- Status
monstercameron
Newcomer
Amd offers mobile bristolridge chips in fp4 platform with ctpd of 7-15W.
Those are a bit power hungry for a truly mobile device but so the x1. It also needs active cooling...even the old shield handheld needed active cooling...heck even ny gs7 needs active cooling
D
Deleted member 13524
Guest
AMD could but there's a good chance they simply didn't want to allocate their limited resources into developing one, so it would never come out as cheap as Nintendo likes.
They're already making new SoCs for microsoft and Sony to be released in 2016/2017. Because microsoft has been promising a smooth transition between xboxes of different generations, there's a large chance AMD is already working with them for the true next-gen console in ~2020 (and maybe Sony too if they follow suit).
They're trying to get back into the gaming PC + professional + server markets with Polaris+Vega+Navi, all different architectures within consecutive years, and Zen CPUs + very powerful Zen APUs with HBM2 (Greenland).
Plus, Zen will supposedly scale down but will start as high-performance cores and they've announced that the cats cores are discontinued, moreover the AMD have been rather silent about their ARM solutions (which were never intended for mobile anyways).
AMD may not have had any easy way to develop a mobile SoC at the moment. Their current ~5W solutions are heavily based off ~5 year 28nm designs with very weak iGPUs (2 CUs IIRC).
The Tegra line, while it may have been largely a failure among portable Android devices (only Google picks them up, maybe due to their philosophy of keeping the SoC market interesting), is already made for mobile and nvidia may be desperate to get additional console developer attention (according to rumors at least), so the price may have been a lot more to Nintendo's liking.
Nintendo would be dumb to go for nVidia, no company who dealt with NV ever did it twice sofar... (MS, Sony)
If correct I'm disappointed, PowerVR Wizard would have been a much better choice.
.
I think Nvidia's "bad" role with consoles is way overblown. The problem with Xbox OG and PS3 wasn't Nvidia, but MS and Sony and the deals they agreed too. It's up to Nintendo to secure a good deal from Nvidia.
IMO, I hope this is some sort of 16nm part, even if it's just a shrink of X1. I don't think a 20nm part makes that much sense.
monstercameron
Newcomer
The latest tegra is as mobile as is amds current mobility products. Look at the power draw of the sheild tv. To make that truly mobile nintendo would have to downclock it severely.
Also i disagree that amd wouldnt have enough resources or care to design another console - read lock in- chip for one of their oldest partners.
The latest fp4 apus have 2-4 cus
Last edited:
Statistically 100% of the other company has tried nvidia at least one time
D
Deleted member 13524
Guest
Can we make guesses about the SoC's performance?
Looking at the PX2, I don't think this will be a Tegra X2, which seems to be very CPU-centric in order to work with external GPUs. The Cortex A57 has way too many power efficiency problems so I think this will have either many (>=8?) moderately clocked Cortex A53, 4 low-clocked Cortex A72 or a new generation of their own Denver cores. As for the iGPU, I'm betting on 3 SM / 384 sp + 24 TMU + 16 ROPs at ~1.2GHz using 128bit 3200MT/s LPDDR4.
0.92 TFLOPs GPU with 64GB/s.
Enough to warrant PS4 graphics at 720p.
The Tegra X1 in the shield TV is clocked beyond its optimal efficiency values because it's a set-top-box. There is a handheld in the market with a Tegra X1 and the difference to the Shield TV is only ~15% in the iGPU core clock.
Oh they certainly would've loved to, but they're at a point where they have to choose their battles. Make no mistake that the 2013 console design wins were partially responsible for AMD not having the resources for properly competing with Maxwell in 2015.
You claim they could have adapted a 15W Bristol Ridge into a ~7W handheld. While true, it's a 28nm design using Excavator cores (with terrible FP output), doesn't use LPDDR and at that TDP its iGPU performance would probably be nowhere near even the current Tegra X1, let alone a 16FF Pascal + ARM + LPDDR4.
Looking at the PX2, I don't think this will be a Tegra X2, which seems to be very CPU-centric in order to work with external GPUs. The Cortex A57 has way too many power efficiency problems so I think this will have either many (>=8?) moderately clocked Cortex A53, 4 low-clocked Cortex A72 or a new generation of their own Denver cores. As for the iGPU, I'm betting on 3 SM / 384 sp + 24 TMU + 16 ROPs at ~1.2GHz using 128bit 3200MT/s LPDDR4.
0.92 TFLOPs GPU with 64GB/s.
Enough to warrant PS4 graphics at 720p.
The Tegra X1 in the shield TV is clocked beyond its optimal efficiency values because it's a set-top-box. There is a handheld in the market with a Tegra X1 and the difference to the Shield TV is only ~15% in the iGPU core clock.
Oh they certainly would've loved to, but they're at a point where they have to choose their battles. Make no mistake that the 2013 console design wins were partially responsible for AMD not having the resources for properly competing with Maxwell in 2015.
You claim they could have adapted a 15W Bristol Ridge into a ~7W handheld. While true, it's a 28nm design using Excavator cores (with terrible FP output), doesn't use LPDDR and at that TDP its iGPU performance would probably be nowhere near even the current Tegra X1, let alone a 16FF Pascal + ARM + LPDDR4.
Also I think it is more likely to be Tegra X2 (Pascal orientated) rather than the older SoC limited Maxwell X1.
Emily Roges called this out some time ago it was not x86,not AMD, but was using the very latest generation tech of another IHV.
Also the Tegra X1 really does not have enough power.
This is what Emily (Nintendo 'insider' reporter) said: https://arcadegirl64.wordpress.com/
If it is X2 orientated, then it could be an interesting product, but I agree it gives developers possibly some headaches although they seem to be pretty positive about the NX capabilities and what they can do.
Cheers
Emily Roges called this out some time ago it was not x86,not AMD, but was using the very latest generation tech of another IHV.
Also the Tegra X1 really does not have enough power.
This is what Emily (Nintendo 'insider' reporter) said: https://arcadegirl64.wordpress.com/
If it is X2 orientated, then it could be an interesting product, but I agree it gives developers possibly some headaches although they seem to be pretty positive about the NX capabilities and what they can do.
Cheers
To be fair the design was one time only that lasted a couple years, but Nintendo opted to keep using it for those 20 years.
We still don't know if NX has different mode for home use. Let's talk about mobile use first.
http://www.anandtech.com/show/8811/nvidia-tegra-x1-preview/3
The tegra x1 consumes 1.5W of GPU power with iPad Air2 performance. 1.5W is a reasonable limit of a handheld GPU.
If NX has home-mode with high frequency, the GPU performance may have 2x improvement considering reasonable cooling solution of a tablet-size unit.
http://www.anandtech.com/show/8811/nvidia-tegra-x1-preview/3
The tegra x1 consumes 1.5W of GPU power with iPad Air2 performance. 1.5W is a reasonable limit of a handheld GPU.
If NX has home-mode with high frequency, the GPU performance may have 2x improvement considering reasonable cooling solution of a tablet-size unit.
Silent_Buddha
Legend
Interesting. So 32 GB carts. Presumably since this is still primarily a handheld console, they won't likely deviate too far from their current handheld game pricing. So ~40 USD game carts. Nintendo sees that as profitable for a 32 GB cart. So a 64 or 96 GB cart would likely be profitable at 60 USD. Interesting. 
Regards,
SB
Regards,
SB
I think we should speculate backwards from BOM and power consumption.
I'm going to guess the NX will cost (as in BOM) around $150 and will provide a reasonable continuous 4-5 hour gaming time. I think a lot of "high-end" mobile SOC's consume too much power when fully engaged. They're great at varied workloads, but under a continuous work load they suffer (like gaming, see Pokemon)
Therefore, I don't expect the Nintendo Tegra SOC to be very high spec'ed. I think something it could definitely be Parker based and on 16ff. That makes senses given the power consumption savings and the time when NX is supposed to release.
But I could see it coming with simple CPU's (A53's only) and a very low clock GPU (maybe 128 shaders @ 300 MHz) .
I'm going to guess the NX will cost (as in BOM) around $150 and will provide a reasonable continuous 4-5 hour gaming time. I think a lot of "high-end" mobile SOC's consume too much power when fully engaged. They're great at varied workloads, but under a continuous work load they suffer (like gaming, see Pokemon)
Therefore, I don't expect the Nintendo Tegra SOC to be very high spec'ed. I think something it could definitely be Parker based and on 16ff. That makes senses given the power consumption savings and the time when NX is supposed to release.
But I could see it coming with simple CPU's (A53's only) and a very low clock GPU (maybe 128 shaders @ 300 MHz) .
D
Deleted member 13524
Guest
I'm not sure I'd trust those numbers. The article's authors also advise to take those numbers with a grain of salt, since they were readings presented by nvidia in a very controlled environment ("oh, we're reading this thing which is totally the GPU's power rail").
32GB micro-SD cards are selling for $10 directly to the consumer. 32GB of flash storage must be really cheap nowadays, and it'll blast any current optical solution in terms of performance.Interesting. So 32 GB carts. Presumably since this is still primarily a handheld console, they won't likely deviate too far from their current handheld game pricing. So ~40 USD game carts. Nintendo sees that as profitable for a 32 GB cart. So a 64 or 96 GB cart would likely be profitable at 60 USD. Interesting.
Regards,
SB
And for a handheld, there's really no other option.
A 76 GFLOPs GPU? That would be ridiculous even for Nintendo, and I'm being serious here.
That's less than half the theoretical performance of the Wii U and the equivalent of a 3.5 year-old Snapdragon 600.
For that kind of performance, they would be better off just buying an off-the-shelf SoC from Mediatek or Rockchips, whose <$20 mid-ranges are largely surpassing that performance.
nVidia wouldn't want to get attached to such a low-performing device.
Do not forget the VR capability being added to this console.
Suggesting again this is the Pascal Tegra X2 type SoC, and backing up Emily Rogers claim when she says "The chips are industry leading because they are very modern chips".
She is pretty well known for insight into Nintendo..
Cheers
Suggesting again this is the Pascal Tegra X2 type SoC, and backing up Emily Rogers claim when she says "The chips are industry leading because they are very modern chips".
She is pretty well known for insight into Nintendo..
Cheers
Last edited:
D
Deleted member 13524
Guest
I don't know how a handheld in 2017 can provide any worthy VR experience besides movies.
I'm saying this as I don't think GearVR's games, from what I experienced, are worthy of console-level investments.
That and the BoM would balloon if they put a high-DPI + low-latency screen in the console.
Besides, Reggie has recently put the idea of VR away AFAIK.
Even the idea of a 3DS-like auto-stereoscopic screen seems a bit unlikely at the moment, to be honest.
Don't you mean Nintendo?
LOL yes regarding Emily (even linked her earlier as well) doh
Will edit.
Remember this 'console' is being presented as both mobile and home.
The VR aspect-rumour seems pretty solid for the Nintendo NX, just like the info from Emily.
Also worth considering the delay to March 2017.
I think we get to see more info about Tegra X2 as well in August.
Cheers
Last edited:
Silent_Buddha
Legend
Unless it features extremely simplistic and low resolution VR, I don't see it being all that useful for VR gaming. GTX 1060 level performance in a less than 5 watt power envelope? Somehow I find that extremely unlikely. Even half a GTX 1060 (performance, not actual configuration) wouldn't fit in a handheld gaming device power envelope. And all current VR solutions are already ridiculously low resolution for VR. Going even lower than that would be painful.
Regards,
SB
- Status
Similar threads
