Despite the hype leading to a frequent lack of stock online, Nintendo's newest home console ended up selling 906,000 units in the month of its launch. This makes the Switch one of the fastest selling Nintendo consoles in the company's history. Console sales weren't the only thing reported, either, as sales figures for The Legend of Zelda: Breath of the Wild were revealed as well. The Switch version of the game moved 925,000 copies, oddly outselling the target console itself. In the case of the Wii U version, there were 460,000 games sold.

Source: GBAtemp Press Inbox

 

[WiiUBricker]

Really? I could've sworn other sources said the Switch sold 1.5 million in it's launch week...

Either way, that's not bad sales-wise, a little under both the Xbone and the PS4 if these numbers are right, but launch sales mean jack. If Nintendo can get a properly good E3 going and announce more than 2 games for 2017 that'd be a good start.

I don't know what to believe anymore.

 

[Tom Bombadildo]

I don't know what to believe anymore.

My link was just for the launch week, this thread is for total NA sales in March, and that link is for total world wide sales in March lol.

 

[TheDarkGreninja]

You're making your own life difficult, the shader formula exists for a reason.

256 shaders * 2 Ops (Multiply + Accumulate) * 1GHz = 512 GFLOPS FP32 on stock X1

The Switch peaks at 768MHz and compute power drops linearly.

I know....

However maxwell is still 50% more efficient than GCN.

 

[Foxi4]

I know....

However maxwell is still 50% more efficient than GCN.

What good is that if you're dealing with 1/3rd of compute performance? You can't conjure calculations out of a hat.

 

[TheDarkGreninja]

What good is that if you're dealing with 1/3rd of compute performance? You can't conjure calculations out of a hat.

I know, however, it should be more like 500 Gflops in comparison to GCN even with lower clocks.

 

[Foxi4]

I know, however, it should be more like 500 Gflops in comparison to GCN even with lower clocks.

That's not how efficiency works. I don't know why you're bringing GCN into the conversation when official NVidia documentation lists the X1 at 1TFLOP FP16/512GFLOPS FP32 @ 1GHz and the shader-based math supports that. That's the peak performance in the two categories, it doesn't go higher beyond that point without overclocking.

The reason for the efficiency difference in real-life applications is the fact that, to my knowledge, the baseline GCN used in the old PS4 cannot perform FP16 calculations - it doesn't support half-floats, all calculations are treated as FP32 on the silicon level and take up the same amount of resources as full floats regardless of the intended precision. The X1 can, thus lower precision calculations are performed quicker. This was corrected in the PS4 Pro which includes customised GCN, borrowing features from Polaris, and is also capable of performing two FP16 calculations instead of one FP32. If we're talking in those terms, the Pro is an 8TFLOP+ machine which we both know is false, or at least misleading.

"Finally, there's better support of variables such as half-floats. To date, with the AMD architectures, a half-float would take the same internal space as a full 32-bit float. There hasn't been much advantage to using them. With Polaris though, it's possible to place two half-floats side by side in a register, which means if you're willing to mark which variables in a shader program are fine with 16-bits of storage, you can use twice as many. Annotate your shader program, say which variables are 16-bit, then you'll use fewer vector registers. (...)
One of the features appearing for the first time is the handling of 16-bit variables - it's possible to perform two 16-bit operations at a time instead of one 32-bit operation (...) In other words, at full floats, we have 4.2 teraflops. With half-floats, it's now double that, which is to say, 8.4 teraflops in 16-bit computation. This has the potential to radically increase performance."

http://www.eurogamer.net/articles/d...tation-4-pro-how-sony-made-a-4k-games-machine

If the Switch is a 1TFLOP machine then we have to automatically treat the PS4 Pro as an 8.4TFLOP machine, except we both know that's ridiculous - in real life we mostly use full precision and half precision support only increases efficiency if half-floats are used.

 

[TheDarkGreninja]

That's not how efficiency works. I don't know why you're bringing GCN into the conversation when official NVidia documentation lists the X1 at 1TFLOP FP16/512GFLOPS FP32 @ 1GHz and the shader-based math supports that. That's the peak performance in the two categories, it doesn't go higher beyond that point without overclocking.

The reason for the efficiency difference in real-life applications is the fact that, to my knowledge, the baseline GCN cannot perform FP16 calculations - all calculations are treated as FP32 on the silicon level regardless of intended precision. The X1 can, thus lower precision calculations are performed quicker. This was corrected in the PS4 Pro which includes customised GCN and is also capable of performing two FP16 calculations instead of one FP32. If we're talking in those terms, the Pro is an 8TFLOP+ machine which we both know is false, or at least misleading.

I brought GCN into this because of the fact both PS4 and XBONE use GCN and as such architectural differences should be accounted for also, not just numbers. Thats why a Pascal GPU at 4Tflops still outperforms the same 4.5Tflops Polaris equivalent.

 

[Foxi4]

I brought GCN into this because of the fact both PS4 and XBONE use GCN and as such architectural differences should be accounted for also, not just numbers. Thats why a Pascal GPU at 4Tflops still outperforms the same 4.5Tflops Polaris equivalent.

You're pulling a number out of your ass though, you have no actual data to support it and have no idea how much of raw compute power is wasted on the Switch since no actual benchmarks exist. We're talking about two custom machines, so we can only talk about silicon, the rest is pure conjecture. That kind of a performance difference wouldn't be counted in TFLOPS anyways, the TFLOPS stay the same, it's how efficiently the math is used in rendering and physics that's the subject here. You're confusing raw performance with algorithms

 

[TheDarkGreninja]

You're pulling a number out of your ass though, you have no actual data to support it and have no idea how much of raw compute power is wasted on the Switch since no actual benchmarks exist. We're talking about two custom machines, so we can only talk about silicon, the rest is pure conjecture. That kind of a performance difference wouldn't be counted in TFLOPS anyways, the TFLOPS stay the same, it's how efficiently the math is used in rendering and physics that's the subject here. You're confusing raw performance with algorithms

... I might be. The X1 is pretty off the self and there are quite a lot of benchmarks of it, and Ive seen a few simulated PS4 benchmarks based on similarily spec'd PC's usng comparable GPU's, but thats not the point.

Also thats what I was talking about when I said efficiency.

Anyway, weve digressed enough I think.

 

[Foxi4]

... I might be. The X1 is pretty off the self and there are quite a lot of benchmarks of it, and Ive seen a few simulated PS4 benchmarks based on similarily spec'd PC's usng comparable GPU's, but thats not the point.

Also thats what I was talking about when I said efficiency.

Anyway, weve digressed enough I think.

Sorry about that, the subject was beaten to death indeed. My point was that the gap is significant and it'll only grow wider as the generation goes on, which can be an issue. Just the fact that the X1 in the Switch is downclocked widens it past the point of any architectural gains, but that's neither here nor there - it can be optimised for if it succeeds as a platform. If games have to run at 720p 30FPS, that's just life - I was never a resolution buff anyways.

 

[TheDarkGreninja]

Sorry about that, the subject was beaten to death indeed. My point was that the gap is significant and it'll only grow wider as the generation goes on, which can be an issue. Just the fact that the X1 in the Switch is downclocked widens it past the point of any architectural gains, but that's neither here nor there - it can be optimised for if it succeeds as a platform. If games have to run at 720p 30FPS, that's just life - I was never a resolution buff anyways.

Dude, Im not disagreeing with you. If this were a perfect world the switch would have A72's instead of the far inferior A57's, but its not. I feel nintendo have rushed the switch to market.

 

[Captain_N]

Is this enough to get through to the 3rd parties thick heads..... I bet its not. It has to be the power of the scorpio or they wont touch it. They really have not excuse when a dev kit is $500.

 

[Foxi4]

Dude, Im not disagreeing with you. If this were a perfect world the switch would have A72's instead of the far inferior A57's, but its not. I feel nintendo have rushed the switch to market.

I'm inclined to believe that, as the system has some glaring design mistakes that are blatantly obvious even to someone with no engineering background. For instance, if you want to market the device by stressing the set-top feature and the integrated kick stand, you probably don't want to put the charger port on the bottom of the system, where it's inaccessible, so that the console can't be charged simultaneously. It's a bunch of little things that seem bizzare to me.

 

[TheDarkGreninja]

I'm inclined to believe that, as the system has some glaring design mistakes that are blatantly obvious even to someone with no engineering background. For instance, if you want to market the device by stressing the set-top feature and the integrated kick stand, you probably don't want to put the charger port on the bottom of the system, where it's inaccessible, so that the console can't be charged simultaneously. It's a bunch of little things that seem bizzare to me.

I could bet my house that they'll release a revision in the vain of the Xbox One S (4k capabilities, additional charging port, etc/)

 

[Kioku]

I'm inclined to believe that, as the system has some glaring design mistakes that are blatantly obvious even to someone with no engineering background. For instance, if you want to market the device by stressing the set-top feature and the integrated kick stand, you probably don't want to put the charger port on the bottom of the system, where it's inaccessible, so that the console can't be charged simultaneously. It's a bunch of little things that seem bizzare to me.

Is it really such an inconvenience though? Granted, they could very well have placed it on the top, and the vents on the bottom.. So, it'd be upside down while docked, not that it matters.

 

[TheDarkGreninja]

Is it really such an inconvenience though? Granted, they could very well have placed it on the top, and the vents on the bottom.. So, it'd be upside down while docked, not that it matters.

Or you know, just had two ports.

 

[Kioku]

Or you know, just had two ports.

Why waste that much circuitry..

 

[grossaffe]

Not at all, you ween just enough to each store causing artifical demand. It's human instinct to want to have what others don't. That and it makes it seem like a sellout product to the general public, making them seek it out more. Hence why a Zelda box pushed so many units

Try taking a class on economics and learn how supply and demand really works.

 

[TheDarkGreninja]

Why waste that much circuitry..

I wouldnt consider it a waste. That USB-C couldve been used for more than just charging.

 

[Foxi4]

I could bet my house that they'll release a revision in the vain of the Xbox One S (4k capabilities, additional charging port, etc/)

Of course they will, Nintendo invented pointless revisions. Taking one look at the 3/DS line makes the future of the Switch blatantly obvious, provided the system is successfully adopted.

Is it really such an inconvenience though? Granted, they could very well have placed it on the top, and the vents on the bottom.. So, it'd be upside down while docked, not that it matters.

Yes, because the port placement makes it impossible to use it. The vent is placed correctly, but the USB-C port is not - it should be on the top, right next to the vent, where it'd be always accessible, while the bottom should be occupied by a proprietary docking rail specifically designed for the Switch dock for proper seating instead of the wobbly nonsense that was implemented in the final unit. The heatsink and vent placement matters because of physics - heat rises to the top, so the sink and vent need to be as high as possible to prevent heat from affecting the case and other components. It's placed correctly as it is and shouldn't be moved.

Why waste that much circuitry..

It's a couple traces and a port that costs less than a dollar.