Talk:Latte (graphics chip)

On sourcing specifications
To other editors: please do not add unsourced specs to the article. The vast majority of the specs listed (which I removed) were either guesswork by the community, or plain and outright wrong (e.g. the claims of the existence of a Cortex-A8 - which I personally know to be untrue - or a Cortex-M3 DSP - there's a DSP, but it's based on the Gamecube DSP and has nothing to do with ARM).

In particular, to User:Eyeofcore: I know you love wild speculation and outlandish theories, but those do not belong on Wikipedia. Please keep your theories elsewhere unless you can find reliable research that supports them first. Please do not attempt to argue for your theories here, as Wikipedia is not a platform for original research.

To other editors: this reference essentially summarizes what we know about the Wii U hardware as determined though reverse engineering. Almost everything else out there (there is very little official information, so the vast majority is unofficial) is based on guesswork and does not belong here unless it can be sourced from reliable external research (e.g. reverse engineering or reliable reports from official developers). Although there have been extensive attempts at guessing at the detailed capabilities of the Latte based on a number of indirect factors such as die size, die shot analysis, and power consumption, (e.g. discussions on NeoGAF) and although those attempts are commendable and very interesting, they don't really qualify as reliable sources and should not be cited.

To address one particular point of contention, the GPU generation: as mentioned in the above link, we (myself and other reverse engineers) have strong evidence (from firmware references) that the core is R600-derived, and, by the implication that the R700 is itself a minor evolution of the R600, and a few scattered references to the latter in firmware, are rather confident that the R700 is indeed the Latte's nearest relative. If other editors believe that the research in the above linked article is not reliable enough to qualify, please feel free to entirely remove the mention of the GPU family. However, claims that the GPU is R800-based or newer were solely based on the reports by developers that the GPU has support for tesselation and Stream Out, which were erroneously claimed to be present only on newer generations. The reality is that those features already existed on the R7xx family (if perhaps not as heavily advertised), which can be verified through some simple Google searches. Therefore, there is absolutely no evidence that the GPU is based on R800 or newer - please do not make that claim unless you have a reliable source for it.

Marcan (talk) 20:23, 28 July 2014 (UTC)