Article [SA] "Samsung kills off a major silicon design program" ... RIP Mongoose?

[NTMBK]

Samsung kills off a major silicon design program

Samsung just knifed a major silicon development program, one of their key halo products.

www.semiaccurate.com

Story is behind a paywall, but speculation and rumours on Twitter suggest that it might be the Austin CPU team that has been shut down. If true, sad to see a CPU architecture bite the dust. NVidia's CPU seems to have gone embedded only, Samsung's is dead, Qualcomm are just tweaking ARM designs instead of making their own... are Apple the only ones left making a fully custom CPU in phones?

 

[DrMrLordX]

I don't like hearing about any CPU design team being shut down. Less interesting tech for consumers to play with. What does that mean for Exynos moving forward? Bone stock ARM designs? A76/A77/A78? Which is basically what Huawei and Qualcomm are doing?

 

[Nothingness]

Story is behind a paywall, but speculation and rumours on Twitter suggest that it might be the Austin CPU team that has been shut down.

That's true.

If true, sad to see a CPU architecture bite the dust. NVidia's CPU seems to have gone embedded only, Samsung's is dead, Qualcomm are just tweaking ARM designs instead of making their own... are Apple the only ones left making a fully custom CPU in phones?

CPU development costs have increased a lot and ARM Ltd teams have been doing better the last years. So that was alas predictable.

 

[NTMBK]

I don't like hearing about any CPU design team being shut down. Less interesting tech for consumers to play with. What does that mean for Exynos moving forward? Bone stock ARM designs? A76/A77/A78? Which is basically what Huawei and Qualcomm are doing?

There's the AMD GPU deal they announced a few months ago, so presumably we'll see A77+RDNA.

 

[moinmoin]

I don't like hearing about any CPU design team being shut down. Less interesting tech for consumers to play with. What does that mean for Exynos moving forward? Bone stock ARM designs? A76/A77/A78? Which is basically what Huawei and Qualcomm are doing?

What advantageous customization did the Exynos chips have? In the comparisons with Qualcomm's offerings they nearly always looked worse in Samsung's own phones where they picked Exynos or Snapdragon based on region of sales.

 

[abufrejoval]

Samsung kills off a major silicon design program

Samsung just knifed a major silicon development program, one of their key halo products.

www.semiaccurate.com

Story is behind a paywall, but speculation and rumours on Twitter suggest that it might be the Austin CPU team that has been shut down. If true, sad to see a CPU architecture bite the dust. NVidia's CPU seems to have gone embedded only, Samsung's is dead, Qualcomm are just tweaking ARM designs instead of making their own... are Apple the only ones left making a fully custom CPU in phones?

Well, perhaps AMD bled them of so much talent, that they couldn't continue (and will take who's left).

If Samsung prices hadn't skyrocketed beyond the reasonable, I'd have been angry for selling a sub-par product to us poor Europeans. Perhaps they can now afford to deliver an affordable product with less bling-bling and better specs.

Some of us do actually travel, Samsung, so we really need a global phone, not an EU cut-down.

 

[NostaSeronx]

There's the AMD GPU deal they announced a few months ago, so presumably we'll see A77+RDNA.

What we should hope for is Zen5/RDNA4 on 3nm GAAP by 2022 in mobile phones.

 

[darkswordsman17]

This makes sense as ARM's reference designs have outstripped everyone but Apple in the CPU (well ARM CPU, although ARM has started pushing Intel and AMD). Nvidia, who desperately needs CPU now that AMD and Intel are making GPUs (which means they'll be able to start squeezing Nvidia out), haven't had any luck getting traction on their CPU design (of course they took a more drastic approach). But Qualcomm couldn't either so it was becoming kinda futile. On top of that, custom ARM designs in many ways was fracturing the market, causing software related issues that prevent them from maximizing development (its why all the ARM GPUs have been chasing feature set of the PC GPUs, meet the standards to maximize compatibility). With ARM growing outside of mobile, it kinda needs to converge some to aid software development.

I'd guess another elephant in the room is IP issues. I think that was the major impetus behind the AMD GPU deal, but like with that, if they work with others they can pool R&D money and Samsung can trim some fat. Now Samsung lets others do the grunt of the chip design, and they can focus on the production engineering. Seems like a smart move for Samsung.

Heck, maybe AMD would absorb the team and put them to work developing the Samsung designs (pairing ARM reference with AMD GPU)? I think AMD could make the case, especially for their semi-custom unit as not having a track record with ARM very possibly cost them the Nintendo Switch deal (and there's definitely going to be others that want ARM designs, as streaming and VR/AR begins to grow).

Seemingly the team would be complete so AMD wouldn't need to do much more than give them some resources (money mostly, and if they use the ARM reference stuff mostly then ARM will be doing most of the chip design work) and since its still some time before the Samsung chips will be showing up, it shouldn't upset anything currently (i.e. steal production from AMD's other stuff; while giving them some help in using Samsung's fabs) and AMD could possibly work a deal with Samsung to get the team for cheap due to their established deal.

 

[beginner99]

What advantageous customization did the Exynos chips have? In the comparisons with Qualcomm's offerings they nearly always looked worse in Samsung's own phones where they picked Exynos or Snapdragon based on region of sales.

Agree thats probably why it gets canned. Why spend the money if a tweaked core is better than your custom one? Still, you will need a good team to tweak default core.

Another reason might be that CPU is more and more and afterthought, at least marketing wise. It's about gpu, AI, camera / image processing, sound, etc. and not boring CPU power. So the fixed function / enhancemnt blocks matter more and more and due to power constraint that actually makes some sense.

 

[darkswordsman17]

What advantageous customization did the Exynos chips have? In the comparisons with Qualcomm's offerings they nearly always looked worse in Samsung's own phones where they picked Exynos or Snapdragon based on region of sales.

I can't speak to what Samsung's customization was, but I got the impression that a lot of the issue with performance was software related. And I think that's easy to overlook. With ARM expanding (they're pushing into laptops, and the overall mobile market is changing as more smart devices and AR/VR develop as markets), I think there's benefits to the industry converging some, especially when most companies were not doing well on their custom ARM stuff (Qualcomm's is mostly integrating their modem; Nvidia's is their GPU).

Frankly, I think Exynos was mostly just so that Samsung would have a small chip to produce as they developed their fab processes, and as a bit of a safety net due to some of the IP issues, but I think things were starting to change there such that Samsung's custom stuff was possibly going to be more problematic from an IP standpoint (which is why I think they worked the GPU deal with AMD).

 

[soresu]

What we should hope for is Zen5/RDNA4 on 3nm GAAP by 2022 in mobile phones.

I do like your enthusiasm - but no, just no.

Intel went down that rabbit hole to no avail, ARM simply got there first and dug in deep.

On the other hand, such a chip in a standalone VR headset is less of a risky prediction - AMD already had a Carrizo based headset made by a Chinese company.

 

[soresu]

I think there's benefits to the industry converging some, especially when most companies were not doing well on their custom ARM stuff (Qualcomm's is mostly integrating their modem; Nvidia's is their GPU)

nVidia's current custom v8-A core Carmel is clearly aimed at AI/ML applications rather than general use, I doubt we shall see it in a Shield someday, it trades blows with A76, but A77 was announced 4 months ago, and A78 will add at least 15% IPC on top of that in 2020-21.

I had heard that their efforts to snag contracts for driverless car compute solutions were not doing as well as hoped - Tesla is going with its own design, and nVidia were banking on them I think.

I'm still hopelessly waiting for a refreshed Shield TV with A76-77 and a fresh GPU in it to at least match PS4 performance, but alas.....

 

[soresu]

a lot of the issue with performance was software related

Sounds like a VLIW issue if so - they are great on paper, but compiling optimally for them is not so easy.

 

[Yotsugi]

perhaps AMD bled them of so much talent

I doubt Mongoose team (formerly Cat cores team) would go back to AMD, but you'll never know!

nVidia's current custom v8-A core Carmel is clearly aimed at AI/ML applications rather than general use

It is for automotive, mostly.
It also just sucks on top of all that.

 

[soresu]

Another reason might be that CPU is more and more and afterthought, at least marketing wise. It's about gpu, AI, camera / image processing, sound, etc. and not boring CPU power. So the fixed function / enhancemnt blocks matter more and more and due to power constraint that actually makes some sense.

The ever expanding repertoire of vision/ML/AI means that accelerators have a very interesting place in the future of mobile - voice/face/gesture/expression/handwriting recognition, language translation, voice synthesis, denoising camera image/video, motion compensation, fog/mist reduction in images....

The list goes on and on, as long as you have the necessary space to store the various code/NN's on the device (and Google seem to be on the ball with reducing that from the last IO conference).

Computational imaging is another important one - the recent flagships have a ridiculous amount of image sensors, combining their inputs effectively with computation can offer more choice to the user/artist, perhaps even simple VR video recording with some creative physical design of the phone/device.

Certainly with the right sensors or combination of them, you can 3D map the environment efficiently with such accelerators.

 

[Yotsugi]

The ever expanding repertoire of vision/ML/AI means that accelerators have a very interesting place in the future of mobile - voice/face/gesture/expression/handwriting recognition, language translation, voice synthesis, denoising camera image/video, motion compensation, fog/mist reduction in images....

Camera and gimmicks, yes.

The last bastion of differentiation in a mature, long overdue for a decline market.

 

[soresu]

It is for automotive, mostly.

That's where the AI/ML I mentioned comes in, though obviously aided by Volta GPU power too, so somewhat pointless, and yes likely neither will be as efficient as a custom solution designed from the ground up for ML acceleration/efficiency.

Camera and gimmicks, yes.

True, but those that use smartphone cameras a lot, rather than something like a DSLR/Cinecam/Camcorder don't want to spend time on a PC in Photoshop, or Premiere to get the right look, they want it fast and efficient on their phone.

Not the way I would go, but I'm in a visual field anyway, the long way just seems better - I'm starting to sound like my dad talking about vinyl LP's, I know it......

The last bastion of differentiation in a mature, long overdue for a decline market.

Pretty much.

IMHO as standalone VR/AR matures in features and compact/lightweight form factor, I think it will pick up some people less than excited by the lackluster visual improvement (by which I mean easily visible) in recent smartphones.

Though I believe that if the rising 'gaming' phone market gets its act together and creates a compelling hardware/controller package similar to the Switch, and sponsors some Android AAA games development, then smartphones may limp on some time longer.

 

[Yotsugi]

That's where the AI/ML I mentioned comes in

Not on CPU, Xavier uses Volta GPU + PVA.

True, but those that use smartphone cameras a lot rather than something like a DSLR don't want to spend time on a PC in Photoshop or Premiere to get the right look, they want it fast and efficient on their phone.

True, but computational photography quickly turns to diminishing returns at iso sensor sizes.

IMHO as standalone VR/AR matures in features and compact/lightweight form factor

It's been years since VR hype started and we've yet to see at least a single killer usecases.

Though I believe that if the rising 'gaming' phone market gets its act together and creates a compelling hardware/controller package similar to the Switch, and sponsors some Android AAA games development, then smartphones may limp on some time longer.

Oh no, no, definitely no, for phone vendors have even less devrel than the likes of Google.
Phone game market is either gacha or proper gameplay coated with ludicrous amounts of gacha.

 

[soresu]

It's been years since VR hype started and we've yet to see at least a single killer usecases.

Film and TV, it always been that for me - I just won't pull the trigger while they are bulky, low FOV, heavy and less than 4K (I have a small room, a TV big enough to appreciate 4K would deprive me of necessary space).

It's already offered by some airlines I believe, it makes sense in space constrained environments - IMHO it does also offer some light at the end of the tunnel for ridiculous multi monitor desktop set ups, as long as you can see the keyboard and mouse that is.

I also think that the 3D audio headset problem does need to be fixed - hopefully MEMS speakers can rescue us from the deluge of 'Virtual 7.1' products that have saturated the market by providing real multi driver surround sound in a headset.

 

[ehume]

... are Apple the only ones left making a fully custom CPU in phones?

IIRC the Apple CPUs are upgraded ARM CPUs.

 

[Yotsugi]

IIRC the Apple CPUs are upgraded ARM CPUs.

No they're not, both big and LITTLE Apple cores are fully custom.

 

[ehume]

No they're not, both big and LITTLE Apple cores are fully custom.

Thanks for the information.

 

[senseamp]

Samsung CPU team layoff

Seems to be huge layoff happening at samsung CPU division. If you can, help those affected by it

www.teamblind.com

CPUs are becoming commodity chips, up against the wall of diminishing returns, and all serious compute is moving to accelerators for speed or efficiency reasons. CPU's job now is to do miscellaneous tasks, host accelerators, and move data to them. So it makes sense to just use ARM cores if the financial terms are good.

 

[beginner99]

Film and TV, it always been that for me - I just won't pull the trigger while they are bulky, low FOV, heavy and less than 4K (I have a small room, a TV big enough to appreciate 4K would deprive me of necessary space).

Agree. VR has 2 problems: hardware. To do it right you need probably at least 2 2k screen, better 4k each at 120 fps. In a small form factor. That is literally at least a decade away. I haven't tried VR much but in all cases I got disoriented somewhat with time and a little sick. too much stutter.

Second problem is social. It's geeky and nerdy as hell.

 

[ThatBuzzkiller]

Sounds like a VLIW issue if so - they are great on paper, but compiling optimally for them is not so easy.

How so ? Adreno GPUs and even most popular Nvidia GPUs currently in use today use VLIW as well ...