AMD "Ryzen C7" Smartphone SoC Specifications Listed

[yeeeeman]

Wtf is this? Samsung launches a new SoC and calls it Ryzen? The ryzen exynos? LOL
Anyways, besides the fact that it has nothing to do with...Ryzen, the GPU is nothing spectacular. By the time this launches Qualcomm will also have a new SoC, the SD875 which will bring considerable improvements given the transition to 5nm...
Also, I don't think there was something wrong with the Mali GPUs and I wouldn't be too surprised if Samsung would go from its SoC having a shitty custom CPU to a shitty custom GPU. By shitty I mean either having crap efficiency or crap drivers. Anyways, while this exotic combination of desktop gpu tech and mobile soc is nice, I don't think they can beat the efficiency of a pure mobile GPU.

 

[Vayra86]

How about Rexyzon?

 

[Fouquin]

I swear this was already confirmed fake on other outlets, but it's Monday and my brain may be tricking me.

 

[TheLostSwede]

I would think that given the cost of an engineering team capable of putting together a chip design - even from licenced designs - and getting it made would be high enough to significantly diminish the effect of licencing prices, though that of course depends on the intended market and volume of the chip. If you for example are Samsung, a couple million dollars in licencing cost for a flagship SoC would likely be negligible - it's going to sell in tens if not hundreds of millions of units after all, and at very high prices. The same would obviously be true for Qualcomm - even if they are reliant on selling SoCs for profit rather than phones, they sell hundreds of millions of flagship SoCs a year for $50-100 apiece, so a couple million dollars in extra cost for one of those designs will also be negligible as long as the performance is there. They dropped the expensive custom licencing as you said due to it no longer gaining them anything - the last time a custom Qualcomm core clearly beat the matching ARM core (and wasn't just a lightly modified ARM core with more cache) was IIRC before the Snapdragon 800-series launched. The wording around this licence makes it sound like it's a necessity to get access to the X1 core, which makes sense when it represents a new branch of SoC development for ARM and thus let's then recoup the cost of what might otherwise be a risky endeavor, while laying the groundwork for more competitive high end solutions down the line that will do better in fast tablets and laptops that will expand the ARM device market. Any chip with an X1 core should also be scalable enough to be useful in a broad range of devices, meaning chip vendors will get a lot of bang for their buck even if the cost of entry is higher than normal.

The trick with ARM is that you pay a licence per "core design", so if you paid a couple of million for say the Cortex-A76, the same applies to the Cortex-A77 and Cortex-A78...
At some point you're going to have to gauge the benefits of your customisations compared to the extra expenses in paying more for a certain license type. This is in part why there are so many Cortex-A7 and Cortex-A53 designs out there, since the companies making those chips, only had to pay for the license once.
Then there's the complexity difference, a quad core Cortex-A53 is relatively simple compared to a big.LITTLE design where you end up with larger caches and most likely difference SoC interfaces as well.

The Snapdragon 805 with the Krait 450 was the last custom core they made, so that's 2014.

The X1 actually looks like ARM doing a reverse Intel, as they're strapping one fast(er) core to two sets of slower cores, whereas Intel strapped several slower cores to a fast core with Lakefield.

Wtf is this? Samsung launches a new SoC and calls it Ryzen? The ryzen exynos? LOL
Anyways, besides the fact that it has nothing to do with...Ryzen, the GPU is nothing spectacular. By the time this launches Qualcomm will also have a new SoC, the SD875 which will bring considerable improvements given the transition to 5nm...
Also, I don't think there was something wrong with the Mali GPUs and I wouldn't be too surprised if Samsung would go from its SoC having a shitty custom CPU to a shitty custom GPU. By shitty I mean either having crap efficiency or crap drivers. Anyways, while this exotic combination of desktop gpu tech and mobile soc is nice, I don't think they can beat the efficiency of a pure mobile GPU.

If this in indeed a real thing, then why couldn't AMD have their own SKU name for it, that would be different from whatever Samsung will call it, assuming this is a collaboration between the two?

 

[Totally]

Assuming this is real, I can see this being a custom part for Microsoft and their attempt to bring Windows on ARM.
I mean, AMD has already made some custom parts for Microsoft, so why not for a different kind of product as well?
I doubt this will be for phones.

Windows on ARM has already happened but not by AMD. Microsoft SQ1 found in the Pro X that was released last fall.

 

[TheLostSwede]

Windows on ARM has already happened but not by AMD. Microsoft SQ1 found in the Pro X that was released last fall.

And? Microsoft is only allowed one hardware partner?

 

[Totally]

And? Microsoft is only allowed one hardware partner?

Do they move enough hardware to attract attention from more than one?

 

[Valantar]

The trick with ARM is that you pay a licence per "core design", so if you paid a couple of million for say the Cortex-A76, the same applies to the Cortex-A77 and Cortex-A78...
At some point you're going to have to gauge the benefits of your customisations compared to the extra expenses in paying more for a certain license type. This is in part why there are so many Cortex-A7 and Cortex-A53 designs out there, since the companies making those chips, only had to pay for the license once.
Then there's the complexity difference, a quad core Cortex-A53 is relatively simple compared to a big.LITTLE design where you end up with larger caches and most likely difference SoC interfaces as well.

The Snapdragon 805 with the Krait 450 was the last custom core they made, so that's 2014.

The X1 actually looks like ARM doing a reverse Intel, as they're strapping one fast(er) core to two sets of slower cores, whereas Intel strapped several slower cores to a fast core with Lakefield.

IIRC the Snapdragon 805 was where things started going downhill for QC, no? The 800/801 were decent enough, the 808 and 810 were downright disastrous (were those the 22nm ones?), though they managed to stay ahead just by sheer momentum, with the 820/821 being okay if unremarkable (my Oneplus 3T was perfectly fine), with the 835 onwards being back to actually delivering good chips thanks to doubling down on good implementations of Arm's designs rather than spending massive sums on customized designs that barely outperformed their stock counterparts.

And well, yes, this is very Lakefield in reverse, but then Lakefield is kind of Intel's take on big.LITTLE, so it goes around I guess?

I didn't know the licencing was per design (frankly I had assumed some sort of entry price + per shipped chip or some such), but IMO that further makes the case for ratcheting up the price of entry for a relatively limited market part like the X1 - it's the best, but if you want it, you pay for the privilege. In return you get the fastest chip around, plus a design that can likely scale (depending on your silicon design) from flagship phones to tablets and even laptops. A 2x X1 + 2x A78 + 2x A55 chip would likely make for a very snappy device for most tasks and likely scale from ~3-4W sustained at the low end to 10+W at the high end, even if WoA emulation overhead will make that less than optimal.

I'll still hold off on believing this until we see some better sourced info, but it would sure be interesting.

If this in indeed a real thing, then why couldn't AMD have their own SKU name for it, that would be different from whatever Samsung will call it, assuming this is a collaboration between the two.

That would actually make perfect sense. A deal for a part like this might for example say that Samsung gets all low-power mobile bins, AMD gets all high performance WoA laptop bins, and there'll likely be some overlap in the tablet/convertible space, with each selling their allotment of chips under their own branding.

 

[TheLostSwede]

Do they move enough hardware to attract attention from more than one?

If they pay for it, sure. I mean take the Xbox as an example, it's entirely custom hardware, so why not get a custom SoC they can use for more than one thing?

 

[mtcn77]

- "65"% – spare me the pleasantries...

I don't think this gets the due credit. Arm is different. You don't get memory buffer overflow with arm. Everything can shuffle the same amount of work. Quite a different environment where there is room for premium since it is open for comparison on a seperate chip with a seperate power target. Qualcomm always wins in power efficiency for instance - never fails. So it means something to win at a benchmark, quite so different on x86. You can have different levels of native performance depending on the software adoption level on x86. On Arm it is dependent on hardware adoption level, imo.

 

[TheLostSwede]

IIRC the Snapdragon 805 was where things started going downhill for QC, no? The 800/801 were decent enough, the 808 and 810 were downright disastrous (were those the 22nm ones?), though they managed to stay ahead just by sheer momentum, with the 820/821 being okay if unremarkable (my Oneplus 3T was perfectly fine), with the 835 onwards being back to actually delivering good chips thanks to doubling down on good implementations of Arm's designs rather than spending massive sums on customized designs that barely outperformed their stock counterparts.

And well, yes, this is very Lakefield in reverse, but then Lakefield is kind of Intel's take on big.LITTLE, so it goes around I guess?

I didn't know the licencing was per design (frankly I had assumed some sort of entry price + per shipped chip or some such), but IMO that further makes the case for ratcheting up the price of entry for a relatively limited market part like the X1 - it's the best, but if you want it, you pay for the privilege. In return you get the fastest chip around, plus a design that can likely scale (depending on your silicon design) from flagship phones to tablets and even laptops. A 2x X1 + 2x A78 + 2x A55 chip would likely make for a very snappy device for most tasks and likely scale from ~3-4W sustained at the low end to 10+W at the high end, even if WoA emulation overhead will make that less than optimal.

I'll still hold off on believing this until we see some better sourced info, but it would sure be interesting.

That would actually make perfect sense. A deal for a part like this might for example say that Samsung gets all low-power mobile bins, AMD gets all high performance WoA laptop bins, and there'll likely be some overlap in the tablet/convertible space, with each selling their allotment of chips under their own branding.

The 800 and 801 were VERY toasty, in fact, they were the hottest running mobile phones SoCs made to date at the time, the only thing worse was the 810 with the toasty Cortex-A57 in it. I had an HTC something or the other with both the 801 and 810 and they were the only phone's I've owned where you could feel it heat up in your hand doing almost nothing. So no, I don't really agree with you there, as they were clearly doing something wrong with their designs at the time.
They were 28nm parts https://en.wikipedia.org/wiki/List_...on-chip#Snapdragon_800,_801_and_805_(2013/14)

Oh, I didn't infer that ARM was trying to copy Intel, but it's like we now have Huge.big.LITTLE, which is getting a bit confusing, but it was clearly something that Samsung and Qualcomm got into first with the one or two faster cores.

The license isn't per design, it's per core SKU you want from ARM, but on top of that, if reading the ARM site correctly, depending on the license you have, you also have to pay them at tape-out and a per unit made/sold royalty as well...

I guess we're going to have to wait and see with regards to performance, but this is why I said there's likely to be custom implementations, as we'll most likely see single X1 cores in high-end phones and dual X1 cores in something like what you're proposing.

As for the supposedly AMD part, who knows, right now it's an image on the internet...

That assumption of yours makes sense as well, if they can bin them like that, or at least make a couple of very similar SKUs that doesn't require too much change in terms of manufacturing.

 

[zlobby]

even so it might be limited to samsung device only. for mobile market having qualcomm modem/baseband is almost mandatory for other smartphone maker due to legal cost. this is what kills nvidia tegra beyond tegra 3 and also intel despite having their own modem. and in intel case they also do contra revenue which cost them billions and yet they still fail to make even slight dent on qualcomm dominance. in smartphone world CPU/GPU performance comes as seconds. you need hardware based modem/baseband first and foremost.

I think Samsung will gladly sell these if they have enough inventory. This way they will land a huge blow on Qualcomm.

 

[renz496]

This will be really interesting. Its going to be AMD vs AMD since the Adreno is also based off an AMD design previously. Anyway, it is good to see a good graphic leader in the SOC GPU space again.

This may also put the Nvidia Tegra SOC to shame with the old ARM processor and Nvidia graphics.

This probably won't even reach the success nvidia once had with tegra. The mobile market is not like pc market. Good cpu and gpu does not guarantee success. But having hardware based modem is the primary key feature. Qualcomm have very strong lockdown in this market to the point they make something that seems unfair from competition standpoint into a legal advantage. This samsung partnership with AMD is their latest effort to wrestle their way through qualcomm dominance. What samsung hate the most with qualcomm probably the fact that they can't no longer license qualcomm modem to be included in their exynos design. But their exynos development itself also in rough spot.

 

[SIGSEGV]

So happy that the collaboration between two giants will challenge the domination of qualcomm's snapdragon (intel version on mobile soc).
let's see the battle begins in early 2021.
Whoever the winner please shut up and take my money lol
I am ready...

 

[Mouth of Sauron]

Assuming this is real, I can see this being a custom part for Microsoft and their attempt to bring Windows on ARM.
I mean, AMD has already made some custom parts for Microsoft, so why not for a different kind of product as well?
I doubt this will be for phones.

I think this is very realistic. Much more than "AMD designs complete ARM SoC for Samsung" or stuff like that...

 

[xBruce88x]

Nice, tho not surprised that a new amd gpu beats old radeon. Adreno = Radeon.

Wouldn't be surprised to see this Ina future Asus ROG phone or the like

Would also be interesting to see Gpd do something with it

 

[THANATOS]

In my opinion this is fake. Look at the GPU!
Who would believe a 4CU RDNA2 at 0.7GHz with 360GFLOPs could be 45% faster than Adreno 650, which has 3-4x more GFLOPs. It's true, GFLOPs between different architectures don't mean the same performance(RDNA vs Vega), but this is just too much of a difference to be real.

 

[zlobby]

This probably won't even reach the success nvidia once had with tegra. The mobile market is not like pc market. Good cpu and gpu does not guarantee success. But having hardware based modem is the primary key feature. Qualcomm have very strong lockdown in this market to the point they make something that seems unfair from competition standpoint into a legal advantage. This samsung partnership with AMD is their latest effort to wrestle their way through qualcomm dominance. What samsung hate the most with qualcomm probably the fact that they can't no longer license qualcomm modem to be included in their exynos design. But their exynos development itself also in rough spot.

True, but check out how Samsung's 990 SoC spelled disaster for their latest S20 series.

 

[Xajel]

That's not what ARM said. The X1 is "off roadmap" and seemingly any ARM licensee can make their own X1 implementation, so expect a lot of different implementations to show up.

What ARM said was that X1 was off roadmap, and only those who were on the X1 project from the beginning will be able to use it, so the first X1 will only be used by selected licensees, and if this rumor is correct then AMD is one of these. If I'm guessing also, I think Samsung is another one and maybe Qualcomm, Both Samsung and Qualcomm stopped developing their custom High-Performance ARM awhile ago.

 

[TheLostSwede]

What ARM said was that X1 was off roadmap, and only those who were on the X1 project from the beginning will be able to use it, so the first X1 will only be used by selected licensees, and if this rumor is correct then AMD is one of these. If I'm guessing also, I think Samsung is another one and maybe Qualcomm, Both Samsung and Qualcomm stopped developing their custom High-Performance ARM awhile ago.

Most likely Huawei/HiSilicon as well. I doubt MTK is in on this.

 

[ScaLibBDP]

It looks like an Absolutely Fake news because:

1. Adreno is a transformed Radeon word, and
2. I do Not see any reason why AMD should use ARM technology instead of Ryzen technology for mobile market

 

[THANATOS]

It's a fake but I disagree with your examples of why It's fake.
1. Adreno is a GPU brand from Qualcomm and they bought Imageon from AMD, that's why It's an Anagram of Radeon.
2. Maybe because ARM provides better performance at very very low TDP <=5W? And I am not sure If Ryzen can use Android without emulating It.

 

[Valantar]

It looks like an Absolutely Fake news because:

1. Adreno is a transformed Radeon word, and
2. I do Not see any reason why AMD should use ARM technology instead of Ryzen technology for mobile market

I'm not saying this is real, but both of those arguments fall flat.
1: It is public knowledge at this point that there will be an Radeon RDNA (2?) GPU in an upcoming Samsung mobile SoC. It has been stated publicly by both companies.
2: It is quite unlikely that Zen can scale low enough in power for applications like this. While Zen 2 cores are extremely efficient for PC chips, and can run on just a couple of watts at relatively high speeds, that's not enough for a smartphone or tablet SoC. It would need at least 4 cores (ideally 8) and peak at ~5W with sustained power draw at ~3W, which would likely be impossible with Zen. They also couldn't use infinity fabric, again due to power draw.

 

[john3825]

Gaugin ? its gauguin
Amd arm cpu ? lol no

Samsung soc make by tsmc ? lol no

There are miss spellings. Who did proofread the leaked information. Gaugin, Arm microarchitecture is nothing to do with Ryzen made of ZEN microarchitecture. Gaugin, arm microarchitecture can not put on Ryzen made of ZEN microarchitecture.

There are miss spellings. Who did proofread the leaked information. Gaugin, Arm microarchitecture is nothing to do with Ryzen made of ZEN microarchitecture. Gaugin, arm microarchitecture can not put on Ryzen made of ZEN microarchitecture. Even an amateure can understand that!!! Do not insist something with wrong information!!!

A fabless ccompany went woo wrong. They should consider doing something without mistakes.

 

[davideneco]

There are miss spellings. Who did proofread the leaked information. Gaugin, Arm microarchitecture is nothing to do with Ryzen made of ZEN microarchitecture. Gaugin, arm microarchitecture can not put on Ryzen made of ZEN microarchitecture.

There are miss spellings. Who did proofread the leaked information. Gaugin, Arm microarchitecture is nothing to do with Ryzen made of ZEN microarchitecture. Gaugin, arm microarchitecture can not put on Ryzen made of ZEN microarchitecture. Even an amateure can understand that!!! Do not insist something with wrong information!!!

A fabless ccompany went woo wrong. They should consider doing something without mistakes.

2 year laterstill fake