Heatsink optimization for Mining? Thoughts?

[opieum2]

One thing I have consistently noticed with miners received, is the incredibly poor milling quality of the heatsinks. Without singling out any one vendor as most if not all have this issue, the heatsinks are poorly milled on the contact point with the chips. From my own experience this actually can create a fair bit of heat.

Sanding them down to an almost mirror finish helps quite a bit with better contact and therefore better temperatures in the chips themselves. Any reason why no manufacturer does this?

Usually this is an issue that would be relegated to the source they get the heatsinks from, but should be noted.

Additionally is there a reason why heatpipes or copper heatsinks are not used in ASIC hardware? Cost seems to be one part, but considering the cost of some of this hardware, it seems to be better to just add it in.

Seems like it would be simpler to just have the excess heat drawn upwards into the heatsink using the liquid closed in the pipes and have it cooled out. It is used with great effect on PC hardware, this is also high performance hardware that generates intense heat.

[1715294905]

1715294905

[1715294905]

1715294905

[1715294905]

1715294905

[Blazed]

My guess is just costs for the nicer heat sinks. I do not think ASIC manufactures have a very good margin these days so they cut costs everywhere possible.

[opieum2]

Honestly if they provided specs on the heatsinks alone, then we could get them made to order from their suppliers. Least offering the option at a premium would not be so bad. Cooling is one of those real important things for the hardware to last as long as possible. Also opens the door to some manufacturer overlocking options so more hash power can be gained from the hardware itself. Makes it profitable a bit longer at any rate.

[dogie]

Sanding them down to an almost mirror finish helps quite a bit with better contact and therefore better temperatures in the chips themselves. Any reason why no manufacturer does this?
...
Seems like it would be simpler to just have the excess heat drawn upwards into the heatsink using the liquid closed in the pipes and have it cooled out. It is used with great effect on PC hardware, this is also high performance hardware that generates intense heat.

Because 1) its not required, no ASIC chip is heat limited, 2) it would increase unit costs by $5-15 a heatsink for which you already complain about prices enough.

[opieum2]

Sanding them down to an almost mirror finish helps quite a bit with better contact and therefore better temperatures in the chips themselves. Any reason why no manufacturer does this?
...
Seems like it would be simpler to just have the excess heat drawn upwards into the heatsink using the liquid closed in the pipes and have it cooled out. It is used with great effect on PC hardware, this is also high performance hardware that generates intense heat.

Because 1) its not required, no ASIC chip is heat limited, 2) it would increase unit costs by $5-15 a heatsink for which you already complain about prices enough.

Well I dont mind the premium if it gets more performance out of the hardware, thats just me...not to say I would not like cheaper either but I get the pragmatism as well behind cheaper heatsinks. But again with cooling being such a factor, also factory overclocks would be a nice win and an extension of shelf life of some of this hardware if it can be pushed a bit harder.

[klondike_bar]

Sanding them down to an almost mirror finish helps quite a bit with better contact and therefore better temperatures in the chips themselves. Any reason why no manufacturer does this?
...
Seems like it would be simpler to just have the excess heat drawn upwards into the heatsink using the liquid closed in the pipes and have it cooled out. It is used with great effect on PC hardware, this is also high performance hardware that generates intense heat.

Because 1) its not required, no ASIC chip is heat limited, 2) it would increase unit costs by $5-15 a heatsink for which you already complain about prices enough.

Well I dont mind the premium if it gets more performance out of the hardware, thats just me...not to say I would not like cheaper either but I get the pragmatism as well behind cheaper heatsinks. But again with cooling being such a factor, also factory overclocks would be a nice win and an extension of shelf life of some of this hardware if it can be pushed a bit harder.

most of the limits on the hardware are not just heat, but the amperage. DC/DC regulators are semi-pricy, and at 0.75V on a lot of asics, you get very high amperage draws when you have hundreds of watts of power draw.

you could charge $20 more for a 10% better heatsink, but youd likely see only 2-3% extra overclocking space, and likely more powerdraw/GH doing so. not worth it. cheap components are the best ROI

[opieum2]

Sanding them down to an almost mirror finish helps quite a bit with better contact and therefore better temperatures in the chips themselves. Any reason why no manufacturer does this?
...
Seems like it would be simpler to just have the excess heat drawn upwards into the heatsink using the liquid closed in the pipes and have it cooled out. It is used with great effect on PC hardware, this is also high performance hardware that generates intense heat.

Because 1) its not required, no ASIC chip is heat limited, 2) it would increase unit costs by $5-15 a heatsink for which you already complain about prices enough.

Well I dont mind the premium if it gets more performance out of the hardware, thats just me...not to say I would not like cheaper either but I get the pragmatism as well behind cheaper heatsinks. But again with cooling being such a factor, also factory overclocks would be a nice win and an extension of shelf life of some of this hardware if it can be pushed a bit harder.

most of the limits on the hardware are not just heat, but the amperage. DC/DC regulators are semi-pricy, and at 0.75V on a lot of asics, you get very high amperage draws when you have hundreds of watts of power draw.

you could charge $20 more for a 10% better heatsink, but youd likely see only 2-3% extra overclocking space, and likely more powerdraw/GH doing so. not worth it. cheap components are the best ROI

Good points, in other situations was also just trying to see if I can get better temps. Helps with cooling issues overall. If not the overclocking situation. That would have been a nice to have. But focusing on cooler systems has its uses. Means less overall effort to keep them cool small but can be significant powersavings as alternative means to cooling can be applied.

[dogie]

Good points, in other situations was also just trying to see if I can get better temps. Helps with cooling issues overall. If not the overclocking situation. That would have been a nice to have. But focusing on cooler systems has its uses. Means less overall effort to keep them cool small but can be significant powersavings as alternative means to cooling can be applied.

If you want better temps, wack a more powerful fan over it. Convective heat transfer scales with air speed as a root, ie 2x your airflow and you get root(2)x more heat transfer [assuming constant air temp].

The difference between fan models are pretty much $1 wholesale, so its far more efficient to put on a stronger fan than spend $10 on a larger, heavy heatsink.

[unrealistic]

If you want better temps, use water cooling.