What’s the fate of K12/ARM at AMD?

treevor

Junior Member
Jun 1, 2019
17
5
36
Was revisiting Ian’s interview with Jim Keller and completely forgot he was working on the K12 architecture before he left for Tesla/Intel and was kinda curious if AMD will ever revive it or where it could of from there.

It would be really interesting to see the Ryzen team incorporate ARM/RISC-V into their laptop platform as an ultra low power alternative to zen or as some sort of coprocessor. The possibilities are endless but all we can do is speculate so hey, why not?!
 

ubern00b

Member
Jun 11, 2019
171
75
61
There are fully functioning x86 Zen processors for mobile with a 15w TDP, what could ARM/RISC do that could improve upon this?
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
There are fully functioning x86 Zen processors for mobile with a 15w TDP, what could ARM/RISC do that could improve upon this?

Not to be a stick in the mud, but there are some compelling ARM offerings at or below that power envelope. A12 and A12x are seriously fast. Ignore the GB4 results and look at SPEC scores instead. 8cx may also be a winner but I haven't seen many good benchmarks on it yet. These chips happen to be coming from companies that are not AMD. I don't think it's rational to conclude that AMD can dust off K12 and compete with those guys immediately, or perhaps ever.
 

moinmoin

Diamond Member
Jun 1, 2017
4,934
7,619
136
K12 and Zen were worked on by the same team at the same time, so aside ARM vs. x86 cores (possibly even only the decoder frontend stage) there likely was a very big overlap. Note that every Zen chip still always includes ARM for TrustZone, likely also for the scalable control fabric (SCF) part of IF as well as management tools of the PRO chips.
 

Thala

Golden Member
Nov 12, 2014
1,355
653
136
Was revisiting Ian’s interview with Jim Keller and completely forgot he was working on the K12 architecture before he left for Tesla/Intel and was kinda curious if AMD will ever revive it or where it could of from there.

It would be really interesting to see the Ryzen team incorporate ARM/RISC-V into their laptop platform as an ultra low power alternative to zen or as some sort of coprocessor. The possibilities are endless but all we can do is speculate so hey, why not?!

It is not so interesting of having yet another low power SoC but why not a real high performance server/desktop part with SVE and all the bells and whistles? Thing is the highest performant ARM consumer SoCs maxing out at about 7W TDP like the Qualcomm 8CX - this puts quite a limit on the frequencies. Still the higher TDP of 8CX compared to Snapdragon 855 should enable roughly 3GHz all-core.
Now imagine what could be done at 15W and higher.
Or maybe an implementation of ARM Neoverse N1 platform or similar? We are talking 64 cores with around 100W TDP!
 
Last edited:

treevor

Junior Member
Jun 1, 2019
17
5
36
I see more opportunity for a coprocessor than a replacement chip. Maybe it’s the tinkerer in me but i would just like to some sort of secondary chipset to revive serial analog connections and interface with various new/legacy hardware. I hate that laptops are going to all usb all the time, I have to service too much old machinery that needs legacy connectors and if i had a configurable serial interface i could slap together a port monitor, read and format the data i need without the need of old legacy equipment..... idk it’s a bit of a pipe dream
 
Last edited:

NostaSeronx

Diamond Member
Sep 18, 2011
3,683
1,218
136
My thoughts are K12/ARM was shuffled away for a full x86 plan.

Zen => High cost, High peformance ==> Higher : same ASPs/Margins per generation.
Excavator => Low cost, Low power ==> Lower : same ASPs per generation, with growing margins as FDSOI processed wafer prices drop off.

post-Zen(ex:Zen2, Zen3, Zen4, Zen5) successive architectures would have taken the FinFET/GAA roadmap; 14LPP -> 7LP -> 5LP -> 3LP "high-performance, power-efficiency at all costs"
post-Excavator(ex: NotExcavator2, etc) successive architectures would have taken the FDX roadmap; 28nm BLK -> 22FDX -> 12FDX -> 7FDX "right balance of performance, power and cost"

For this to work out there needs to be a successor to 2019's A6-9220C/A4-9120C. Someone that architected for Bobcat/Jaguar is supposedly doing the HE(Highly Efficent)/ULP(Ultra Low Power) x86 core. As far as it goes, it will supposedly stay Cluster-based Multithreading. Do to Stoney Ridge's success as a low-cost SoC over the previous Bobcat/Jaguar generations which bombed(Bhavani(AM1)/Beema(FT3b)/Carrizo-L(FT3b) all 28nm HKMG (FDX)).
 
Last edited:
  • Like
Reactions: BHZ-GTR and amd6502

NTMBK

Lifer
Nov 14, 2011
10,208
4,940
136
K12 appears to be shelved. I don't think it is in AMD's interests to encourage ARM servers. Right now the server AMD64 CPU market is basically a duopoly, and AMD and Intel aren't giving out IP licenses to allow for new competitors- whereas ARM will happily sell a license to anyone with a big enough cheque book. Better for AMD to let the ARM market flounder and die, and keep a nice big chunk of a restricted market.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
@NTMBK

Not sure it will flounder and die. Flounder, sure. Die? Nah. If AMD prices their offerings the same way Intel has Xeons in the past, the anti-Intel/pro-ARM crowd will noisily call for AMD's ouster.
 
  • Like
Reactions: lightmanek

moinmoin

Diamond Member
Jun 1, 2017
4,934
7,619
136
Right now the server AMD64 CPU market is basically a duopoly, and AMD and Intel aren't giving out IP licenses to allow for new competitors- whereas ARM will happily sell a license to anyone with a big enough cheque book. Better for AMD to let the ARM market flounder and die, and keep a nice big chunk of a restricted market.
Indeed. And AMD's business is directly competing against the ARM ecosystem as well. Whereas Intel essentially only offer off the shelf chips in huge contrast to ARM's licensing allowing fully customizing chips, AMD through its semi-custom business enables a middle way, using high performance "industry standard" x86 cores while still combining it with custom IPs and platform configurations (see the Frontier exascale supercomputer being confirmed customized). That's a competitive advantage no other player can offer.
 

amd6502

Senior member
Apr 21, 2017
971
360
136
Well, they have tested the waters with the A1100 series and it seems acorn server ecosystem system still needs a lot of time to mature. They will probably stand by and watch players like xilinx. If they get an offer for a custom chip that fits K12, they may move ahead, or if the ARM market starts to look lucrative they will move ahead with a low risk medium low budget project.

It would be nice if they could reuse the Rome or Milan IO hub. If it shares components from Zen2 or Zen3 then unfortunately FDX chiplets doesn't look likely. (though I don't even know if it's likely whether ARM server will really catch on.)

My thoughts are K12/ARM was shuffled away for a full x86 plan.
[...]
For this to work out there needs to be a successor to 2019's A6-9220C/A4-9120C. Someone that architected for Bobcat/Jaguar is supposedly doing the HE(Highly Efficent)/ULP(Ultra Low Power) x86 core. As far as it goes, it will supposedly stay Cluster-based Multithreading. Do to Stoney Ridge's success as a low-cost SoC over the previous Bobcat/Jaguar generations which bombed(Bhavani(AM1)/Beema(FT3b)/Carrizo-L(FT3b) all 28nm HKMG (FDX)).

Maybe a single core Zen3 would have almost all the low power advantages that a Stoney++ 4 thread APU would have. The only disadvantage might be cost of a minimum die area 7nm+ or 5nm APU. There 22FDX or maybe even 12FDX non-Zen core APU may have an advantage.
 
Last edited:

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
There probably isn't much of a reason to revive K12 at this point, unless Intel implodes.
 

treevor

Junior Member
Jun 1, 2019
17
5
36
Yeah I did a little digging and the only traces of a single K chip of any kind in the recent product releases is the x470 k17 north bridge. Looks like my dream of an x86less society is at bay for another day
 

treevor

Junior Member
Jun 1, 2019
17
5
36
Fair, we've already seen some rumors of Apple looking into ARM for up and coming macbooks (maybe replacing the core M models)
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
Fair, we've already seen some rumors of Apple looking into ARM for up and coming macbooks (maybe replacing the core M models)

And Qualcomm has 8cx-based win10 laptops coming. 7cx too (for the budget crowd).
 

soresu

Platinum Member
Dec 19, 2014
2,617
1,812
136
Fair, we've already seen some rumors of Apple looking into ARM for up and coming macbooks (maybe replacing the core M models)

The new iPad OS is a very strong push towards Mac OS functionality, I would not be entirely surprised to find that it has even more in common with the Mac OS codebase than iOS does.

Especially considering it addresses a lot of the productivity functionality that was missing in iOS that would be necessary for migrating Mac apps to iOS/iPadOS and ARM.
 

soresu

Platinum Member
Dec 19, 2014
2,617
1,812
136
There probably isn't much of a reason to revive K12 at this point, unless Intel implodes.
Nonsense, I wouldn't be surprised if Nintendo are not especially happy with nVidia after the Tegra X1 bug fiasco that compromised Switch security measures.

A high performance core would fit in well with a Switch successor, and I doubt Nintendo want to switch away from ARM anytime soon now that they are essentially divorced from PPC.

Of course, they could probably just fit an off the shelf ARM Cortex-A core - but the benefit of going custom is that they can be very exacting for their security needs, which I imagine are much higher up their list of priorities than they used to be.

Samsung's vote of confidence in RDNA for mobile graphics also implies that it would fit well for a Switch successor.

Edit: The main problem is whether the blocks of K12 design were modular enough that they could apply the R&D of Zen 2 and onwards to K12 without a massive extra effort, which depends on how similar Zen and K12 were to begin with.
 

Shivansps

Diamond Member
Sep 11, 2013
3,835
1,514
136
And Qualcomm has 8cx-based win10 laptops coming. 7cx too (for the budget crowd).

And RPI 4... it all depends on Microsoft at this point. Running W10 ARM on RPI 3 and RPI 4 is petty much illegal now. RPI4 has enoght power if a gpu driver is ever developed. I mwould not bet on it but ill like to see RPI 4 to AM1 cpus comparisons.
 

gdansk

Golden Member
Feb 8, 2011
1,988
2,357
136
Not to be a stick in the mud, but there are some compelling ARM offerings at or below that power envelope. A12 and A12x are seriously fast. Ignore the GB4 results and look at SPEC scores instead. 8cx may also be a winner but I haven't seen many good benchmarks on it yet. These chips happen to be coming from companies that are not AMD. I don't think it's rational to conclude that AMD can dust off K12 and compete with those guys immediately, or perhaps ever.
Wasn't K12 planned to have the same backend as their x86-64 cores? How then do you imagine it would end up much faster? Maybe smaller but Zen is already minuscule.

They developed a high performance x86-64 core around 4mm^2 with up to 64 in a single socket. I can see why they abandoned the ARM route. Instead they'll bring more cores to the x86-64 market. It seems like a good strategy when ARM servers are still a rounding error in the server market. Put another way: they could develop an ARM chip similar to Rome, but how much smaller would that market be?
 
Last edited:
  • Like
Reactions: Tlh97 and amd6502

Thala

Golden Member
Nov 12, 2014
1,355
653
136
Wasn't K12 planned to have the same backend as their x86-64 cores? How then do you imagine it would end up much faster? Maybe smaller but Zen is already minuscule.

Jim Keller himself mentioned in an interview, that he can get more performance out of K12 compared to Zen at same size due to the more modern ISA. Today however K12 would have to compete with Cortex A77 - which is quite a bit smaller than Zen 2 - I assume less than half the size of Zen 2.
 

gdansk

Golden Member
Feb 8, 2011
1,988
2,357
136
Jim Keller himself mentioned in an interview, that he can get more performance out of K12 compared to Zen at same size due to the more modern ISA. Today however K12 would have to compete with Cortex A77 - which is quite a bit smaller than Zen 2 - I assume less than half the size of Zen 2.
Did he say how much? Most studies have shown that ISA is nearly irrelevant for modern compiler-produced machine code. Although I suppose a smaller decoder would allow for larger buffers, queues and caches.
 

Thala

Golden Member
Nov 12, 2014
1,355
653
136
And RPI 4... it all depends on Microsoft at this point. Running W10 ARM on RPI 3 and RPI 4 is petty much illegal now. RPI4 has enoght power if a gpu driver is ever developed. I mwould not bet on it but ill like to see RPI 4 to AM1 cpus comparisons.

Yup, missing GPU driver is pretty much a showstopper if you want to run anything graphic intensive.
Now development of a DX12 compliant GPU driver is very work intensive - but nothing Microsoft can not solve if they want to.