4870 is a heater too?

stonewalker

stonewalker

Gawd
Joined
May 26, 2007
Messages
765
lmao so i buy this card which is pretty fucking strong i must say but this thing is pushing so much hot air out the case that its heating up my room!lol, what options do i have.does the new nvidia gtx200's push out as much heat as the 4870 thx.
 
It only puts out the equivalent to a 60 watt bulb over what a 3850 puts out. Any high end new video card will put out a fair amount of heat.
 
my 8800gts 640 had it seems the same type of cooler but didnt put out this much hot air
 
There's not a lot of difference in the real world heat that these cards put out. A properly functioning GTX seems to be much better at heat and power management. When I'm running 3x SLI on a a demanding game, yes, my office windows AC unit is turned on since its summer. But I've had hotter systems. I had a Athlon X2 4800+ / 2x SLI 7800 512 GTX system and that puppy heated up I think a little more than the sig rig.
 
LOL, watch what you say around here or you'll get yourself reamed; check out THIS Thread
 
stonewalker said:
my 8800gts 640 had it seems the same type of cooler but didnt put out this much hot air
Click to expand...

That's correct. The power consumption of the 4870 is higher than the old 8800GTX, let alone the 8800GTS 640. It's amazing really, how well-tuned TSMC's 90nm process was; the 65nm and 55nm processes are leaky as hell.

I mean, cards barely twice as fast as your old 8800GTS are using MORE power, despite 1.5 process shrinks. That doesn't bode well for the future of video cards, because unless they fix the leakage problem, consumption will only become more pronounced.
 
defaultluser said:
That's correct. The power consumption of the 4870 is higher than the old 8800GTX, let alone the 8800GTS 640. It's amazing really, how well-tuned TSMC's 90nm process was; the 65nm and 55nm processes are leaky as hell.

I mean, cards barely twice as fast as your old 8800GTS are using MORE power, despite 1.5 process shrinks. That doesn't bode well for the future of video cards, because unless they fix the leakage problem, consumption will only become more pronounced.
Click to expand...

The 4850 is faster than the 8800GTX and uses less power, not sure what you are really complaining about. Which, of course, makes your argument that it is the 55nm process leakage at fault incorrect, as the 4850 and 4870 use the same GPU.
 
Meh, you will be thanking the 4870 come winter time =)

wtf, sorry for the double post, sight wouldnt load.
 
Some people need to understand that Pcs in general give off heat like any electrical device. The more power it needs, the more heat it will give off. The same thing happens with high performace GPUs. The more performing they are, the more heat they will give off generally speacking due to the higher TDP. Thats the sacrifice you must make when you go high-end.

The same thing happens with cars. Its you as the consumer who decides if you want brute force like a Ford Mustang GT 5.0 Lt or a Toyota Echo. One will give you more power, but cost a heck of lot more in gas.

The same thing applies to GPUs. So please lets stop complaining about how hot these cards get.. If hot isnt what you want, then go buy something less performing like a 3850 or something.
 
defaultluser said:
I mean, cards barely twice as fast as your old 8800GTS are using MORE power, despite 1.5 process shrinks. That doesn't bode well for the future of video cards, because unless they fix the leakage problem, consumption will only become more pronounced.
Click to expand...

Just wait a year... We'll have something faster and with all the power consumption problems solved. Just look what we had a year earlier...
 
I swear, sometimes, these hardcore pc forums are starting to sound like greanpeace forums lol. We live in truly surprising times when you see places lile these complain about video cards costing them too much on there electricity bill or over heating thier room.

What ever happended to the good old I NEED MORE POWER for my PC so I can rule the charts/gaming.....

More perforamance= more power=more heat output.
 
Spyhawk said:
The more power it needs, the more heat it will give off.
Click to expand...

Not necessarily.

Just look at Prescott P4's versus early Core 2's.

Both used about the same amount of power, but the fact that the Core 2 runs so much faster, and that it was built on a more power efficient platform (65nm too), Core 2 still made much less heat.

It's all about efficiency, the more heat given off = the more energy wasted. If something is more efficient, it can still use the same amount of power, but still run cooler.

Just like with light bulbs... compare a 100watt Incandescent light bulb vs a Compact fluorescent light bulb, both use power, but the fluorescent light creates less heat = more efficient with power.
 
if you take some duct tape you can cover up the vents on the back then it won't vent into your room..
 
Yea, the incandescent uses way more power and therefore wastes more as heat but that's comparing apples and oranges. With video cards it's more like 60w incandescent vs 100w incandescent.

If it has the SAME TDP then it puts ~ the same amount of heat into the room and therefore will heat your room about the same amount.
 
kllrnohj said:
The 4850 is faster than the 8800GTX and uses less power, not sure what you are really complaining about. Which, of course, makes your argument that it is the 55nm process leakage at fault incorrect, as the 4850 and 4870 use the same GPU.
Click to expand...

You don't understand how semiconductors work.

If you can ignore leakage, the power consumption should drop by half with every process revision (%100 improvement). This is because the capacitance required to charge a gate drops by half when you move down one process revision (e.g. 90nm to 65nm to 45nm). Yeah, you never see this, but in the past it was not unheard-of to hit %80 improvement with a process revision.

Thus, a card with similar horsepower to an 8800GTX on 55nm would use about half the power, or deliver twice the performance. But we're not seeing that. The 4850 (110w), despite the low voltage, uses just %80 of the power of the 8800GTX, and delivers about %20 more performance. That's a total improvement in both categories combined would be:

sqrt( %120^2 (performance) + %120^2 (lower power) ) = %170 (total improvement).

I was expecting around %200 total from 1.5 process shrinks. That %30 lost is due to increasing leakage. The fact is, leakage kicks in as your process gets finer, erasing your dynamic power gains.

In actuality, the two reasons the 4870 has high power consumption are (1) higher leakage and (2) higher voltage than the 4850. Reducing the voltage not only reduces dynamic power, it also reduces leakage.
 
colinstu said:
Just like with light bulbs... compare a 100watt Incandescent light bulb vs a Compact fluorescent light bulb, both use power, but the fluorescent light creates less heat = more efficient with power.
Click to expand...
I don't really understand what you're trying to say.

The amount of heat produced is directly linked to consumed power.
The fluorescent light bulbs consume much less power therefore run much cooler.
The incandescent bulbs are way more power hungry and run way hotter.

It doesn't matter what generation the card belong to or on what processed it's been build.

Two parts of the same power ratings will produce the same amounts of heat.
 
defaultluser said:
You don't understand how semiconductors work.

If you can ignore leakage, the power consumption should drop by half with every process revision (%100 improvement). This is because the capacitance required to charge a gate drops by half when you move down one process revision (e.g. 90nm to 65nm to 45nm). Yeah, you never see this, but in the past it was not unheard-of to hit %80 improvement with a process revision.

Thus, a card with similar horsepower to an 8800GTX on 55nm would use about half the power, or deliver twice the performance. But we're not seeing that. The 4850 (110w), despite the low voltage, uses just %80 of the power of the 8800GTX, and delivers about %20 more performance. That's a total improvement in both categories combined would be:

sqrt( %120^2 (performance) + %120^2 (lower power) ) = %170 (total improvement).

I was expecting around %200 total from 1.5 process shrinks. That %30 lost is due to increasing leakage. The fact is, leakage kicks in as your process gets finer, erasing your dynamic power gains.

In actuality, the two reasons the 4870 has high power consumption are (1) higher leakage and (2) higher voltage than the 4850. Reducing the voltage not only reduces dynamic power, it also reduces leakage.
Click to expand...


Thats becuase thier using the extra room to add more performance. Thats why you dont see so much lees power needs with a shrink. In order to stay competitive AtI or NV will always puch the TDP envelope to extremes...they cant afford not to in the high end.
 
Spyhawk said:
Thats becuase thier using the extra room to add more performance. Thats why you dont see so much lees power needs with a shrink. In order to stay competitive AtI or NV will always puch the TDP envelope to extremes...they cant afford not to in the high end.
Click to expand...


8800GTX number of transistors: 681 million

4850 number of transistors: 666 million.

This is actually why I took the comparison mentioned above and ran with it, because they have approximately the same number of transistors.

Also, I was comparing the 8800GTX with the 4850, not the 4870. Doing this actually takes the higher voltage of the 4870 OUT OF THE EQUATION. Thus, there is no excuse for the poor gains in efficiency moving from 90nm to 55nm except leakage.

You're also forgetting that I took into account the COMBINED improvements in power consumption and performance, for a TOTAL of %70 improvement. You can combine performance gains and power consumption reductions because they are directly related by the equation P = fCV^2.

What, am I the only electrical engineer on this forum? Read my entire post before you reply next time.
 
all i know is i feel hot with my x1800xt --> 8800gts512 in april--->4850 june. Heat is always going to be there its a gaming computer not a HTPC.
 
I have 2 of these kicking out heat into my room, feel the exahust fans at all other points of the case and it's very cool but behind these cards are kicking out a load of heat, my kitten has stopped sleeping on my AV reciever which heats up and keepers her tummy warm, and now sleeps beside the PC where the hot air builds up.
 
I noticed a difference between the 8800GTS 512 and 9800GTX. If I could do it over again, I wouldn't've wasted the $15 and time of stepping up. Both cards I feel never have cooling problems, but I do notice that the 9800GTX is warming my room up a little faster.
 
defaultluser said:
8800GTX number of transistors: 681 million

4850 number of transistors: 666 million.

This is actually why I took the comparison mentioned above and ran with it, because they have approximately the same number of transistors.
Click to expand...

The R770 (4850/4870) has 965 million transistors. The R600 (3870) has 666 million.

I think we just found out why the R700 has a TDP of the 90nm part...or close to it. If you run your math again to show the 3870 vs. the 8800...it probably won't do you any good on the performance side since its obviously not near an 8800GTX...but if you run it on the power end of things with TDP...you will see a larger number I imagine than what you are coming up with for a part that has 50% more transistors packaged.
 
In the summer, my computer room can get pretty hot..but I don't think my computer is heating up the room...Last summer I had a Shuttle xPC with an Athlon64 3500+ and X800XT.. this summer I've had the system in my sig. No difference. If it gets hot, it's simply because my apartment is on the 8th floor of a brick and concrete building, with most of the windows facing south-west..the little extra heat that the computer puts out is nothing compared to the baking sun during the 2-3 weeks of truly warm weather that we get in rainy, cold Sweden.

I'm sure that if I got some AC (which is a waste of money, because 95% of the year you need heating, not AC here), a computer wouldn't overpower it, no matter if it was a low-end PC or a power hungry Phenom with two GTX280's...

I've seen quite a few people complaining about the 4870 in particuloar, even though it doesn't get that much hotter than previous cards, let alone the GTX280.. Maybe the heatsink is more efficient, causing more of the hot air that the GPU generates to get exhausted out into the room.. With my 4850, I guess most of the hot air stays inside the case, causing it to heat up, but not affecting the temperature of the rest of the room much.
 
Kluden said:
The R770 (4850/4870) has 965 million transistors. The R600 (3870) has 666 million.

I think we just found out why the R700 has a TDP of the 90nm part...or close to it. If you run your math again to show the 3870 vs. the 8800...it probably won't do you any good on the performance side since its obviously not near an 8800GTX...but if you run it on the power end of things with TDP...you will see a larger number I imagine than what you are coming up with for a part that has 50% more transistors packaged.
Click to expand...

This is true, but it does not invalidate my analysis above. The problem is, we are comparing two different architectures, so you have to ask yourself: what should we compare?

Should we compare just the transistors? If we do that, we still come up short, with about %190 accounted for. If you compare just performance, you only get around %170 accounted for, showing possible inefficiencies in ATI's architecture.

But let's say for a moment that the number of transistors is the best choice for comparison. This gives you:

sqrt( %150^2 (performance) + %120^2 (power reduction) ) = %190

Now, I was actually being incredibly generous targeting only %200 in my previous posts. A 1.5 shrink step should net you a %267 improvement on-paper:

(90nm^2/55nm^2) = 2.67

I would be generous to expect only %220 on-paper (that's still a lot of overhead), which you can't approach even when you take into account the transistors.

Did I miss anything here? Sorry about missing the transistor count, it's the main reason I went on the performance track.
 
GlacierFreeze said:
Doesn't more heat output mean less heat on the GPU? That's a good thing...
Click to expand...

This is the first time in a long time I've read a rational comment about this topic. Personally, if somebody gave me two machines that both had the same performance and consumed the same power from the wall, I would take the one that exausted warmer air. Why? That means the temps inside the case (and the devices with it), "should" be lower. However, once steady state is achieved, both machines will "heat" the room equally.

colinstu said:
Not necessarily.

Just look at Prescott P4's versus early Core 2's.

Both used about the same amount of power, but the fact that the Core 2 runs so much faster, and that it was built on a more power efficient platform (65nm too), Core 2 still made much less heat.

It's all about efficiency, the more heat given off = the more energy wasted. If something is more efficient, it can still use the same amount of power, but still run cooler.

Just like with light bulbs... compare a 100watt Incandescent light bulb vs a Compact fluorescent light bulb, both use power, but the fluorescent light creates less heat = more efficient with power.
Click to expand...

150W is a 150W. You cannot violate the laws of thermodynamics. The only thing that isn't converted to heat in a computer system is the light coming out of the monitor...and that is only until it hits a surface and is well...converted to heat.

No matter what....no matter how much you want it not to be....all work done by the CPU and the GPU is converted to HEAT (and pretty much everything else inside the case). .PERIOD.
 
What size room is your computer in? I wouldn't think the amount of hot air a PC puts out is enough to heat a room with some air flow. Am I the only one who finds it hard to believe switching to a card that produces less heat will make that much of a difference?
 
cnick79 said:
What size room is your computer in? I wouldn't think the amount of hot air a PC puts out is enough to heat a room with some air flow. Am I the only one who finds it hard to believe switching to a card that produces less heat will make that much of a difference?
Click to expand...

it actually does make quite a difference.. i have a fairly big room and within an hour of turning on the computer the temperature in the room will increase by 1 or 2 degrees..
 
I live in a 650 sqft 1bdrm apartment and when i get home from work, my apt will go from 75F to about 80 in under an hour from turning my computer, tv and receiver on. But then again i'm running a 42" rear projection LCD, a q6600 @3.6ghz and crossfire 4870s. I will say this much about the "cooler" that the cards come with. Using just the stock cooler, the backside of the card used to get so hot it was almost untouchable, now that i have the cards on water, the backside of the card doesnt even heat up. The cooker cages that come with the cards do exhaust a lot of hot air, but i still think they are insufficient.
 
veedubfreak said:
I live in a 650 sqft 1bdrm apartment and when i get home from work, my apt will go from 75F to about 80 in under an hour from turning my computer, tv and receiver on. But then again i'm running a 42" rear projection LCD, a q6600 @3.6ghz and crossfire 4870s. I will say this much about the "cooler" that the cards come with. Using just the stock cooler, the backside of the card used to get so hot it was almost untouchable, now that i have the cards on water, the backside of the card doesnt even heat up. The cooker cages that come with the cards do exhaust a lot of hot air, but i still think they are insufficient.
Click to expand...
what kind of water setup do you have and how much does it cost.
 
veedubfreak said:
I live in a 650 sqft 1bdrm apartment and when i get home from work, my apt will go from 75F to about 80 in under an hour from turning my computer, tv and receiver on. But then again i'm running a 42" rear projection LCD, a q6600 @3.6ghz and crossfire 4870s. I will say this much about the "cooler" that the cards come with. Using just the stock cooler, the backside of the card used to get so hot it was almost untouchable, now that i have the cards on water, the backside of the card doesnt even heat up. The cooker cages that come with the cards do exhaust a lot of hot air, but i still think they are insufficient.
Click to expand...


maybe ur just full of hot air :p

seriously sounds like u don't have a good enough AC.
 
Well my video card has inspired me to create the new "My Damn Video Card Is Making My Damn Room Too Damn Hot So I'm Going to Lose Some Damn Weight So I Give Off Less Damn Body Heat" Diet.

Still working on a more market friendly name.
 
ALL4AMD said:
maybe ur just full of hot air :p

seriously sounds like u don't have a good enough AC.
Click to expand...

AC works fine, i'm just too cheap to run it :) Also doesnt help that my apt is above the laundry room. Good news is i never have to turn my heater on during the winter in colorado :D

As to my water set up, i'm running a swiftech apogee cpu block, DD universal chipset block, and swiftech 320 radiator on one loop, on the other loop is another swiftech 320 rad, and a DD maze5 block on each of the 4870s.
 
stonewalker said:
what kind of water setup do you have and how much does it cost.
Click to expand...

Wouldn't water cooling be a good solution if your room gets too hot? A big water tank should be able to accumulate alot of heat energy before it starts to dissipate it to the surrounding air..

Another solution might be to simply open the door and/or buy a small floor/pedestal fan..I got one running right now.. it only cost like 50 euros and it's nearly silent.. It might not drop the actual temperature in the room, but it does make it "feel" cooler and easier to breathe :)

Technophobe said:
Its getting close to the stage where PC's will make that hot cup of coffee for you. :)
Click to expand...

I'm thinking about returning my new next-gen coffee maker.. It makes great coffee, but it heats up the kitchen by at least 5C. :p
 
unleashed said:
I don't really understand what you're trying to say.

The amount of heat produced is directly linked to consumed power.
The fluorescent light bulbs consume much less power therefore run much cooler.
The incandescent bulbs are way more power hungry and run way hotter.

It doesn't matter what generation the card belong to or on what processed it's been build.

Two parts of the same power ratings will produce the same amounts of heat.
Click to expand...



Glade you could give us an explanation of the first law of thermodynamics :)

For those people that think their 4870 run's way too hot, modify the fan speed.
 
You must log in or register to reply here.
Back
Top