That's not how it works. The last 3 generations were all done on the same process, a cheaper process would not offer the same performance. The biggest advancement in the silicon industry is using the best possible manufacturing process. 28nm is cheaper than 14/16nm, but just compare the Titan X to the 1080. It's twice the size (probably around the limit when it comes to chip size) of a 1080 chip. The 1080 clocks higher and offers the same amount of transistors at a lower size.
There are no alternatives besides using the most expensive process tech.
If that's the case, how do they have define a lower bound of performance for the chip? How do they eve have a standard for the set of internal components of the chip for a certain tier, say 1070, if the chip's damage (and its degree, e.g., # of damaged components) is random (it probably is)?
Exactly which components turn out damaged in some process of manufacture?
- - - Updated - - -
Wow.... the greed and the capitalism is strong with gpu manufacturers.
Last edited by Kuntantee; 2016-07-02 at 05:46 PM.
https://www.overclockers.co.uk/sapph...gx-37b-sp.html
Actually a good looker compared to most AIBs, and depending on how this performs, might be good value, but considering the ref 4GB models did release at £175 before the price went upto £188 for the cheapest model.
Also to note, seems like other gpus have exceeded the PCI express slot power before, also apparently according to Gibbo, they sold around 2000 units so far of the RX480, which is pretty phenomenal, this card might just be doing what AMD needed and thats regaining market share.
Will there be a 4GB version of Sapphire custom board? I am planning to get a 4gb, and upgrade to 1080 when I have some spare money.
According to gibbo AMD stopped producing the 4GB model, though unsure how true that is, think he said the 4GB models were there best sellers, if it prints money, don't see why not at some point.
@Judy Hopps
I am not willing to transform this discussion to one about the business model of GPU manufacturers but know this: Their current model is optimizing their own profits. There exist a model which they do not damage proper chips to keep high-end card's price stable.
They don't lower their yield, they just adjust for demand. For example, if AMD made 300 Polaris chips but only 100 of them are fully working. The other 200 are defective in some way, and engineers put the chips through testing to see which chips can be salvaged into RX 370's. Lets say about 150 of them, and the other 50 are thrown away cause they couldn't be saved. But AMD sells more RX 370's than 380's, so they may end up taking some perfectly working Polaris 10 chips and disabling them into a working RX 370.
Why not sell more 380's for a lower price? Cause AMD doesn't want to devalue the 380's for the customers benefits. Like I said, capitalism. Not all the chips from this wafer is a 100% working chip. Some have defects.
Sections of the chip just end up being damaged from manufactuering. This is just normal bad luck. Even still, AMD/Nvidia can hand pick the chips for better stability and overclocks. Those end up being FirePro's or Tesla GPUs for servers.
You can make it work in your favor. In the past they would disable those sections with drivers, which people got smart and hacked the drivers to turn cheap cards into expensive ones. Going even further, you can flash the bios to get around hacking the drivers. Nowadays AMD/Nvidia laser cut their chips physically to avoid this, but with 14nm and 16nm I wonder if they can continue to do this? It is an added expense to laser cut chips to prevent people from getting more for their money.Wow.... the greed and the capitalism is strong with gpu manufacturers.
- - - Updated - - -
Doesn't work like that. If you sell $1500 GTX 1080's, who would buy them? Nobody would. We'd be using the CPU to render graphics. When companies do these things, they're not making it cheap enough for you to buy, they're just pocketing more money.