Leak shows AMD "Ryzen C7" SoC for smartphones with impressive specs

[Vode]

6 hours ago, leadeater said:

But why? Do we need this in cellphones?

Be real. The majority of gamers is on mobile now.

[Bananasplit_00]

6 hours ago, leadeater said:

But why? Do we need this in cellphones?

Ryzen does what Inteldon't I guess

[leadeater]

46 minutes ago, Vode said:

Be real. The majority of gamers is on mobile now.

By count sure, but none of them need Ray Tracing 

 

Spoiler

Neither would hardware assisted RT on a cellphone be of any actual use anyway, increased power and die area and lower general performance, not a good idea.

 

[mr moose]

4 minutes ago, leadeater said:

By count sure, but none of them need Ray Tracing 

 

  Hide contents

Neither would hardware assisted RT on a cellphone be of any actual use anyway, increased power and die area and lower general performance, not a good idea.

 

But RT is about all that's left for candy crush, unless they release a Disney reskin...   again...

[SpaceGhostC2C]

8 hours ago, leadeater said:

But why? Do we need this in cellphones?

When your whole life flashes before your eyes, how much of it do you want to be spent on a non ray traced mobile screen?

 

 

 

[suicidalfranco]

55 minutes ago, leadeater said:

By count sure, but none of them need Ray Tracing 

wait till you see RT Clash of Clans, it will be glorious 

[RejZoR]

This is good, AMD expanding to other fields. If anything, this might bring me back to Android. A Ryzen powered smartphone. I'd want it just for that alone lol

[LAwLz]

So it seems like this is fake, but my guess is that it will be pretty similar to what we will see from the next Samsung SoC (for the next Galaxy S device, not the Note).

Cortex-X1 for the big cores, Cortex-A78 for the middle cores and Cortex-A55 for the small cores. The clock speeds seems about right as well.

TSMC 5nm might be ready by next year.

Samsung and AMD have collaborated on building a GPU for mobile devices so the "RDNA_2 Mobile" could be real. Not sure about the other details though.

Probably won't use MediaTek's modem since Samsung has their own.

Memory seems believable.

Display support, sure.

The other stuff? Yep, seems pretty standard.

 

I'd say 90% of this will turn out to be more or less accurate, but that it's the Samsung SoC instead.

With that being said, I think this is a completely fabricated "leak". But the stuff here isn't that hard to think of, and will probably be true.

 

 

 

9 hours ago, Commodus said:

If this is accurate (it's easy to create a bogus wall of text, so be careful), it'd finally end some of the stagnation in Android phone CPUs and might give Apple some real competition performance-wise.  Raytracing probably won't get much use in mobile games for a while, but I'd rather have it than not.

CPU wise, it seems like the X1 will be comparable to the big cores in the A13. The medium cores (A78) will be faster than the small cores in the A13 though. And then we got 4 shitty A55 cores. So overall, if this SoC were to be true and be released today, it would stack up really well against the A13. This would probably be slightly faster at peak CPU performance.

But by the time this gets released, it will be up against the A14 and Apple will probably be in the lead again. But Apple's lead will be way smaller than it has been before.

 

 

8 hours ago, dizmo said:

Samsung SoCs are used in a lot of other products, it's just most people pick Qualcomm because they're more widely known. As Samsung's chips get better and better, they might be seen in more and more devices. I think Samsung's flagship SoCs aren't made in enough numbers to be used in anything but their own devices.

You sure?

I've only ever seen Samsung's SoCs (and by that I mean Exynos) used in Samsung devices and then a few Meizu devices.

I think one of them problems has been licensing for cellular connectivity (because Qualcomm are dicks), primarily CDMA.

[Vode]

3 hours ago, leadeater said:

By count sure, but none of them need Ray Tracing 

 

  Hide contents

Neither would hardware assisted RT on a cellphone be of any actual use anyway, increased power and die area and lower general performance, not a good idea.

 

I don't think we have the capabilties to assess, data or ability to decide what other may or may not need.

 

The market usually decides that. We enthusiasts are a fraction of people.

 

On the tech aspect. I think I agree with you although bleeding edge tech often seems to be useless and/or wasteful at first. But without making the first step progress is hampered. Innovation, selection and creative destruction feeds progress.

[leadeater]

27 minutes ago, Vode said:

I don't think we have the capabilties to assess, data or ability to decide what other may or may not need.

Well I think we do as we know hardware RT requires actual die area and does not add performance to any other aspect and the extra transistors require power gating when not in use but power gating is not perfect so you are taking die area and increasing power usage, in a mobile part optimized for low power, so gain what? You can only use it to do BVH calculations so unless everything else in the SoC is capable of supporting it then it's literally useless.

 

So why increase the cost of the SoC by making it bigger and also increase idle power usage for a feature you can't actually use, the GPU raster pipeline is too slow, and wouldn't be effectively utilized on smaller screens. This would only make sense on a large form factor tablet, that's it.

 

Taking the very same die area and transistors and using that for CPU or GPU would actually result in a hardware performance gain that people actually use and would benefit from.

 

Not only is this rumor likely fake Ray Tracing is just a current buzz word often used to peddle rumors because it's the in thing and people want to believe it's getting wider usage. This is just a rumor, as with all rumors do a sensibility check before believing it, hardware Ray Tracing in a mobile SoC from a company that hasn't deployed any such technology yet in a market they aren't currently even in yet is a hard fail to me.

[RejZoR]

@leadeater

I mean, there isn't any "fake" RT. It is or it isn't. You do have some room with precision, but you have to calculate ray bounces. Precision is where you define how many rays per pixel you use and how many bounces each ray does. You can go as low as 1 ray per pixel and do 1 bounce. I guess you could do ray approximation and maybe fake it to get similar effect with little performance hit, but you'd still have to do some ray tracing.

[Bombastinator]

23 minutes ago, RejZoR said:

@leadeater

I mean, there isn't any "fake" RT. It is or it isn't. You do have some room with precision, but you have to calculate ray bounces. Precision is where you define how many rays per pixel you use and how many bounces each ray does. You can go as low as 1 ray per pixel and do 1 bounce. I guess you could do ray approximation and maybe fake it to get similar effect with little performance hit, but you'd still have to do some ray tracing.

Definition of terms. Totally fake ray tracing can be done. It’s been a thing since the 15th century in western art.  Apparently one can use bits of this to dramatically enhance the effect of small amounts of actual ray tracing. “Faked up” maybe.

[leadeater]

32 minutes ago, RejZoR said:

@leadeater

I mean, there isn't any "fake" RT. It is or it isn't. You do have some room with precision, but you have to calculate ray bounces. Precision is where you define how many rays per pixel you use and how many bounces each ray does. You can go as low as 1 ray per pixel and do 1 bounce. I guess you could do ray approximation and maybe fake it to get similar effect with little performance hit, but you'd still have to do some ray tracing.

I said the rumor is fake bouncing of the hype of RT due to people wanting to believe anything around that. Probably needed a comma.

[Bombastinator]

The thing about RT is it’s simultaneously the most computationally intensive and least useful thing that can be done.  It’s difficult to even see the effects of RT on something quite frequently, while using huge sections of compute capacity.  Existence of RT says “we’ve run out of really useful stuff so we’re looking for scraps”.  This might be where phones are now I guess.  Kinda doubting it.

[leadeater]

1 hour ago, Bombastinator said:

The thing about RT is it’s simultaneously the most computationally intensive and least useful thing that can be done

Well it's only not useful right now because we can't, due to hardware performance or not found superior methods, do an entire RT render path in real time to replace everything we do now. Remember when we didn't have dynamic shadows and lighting? There was a reason for that, similar reason to this. We'll likely make that transition but I wouldn't put money on the next 2 hardware/architecture generations being that time.

 

If the whole RT thing isn't a flop on consoles I suspect we'll see a lot more noticeable and useful RT usage in games, whether PC gamers want to admit it consoles do dictate industry trends and technology development that flows all the way down to game engine design and supporting tool-sets and all the way up to what we get on the PC. People wonder why graphics sliders and settings don't do much after medium or high, consoles is your answer (diminishing returns on fidelity improvements on features limited to what consoles support, rare usage of sponsored features like RTX and GameWorks).

[WereCat]

12 hours ago, Taf the Ghost said:

Hasn't this been conclusively debunked as a fake?

Yes 

[Tech_Dreamer]

14 hours ago, leadeater said:

But why? Do we need this in cellphones?

[Bombastinator]

So the actual article is apparently almost certainly garbage, but the discussion is now about processing power increases in cellphones.   
 

cellphones will increase in processing power as their battery power envelope allows.  One issue is RT, which has gigantic processing requirements for a return that is pretty questionable currently.   Will cellphone power envelopes eventually get to a point where processing power becomes gratuitous and RT becomes something to add? Probably, eventually.  Has this already happened?  Maybe.  I doubt it.    will RT develop enough to the point it is worth making space for amongst more mission critical things?  Possibly.  Not yet though.

[Doobeedoo]

Saw this coming eventually. Well maybe not the actual x86 though Ryzen name reminded me about it. We knew they work with Samsung to use RDNA based GPU for the new SoC but I knew we'd see Cortex X1 cores for this SoC though. We know that those cores are finally a new high performance tier aka Apple big cores, so this will be very fun to see. The 4 cores config of those will probably be for tablets and some 2in1's or such. 

[Bombastinator]

18 minutes ago, Doobeedoo said:

Saw this coming eventually. Well maybe not the actual x86 though Ryzen name reminded me about it. We knew they work with Samsung to use RDNA based GPU for the new SoC but I knew we'd see Cortex X1 cores for this SoC though. We know that those cores are finally a new high performance tier aka Apple big cores, so this will be very fun to see. The 4 cores config of those will probably be for tablets and some 2in1's or such. 

It’s why the fake got as far as it did I think.  It’s not specifically impossible, just bleeding edge.  The 5nm thing is done because there isn’t much 5nm out yet so there isn’t any data and it makes the power draw stuff less implausible.  Standard mountebank technique. Fraudulent science lives in the grey area of just outside proven things where stuff might be possible.

[JoostinOnline]

This is possibly one of the most ridiculous rumors I've ever seen on here. There are soooooo many red flags. I'd give a Nigerian Prince a loan before I bet on this being true.

[Bombastinator]

35 minutes ago, JoostinOnline said:

This is possibly one of the most ridiculous rumors I've ever seen on here. There are soooooo many red flags. I'd give a Nigerian Prince a loan before I bet on this being true.

It can be debunked using non science stuff but even the science doesn’t really hold up.  Shows how people will believe even a really weak lie as long as it’s pretty enough.  Been a whole lot of that going around for years lately

[dizmo]

7 hours ago, LAwLz said:

You sure?

I've only ever seen Samsung's SoCs (and by that I mean Exynos) used in Samsung devices and then a few Meizu devices.

I think one of them problems has been licensing for cellular connectivity (because Qualcomm are dicks), primarily CDMA.

They're used in some Motorola and Vivo phones, some older Lenovo (maybe still is, haven't looked) as well as quite a few DAPs.

[williamcll]

The CPU performance doesn't look too impressive considering it's still using a similar layout to what Qualcomm has to offer.

 

Curious to see how it hold up against Apple.