Baking GCN - The card with no fans

[Erenhardt]

How far do you push your GCN? Why not just roast it?!

full review:
http://hexus.net/tech/reviews/graphics/55413-powercolor-scs3-hd7850/
Is it the fastest passive cooled card?
It looks like the card starts to throttle down above 98'C

What you say about that Kepler?
I guess I'm going to disable fan on my hd7770

 

[BallaTheFeared]

102C, do not want.

I was mining in my garage at 1100MHz and 30% fan speed (somehow managed to disable user defined curve with remote desktop), but I also have a 3000rpm scythe kaze hitting the backs of both cards and it was hitting 93C...

I died a little on the inside, almost as bad as seeing my 470 boiling the water in my loop when the pump died.

 

[Rvenger]

Typical of power color.

 

[Stuka87]

Typical of power color.

Huh?

 

[SlowSpyder]

Typical of power color.

It looks like a decent enough design as long as you have case fans running (if you're into passively cooled cards that is - I'm not).

 

[Imouto]

Man, removing that stupid sticker should lower the temps a lot.

 

[birthdaymonkey]

I was hoping this would be a thread about reviving poorly soldered cards by baking them in the oven like I did with my GTX 280. Oh, the memories...

 

[Dankk]

Dem temperatures.

I would definitely remove that plastic sticker thing on the heatsink.

 

[Erenhardt]

Dem temperatures.

I would definitely remove that plastic sticker thing on the heatsink.

IMHO, heatsink should be above PCB when the card is pluged into PCIe slot. Heat goes up, so it feels counter-intuitive having radiator below heat source. They should plug it through flexible PCIe extension and mount the card with heatsink facing up.

 

[Imouto]

IMHO, heatsink should be above PCB when the card is pluged into PCIe slot. Heat goes up, so it feels counter-intuitive having radiator below heat source. They should plug it through flexible PCIe extension and mount the card with heatsink facing up.

I don't know why GFX cards have been facing downward for so long. It's stupid in every metric: GPU cooling, fan longevity and circuitry protection.

 

[24601]

102C, do not want.

I was mining in my garage at 1100MHz and 30% fan speed (somehow managed to disable user defined curve with remote desktop), but I also have a 3000rpm scythe kaze hitting the backs of both cards and it was hitting 93C...

I died a little on the inside, almost as bad as seeing my 470 boiling the water in my loop when the pump died.

I was actually thinking that a passive bucket style water cooling system with just a bucket, loop and a pump for 7970/7950 would be the cheapest quiet solution.

Ghetto Phase change system ^_^

My Sapphire 7970 doesn't crash up to 105C so it might work.

Maybe I can get it to recapture some of the energy as well to lower power costs since its essentially a boiling water reactor .

Full water-blocks for 7970 are expensive though, wish they would go down to 50 USD.

 

[thilanliyan]

^Just use a universal block and use alternate cooling for the VRM and VRAM. I've been using this DD Maze4 GPU block since my X1800XL lol...I had to finally modify it for the 7970s rotated GPU, but it is still chugging along.

 

[24601]

^Just use a universal block and use alternate cooling for the VRM and VRAM. I've been using this DD Maze4 GPU block since my X1800XL lol...I had to finally modify it for the 7970s rotated GPU, but it is still chugging along.

The thing is I'm not going to do water cooling unless I can get rid of all fans on the product itself and do it for 50 USD or less.

My use case has the VRM and VRAM heating up massively (Litecoin mining) and 100C on both is not uncommon at 78-80F ambient with best case at 100% current Sapphire Dual-X fan.

There is no way that a setup that you describe will be able to dissipate the required heat from those components.

 

[BallaTheFeared]

Cheapest way you could do it would be a universal block and a rasa cpu kit, they're like $120 and you can sell the cpu block for probably $35 on the forum.

Still you'd need some kind of air movement over the rad, otherwise the water will just keep absorbing the heat until it begins to boil. Unless you have a seriously huge drum of water that would dissipate the heat well enough to keep the overall water temp low it's a lost cause.

 

[Cerb]

I don't know why GFX cards have been facing downward for so long. It's stupid in every metric: GPU cooling, fan longevity and circuitry protection.

Because BTX was an attempt to make ever-hotter CPUs work, instead of being an overall improvement. You have to get the somewhat rare inverted cases, to do it w/ ATX.

 

[Rvenger]

Huh?

Sorry my phone separated PowerColor. I meant this is a typical design of powercolor to lack the necessary cooling elements for their cards. They usually use OK coolers but lack in the VRM/Memory cooling areas massively.

 

[thilanliyan]

The thing is I'm not going to do water cooling unless I can get rid of all fans on the product itself and do it for 50 USD or less.

My use case has the VRM and VRAM heating up massively (Litecoin mining) and 100C on both is not uncommon at 78-80F ambient with best case at 100% current Sapphire Dual-X fan.

There is no way that a setup that you describe will be able to dissipate the required heat from those components.

I'm LTC mining on my MSI 7950 with the Maze4 block and half-baked custom VRM cooling. My card is undervolted though to 0.988v. If you want to make it work, you can...biggest issue are the VRMs, and I could do a better job than I have, but it is enough for my purposes.

 

[Imouto]

Because BTX was an attempt to make ever-hotter CPUs work, instead of being an overall improvement. You have to get the somewhat rare inverted cases, to do it w/ ATX.

I'm not that concerned about the matter. It's just that you could have pretty things in your PC if it wasn't stuck in the 90's.

 

[24601]

Cheapest way you could do it would be a universal block and a rasa cpu kit, they're like $120 and you can sell the cpu block for probably $35 on the forum.

Still you'd need some kind of air movement over the rad, otherwise the water will just keep absorbing the heat until it begins to boil. Unless you have a seriously huge drum of water that would dissipate the heat well enough to keep the overall water temp low it's a lost cause.

Yes, the boiling of the water to keep the temps at around 100C is what I'm after, since I'm told 100C is the temps of boiling water

Phase change for efficiency . The bucket will literally be a reservoir specifically for more water to boil.

Also would make a kick-ass humidifier.

I'm LTC mining on my MSI 7950 with the Maze4 block and half-baked custom VRM cooling. My card is undervolted though to 0.988v. If you want to make it work, you can...biggest issue are the VRMs, and I could do a better job than I have, but it is enough for my purposes.

That's why yours is working for you.
I'm running at specs listed in my sig:
Sapphire Dual-X 7970 Boost 70.1% ASIC Quality 1.175v@1140 core 1.700v@1870 ram

With those volts and clocks there's no way a passive cooling of VRMs and VRAM would work out.

 

[BallaTheFeared]

My 470 actually shut down, though that's with a broken pump so no water movement. I really have no idea how well it would work with the pump running but the rad fan turned off (or a bucket of water in your case).

How much heat are you looking to dissipate?

 

[24601]

My 470 actually shut down, though that's with a broken pump so no water movement. I really have no idea how well it would work with the pump running but the rad fan turned off (or a bucket of water in your case).

How much heat are you looking to dissipate?

200-300W per 7970/7950, looking to hook up 1-3 of them to this setup if I can get the full block water blocks for cheap.

And the way I see it, the water won't let the temps get very far above 100C, especially since the boiling point of water where I live is ~98.7C

http://www.engineeringtoolbox.com/boiling-points-water-altitude-d_1344.html

If the temp readings on the GPU aren't lieing I can run these particular GPUs at 100-101C with slightly lowered clocks rock solid stably.

The sound of boiling water (basically a humidifier) from the bucket is not going to be anywhere near the absolutely earsplitting sound of the combined 100% fan speed from 3 7970/7950s

Would also have the side benefit of not allowing much of anything to grow in the water, since even 60-70C should be enough to stop most microbial growth as long as its constant.

1 kg * 2261 KJoules/kilogram = 2261 kJoules

Watt = Joule / Second

Assuming each card was running at 300W (powertune limit)

60 x 300 = 18000 joules per minute x 60 = 1080000 joules per hour

1080000 / 2261000 = 0.47766475011 Liters per hour per 300w assuming only phase change cooling excluding other kinds of dissipation.

http://forums.bit-tech.net/showthread.php?t=230772

I don't know why these things cost 100+ USD
You don't need to mill water-blocks from a solid block of metal, you could just easily mill cast water-blocks which would require far less milling.

 

[BrightCandle]

Using boiling water has a couple of problems.

1. Water cooling loops aren't designed to be open. We use distilled water as any particles in the liquid will clog up the blocks which have incredibly fine metal work on the inside that clogs easily.

2. The water will be under pressure meaning it will boil at a higher temperature.

3. Just because the water is around 100C does not mean your components will be. Under load expect them to be at least +20C from the water temperature.

You can achieve a very quiet system by using the appropriate amount of radiators and low speed fans. 800rpm is almost silent and given that a single 120mm slot of a thick radiator can dissipate around 110W to a water delta of 10C. That means each card needs a 3x120mm radiator. 3 of those and all your cards will come down to 40-50C. You likely don't need that good so you could drop one of the rads, tar getting roughly 15-20C water deltaand run a hot loop. Just bare in mind you can't put a CPU in a hot loop.

Getting a machine quiet with water cooling is mostly about fan area, if you want really quiet then you just need more rads and slower fans.

 

[24601]

Using boiling water has a couple of problems.

1. Water cooling loops aren't designed to be open. We use distilled water as any particles in the liquid will clog up the blocks which have incredibly fine metal work on the inside that clogs easily.

2. The water will be under pressure meaning it will boil at a higher temperature.

3. Just because the water is around 100C does not mean your components will be. Under load expect them to be at least +20C from the water temperature.

You can achieve a very quiet system by using the appropriate amount of radiators and low speed fans. 800rpm is almost silent and given that a single 120mm slot of a thick radiator can dissipate around 110W to a water delta of 10C. That means each card needs a 3x120mm radiator. 3 of those and all your cards will come down to 40-50C. You likely don't need that good so you could drop one of the rads, tar getting roughly 15-20C water deltaand run a hot loop. Just bare in mind you can't put a CPU in a hot loop.

Getting a machine quiet with water cooling is mostly about fan area, if you want really quiet then you just need more rads and slower fans.

Lets continue this here:

http://forums.anandtech.com/showthread.php?p=35030569#post35030569

 

[joshhedge]

200-300W per 7970/7950, looking to hook up 1-3 of them to this setup if I can get the full block water blocks for cheap.

And the way I see it, the water won't let the temps get very far above 100C, especially since the boiling point of water where I live is ~98.7C

http://www.engineeringtoolbox.com/boiling-points-water-altitude-d_1344.html

If the temp readings on the GPU aren't lieing I can run these particular GPUs at 100-101C with slightly lowered clocks rock solid stably.

The sound of boiling water (basically a humidifier) from the bucket is not going to be anywhere near the absolutely earsplitting sound of the combined 100% fan speed from 3 7970/7950s

Would also have the side benefit of not allowing much of anything to grow in the water, since even 60-70C should be enough to stop most microbial growth as long as its constant.

1 kg * 2261 KJoules/kilogram = 2261 kJoules

Watt = Joule / Second

Assuming each card was running at 300W (powertune limit)

60 x 300 = 18000 joules per minute x 60 = 1080000 joules per hour

1080000 / 2261000 = 0.47766475011 Liters per hour per 300w assuming only phase change cooling excluding other kinds of dissipation.

http://forums.bit-tech.net/showthread.php?t=230772

I don't know why these things cost 100+ USD
You don't need to mill water-blocks from a solid block of metal, you could just easily mill cast water-blocks which would require far less milling.

Your maths is wrong for calculating the number of litres per hour. 0.48 Litres per hour is a really small amount. Work out the temperature you wish the GPU to be cooled to, such as 80 degrees from 100, so you have your delta T, 20. The specific heat capacity of water being 4.18 joules per kelvin per gram, baring in mind that the delta T for kelvin and degrees Celsius is the same.

Using the formula Q=MCwdeltaT;

1080000 = 4.18*M*20

M = 12918g - 12.9Kg of water - 12.9 Litres of water per hour.

 

[24601]

Your maths is wrong for calculating the number of litres per hour. 0.48 Litres per hour is a really small amount. Work out the temperature you wish the GPU to be cooled to, such as 80 degrees from 100, so you have your delta T, 20. The specific heat capacity of water being 4.18 joules per kelvin per gram, baring in mind that the delta T for kelvin and degrees Celsius is the same.

Using the formula Q=MCwdeltaT;

1080000 = 4.18*M*20

M = 12918g - 12.9Kg of water - 12.9 Litres of water per hour.

I wasn't using the specific heat of water.

The liters per hour was to make sure I could have enough water in the bucket for extended runs without all of it boiling away too quickly to be manageable (the liters per hour were water loss due to evaporation from the system per hour).

I was using the heat of vaporization of water as I was planning on having only the vaporization of the water as the primary heat dissipation method out of the system.

Anyway I abandoned the idea as the huge expense of the full-length water blocks makes the whole cost savings pretty pointless compared to a traditional loop.

Continued here:

http://forums.anandtech.com/showthread.php?p=35030569#post35030569