Barkley said: Am I missing something or having now actually looked at the Data in the OP propperly isn't it completely illogical? The GFlops don't make any sense. Docked: 1005mhz (875.77 GFlops) Undocked: 275mhz (375 gflops) ??? Look at the clock speeds, with that underclock the Handheld GFlops should be 240GFlops not 375 right? It's a bigger underclock in HH then the EG rumour. And the CPU clocks in HH mode being SO much lower is crazy.

If my suspition is right and this is a benchmark, the results wouldn't nessesarily reflect theorical numbers.

Put it simply, if the GPU is clocked at 1ghz but the software doesn't actually ask it to make a single floatin point operation then the average of flops will be 0. So we need to know what this software is and what it is asking of the GPU.

Wyrdness said: Barkley said: Am I missing something or having now actually looked at the Data in the OP propperly isn't it completely illogical? The GFlops don't make any sense. Docked: 1005mhz (875.77 GFlops) Undocked: 275mhz (375 gflops) ??? Look at the clock speeds, with that underclock the Handheld GFlops should be 240GFlops not 375 right? It's a bigger underclock in HH then the EG rumour. And the CPU clocks in HH mode being SO much lower is crazy. 275MHZ is standby mode not when the platform is running check the GAF link someone posted.

Ah I see.

It'll be nice when the Switch embargo's finally lift and maybe DF can post some definitive information. It would be nice if this is true but I have strong doubts.

setsunatenshi said: 1- No 2- if anything those are FP16 numbers, so if you want any sort of apples to apples comparison, divide by 2 3- especially in handheld mode, theoretical peak performance is not the same as what you can expect in normal use, especially during longer play sessions (due to thermal issues) 4- just wait 2 weeks more and DF will have the full breakdown

na, it's FP32, but like I said, I wouldn't read too much into the rumor.

NVidia flops>AMD flops.

In addition to improvements over the architecture itself. It's useless looking at TFlops between the two and making assumptions

The weaker Nintendo's Switch hardware is, the more fun it's games are going to be.

More powerful hardware = Less fun games.

Here's hoping for the 150 handheld mode and 500 docked, that makes it's games around 2 or 3 times more fun than PS4/Xone games.

aLkaLiNE said: NVidia flops>AMD flops

Wrong.

It depends on the GPU architecture in question. If AMD has the more efficient architecture, then it will have an advantage.

Vega should beat Fermi for example.
Regardless, the flops are irrelevant, it's a theoretical number, the other parts of the chip are arguably just as much, if not more important and have an influence in total performance as well.

Valdath said: The weaker Nintendo's Switch hardware is, the more fun it's games are going to be. More powerful hardware = Less fun games. Here's hoping for the 150 handheld mode and 500 docked, that makes it's games around 2 or 3 times more fun than PS4/Xone games.

How do you come to that conclusion?

Power means new gaming potential, once we reached a point where physics was possible... We got Half Life 2 with it's gravity gun which was pretty ground breaking at the time.

More power can also mean more realism which assists in immersion which can assist you in becoming completely immersed within a games world, driving it's story and believability.

shenlong213 said: Architecture 516 ARMv8 Cores (512 CUDA Cores + 4 ARMv8 maybe?) Not sure what does it means, They'are talking about 386 CUDA Cores or maybe 512 Cores CUDA Cores in the comments section. Handheld mode GPU 275mhz GFLOPS Benchmark : 375 gflops FP32

Single Precision (FP32) Gflops is counted as Shaders * 2 instructions * Clockrate.

Ergo. 512 * 2 * 275 = 281,600. (281 Gflop)
If it had 386 Cuda cores that would mean. 386 * 2 * 275 = 212,300 (212 Gflop.)

516 ARMv8 cores is stupid and nonsensical.

shenlong213 said: Docked mode GPU is 1005mhz GFLOPS Benchmark : 875.77 gflops FP32

With the same math.
512 * 2 * 1005 = 1,029,120. (1,029 Gflop.)
386 * 2 * 1005 = 775,860. (775 Gflop.)

And in one fell swoop I have debunked this rumor.
It can't even remain accurate with itself.

And you still ignore the rest of the chip, which is just as important as single precision floating point and which the Xbox One still has a big advantage in.  

Valdath said: The weaker Nintendo's Switch hardware is, the more fun it's games are going to be. More powerful hardware = Less fun games. Here's hoping for the 150 handheld mode and 500 docked, that makes it's games around 2 or 3 times more fun than PS4/Xone games.

You're awesome. I like you.

Valdath said: The weaker Nintendo's Switch hardware is, the more fun it's games are going to be. More powerful hardware = Less fun games. Here's hoping for the 150 handheld mode and 500 docked, that makes it's games around 2 or 3 times more fun than PS4/Xone games.

Sarcasm aside, I do think Devs try harder to make the games fun when they are not focusing so much on graphics. Or maybe since many of them can't compete in the technical side of graphics they put their effort elsewhere to make their product attractive.

Barkley said: zippy said: I've played the remastered version of dishonoured on PS4 and that suffers from huge frame dips. A game ported from last gen hardware to a far superior machine, which is no benchmark to judge the machines power. Because of improved assets (which BOTW doesn't have) and shoddy development, are you insinuating Nintendo did a crap job with the Switch version of Breath of the Wild? But I admit it's not the best gauge of performance, doesn't change my opinion that the Switch is no where near 0.9TF docked. I'll stick with the Foxconn/EG leaks that are around the same ballpark.

 

I'm not insinuating Nintendo did a crap job, but I imagine they had a very tight deadline to make sure Zelda is ready for Switch. Therefore I'm sure there are areas Nintendo could have done more improvements on with a more relaxed time scale, I know Nintendo like to take their time but in this instance Zelda HAD to be ready for launch. Looking forward to seeing something really push the system, maybe a Metroid or an ambitious 3rd party game.

Pemalite said: shenlong213 said: Docked mode GPU is 1005mhz GFLOPS Benchmark : 875.77 gflops FP32 With the same math. 512 * 2 * 1005 = 1,029,120. (1,029 Gflop.) 386 * 2 * 1005 = 775,860. (775 Gflop.) And in one fell swoop I have debunked this rumor. It can't even remain accurate with itself. And you still ignore the rest of the chip, which is just as important as single precision floating point and which the Xbox One still has a big advantage in.

Not really, you didn't debunked anything. Not saying that I trust this "leak" but the results of a benchmark don't nessesarily need to match the maximun theorical performance of a chip, which is what you get with the fancy math.

Also the tegra x1 and XBO GPU both have 16 ROPs, so assuming the tegra X1 is clocked higher it will result in higher theorical pixel fillrate. Thats one potential advantage the tegra has over the XBO. So the XBO doesn't nessesarily wins in everything.