Discussion AMD's Soundwave ARM APU: The Beginning of Transformation !!!

[Nothingness]

But they did ship the Seattle cores in 2016.

Seattle was using a standard Arm Cortex-A57, not an AMD designed one.

 

[LightningDust]

Seattle was using a standard Arm Cortex-A57, not an AMD designed one.

And it was a toy, filled with an impressive number of errata, with no real customers and no particular use case.

"Ship" needs to be defined very broadly in turtile's post.

 

[igor_kavinski]

And it was a toy, filled with an impressive number of errata

The better to test your engineering workforce's capabilities, my dear

After all, the difference between a good and bad engineer is how they react to bad hardware.

Good one starts documenting the issues and suggests fixes.

Bad one can't even tell where the issue lies.

 

[LightningDust]

The better to test your engineering workforce's capabilities, my dear

After all, the difference between a good and bad engineer is how they react to bad hardware.

Good one starts documenting the issues and suggests fixes.

Bad one can't even tell where the issue lies.

"Errata are Good Actually" is certainly a perspective. I'll have to remember that one for future workplace use.

 

[Nothingness]

The better to test your engineering workforce's capabilities, my dear

You don't exercise your CPU design skills that way, which was the point of the discussion 😉

 

[LightningDust]

You don't exercise your CPU design skills that way, which was the point of the discussion 😉

Nah. You don't understand. Real men (and women, in my case) live life on the edge. "The design works in the sim and on the Veloce and the power and clock physicals look fine in ansys, ship it as soon as we get silicon back" is a perfectly valid flow!

(Jesus, just typing that is giving me hives.)

 

[Nothingness]

Nah. You don't understand. Real men (and women, in my case) live life on the edge. "The design works in the sim and on the Veloce and the power and clock physicals look fine in ansys, ship it as soon as we get silicon back" is a perfectly valid flow!

(Jesus, just typing that is giving me hives.)

s/ansys/synopsys/

 

[LightningDust]

s/ansys/synopsis/

Does Synopsys even make equivalents to stuff like Ansys Redhawk or Exalto?

Eh, hell with it, who cares? It's a problem for whoever gets the chip. Errata build character.

 

[Nothingness]

Does Synopsys even make equivalents to stuff like Ansys Redhawk or Exalto?

Eh, hell with it, who cares? It's a problem for whoever gets the chip. Errata build character.

Ansys is being bought by Synopsys 😉

 

[LightningDust]

Ansys is being bought by Synopsys 😉

Oh yeah. Huh. That is going to be so weird. I had read that when it was announced but I think I failed to internalize it because I don't do much physical-design stuff.

They're running ARC into the ground, but their DRAM and PCIe IP seems decent enough. Wonder if they'll raise Ansys tool prices from "exorbitant" to "psychotic."

 

[Nothingness]

Oh yeah. Huh. That is going to be so weird. I had read that when it was announced but I think I failed to internalize it because I don't do much physical-design stuff.

I don't do any physical stuff, not even synthesis, and RTL simulation only once every other year. But I used to work in a company that was later acquired by Ansys and still have one close friend there.

They're running ARC into the ground

This reminds me Jez San had offered me a job to work on GNU compiler last century. I finally went to TI and wonder where I would be now if I had accepted...

Sorry for the very personal tone of the post

 

[LightningDust]

I don't do any physical stuff, not even synthesis, and RTL simulation only once every other year. But I used to work in a company that was later acquired by Ansys and still have one close friend there.

I hope your friend does well amid the inevitable chaos and disruption that comes with getting acquired.

This reminds me Jez San had offered me a job to work on GNU compiler last century. I finally went to TI and wonder where I would be now if I had accepted...

I like a lot of TI stuff (C6k mostly, though I'm looking at the new F29 with some interest) but I've heard some pretty negative things from nearly everyone I know who has worked there.

 

[DAPUNISHER]

GN discussion with Wendell from Level 1 about ARM and if x86 is screwed. Wendell thinks AMD can pivot fairly quickly but that Intel is boned. He says pricing and that it's "good enough" are strong points. Talks about how none of those Amazon Graviton or other ARM servers are ever coming back to x86; that biz is gone for good, just as with Apple.

 

[marees]

GN discussion with Wendell from Level 1 about ARM and if x86 is screwed. Wendell thinks AMD can pivot fairly quickly but that Intel is boned. He says pricing and that it's "good enough" are strong points. Talks about how none of those Amazon Graviton or other ARM servers are ever coming back to x86; that biz is gone for good, just as with Apple.

So hyperscalers & Apple will continue to have their own CPU teams ??

Will ARM be relevant or the vendors go for custom CPU architectures?

 

[adroc_thurston]

He says pricing and that it's "good enough" are strong points

This is a very dumb point since Neoverse communist is reliant on ARM eternally subsidizing design costs wholly.
Which is not a real business model, more like a trap before ARM itself becomes a merchant vendor.

Talks about how none of those Amazon Graviton or other ARM servers are ever coming back to x86

2-3 Neoverse licensing hikes and they'll be crawling back and begging for mercy.

So hyperscalers & Apple will continue to have their own CPU teams ??

Apple, QC, GOOG. That's it.

Will ARM be relevant or the vendors go for custom CPU architectures?

well yeah they define the ISA.

 

[Gideon]

2-3 Neoverse licensing hikes and they'll be crawling back and begging for mercy.

If this were to happen they would probably transition to RISC-V, though it would take time

 

[Gideon]

GN discussion with Wendell from Level 1 about ARM and if x86 is screwed. Wendell thinks AMD can pivot fairly quickly but that Intel is boned. He says pricing and that it's "good enough" are strong points. Talks about how none of those Amazon Graviton or other ARM servers are ever coming back to x86; that biz is gone for good, just as with Apple.

While this is slighlty offtopic, they also mentioned Nvidia Windows ARM drivers (as a thing one should not mention) This makes it pretty obvious Nvidia is doing Windows ARM SoCs. Might also explain why their drivers suddenly become such dumpster fire (porting).

 

[DrMrLordX]

Apple, QC, GOOG. That's it.

Hmm, thought MS had a team too? Or did they ditch theirs to throw in with Qualcomm?

 

[poke01]

Apple, QC, GOOG. That's it.

What about Nvidia? I thought Vera was custom

 

[MS_AT]

GOOG

Isn't google in the same boat as Microsoft? With the difference being Google bought V2 and M$ N2?

 

[adroc_thurston]

If this were to happen they would probably transition to RISC-V, though it would take time

RISC-V is a poopoo favela ISA that is pure horrors to work with for high perf.
Also, no, you can't just snap fingers and transition anything outside of embedded dirt pits across ISAs that easily.

Hmm, thought MS had a team too?

Nothing came outta that one.

What about Nvidia?

Irrelevant.

Isn't google in the same boat as Microsoft?

Nope, they have a custom core team making *something*.

 

[poke01]

RISC-V is a poopoo favela ISA that is pure horrors to work with for high perf.

Jim Keller looks like he will pull it off

 

[DZero]

As long X86 is strategic, it won't die.

 

[adroc_thurston]

Jim Keller looks like he will pull it off

"looks like" doesn't mean anything is shipping in volume designs.

 

[soresu]

Nope, they have a custom core team making *something*.

Oh dear - at this rate Samsung and Mediatek will be the only big volume players using ARM Ltd's Cortex X CPU IP in consumer hw 😅