By using this site, you agree to our Privacy Policy and our Terms of Use. Close
hsrob on 12 February 2015
Edit Thread
BlueFalcon said:
hsrob said:

Depends on when it's released. The Tegra X1 is delivering 1 teraflop next month in tablet format which is pretty significant. Who knows what will be possible in another 18-24 months.

It also depends on whether Nintendo insists on following the small form factor, low power consumption of the Wiis. Considering how small the Japanese market is these days I just don't think it's a goal worth pursuing.

The 1 TFlop figure has been picked up and ran with by most non-technical sites. All modern GPUs and consoles are measured in 32-bit FP32 flops. 1 TFlop on X1 is only possible at FP16; and that rate drops to just 0.512 TFlops at FP32 vs. 1.84 TFlops for PS4. 

http://www.anandtech.com/show/8811/nvidia-tegra-x1-preview/2

The 2nd mistake almost everyone makes is directly comparing NVidia to AMD floating point performance. It's simply not comparable and never has been a good estimate of actual gaming performance. In fact, floating point performance is a very vague gauge of gaming performance overall even when comparing AMD vs. AMD or NV vs. NV cards:

1. NV GTX 680 is 35-40% faster than a GTX580 but has 3.25 TFlop floating vs. 1.58 TFlops for the 580.

2. NV GTX980 is about 10% faster than an AMD 290X but has just 4.98 TFlops vs. 5.63 TFLops for the 290X. 

3. AMD HD7970Ghz is 70-75% faster than AMD HD6970 but the former only has 4.3 TFlops vs. 2.84 for the 6970, or just 51% more. 

The explanation is very simple -- factors such as instructions per clock cycle, L2 cache, ROPs, Texture full-rate, shader and geometry performance, memory bandwidth and color fill-rate throughput and efficiencies matter a lot more than raw Floating Point performance for games. 

You are right that in 2-3 years performance will be leaps and bounds beyond today's products because GPU makers will adopt 14nm nodes and so on. AMD should also have their brand new Zen architecture made on 14nm node by end of 2016 iirc. That means if Nintendo wanted to, they could make a console at $499 way more powerful than a PS4, with CPU and GPU 50-100% more powerful actually, since 8 Jaguar cores are weak and the GPU is only about HD7850 level of performance. However, if the next Nintendo console is a lot more powerful, but has a low install base, most developers will not spend the extra money to take advantage of all that power since it will be too costly to amortize those costs across a small number of game sales relative to then large PS4/XB1 userbase. 

For this reason, I think N should aims at a $299 console with PS4's power and no more, around late 2016, early 2017. If we project ~ 6 year life for PS4/XB1, next gen consoles would launch Fall 2019. Nintendo has no choice though because the Wii U can't last until 2019. That means Nintendo's best bet is to make 2 consoles between end of 2016 and 2021.  Nintendo then could launch a PS5/XB2 competitor around 2021, albeit 2 years behind. Also, we don't know if Sony/MS will wait until 2020 to launch their next consoles but we do know that after Zelda U and 2015, Nintendo will probably have nothing worthy for the Wii U which means late 2016/early 2017 is a perfect time for them to launch a new console as an intermediary strategy before Ps5/XB2 drop. 

If the console acts as an intermediary stop-gap console, it doesn't even need to be more powerful than PS4 imo. 

Interesting to know because to be honest :) I haven't really been folowing this too closely, just saw the headline. It will still be interesting to see what level of real-world performance the X1 is capable of (i haven't seen any real-world benchmarks)

I think my main question still stands though, will Nintendo go for low power consumption/small form factor or will they forego this approach.  The WiiU could have been a fair bit more powerful, at the same price point, if Nintendo had just used off-the-shelf parts.