Sign in with
Sign up | Sign in
Your question

A principal problem: fans on gfx cards - share your thoughts

Last response: in Graphics & Displays
Share
Anonymous
June 14, 2004 9:46:21 AM

Archived from groups: alt.comp.periphs.videocards.nvidia (More info?)

Greetings all.

*** Warning, this post is a bit long (but not too long). I'd like to
see what people think of the issue, so please post your point of view
on the matter. :) 


PREFACE

I am ashamed to admit, but I have never used an actively cooled
graphics card on a computer of mine. Ideally I'd like to keep it this
way, but not because I stick with low-end cards, but because I'm
hoping companies would start taking notice.


PASSIVE COOLING

A good conduct, as far as I'm concerned, is Sapphire's Radeon 9x00
Ultimate series -- medium and high end cards with passive cooling.
Unfortunately there is no latest-gen (X800) equivalent. (Yes, this is
an nVIDIA newsgroup, but I am not aware of any passively cooled
nVIDIAs that are not very-low-end).

I'm not sure I get the big idea. Instead of using better heatsinks
and keeping it passive, companies just stick a fan, even on low-end
models where a slightly better heatsink would easily solve the
problem.

Perhaps some ppl buy their cards based on how "beefed up" it looks,
but I'm sure there are plenty of others who don't -- why not cater for
them too? Perhaps fans are cheaper than heatsinks, and saving 5 cents
is more important than making sure card function in the longer run.

Providing sufficient passive cooling for higher end models -- through
more exotic and bulky heatsinks and heatpipes -- would cost more, but
there are people who will pay more (me included).

For the very highest end (say 6800 Ultra Extreme [silly name]) passive
cooling may not be enough, that I can accept.


ACTIVE COOLING DONE RIGHT

If a fan is thrown in after all, why not do it right? Every
manufacturer uses a proprietary exotic design, and so, unlike the
situation with CPUs, it's difficult or impossible to find a
replacement in case of a failure. If fans are an integral part of
graphics cards, having mounting standards would only do good.

And there's the issue of thermal protection. I haven't been able to
find much info on that, but what I have read suggests most graphics
chips would fry in the case of a complete fan failure. Is that true,
or perhaps some gfx chips are able to handle it gracefully (a la
Pentium 4)?


Cheers!
Ehud Shapira
Anonymous
June 14, 2004 7:49:45 PM

Archived from groups: alt.comp.periphs.videocards.nvidia (More info?)

I agree that the fan is the most unreliable part of the gfx card,
I've had several die/ make excessive noise.
Problem is,
a fan not only prevents chip failure, it also improves performance
significantly. From a sales and marketing standpoint, there is no
component in a PC that is as sensitive
to performance as a graphics card. Regardless of HS design,
a card will always perform better with a fan, and performance sells.

Also, I notice that the trend is for quiet fan operation, especially
during 2D. For example, I have a BFG5600U. The fan comes on until
I boot into windows, then turns off. It comes on again only if
I launch a game. It stays on about 5 minutes after I exit,
turns off.
Currently I'm using an ATI x800pro. It comes on at high fan
speed (noisy), but part way thru POST, fan slows down and stays that way
until I turn off the computer. Fan is always on, but so quiet
I can't hear it.



> And there's the issue of thermal protection. I haven't been able to
> find much info on that, but what I have read suggests most graphics
> chips would fry in the case of a complete fan failure. Is that true,

no.

> or perhaps some gfx chips are able to handle it gracefully (a la
> Pentium 4)?

Yes, here's why: GPUs are CMOS devices so the only time
current flows is during switching (displacement current due to
capacitance of the interconnects.) So if the chip 'locks up'
due to a heat related failure, then current stops, and no more
heat is generated.

Jeff B
Anonymous
June 15, 2004 1:02:17 AM

Archived from groups: alt.comp.periphs.videocards.nvidia (More info?)

I have now a Aopen GF 5800 FX , and earlier i had GF 3 ti200 and GF ti4200
(and before that a voodoo 3 3000 and voodoo 2 12mb and voodoo 1 4mb). I have
never used a fan on any of these, and will not use fans in the future
either. I demand a quit pc (without loss in performance) and make it so. I
have no fan on the cpu either, only a 120 mm fan blowing gently over the gpu
and cpu, and one in front of the case and one in the psu. These fans are
manually adjustable so i can keep things cool enough in games (like far
cry), but most of the time i cant hear my pc. Earlier i had western digital
hd-drives and they made some noise even when idle, so to make my pc quiet i
switced these for a couple of seagates. I have overclocked my cpu and gpu
heavily and have a nice performance (ca. 5400 in 3Dmark 03). I does not have
to be loud. I have Zalman heatsinks one both gpu and cpu, and are very
pleased with this (cheap also).
Related resources
Anonymous
June 15, 2004 10:27:12 PM

Archived from groups: alt.comp.periphs.videocards.nvidia (More info?)

Vegar Skogland wrote:
> I have now a Aopen GF 5800 FX , and earlier i had GF 3 ti200 and GF ti4200
> (and before that a voodoo 3 3000 and voodoo 2 12mb and voodoo 1 4mb). I have
> never used a fan on any of these, and will not use fans in the future
> either. I demand a quit pc (without loss in performance) and make it so. I
> have no fan on the cpu either, only a 120 mm fan blowing gently over the gpu
> and cpu, and one in front of the case and one in the psu. These fans are
> manually adjustable so i can keep things cool enough in games (like far
> cry), but most of the time i cant hear my pc. Earlier i had western digital
> hd-drives and they made some noise even when idle, so to make my pc quiet i
> switced these for a couple of seagates. I have overclocked my cpu and gpu
> heavily and have a nice performance (ca. 5400 in 3Dmark 03). I does not have
> to be loud. I have Zalman heatsinks one both gpu and cpu, and are very
> pleased with this (cheap also).
>
>

Like you, I demand a very quiet PC, but I did it a little differently,
I run open case. This grants you a 'free' performance/quiet bugdet
that you can spend on quiet or performance in any proportion you like.
Because open case, I have no case fans, no case fan noise. The other
secret is to run as large a CPU fan possible because the larger the
fan, the higher the cooling/noise ratio. I run a $11 120 mm case fan
on my CPU, turning at 1700 RPM...very quiet. My third secret: I
removed my PS, took off the cover, and tucked it away inside my
case as well as I could. I repositioned the fan so it blows straight
down on the now exposed components. Although the fan is running at the
same speed as before, I have no 'wind tunnel' effect, and the noise is
greatly reduced with even better cooling. How ofetn can you do a
no-cost mod that gives better performance AND less noise? One could go
further and put a fan control in series with the PS fan for still less
noise.
To review: CPU runs cooler cause of big fan and heat isn't trapped.
All mobo components run cool cause heat isn't trapped.
This offers more cooling than running closed case with noisy
case fans.
PS is super quiet cause no wind tunnel effect.
The only other fan is the vid card fan, which is very quiet
on my x800pro.
Of course with the guts exposed, perhaps some would argue the computer
doesn't look so good. But it's tucked away under my desk, so this is
not an issue, I don't see it anyway. But I DO hear the noise from not
doing as I described, which is a real problem

Jeff B
Anonymous
June 16, 2004 12:38:54 PM

Archived from groups: alt.comp.periphs.videocards.nvidia (More info?)

--- @ jeff:

About fans improving performance:

More cooling certainly gives more headroom, but fans are still used
too much right now, and without a real mounting standard.

Sapphire's passively cooled cards show that it is possible to have
both passive cooling AND cutting edge performance (though the 9800XT
was dethroned by now).

Sure, overclocking is more likely with a fan, but I am so much more
concerned with quietness and reliability.


Temperature controlled fans:

That's not too useful when cards kick into full speed while playing
games. I want quiet environment for games too.


Your X800 pro:

Is it more quiet while playing games than is it at startup?


About cards frying:

I got the impression that quiet a few ppl got their cards permanently
damanged when the fan failed. I have a fan-failed MX440 which seems
damaged (although I can't absolutely be sure because I don't have a
reliable motherboard to test it on).

How are GPUs different from CPUs? They are targeted at 3D and not
general purpose computing, but otherwise should be pretty similar, no?
I still remember with horror the Tom's Hardware videos of HSF-less
Athlons going to 240C after 1 second and starting to smoke. Would gfx
chips be any different without specific safeguards?

Can you point me to someplace elaborating on what you said earlier?


Open cases:

This may not always lead to better cooling, depending on your case
design, cabling and fans.

Open cases collect more dust, too. :) 


--- @ Vegar:

Hand picking 3rd party heatsinks for computer parts is annoying, costs
more, and availability is a problem. Using them is a potential
warranty voider. But alas, it seems to be the only way.

What models of coolers are/were you using on the various cards? (Are
you looking for a buyer for the Ti4200 or Ti200? :) 

My attemps, two so far, at noise reduction have been futile, or worse.

One was a Thermaltake Volcano 8 for the CPU. I was fooled by false
noise figures specified in the mfg site -- it was the noisiest fan I
have ever witnessed. In the end I had to use a custom-built voltage
regulator to slow the fan down (couldn't find a Zalman Fanmate around
here). After all that, it was about as noisy as the generic cheapo
fan it replaced.

Second was a Vantec Stealth case fan advertised at 20-21dBa. Again,
false claims. Noise was at about the same level as the previous
generic fan.

My computer is far too noisy for me to be concerned about harddrive
idle noise. :) 

And to think I thought in my new computer (at the time) would be more
quiet than the dual-PSUed Pentium...


~ Ehud.
Anonymous
June 16, 2004 5:25:08 PM

Archived from groups: alt.comp.periphs.videocards.nvidia (More info?)

--- @ jeff:

About fans improving performance:

More cooling certainly gives more headroom, but fans are still used
too much right now, and without a real mounting standard.

Sapphire's passively cooled cards show that it is possible to have
both passive cooling AND cutting edge performance (though the 9800XT
was dethroned by now).

Sure, overclocking is more likely with a fan, but I am so much more
concerned with quietness and reliability.


Temperature controlled fans:

That's not too useful when cards kick into full speed while playing
games. I want quiet environment for games too.


Your X800 pro:

Is it more quiet while playing games than is it at startup?


About cards frying:

I got the impression that quiet a few ppl got their cards permanently
damanged when the fan failed. I have a fan-failed MX440 which seems
damaged (although I can't absolutely be sure because I don't have a
reliable motherboard to test it on).

How are GPUs different from CPUs? They are targeted at 3D and not
general purpose computing, but otherwise should be pretty similar, no?
I still remember with horror the Tom's Hardware videos of HSF-less
Athlons going to 240C after 1 second and starting to smoke. Would gfx
chips be any different without specific safeguards?

Can you point me to someplace elaborating on what you said earlier?


Open cases:

This may not always lead to better cooling, depending on your case
design, cabling and fans.

Open cases collect more dust, too. :) 


--- @ Vegar:

Hand picking 3rd party heatsinks for computer parts is annoying, costs
more, and availability is a problem. Using them is a potential
warranty voider. But alas, it seems to be the only way.

What models of coolers are/were you using on the various cards? (Are
you looking for a buyer for the Ti4200 or Ti200? :) 

My attemps, two so far, at noise reduction have been futile, or worse.

One was a Thermaltake Volcano 8 for the CPU. I was fooled by false
noise figures specified in the mfg site -- it was the noisiest fan I
have ever witnessed. In the end I had to use a custom-built voltage
regulator to slow the fan down (couldn't find a Zalman Fanmate around
here). After all that, it was about as noisy as the generic cheapo
fan it replaced.

Second was a Vantec Stealth case fan advertised at 20-21dBa. Again,
false claims. Noise was at about the same level as the previous
generic fan.

My computer is far too noisy for me to be concerned about harddrive
idle noise. :) 

And to think I thought in my new computer (at the time) would be more
quiet than the dual-PSUed Pentium...


~ Ehud.
Anonymous
June 17, 2004 12:45:54 AM

Archived from groups: alt.comp.periphs.videocards.nvidia (More info?)

> Sapphire's passively cooled cards show that it is possible to have
> both passive cooling AND cutting edge performance (though the 9800XT
> was dethroned by now).

But add a fan to the 'passive cooled cutting edge performance'
card, and the performance will improve. That's my point, performance sells.


>
> Sure, overclocking is more likely with a fan, but I am so much more
> concerned with quietness and reliability.

Reliability isn't an issue, but noise is. That's why many trade some
performance for low noise.

> Temperature controlled fans:
>
> That's not too useful when cards kick into full speed while playing
> games. I want quiet environment for games too.

Fan noise in games is typically drowned out by the inherent game
noise.

>
>
> Your X800 pro:
>
> Is it more quiet while playing games than is it at startup?

I guess I didn't make that clear....startup noise lasts 3 seconds,
is gone long before windows slash screen. Then, the noise is
virtually inaudible. The fan noise does not change after that, even if
a game is launched. Remember, I have a accurate way of determining
the source of noise cause I run open case. Since I have exactly 3 fans
total, and I can easily stop them one at a time, it's trival to
determine which one is making whatever noise I hear, however low level
that noise is. That's why I know that the GPU fan is the least noise
of a 3 very quiet sources. The PS fan is next loudest, the CPU fan
the most loud.

>
>
> About cards frying:
>
> I got the impression that quiet a few ppl got their cards permanently
> damanged when the fan failed.

I never heard of a GPU being damaged by fan failure.


> How are GPUs different from CPUs? They are targeted at 3D and not
> general purpose computing, but otherwise should be pretty similar, no?

Yes, correct. All are CMOS.

> I still remember with horror the Tom's Hardware videos of HSF-less
> Athlons going to 240C after 1 second and starting to smoke. Would gfx
> chips be any different without specific safeguards?

Can't speak for Tom or Athlons, but I have much experience with failed
CPU fans. On two different Pentiums, I've had failures due to
intermitant wiring open circuits. In all cases, the CPU worked fine for
a few minutes, then locked up. After cool down, everything was normal.

>
> Can you point me to someplace elaborating on what you said earlier?

I talked about many things, what specifically are you refering to?


> Hand picking 3rd party heatsinks for computer parts is annoying, costs
> more, and availability is a problem. Using them is a potential
> warranty voider. But alas, it seems to be the only way.

Since my x800pro is so quiet, I no longer have motive to experiment
with various GPU cooling ideas. However, if I had a noisy card
and wanted the highest (performance)*(quietness) product, I would do the
following: First, I'd get a Zalman all-copper HS, they make one
specifically for GPUs. Since there is no fan with them, you don't
have the problem of removing the fan. Then, I would get a 120 mm case
fan such as the one I have. It's rated at 32 db at 94 CFM airflow.
I would add a fan speed controller (or a 50 ohm resistor) in series with
one of the leads to cut the fan speed in half. At that point, noise
would be virtually inaudible. I would then place the fan on an adjacent
PCI card pointed upwards to the AGP card. This would blow about
40 CFM of air on the huge copper HS and result in great cooling.

>
> What models of coolers are/were you using on the various cards?

I have used only stock coolers except on my Geforce 3 card (Gainward).
When the fan failed, I did the trick above, only I didn't use the
Zalman HS. I just removed the defective fan and left the stock HS
in place.

(Are
> you looking for a buyer for the Ti4200 or Ti200? :) 

I have already sold those cards. All I have right now is the nvidia
5600u which I am using in my second computer which is dedicated
to recording/playback of HDTV files (off the air), i.e. I have
turned my backup computer into a giant TiVo for hi def television.

>
> My attemps, two so far, at noise reduction have been futile, or worse.
>
> One was a Thermaltake Volcano 8 for the CPU. I was fooled by false
> noise figures specified in the mfg site -- it was the noisiest fan I
> have ever witnessed. In the end I had to use a custom-built voltage
> regulator to slow the fan down (couldn't find a Zalman Fanmate around
> here). After all that, it was about as noisy as the generic cheapo
> fan it replaced.

Yeah, I have found the Thermal Take products to be very cheesy, very
loud. They can offer relativly good cooling because they
neglect the noise side of the equation. You can get either good
cooling or low noise easily. But getting both at the same time is
hard and therefore expensive (unless you do it my way of course).

>
> Second was a Vantec Stealth case fan advertised at 20-21dBa. Again,
> false claims. Noise was at about the same level as the previous
> generic fan.

Not surprised. That's why I would NEVER buy an integrated fan+HS
solution. The best thing is to get the best HS you can find (sans fan)
and add your own 120 mm fan. After much research, I found that the
Thermalright sk94 HS for the P4 is the best HS available today. Because it
isn't burdened by a worthless fan, it goes for only $50. It is
extremely high quality, all copper, 3 heatpipes. It cools
measureably better than the all copper Zalman it replaced, and is
less than half the mass. It has a very simple fan mounting scheme,
I added the $11 dollar 120 mm case fan plus a 50 ohm resistor mentioned
above spinning at
1700 RPM. Very quiet.

Jeff B
June 17, 2004 5:15:36 AM

Archived from groups: alt.comp.periphs.videocards.nvidia (More info?)

All a fan does is move air, which increase the heat convection rate. Heat
conduction is what controls the temperature at the die surface. If the fan
is properly used with the right heat sink (copper, silver) it will not have
to move allot of air and will remain silent. The trick is to spend the time
on the design of the heat sink, which most video card manufactures don't do.
Video card manufactures look at the bottom line. An overclockable card will
mean that you might not buy the next line of cards they produce, therefore
additional performance in the direction of proper cooling appears to be to
costly in future sale revenue. The only time a video card manufacture will
spend design time on the cooling of the card is when they have to in order
to keep it stable at their rated speeds. Marketing gimmicks saturate the
video card market allot of the bells and whistles on the card are just for
show, for example the leadtech two fan heat sink. Do the research before
you buy or modify a video card. A copper U1 heat sink on a video card with
copper ram sinks will do wonders to the performance of the most nvidia based
cards and there cheap/silent.

"Ehud Shapira" <ehudshapira@hotmail.com> wrote in message
news:fae70eb9.0406140446.5c478e0f@posting.google.com...
> Greetings all.
>
> *** Warning, this post is a bit long (but not too long). I'd like to
> see what people think of the issue, so please post your point of view
> on the matter. :) 
>
>
> PREFACE
>
> I am ashamed to admit, but I have never used an actively cooled
> graphics card on a computer of mine. Ideally I'd like to keep it this
> way, but not because I stick with low-end cards, but because I'm
> hoping companies would start taking notice.
>
>
> PASSIVE COOLING
>
> A good conduct, as far as I'm concerned, is Sapphire's Radeon 9x00
> Ultimate series -- medium and high end cards with passive cooling.
> Unfortunately there is no latest-gen (X800) equivalent. (Yes, this is
> an nVIDIA newsgroup, but I am not aware of any passively cooled
> nVIDIAs that are not very-low-end).
>
> I'm not sure I get the big idea. Instead of using better heatsinks
> and keeping it passive, companies just stick a fan, even on low-end
> models where a slightly better heatsink would easily solve the
> problem.
>
> Perhaps some ppl buy their cards based on how "beefed up" it looks,
> but I'm sure there are plenty of others who don't -- why not cater for
> them too? Perhaps fans are cheaper than heatsinks, and saving 5 cents
> is more important than making sure card function in the longer run.
>
> Providing sufficient passive cooling for higher end models -- through
> more exotic and bulky heatsinks and heatpipes -- would cost more, but
> there are people who will pay more (me included).
>
> For the very highest end (say 6800 Ultra Extreme [silly name]) passive
> cooling may not be enough, that I can accept.
>
>
> ACTIVE COOLING DONE RIGHT
>
> If a fan is thrown in after all, why not do it right? Every
> manufacturer uses a proprietary exotic design, and so, unlike the
> situation with CPUs, it's difficult or impossible to find a
> replacement in case of a failure. If fans are an integral part of
> graphics cards, having mounting standards would only do good.
>
> And there's the issue of thermal protection. I haven't been able to
> find much info on that, but what I have read suggests most graphics
> chips would fry in the case of a complete fan failure. Is that true,
> or perhaps some gfx chips are able to handle it gracefully (a la
> Pentium 4)?
>
>
> Cheers!
> Ehud Shapira
Anonymous
June 18, 2004 9:12:42 AM

Archived from groups: alt.comp.periphs.videocards.nvidia (More info?)

An overclockable card will
> mean that you might not buy the next line of cards they produce, therefore
> additional performance in the direction of proper cooling appears to be to
> costly in future sale revenue.

I don't get it, seems to me the opposite is true. If a manufacturer
gives me an OCable card, then naturally their reputation is
enhanced and ppl will be more likely to buy another card from them,
not less likely.

Jeff B
Anonymous
June 18, 2004 10:28:34 AM

Archived from groups: alt.comp.periphs.videocards.nvidia (More info?)

ehudshapira@hotmail.com (Ehud Shapira) wrote in message news:<fae70eb9.0406140446.5c478e0f@posting.google.com>...
> Greetings all.
>
> *** Warning, this post is a bit long (but not too long). I'd like to
> see what people think of the issue, so please post your point of view
> on the matter. :) 

>
>
> Cheers!
> Ehud Shapira

Well I personly would dream of sufficient passive cooling, but ...
it's not possible in my case. This 9800 Pro ( running at 9800 NP
speeds ) still makes to much heat to keep cool even with the mega
heatsinks Zalman provides...
Anonymous
June 19, 2004 7:32:49 AM

Archived from groups: alt.comp.periphs.videocards.nvidia (More info?)

I am a good example of what Jeff is talking about. I purchased a GF4 Ti4200
128mb Golden Sample from Gainward. It ran very nicely somewhere between
4400 and 4600 speeds. I didn't buy another card (skipping a couple of
generations) until the GF5900 was released. And as an example of what you
said, I also bought my new card (GF 5900XT) from Gainward, and it is also
the Golden Sample version which is running along very happily at better than
5900 speeds. So Gainward lost me as a customer for a couple of newer
generation cards, but kept me as a customer due to the quality and
overclockability of their cards.
Silvertip
"jeff b" <fakeaddy@fjyfj.com> wrote in message
news:D RuAc.48948$Hg2.20739@attbi_s04...
> An overclockable card will
> > mean that you might not buy the next line of cards they produce,
therefore
> > additional performance in the direction of proper cooling appears to be
to
> > costly in future sale revenue.
>
> I don't get it, seems to me the opposite is true. If a manufacturer
> gives me an OCable card, then naturally their reputation is
> enhanced and ppl will be more likely to buy another card from them,
> not less likely.
>
> Jeff B
>
!