Why does the Vega 64 cost so much?

alkatraz333.jh

Commendable
Jan 14, 2018
151
0
1,710
I dont understand it at all. Its performance is so much worse than a 1080ti, yet it costs more, or nearly the same. Could someone please educate me? I mean, people arent actually buying these, are they? You'd think AMD would see that they cant sell them and lower the price somehow.
 
Solution
D
HBM was first used in the Fury cards and it gave a huge memory bandwidth boost over what was possible with GDDR. Unfortunately for AMD that bandwidth didn't translate to raw performance so the 980 Ti ended up as the faster card. AMD spent a huge amount of money developing the tech with Hynix and not only were they were contractually obligated to keep using it but new card designs take years to go from the drawing board to the fab. So Vega ( which was delayed over a year ) was already being designed to use HBM 2. They couldn't just redesign the memory controller.

Turns out they still weren't able to implement it very well even with a second generation of cards. Vega also had process problems. The architecture was designed for mobile...


The release price was 500$ and 700$ for the liquid cooled and I believe they are much more higher now
 
Mining is only part of it. HBM is expensive and at release Vega 64 cost about the same as a 1080 Ti while only performing like a 1080. So Vega started with a disadvantage. AMD had a hard time with the architecture. It uses way more power than current Nvida cards and doesn't clock nearly as high. Vega 56 was competitive though. Vega 56 made Nvidia release the 1070 Ti.

I think at release 8GB HBM stacks cost over $150 each and that was from the factory. Vega 64 uses 2 of them. Then AMD had problems because they used 2 fabrication facilities to make the GPUs before sending them to board partners ( Asus, MSI, Gigabyte etc. ) to finish. Those different fabs didn't make identical GPUs. The HBM stacks were just slightly different heights making it almost impossible for the partners to make a single cooler that fit all the cards. Screw ups like that also affect price.
 


That all makes sense, but why use the more expensive memory if it doesnt provide a considerable performance boost?
 
HBM was first used in the Fury cards and it gave a huge memory bandwidth boost over what was possible with GDDR. Unfortunately for AMD that bandwidth didn't translate to raw performance so the 980 Ti ended up as the faster card. AMD spent a huge amount of money developing the tech with Hynix and not only were they were contractually obligated to keep using it but new card designs take years to go from the drawing board to the fab. So Vega ( which was delayed over a year ) was already being designed to use HBM 2. They couldn't just redesign the memory controller.

Turns out they still weren't able to implement it very well even with a second generation of cards. Vega also had process problems. The architecture was designed for mobile parts and doesn't take voltage well without getting HOT. So clock speed is limited even with liquid cooling.

In my opinion AMD didn't want to release Vega 64 at all but they basically had to. So they released as few as possible and are focusing ( I hope ) on something new.


This goes into it better than I can.

https://www.gamersnexus.net/guides/3032-vega-56-cost-of-hbm2-and-necessity-to-use-it
 
Solution