GPUs.  Most of the recent ATI cards will happily run at 70-75C and will even stay stable to 80C.  The clock throttle on most BIOSes are set to around 90C (194F).  "No more than 140F" is pretty low.  Not only is it low, but it's lower than the manufacturer sets the auto-fan profile to.  They are certainly designed to run hotter.  Longevity is a different argument altogether.

Are you sure that you're not confusing the term "over ambient"?  This applies to water coolant temperature, but doesn't apply to air cooling (or air cooling a radiator) in the way that you are stating...  Engineers don't design an air cooled device (like a video card) to run at a certain temperature over ambient.  They design what temperature/temperature range the device should run at and then specify the minimum/maximum ambient temperatures that will allow it to stay within the set thermal threshold.

[DeathAndTaxes]

I'm done.  You indicated you can aircool at 100F ambient temps.   So yes ambient temp would = 100F.  40F OVER AMBIENT would be (here let me help you)

100F + 40F = 140F.

Keeping the GPU temp below 140F would require a cooling system that can achieve 40 deg F OVER AMBIENT when the Ambient temp is 100F.

It doesn't really matter what happens on the coldest day does it troll?  Since despite 5 posts you don't seem to grasp that I will take it you intent is just to troll and let you troll on.

[chiropteran]

Car = 200F operating temp = 100F Delta T to ambient.
GPU = <140F operating temp = <40F Delta T to ambient.

There is a problem with your well thought out argument.  Most GPU run fine up to 95C.  VRMs, which can sometimes be the limiting factor, function up to 120C.  The delta difference just isn't as big as you think it is.


Now, a video card will almost certainly last longer at 60C than it will at 90C, this is true.  But to make any accurate measure of value you need to look at the cost of the card over it's lifespan in each case and determine if the increased GPU lifespan is worth spending additional funds on cooling.  Then there are other variables, for example you might be better off and end up with lower temperatures AND lower overall cost by using a water-cooling system with outdoor air vs an indoor air cooling system with AC.

[sveetsnelda]

*shrug*.  Whatever.  You're missing the point.  An analogy is meant to be just that -- an analogy...  not a perfect comparison of every aspect.  It's done for simplicity in order to get someone to connect with one aspect of the analogy.  By your logic, the only valid analogy to cooling a GPU is cooling a GPU.

Yes, I'm obviously just a troll.    Look at my post history.  I'm an obvious troublemaker.  I try to correct someone's misinformation and that's trolling?  You correct misinformation in many threads, so you must be a troll too.    Your posts are generally informative and logical.  These have been derogatory and angry like you have a grudge.  Yes, feel free to ignore me if your responses are going to continue to be that way.  I'll go back to my 45Ghash of "magical" GPUs and continue to scale it with my air cooled "sorcery".

[DeathAndTaxes]

There is a problem with your well thought out argument.  Most GPU run fine up to 95C.  VRMs, which can sometimes be the limiting factor, function up to 120C.  The delta difference just isn't as big as you think it is.

A GPU running at 95C core and 120C VRM isn't going to last long.  A farm with dozens of GPU at 95C core and 120C VRM is going to be a complete nightmare to manage.  You will end up making less than minimum wage dealing with the nonstop crashes, hardlocks, equipment failure, and downtime.   I don't know any serious miner which runs the cards at 95C for 24/7 operation.  I shoot for 70C operating temps.  At 80C cgminer is dropping their clocks to keep temps in line.  VRM are pushed hard when mining.  They tend to run 20C+ hotter than core temps.  None of my cards have seen 95C core and hopefully never will.

[k9quaint]

A car with an air cooled engine would stop working during the summer if it became stuck in stop and go traffic.

Analogy fails.

[rjk]

I would love to have 0 degree air to cool my rigs year round, but I don't have that luxury. I need to find something that will work for me in hot humid weather.

[RDarke]

A car with an air cooled engine would stop working during the summer if it became stuck in stop and go traffic.

Analogy fails.

Hmmm. How do people drive VW Beetles(real ones) in the summer then?

[phorensic]

DeathAndTaxes, I would like to see sveetsnelda cool a large, dense mining cluster with 100F intake air.  All he needs is enough airflow, right? lol.  Have fun with the power and noise required to move that much air.  This is where water steps in and becomes way more efficient.  You still have to use forced airflow to dump the heat in the end, but it is NOTHING like designing a large scale air only system.  I'm glad he designs data centers for a living also.  He should tell Facebook they did it all wrong with their latest DC.

Edit:  I missed where he said on the last page in this thread where he lives in a dry climate that gets to 100F and he had to use evap to control the temps. lol

[k9quaint]

A car with an air cooled engine would stop working during the summer if it became stuck in stop and go traffic.

Analogy fails.

Hmmm. How do people drive VW Beetles(real ones) in the summer then?

Overheating on hills or in stopped traffic in hot weather is a common problem with the air cooled bugs & vanagons. They changed the engine to be watercooled to address this. There were also aftermarket oil cooling bolt ons to address this problem. Those engines were designed during an era when stop & go in 100 degree heat was not the usual situation (as it became in Los Angeles).

[cablepair]

i just want to comment here as someone who actually ran a decent size mining operation all summer long last year

unless you live in the arctic circle or some damn place your going to need a powerful and high quality air conditioner to cool the immediate space around your rigs - there's just no way around it

i had about 6ghash over the summer and i had a brand new Frigidaire 12.5k btu ac and on real hot days it struggled to keep up and the only area it was cooling was my mining room - and this is upstate ny, if you live in arizona or something similar your going to need some real cooling power

[RDarke]

A car with an air cooled engine would stop working during the summer if it became stuck in stop and go traffic.

Analogy fails.

Hmmm. How do people drive VW Beetles(real ones) in the summer then?

Overheating on hills or in stopped traffic in hot weather is a common problem with the air cooled bugs & vanagons. They changed the engine to be watercooled to address this. There were also aftermarket oil cooling bolt ons to address this problem. Those engines were designed during an era when stop & go in 100 degree heat was not the usual situation (as it became in Los Angeles).

If you say so. I've never had problems in well over 100 degree heat and left idling with it or my 911. I'd be more worried about my Fieros overheating than either of those. But then most people couldn't find the engine let alone take care of it.

[k9quaint]

A car with an air cooled engine would stop working during the summer if it became stuck in stop and go traffic.

Analogy fails.

Hmmm. How do people drive VW Beetles(real ones) in the summer then?

Overheating on hills or in stopped traffic in hot weather is a common problem with the air cooled bugs & vanagons. They changed the engine to be watercooled to address this. There were also aftermarket oil cooling bolt ons to address this problem. Those engines were designed during an era when stop & go in 100 degree heat was not the usual situation (as it became in Los Angeles).

If you say so. I've never had problems in well over 100 degree heat and left idling with it or my 911. I'd be more worried about my Fieros overheating than either of those. But then most people couldn't find the engine let alone take care of it.

The problem isn't idling, the problem is sustained high output at low speeds. Put 4 people in a Vanagon and then drive up a hill in 100 degree heat. Hook a trailer to your 911 and haul a couple thousand pounds uphill in 100 degree heat, see what happens when you do lots of work and are not going 60 MPH. FYI they switched the 911 to be water cooled in 1998.

Finally, those engines are technically oil-cooled and have radiators to dump the waste heat from the oil. Since they are actually liquid cooled, they are not really appropriate to this analogy.

[RDarke]

A car with an air cooled engine would stop working during the summer if it became stuck in stop and go traffic.

Analogy fails.

Hmmm. How do people drive VW Beetles(real ones) in the summer then?

Overheating on hills or in stopped traffic in hot weather is a common problem with the air cooled bugs & vanagons. They changed the engine to be watercooled to address this. There were also aftermarket oil cooling bolt ons to address this problem. Those engines were designed during an era when stop & go in 100 degree heat was not the usual situation (as it became in Los Angeles).

If you say so. I've never had problems in well over 100 degree heat and left idling with it or my 911. I'd be more worried about my Fieros overheating than either of those. But then most people couldn't find the engine let alone take care of it.

The problem isn't idling, the problem is sustained high output at low speeds. Put 4 people in a Vanagon and then drive up a hill in 100 degree heat. Hook a trailer to your 911 and haul a couple thousand pounds uphill in 100 degree heat, see what happens when you do lots of work and are not going 60 MPH. FYI they switched the 911 to be water cooled in 1998.

Finally, those engines are technically oil-cooled and have radiators to dump the waste heat from the oil. Since they are actually liquid cooled, they are not really appropriate to this analogy.

Actually I have towed with the 911. It was only a small lawn trailer but it was loaded with parts so it was probably about 2k lbs. Not a problem except for the car being squirrely and not being able to get out of it's own way.

My Jeep, Blazer and Fieros all have oil coolers too. Doesn't make them oil cooled.

Try harder or we've derailed this thread enough.

[RandyFolds]

 

A GPU can not operate at 100F over ambient.

In the winter, my inlet air is around 0F (or lower).  My GPUs are regulated by fans to around 160F (71C).  I guess mine are magical because that's a 160 degree difference and I have a few cards that are 3 years old and still doing GPGPU computation 24/7.  I'll just assume that you are trying to say something different here or you are setting some specific value for "ambient" in your head...

So using car as an analogy is as dubious as using a pressurized water reactor as an analogy.

You are completely overcomplicating the argument.  The original statement/arguement was that a GPU cannot be cooled with 90-100F air.  I said that the statement/arguement was wrong and silly because a GPU most certainly can if you provide it with a constant supply of fresh air.  You're arguing my side and apparently not realizing it.

GPU also don't work well at 345C either.

I never said or implied that they do.

If a car needed to operate at no more than 40F over ambient then cooling system would need to be radically changed.

Agreed.  At only 40F over ambient, the fuel would never combust.    The cooling system could just be stripped out entirely.     Jokes aside, I'll assume you're saying "coolant temperature at only 40F over ambient" because you were apparently quoting coolant temperature before when you used the 200F figure.  If you wanted a vehicle's coolant to run at only 40F above ambient on the freeway, just remove the thermostat.    No radical redesign necessary.  Don't say that I didn't warn you that it'll only burn 5 percent of its fuel though. 

GPU need to operate at no more than 140F (worst case scenario that is only 40F over ambient) thus car is a dubious analogy.

Err...  if that's the case, then everyone who fires up Crysis 2, Furmark, OCCT, or Furmark without manually controlling their fan speed will trash their GPUs.  Most of the recent ATI cards will happily run at 70-75C and will even stay stable to 80C.  The clock throttle on most BIOSes are set to around 90C (194F).  "No more than 140F" is pretty low.  Not only is it low, but it's lower than the manufacturer sets the auto-fan profile to.  They are certainly designed to run hotter.  Longevity is a different argument altogether.

Are you sure that you're not confusing the term "over ambient"?  This applies to water coolant temperature, but doesn't apply to air cooling (or air cooling a radiator) in the way that you are stating...  Engineers don't design an air cooled device (like a video card) to run at a certain temperature over ambient.  They design what temperature/temperature range the device should run at and then specify the minimum/maximum ambient temperatures that will allow it to stay within the set thermal threshold.

Cool, bro. You can argue semantics and provide poor analogies.

How about you just explain how 'two window fans' provide an adequate volume of air to compensate for the delta-t of 40F. You do this for a living; I'd love to see the calculations.

[pla]

That article was more theory than proven practice.  They only had 5 servers...in a tent, using way less power than something loaded with GPU's.  Lets try to scale that up and see what happens.
[...]
And yes, it just so happens that I have had to design data centers in my career.  Including power and cooling for them!

Then, as a data center designer, could you kindly explain to me what, exactly, fails to "scale up" when using outside ambient air to cool?  Just how high of a server density do you need before "outside" experiences a significant increase in temperature?

Trust me, Google and Facebook have looked very hard and written great articles about it.

Yes, they have.  And you have apparently read them, because you specifically mention their impressive improvements on the boring ol' "swamp cooler".  So no doubt, you know all about Facebook's newest DC in Lulea, where they expect to need less than two weeks of supplemental active cooling per year.

Now compare the form factor and uptime demands of a hardcore miner against a Facebook datacenter - And try to tell me with a straight face that you don't see just the teensiest difference between a room full of mid-tower PCs loaded with GPUs that can go down for an hour or two in mid-afternoon on the hottest days of the year with no real harm done, vs row after row after freakin' row of 24core x 42U racks with a contractually guaranteed six-nines uptime?