D
Deleted member 887
Guest
I'm taking extensive liberties here, sure.I'm not sure how literally we're meant to take this whole thing so I'll defer making every comment I could make. But:
You're smarter than me, but I don't buy that offhand.
- Plans for T239 have existed since 2020 at the latest, and T239 is Drake, not Dane.
- If some other chip had existed for a year or more and been included as a target for NVN2, we would see some evidence of it in the leak.
Who said it was going to be short? Nintendo has done 3 generations at this point with 2x jumps in power. I'm suggesting that the planned a 2x power jump and to call it something like New Switch, which explains the speculation from folks who heard about it that it was a short lived revision, but it is clear that those people were, in fact, speculating about the intended lifespan.
- The idea of any shorter-lived intermediate revision just doesn't make sense with an Ampere Tegra.
Seems to be a few inconsistencies here.
First, didn't kopite say Dane was 8SMs instead of 4?
I don't recall any leak from kopite about GPU size. In this thread, @Alovon11 was the only significant pusher of 8SMs pre leak. Most of us, myself included, thought it was too far to go on Samsung 8nm, Alovon's argument was that it would be exactly half Orin, and be the simplest design move, consistent with the rest of the hardware.
We don't have the revision history of the files, just copyright dates which are not always updated. But again, the meat of my idea here is just that in 2021 Nintendo and Nvidia saw an opportunity to die shrink their existing product because of other forces in the market and took it.Second, NVN2 code referencing Drake far predates December 2021 I believe. Also doesn't it explicitly say Drake is made at a Samsung foundry in the code?
As for the Samsung reference, LiC has pointed out that the code seems to be copied straight from Orin with no changes, and that there is reason to believe these sorts of references can lag behind, as they don't affect major functionality.
Third, the Wii U was announced at E3 2011 if I'm not mistaken, and released November 2012. A 17 month marketing cycle, not short at all.
Jiminy, this is a thing I completely misremembered! The Wii U era, truly a haze.
I'm not sure if the evidence supports the idea that Dane really was a separate thing but I do admit it would help explain some of the stuff we've heard from rumors.
There is a reason I captioned the theory with "expect to be proved wrong." I generally respect @LiC's analysis more than my own when it comes to anything hardware related, and back in Spring when the "maybe it's a 5nm Dane" had prominence I was one of the folks shooting it down.
But I've long entertained the idea that one of the things that was going on was confusion between a successor and a pro revision, the same way reporting mixed up T239 and OLED. it would neatly explain some things, but all the plausible pro revision chatter has been around DLSS. DLSS can't run on Maxwell.
To vastly shorten my narrative here, I'm not suggesting that one product was scrapped and another built. Just that if Nintendo/Nvidia decided to increase the amount of power in T239 after shipping devkits, that (generally bizzare) move would neatly explain Drake's unusual technical composition and the strange statements of usually competent sources, and that the last three years created an opportunity for said generally bizarre move.