Support NeoGAF

[Hermii]

While the 4310 mAh battery and neon joy-cons are too precise to be a guess, I would still err on the side of believing Eurogamer when it comes to clock speeds. Eurogamer has a good track records with prior hardware leaks and has a reliable way of finding them out (by talking to devs). There's no reason to believe that a Foxconn employee would have any information about clock speeds, fabrication processes or microarchitectures. They're simply shipped a finished SoC by Nintendo and assemble it together with everything else. They probably do some testing of the units once assembled, but I don't imagine it tells them these kinds of specifics.

Spoiler

That said, A72s or A73s would be nice

Edit: Also, as mentioned above by ElTorro, even if the clock speeds are accurate, that doesn't actually mean they'll be used in games, just that they chip can theoretically hit those speeds. Particularly with the CPU clock, which sounds unreasonably high even for A73 cores at 16nm. Nintendo may push the CPU up to higher clocks in specific scenarios (e.g. for Gamecube/Wii emulation) while retaining the 1GHz limit for games.

Good post as always!!

As much as we all want the Switch to be powerful and state of the art, its time to accept the soc is no technological marvel. It will be interesting to find out what customisations Nintendo did to the X1 though.

 

[Mokujin]

Particularly with the CPU clock, which sounds unreasonably high even for A73 cores at 16nm. Nintendo may push the CPU up to higher clocks in specific scenarios (e.g. for Gamecube/Wii emulation) while retaining the 1GHz limit for games.

I had not thought about this, but since you need a lot less grunt on the Gpu side while emulating, the power and thermal budget can be balanced towards the Cpu.

That's a really interesting idea.

 

[Skittzo0413]

https://youtu.be/PzS4LbH5nmA?t=4m5s Says his source's info comes from about 4 to 5 months ago. Said that he didn't run the information then because he was unsure about architecture and any changes until the venture beat article was released earlier in that week.

Ah haven't gotten a chance to watch that but if true, that really makes me believe this Foxconn leak, as it came from late November. Again, I doubt this makes much of a difference for game graphics (20% boost to GPU clocks) but the much improved CPU would be a very, very important and welcome change.

EDIT: And just as a reminder to everyone, Eurogamer/Digital Foundry explicitly admitted that they didn't know the final GPU core configuration/CPU configuration- just the clocks, and apparently from 4-5 months ago. So A72/A73s would not go against what they have reported.

 

[plagiarize]

lol:

There is zero reason to expect Zelda to get a resolution bump at this point. However, I do think it's fair to expect games designed ground up for the Switch to look more impressive than Zelda, for obvious reasons.

For example, the environments and character polycounts are based around what the Wii U could handle, and the textures aren't higher resolution despite the Switch having more RAM to utilize.

But again, whatever the clocks are, at this point, it's basically irrelevant when we've seen footage of actual games, some of which launch in a month and a half.

 

[btrboyev]

The launch games are releasing in 6 weeks. They'll be gold in under 3, especially with Nintendo not being about that post release patch life.

means nothing. Launch games were probably done on not final dev kits. Zelda included. If rumors were true, fianl dev kits didn't even go out to devs until a few weeks ago.

Remember when a bunch of PS4 games shipped at 900p and then had 1080p patches? It's not out of the realm that final hardware is improved from dev kits. That being said, i don't expect Zelda to get any performance patches unless it comes out that it doesn't run well. But I don't think Nintendo would release it if were that bad.

 

[lwilliams3]

https://youtu.be/PzS4LbH5nmA?t=4m5s Says his source's info comes from about 4 to 5 months ago. Said that he didn't run the information then because he was unsure about architecture and any changes until the venture beat article was released earlier in that week.

deleted, sorry.

Hmm. Maybe it's good if we did get Eurogamer to comment on this.

Edit: Btw, would it be best to change the thread title to, "Eurogamer outdated" instead of "Eurogamer wrong?"

 

[KojiKnight]

I had not thought about this, but since you need a lot less grunt on the Gpu side while emulating, the power and thermal budget can be balanced towards the gpu.

That's a really interesting idea.

That's actually a brilliant point and possible use! I hadn't thought of that.

 

[bidguy]

There is zero reason to expect Zelda to get a resolution bump at this point. However, I do think it's fair to expect games designed ground up for the Switch to look more impressive than Zelda, for obvious reasons.

For example, the environments and character polycounts are based around what the Wii U could handle, and the textures aren't higher resolution despite the Switch having more RAM to utilize.

But again, whatever the clocks are, at this point, it's basically irrelevant when we've seen footage of actual games, some of which launch in a month and a half.

i agree, what i dont agree with is the thought that nintendo revealed shoddier versions of games releasing in a few weeks

 

[LordOfChaos]

Something else to chew on:

The psp had a 333 mhz processor that, on launch, developers were limited to 222 mhz and later 266 mhz. 333 wasn't officially supported until psp2000 was out.

Nintendo could be doing something similar, with the higher speeds being denied to developers at launch and that power "released" to all hardware revisions when the inevitable "lite" or "xl" versions of the hardware are released.

BoTW was said to make the battery last 3 hours already, wasn't it? If they unlocked more power later...

 

[Skittzo0413]

i agree, what i dont agree with is the thought that nintendo revealed shoddier versions of games releasing in a few weeks

I wouldn't be surprised to see the framerate smoothed out a bit, but I wouldn't expect more than that.

As a reminder, we don't know how old the builds shown are.

 

[z0m3le]

Hmm. Maybe it's good if we did get Eurogamer to comment on this.

Edit: Btw, would it be best to change the thread title to, "Eurogamer outdated" instead of "Eurogamer wrong?"

Not my thread, although I am sourced for the source lol. I'm working under the assumption they were outdated yeah.

I respect the guys, and I don't mean to look like I'm bashing them, but this foxconn rumor is obviously legit, so it is a head scratcher, unless the final devkits were upgraded, which we heard happened.

 

[KingSnake]

https://youtu.be/PzS4LbH5nmA?t=4m5s Says his source's info comes from about 4 to 5 months ago. Said that he didn't run the information then because he was unsure about architecture and any changes until the venture beat article was released earlier in that week.

deleted, sorry.

Ah haven't gotten a chance to watch that but if true, that really makes me believe this Foxconn leak, as it came from late November. Again, I doubt this makes much of a difference for game graphics (20% boost to GPU clocks) but the much improved CPU would be a very, very important and welcome change.

EDIT: And just as a reminder to everyone, Eurogamer/Digital Foundry explicitly admitted that they didn't know the final GPU core configuration/CPU configuration- just the clocks, and apparently from 4-5 months ago. So A72/A73s would not go against what they have reported.

That's not about the clocks info. That's about the Maxwell specs sheet that it was similar with the rumour from the old thread around here. That is what was practically a normal TX1.

 

[KojiKnight]

BoTW was said to make the battery last 3 hours already, wasn't it? If they unlocked more power later...

While I now think the cpu boost/gpu drop for gcn emulation seems like the more likely explanation now, that is still a possibility. Sony stopped giving a rats ass about the psp 1000 battery life in games once psp 2000 and later came out.

 

[plagiarize]

i agree, what i dont agree with is the thought that nintendo revealed shoddier versions of games releasing in a few weeks

Yeah, there's no way. We've seen what the launch titles will look and run like. I'm holding out hope for improvements to Splatoon 2 in docked mode, but for all the march and april stuff, we know what they're going to look like.

I wouldn't be surprised to see the framerate smoothed out a bit, but I wouldn't expect more than that.

As a reminder, we don't know how old the builds shown are.

for what it's worth this build of Breath of the Wild was labeled 1.0.0 according to digital foundry.

We might see a launch day patch, but I'm sure it's basically gold at this point.

 

[z0m3le]

That's not about the clocks info. That's about the Maxwell specs sheet that it was similar with the rumour from the old thread around here. That is what was practically a normal TX1.

He says that his source was briefed by Nintendo at that time about the specs, that would include "final clocks" info that he is saying. Eurogamer does need to clarify if the clock info was updated after that point because they give us no indication of it. I'm a fan of digital foundry so it's not like I am trying to bury them in sneaky lies here. I'm showing what they said and they said their source's info was dated around 4 to 5 months ago.

 

[Luigiv]

for what it's worth this build of Breath of the Wild was labeled 1.0.0 according to digital foundry.

We might see a launch day patch, but I'm sure it's basically gold at this point.

Really? Because the Japanese stream we saw, which would have I assumed was the same build, was labelled 0.12S.

 

[KojiKnight]

Yeah, there's no way. We've seen what the launch titles will look and run like. I'm holding out hope for improvements to Splatoon 2 in docked mode, but for all the march and april stuff, we know what they're going to look like.



for what it's worth this build of Breath of the Wild was labeled 1.0.0 according to digital foundry.

We might see a launch day patch, but I'm sure it's basically gold at this point.

Day 1 performance and bug fix patches are the norm in the world we live in.

That said, we know the build they were playing was based off of the e3 wii u build. While they certainly updated the code base, it'd still be a vertical slice from an older build while they continued working on the main code.

It shouldn't be neglected either that Zelda seems to have been a last minute decision for launch and that there was likely more they planned/still plan to do and work on.

 

[KingSnake]

He says that his source was briefed by Nintendo at that time about the specs, that would include "final clocks" info that he is saying.

He clearly talks about the specs list that was running around the twitter and ended up also on GAF.

Which specs had different clocks.These are the 4-5 months old specs:

Four ARM Cortex-A57 cores, max 2GHz
NVidia second-generation Maxwell architecture
256 CUDA cores, max 1 GHz, 1024 FLOPS/cycle
4GB RAM (25.6 GB/s, VRAM shared)

Edit: clean it up a bit to keep the main idea for the discussion

 

[LordOfChaos]

This is "X1 hidden dgpu" levels of reaching. The games themselves bear credibility to DFs leaks. Why would a pascal GPU be struggling with 1080p?

Not siding with this rumour here, but the second part of your statement doesn't make sense, Pascal is just as scalable as Maxwell. 256 Pascal ALUs at the same clock speed would have a small performance boost over 256 Maxwell ALUs at the same clock speed. There's no "Pascal GPU = X performance", the architecture is a family.

 

[DrROBschiz]

yeah.....

im gonna wait for system teardown videos in March

im hoping for good news

 

[LordOcidax]

I believe we will.

Yeap.

 

[diaspora]

This is "X1 hidden dgpu" levels of reaching. The games themselves bear credibility to DFs leaks. Why would a pascal GPU be struggling with 1080p?

Isn't it rolling 1080p 60fps for Mario Kart 8?

 

[Hattori]

It's interesting to point out that Tegra X1 works on 76.8mhz increments, so 768mhz is x10 and 921mhz is x12. Pretty lucky guess to pull that number out of thin air IMO.

Also CPU clock of quad A57 20nm @1ghz draws the same power consumption as A72 16nm @ 1.7ghz pretty interesting to say the least.

He said he saw the clocks during the demo iirc, and a lot of the times demos when debugging show the clocks of the GPU and CPU cores. He also confirmed the 1600mhz RAM speed that Eurogamer said as well.

To me it looks like the spring launch allowed for a late audible to 16nm. They are only making 2m for launch, I imagine some off the shelf design could easily be massed produced 2 years after it hit market originally.

now this is great speculation, backed up by numbers and facts 👍

 

[KingSnake]

Isn't it rolling 1080p 60fps for Mario Kart 8?

Which is fitting the story from Eurogamer, being a Wii U port. According to which in handheld mode the clock will allow Switch to reach a similar power to Wii U (with the new architecture advantage) and in the docked mode to allow the increase from 720p to 1080p. At least on the GPU side.

 

[CronoShot]

How on earth could a Foxconn worker know such details about the SoC?

 

[TAS]

This is very interesting and makes a lot of sense. The Switch's higher than expected price point would justify this plus the fact that many people were surprised at the premium level of the device itself including the screen and tech crammed into the JoyCons. It seems like Nintendo didn't hold back this time which is a welcome change. Assuming this info is accurate--how much of a performance increase are we looking at?

 

[Vena]

How on earth could a Foxconn worker know such details about the SoC?

Stability testing of an assembled unit.

 

[Hermii]

This is very interesting and makes a lot of sense. The Switch's higher than expected price point would justify this plus the fact that many people were surprised at the premium level of the device itself including the screen and tech crammed into the JoyCons. It seems like Nintendo didn't hold back this time which is a welcome change. Assuming this info is accurate--how much of a performance increase are we looking at?

Alternately they spent money on screen, controller tech and build quality which is why the device is a bit expensive from a soc viewpoint .

 

[Vena]

The proof is in the pudding. Everyone got to see how games running on Switch looked and played. There's very little reason to go down this road again.

Some people just can't help themselves and will just extrapolate from any tidbit of information endlessly until it fits the narrative they have in their head.

So no, I personally don't think the leak is accurate. We've seen this song and dance a million times before.

These differences in clock would be extremely minor the to the top-end performance ceiling. It a relatively tiny change that would be next to impossible to somehow deduce at a glance of the naked eye on a screen.

This would be more impacting to something like battery life, or emulation (where the CPU is leveraged much more heavily than the GPU).

 

[Hermii]

These differences in clock would be extremely minor the to the top-end performance ceiling. It a relatively tiny change that would be next to impossible to somehow deduce at a glance of the naked eye on a screen.

This would be more impacting to something like battery life, or emulation (where the CPU is leveraged much more heavily than the GPU).

Wouldn't it also mean a lot for ease of porting that the CPU is on par with the other platforms?

 

[KingSnake]

These differences in clock would be extremely minor the to the top-end performance ceiling. It a relatively tiny change that would be next to impossible to somehow deduce at a glance of the naked eye on a screen.

This would be more impacting to something like battery life, or emulation (where the CPU is leveraged much more heavily than the GPU).

We know the battery life, the battery capacity and the rumoured clocks with their estimated power draw. I'm pretty sure somebody (*cough* Thraktor *couch*) could make a calculation. At first sight a battery this big should have a better life with a 16nm based Tegra from this rumour.

 

[ethomaz]

How on earth could a Foxconn worker know such details about the SoC?

They need to test the console after assemble lol

 

[Hoo-doo]

We know the battery life, the battery capacity and the rumoured clocks with their estimated power draw. I'm pretty sure somebody (*cough* Thraktor *couch*) could make a calculation. At first sight a battery this big should have a better life with a 16nm based Tegra from this rumour.

Yeah, wasn't the whole point of Pascal that it would deliver some serious power saving?

Seeing as they introduced the low-end of the battery life (on stage!) as two and a half hours, i'd say they haven't really discovered any benefits of that Pascal so far.

 

[KingSnake]

They need to test the console after assemble lol

The people who should have access to all this info (fab node, SoC type, CPU, GPU and memory bandwidth) should normally know better than to post this kind of info on the internet.

It would be pretty easy to identify who leaked this if the story is true after all.

 

[z0m3le]

He clearly talks about the specs list that was running around the twitter and ended up also on GAF.

Which specs had different clocks.These are the 4-5 months old specs:



Edit: clean it up a bit to keep the main idea for the discussion

He said: he should mention that his source was briefed by Nintendo around the same sort of time frame as their video about NX 5 months ago, but afterwards. He is saying that his sources info from Nintendo is 4 to 5 months old. Yes the maxwell info is primarily what he is talking about, but he tells you when his source was given info from Nintendo, which dates those clocks to August. It's all clearly about the July devkits and he needs to indicate that any new information comes to light, only information he said changed from then specifically is venture beat's article about maxwell architecture.

 

[7Th]

I can't believe people genuinely believe Nintendo would sell a console with with these rumored specs at just 300 dollars.

 

[z0m3le]

We know the battery life, the battery capacity and the rumoured clocks with their estimated power draw. I'm pretty sure somebody (*cough* Thraktor *couch*) could make a calculation. At first sight a battery this big should have a better life with a 16nm based Tegra from this rumour.

Problem is, these specs would draw relatively the same power as the ones we were given by eurogamer. It would be hard to deduce one way or the other.

 

[Stairouais]

Uh, about that 4G thing, what if it's a kindle kind of thing, and the subscription to Nintendo online service would pay for it? (haven't read the whole thread, just thought of this)

 

[darkazcura]

We know the battery life, the battery capacity and the rumoured clocks with their estimated power draw. I'm pretty sure somebody (*cough* Thraktor *couch*) could make a calculation. At first sight a battery this big should have a better life with a 16nm based Tegra from this rumour.

The clock speeds in the OP on a 16nm process would apparently use as much power draw as the rumored clocks on a 20nm process. I don't know the calculations personally, but I think that's the point of the OP. At least that's how I'm reading it..

 

[lwilliams3]

Yeah, wasn't the whole point of Pascal that it would deliver some serious power saving?

Seeing as they introduced the low-end of the battery life (on stage!) as two and a half hours, i'd say they haven't really discovered any benefits of that Pascal so far.

If 4x A57s @ 1GHz on 20nm has roughly the same power comsumption as 4x A72s @ 1.76GHz on 16nm, it sounds to me that Nintendo decided to cram more power in the small form factor instead having a better battery life if this true.

 

[KingSnake]

Problem is, these specs would draw relatively the same power as the ones we were given by eurogamer.

Do you have a calculation to support this?

 

[KingSnake]

If 4x A57s @ 1GHz has roughly the same power comsumption as 4x A72s @ 1.76GHz, it sounds to me that Nintendo decided to cram more power in the small form factor instead having a better battery life of if this true.

This is combined with a Pascal architecture.

 

[lwilliams3]

This is combined with a Pascal architecture.

Oops. Forgot to include that part. Thanks for correcting.

 

[shark sandwich]

I see we have arrived at the "searching for hidden power" phase of the Nintendo hardware speculation process.

 

[MuchoMalo]

This is "X1 hidden dgpu" levels of reaching. The games themselves bear credibility to DFs leaks. Why would a pascal GPU be struggling with 1080p?

I don't think that it's Pascal, but you are VASTLY overestimating the difference between Maxwell and Pascal, as well as how much of an effect architecture has on performance in the first place. Not to mention that this comment is based on one port that runs like crap on Wii U as far as we know.

 

[KingSnake]

He said: he should mention that his source was briefed by Nintendo around the same sort of time frame as their video about NX 5 months ago, but afterwards. He is saying that his sources info from Nintendo is 4 to 5 months old. Yes the maxwell info is primarily what he is talking about, but he tells you when his source was given info from Nintendo, which dates those clocks to August. It's all clearly about the July devkits and he needs to indicate that any new information comes to light, only information he said changed from then specifically is venture beat's article about maxwell architecture.

It's written in stone:

Regardless, behind the scenes, sources inform us that Nintendo continued to brief developers with a spec that is uncannily similar to this Twitter leak that actually surfaced before the official reveal - and in crucial areas, it's a match for a stock Tegra X1.

One source tells us that the 4K30 aspect of the spec was not part of the developer presentation, but everything else was. We can assume that the clock-speeds are theoretical maximums, and not the 768/307.2MHz combo we've confirmed as locked in retail hardware.

 

[z0m3le]

Do you have a calculation to support this?

I'm on my phone so I'll try to accommodate you, here is A57 on 20nm:

1ghz draws 1.83 watts on a quad core.

And here we see A72 doing 1.7ghz at about 1.9 watts.

Battery calculations would be really hard to precisely pin down either cpu or gpu since there is a vague number for battery life and there are other factors beyond the soc that needs to be taken into account.

 

[Skittzo0413]

He said: he should mention that his source was briefed by Nintendo around the same sort of time frame as their video about NX 5 months ago, but afterwards. He is saying that his sources info from Nintendo is 4 to 5 months old. Yes the maxwell info is primarily what he is talking about, but he tells you when his source was given info from Nintendo, which dates those clocks to August. It's all clearly about the July devkits and he needs to indicate that any new information comes to light, only information he said changed from then specifically is venture beat's article about maxwell architecture.

The wording of the article itself seems to contradict that... From reading the article it seems like they very recently got word of the clock speeds from various developers, whereas those developers were briefed on the TX1 specs 4-5 months before. But I admit I still haven't seen the video so I can't be sure.

Some clarification from Eurogamer would indeed be nice.

This is combined with a Pascal architecture.

Wouldn't the Pascal architecture only offer power savings for the GPU? So the 20% increase in GPU clocks might slightly offset the efficiency gains, and the higher increase in CPU clocks makes up for the difference (even though A72/A73 is more efficient than A57). We also still have no idea if there are additional SMs, though the visuals of the games we've seen (like Splatoon primarily) would suggest there are not.

Some calculations would be nice, yeah.

I see we have arrived at the "searching for hidden power" phase of the Nintendo hardware speculation process.

We are contemplating a potential 20% increase in GPU clock, which is a very minor change. The CPU (outside of clocks) has always been unknown, so we're trying to figure out what they've actually gone with.

 

[Lonely1]

I don't know. The performance we have seen from the actual games seems consistent with TX1, down to DOF causing framedrops due to bandwith issues.

 

[tkscz]

This is "X1 hidden dgpu" levels of reaching. The games themselves bear credibility to DFs leaks. Why would a pascal GPU be struggling with 1080p?

Because Pascal and Maxwell are basically the same architecture? This is something I keep seeing but Pascal would only be more power (battery) efficient and be 16nm instead of 20nm. You wouldn't get any sort of frame boost or anything from it.