AMD Vega 10, Vega 11, Vega 12 and Vega 20 Rumors and Discussion

GN review place the 400W Vega in between 1080s but the amount of effort needed to archive that is not something that we can call "normal".
 
It's a detail but I don't get why the liquid cooled version is so long... I get that you need space for the blower one, but the liquid one ? Maybe the RX version will have a Fury size with another pcb ? I like short cards...

The right part of the pcb is nearly empty...

DSC03538%20copy.jpg
 
GN review place the 400W Vega in between 1080s but the amount of effort needed to archive that is not something that we can call "normal".
The alternative appears to be dropping the voltage from 1200mV to 1100mV to achieve a similar effect without all the effort. Lower the power draw to avoid having to increase the power limit in the first place. Vega should drop 20-30W just by cutting the memory pool in half on top of some undervolting gains. AIO or better cooling possibly doing the same with less leakage.
 
Lower the power draw to avoid having to increase the power limit in the first place.
I got the impression that lowering the vcore has little effect on the average clocks, despite lower power consumption and heat output. Looks like the clocks aren't exclusively related to power and temperature.
 
Isn't it more like a 475W Vega? GN was not including PCIe power.

According to Gamers Nexus, PCIe slot power is disproportionally low in default settings (25W tops IIRC).
I don't know how this evolves with overclocking/overvolting though.
 
It's a detail but I don't get why the liquid cooled version is so long... I get that you need space for the blower one, but the liquid one ? Maybe the RX version will have a Fury size with another pcb ? I like short cards...

The right part of the pcb is nearly empty...

DSC03538%20copy.jpg
They probably don't expect for it to have enough volume to bother making another PCB just for the liquid variant.
 
They probably don't expect for it to have enough volume to bother making another PCB just for the liquid variant.

Though if there's a RX Vega with liquid-loop, they could've used the same design.
 
I don't think theres enough black magic in the world to make the car 40% faster via drivers not just in a couple of months but in its entire life span.
 
I don't think theres enough black magic in the world to make the car 40% faster via drivers not just in a couple of months but in its entire life span.
Isn't that what happened with Kepler to Maxwell? Look at Linux driver development and those kind of swings are nothing. A few swings like that in a month aren't always uncommon, but they quickly become situational. If the gains weren't there why even delay Vega's release a month?
 
They previously determined Vega takes very little from PCIe, about 25W.

http://www.3dmark.com/compare/spy/2068702/spy/2065605#
Vega FE liquid @1700MHz tested in TimeSpy, slower than a 1080 by ~8%.

Vega FE air in Prey, slower than a 1080:

But cleary win in Battlefield 1:
and Doom:
and is very close in Witcher 3 @1440p max:


@1600MHz and above Vega RX will have a better performance than a GTX1080 for sure in the vast majority of the cases but whithout a driver overhaul will still be far cry from a GTX 1080 Ti....
 
Last edited:
@1600MHz and above Vega RX will have a better performance than a GTX1080 for sure in the vast majority of the cases but whithout a driver overhaul will still be far cry from a GTX 1080 Ti....
Perhaps, but against GTX 1080's at reference clocks. Custom variants might be a different story with overclocking/similar cooling.
 
Back
Top