Read more.And the AMD Radeon RX Vega pricing debate rumbles on.
Read more.And the AMD Radeon RX Vega pricing debate rumbles on.
What does this mean? Does it mean that they're covering twice the surface area? The clamping load on the chip from the sink is twice as much? The thermal paste gets into twice as many tiny little surface imperfections? If you can easily double the contact with the GPU to the point where it's being doubled then either everyone else is doing it realllly badly or you're probably doing something pointless.•MaxContact Technology that is 2X more contact with GPU for improved thermal transfer.
AMD are helpless IMO, they can't demand retailers sell for X amount (afaik that's illegal) and it's not like supply has been short, at least not going on what this WCCFTech article says...
From what i can tell distributors and retailers are increasing price purely because they can and because cryptocurrency miners are hovering up what supply there is, with AMD being hit hardest because they're better as cryptocurrency mining cards, there's next to nothing manufactures can do to control the price in the shops as increasing supply risks a collapse in cryptocurrency causing thousands of cards to go unsold and we're not talking small numbers hear, a recent article on MarketWatch says...that ‘tens and thousands of Vega GPUs were shipped out’ yet they appear to have run into unforeseen delays which meant that only a fraction of the available quantity was up for sale when the launch window opened.
Sales of add-in cards of AMD and Nvidia hardware were 520,000 units higher in the second quarter compared with the first quarter, according to JPR. Traditionally, we would expect the standard seasonal drop of 10,000-20,000 units. This indicates that upwards of 500,000 total units of high-end graphics were sold into the channel and, indeed, for mining-specific uses. About one in three graphics cards sold at retail, to OEMs or businesses was used for cryptocurrency mining.
Millennium (24-08-2017)
'flat' surfaces aren't perfectly smooth - there will be enough roughness even in a polished surface that the actual contact will be limited to a vanishingly small area of the overall area of the chip. So everyone is making heatsinks really badly, but it's also really hard to do it right.
The ratio of actual metal-to-silicon contact area to apparent contact area can be approximated as equal to the mounting pressure / the microhardness of the softer material [1]. Assuming a mounting pressure on the order of 10kg, a die area for vega around 600mm^2 (for the area of the GPU and memory stacks) and a microhardness of copper on the order of 1GPa[2], then the mounting pressure is a lowly 160KPa. This gives a contact area ratio on the order of just 0.01%, meaning you get just 0.1mm^2 of metal-to-silicon contact for heat conduction and the rest has to go through a few microns of thermal goop.
[1]: https://www.lepten.ufsc.br/publicaco...h_mantelli.pdf This goes into lots more detail on the subject, the Ar/Aa = P / H equation is the simplest one I could see with a quick scan through
[2]: http://citeseerx.ist.psu.edu/viewdoc...=rep1&type=pdf I know, they're looking at a weird alloy, but it's the closest i could find and it should be correct to within a factor of 10
Cheaper graphics cards! Now that would be a bummer...Furthermore, Kumquat asserts that AMD loosening supply "could put the entire GPU market in danger of a pricing bust should cryptocurrency prices take a dive".
There are currently 1 users browsing this thread. (0 members and 1 guests)