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

I still don't get the FE brand. It would make more sense to have a consumer optimized card (gaming) and a pro card (content). This thing seems to try to replicate Nvidias market strategy with the Titan brand but with semi-optimized game drivers... why?

 

This is curious



Why is it recognized as Greenland, when even the AMD drivers separate Greenland from Vega10?

 

"Because the game drivers aren't ready yet" would be the obvious answer

 

There was no need, TitanXp has just a small performance bump over TitanX Pascal, which already got sent to every major media outlet at the time of it's release.

 

Note that AMD themselves are providing and linking to game drivers for the FE dated 6/27

http://support.amd.com/en-us/download/frontier?os=Windows+10+-+64

 

32 bytes for GDDR5 and HBM2, 64 bytes for GDDR5X.

HBM2 does support Pseudo-Channels, if utilized one ca have 16 byte access granularity.
GDDR5X also supports a Pseudo-32-Byte-Mode:
https://www.google.de/url?sa=t&rct=...sg=AFQjCNEz8cqzg64CDApRXm5CXr5IOwn-Hg&cad=rja

 

Then, release it when it's ready?

 

From said driver's INF file:
;-----------------------------------------------
;----------1706261331-17.20.1035-170624a-315622C-CrimsonReLive

; AMD display information file
;
; Installation INF for the AMD display driver.
; Copyright(C) AMD 2007-2017
;-----------------------------------------------

[Version]
Signature="$Windows NT$"
Provider=%ATI%
ClassGUID={4D36E968-E325-11CE-BFC1-08002BE10318}
Class=Display
DriverVer=06/24/2017, 22.19.384.2
;;LayoutFile=layout.inf
CatalogFile=C0315622.CAT

Note how 22.19.384.2 shows up in the posted GPU-z of VEGA FE screenshot as well.

 

Then they'd miss the H1 2017 release goal.

 

These kind of "Funny how..." posts never get old, do they?

 

Well... which is worse for AMD? Bad benchmarks or a month delay with (hopefully) good or at least better benchmarks?

 

I could be wrong, but I thought that the prefetch length increases from 2n to 4n in pseudo channel mode for HBM2. So it's still 32 bytes.

 

Pascal was already proven at that point. Who knows what's the state of Vega exactly is. A press briefing would have done wonders for damage control.

 

Right now it's performing like a OC Fiji with more memory. Something is not right.

 

As ToTTenTranz said, this is effectively a delay and the FE seems to be made specifically for appeasing the shareholders. AMD seems to believe this is the better course of action.

 

Those are forward looking statements. Schedule slips are a fact of life. Unless AMD provably misled investors, there'd be no repercussions at all.

 

You are right, I didn't looked at the prefetch lenght.
It increases from 2 to 4 so effectively it stays at 32 bytes.

 

So apparently there's the "dual pro/game-drivers" and separate "game drivers" now? https://disqus.com/by/klaudiuszkaczmarzyk/

 

According to AMD they have NOT been censoring Reddit or anywhere else