NX power

[sc94597]

haxxiy said:

sc94597 said:

 

10-15% might just be enough to remove bottlenecks. Actually 10-15% extra CPU power is pretty significant in the grand scheme of things. 

I think you are greatly understimating what GPU they can put in for $300, and Nintendo's priorities. While Nintendo has never lit the hardware charts, they've always had consoles that matched the predecessor (Wii matched XBOX, Wii U matches PS3.) For less than a $100 /per console cost of production + AMD profit (in bulk) Nintendo could get Pitcairn, Tobago or Bonaire XTX (surpasses XBOX ONE and matches PS4.) The only thing that will limit this is power consumption, but I am sure they will work out the specifics. That leaves $150-200 for the rest of the components.

The only way I would see Nintendo going for an Oland XT is if they are trying to reduce the price of their console to the $200-250 range. 

You talk about "removing bottlenecks" and "working out the specifics" and "grand scheme of things" without actually pointing out what those mean. I can surely overclock any CPU 10-15% and I doubt it would any significant, even if I coded software specially for it. The X1 CPU is theoretically clocked 10% higher than the PS4's and it doesn't seem like a big deal.The Carrizo benckmarchs point to a nice increase, but it does not break any "bottlenecks".

I agree, though, if Nintendo plays their card well, they could get a bigger piece of silicon (since 28nm is more mature now) and underclock / undervolt it to get to X1 or even better levels of performance, but keeping a smaller case / cooling system. The thing is, Nintendo hasn't played her hand well when it comes to console hardware since the GameCube. And that's being generous for not taking into account the controller and media storage...

If only Nintendo could ditch AMD and get a Maxwell GPU, now we're talking about a mix of good harware and low power consumption / heat. A Geforce GTX 965M consumes less than two thirds of the graphics inside the PS4 and brings an almost even performance. Not to mention it shouldn't be very expensive, since it's basically a GM204 chip with half its units disabled. Yields would be very high...

1. "Removing bottlenecks" - bottlenecks are usually the type of things that if you don't have enough power they limit your system, but the difference to "just enough" could be quite marginal cost vs. a large benefit you gain. The performance gains, when bottlenecks are in the system are not linear/scale evenly by percentage. I am sure you know all of this.

An example in PC gaming:

Notice the marginal differences in computing frequency and/or threads lead to a significant stability in frame-time for the Witcher 3 and then after that it is flat, and before that it slopes somewhat linearly. If you have a 1.3 GHZ 4 threaded CPU you will see gains by increasing that to 1.5 GHZ (+15%), but after that you see no gains. That is what I meant by, 10-15% could be enough to resolve cpu bottlenecks in certain applications.

2. Most advances in CPU IPC from generation to generation (for the same clock-speed) today are quite miniscule. Look at Haswell Refresh -> Skylake. That is what I meant by 10-15% as quite significant.

3. Possibly since they are going to finally ditch the old GameCube architecture they will make wiser decisions this time around. They could increase performance even if they go with AMD. There are low-profile GPU chips that will let them match XBO/PS4 performance for a pricepoint of $300, without leaving their power consumption requirements, and I am sure there will be more options for when they manufacture such a console (probably won't come until the end of 2016/early 2017.)