Whats the source of the 12TF figure anyways?I still see no official information saying xbsx is 12TF.
The latest devkit photo makes me question the theory it's significantly above 200W.
I got lost between the rumors and "leaks"
Whats the source of the 12TF figure anyways?I still see no official information saying xbsx is 12TF.
The latest devkit photo makes me question the theory it's significantly above 200W.
Whats the source of the 12TF figure anyways?
I got lost between the rumors and "leaks"
"We wanted to have a dramatic upgrade from the Xbox One base console. So when we do the math, we're over eight times the GPU power of the Xbox One, and two times what an Xbox One X is."
It was a reasonable conjecture we all made about "when you do the math, it's twice the raw gpu performance of XB1X" but when the journalist asked a simple follow up "you mean the math being 2 x 6TF = 12TF" he refused to answer and evaded.Whats the source of the 12TF figure anyways?
I got lost between the rumors and "leaks"
hmm, I thought measurements (DF and others who have put in real effort) had XSX laid out at 400mm^2 +.It was a reasonable conjecture we all made about "when you do the math, it's twice the raw gpu performance of XB1X" but when the journalist asked a simple follow up "you mean the math being 2 x 6TF = 12TF" he refused to answer and evaded.
There's been more interviews and it's still not clear what he meant. Is it more? Is it less? Is it a performance figure? Who knows?
380 mm2 is kind of short for 60 CUs (4 disabled) and 8 zen2. But it's not impossible the measurement was wrong or the clock higher than we think. Then we get questionable thermals without 7nm+. So if they reach 12TF they must be on 7nm+.
Sure, but what else can it mean? "Twice the raw GPU performance." There's only one way to measure raw GPU performance and that's FLOPS from the FMADDs/FMACs. It was a pretty explicit description of the GPU, as opposed to, say, "it's twice as powerful as an XB1X" which may not be talking about raw power, which would leave room to have less than 2x the Flops but still have 2x the drawing power.It was a reasonable conjecture we all made about "when you do the math, it's twice the raw gpu performance of XB1X" but when the journalist asked a simple follow up "you mean the math being 2 x 6TF = 12TF" he refused to answer and evaded.
I used two different measurement methods and they both ended up at 380, so I am curious why DF gets over 400. Anyway it's unimportant.hmm, I thought measurements (DF and others who have put in real effort) had XSX laid out at 400mm^2 +.
Quite a bit larger than 380mm^2.
As for Phil deflecting. There will be a time to make that announcement. I don't think it was to confuddle the waters about GCN vs Navi flops. But waiting for final confirmation on yield (and thus clockspeed) to discuss final output. Journalists and developers have been given a 12 TF target. That can only be 12 Navi FLOPs. No documentation would point at 12 GCN flops, but here is a Navi chip for development. It's just too confusing for all parties involved.
I used two different measurement methods and they both ended up at 380, so I am curious why DF gets over 400. Anyway it's unimportant.
When did journalists and devs received a 12TF target? That would definitely be navi flops.
Microsoft hasn't even released the next-gen Xbox logo. There's so much unknown.
You want to claim it's exactly 12TF the burden of proof is on you. I see nobody claiming to know it's less.The below 12TF claims hold as much water as the claim that MS won't push next gen graphics, which they have done every generation since OG Xbox.
You want to claim it's exactly 12TF the burden of proof is on you. I see nobody claiming to know it's less.
It was a reasonable conjecture we all made about "when you do the math, it's twice the raw gpu performance of XB1X" but when the journalist asked a simple follow up "you mean the math being 2 x 6TF = 12TF" he refused to answer and evaded.
The die size wouldn't support a 9TF theory anyway even on 7nm, unless RT requires a lot of silicon space. The die size is pretty much the best indication of power.Offcourse i mean ballpark 12TF, not the difference between 9 and 12 which isn't exactly 'very close'
One of the podcast someone pressed him on it, it's so frustrating how they do interviews in podcasts, they are not searchable so trying to find something back you have to watch hours of crap. And we misquote them from memory. I might be as badly quoting as anyone else. I really prefer the long form text interviews with gamasutra.link?
I’ll look around for it again but I recall DF, Thurott, a Windows Central all citing a targeted 12 TF; Targeting was the verbiage. They corroborated each other and then the github leaks came out; and they used the 12 TF corroborated figure to figure out the number of CUs or Clock-speed (this part is muddy to me). And that’s approximately how it came to 56 CU.I used two different measurement methods and they both ended up at 380, so I am curious why DF gets over 400. Anyway it's unimportant.
When did journalists and devs received a 12TF target? That would definitely be navi flops.