• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Discussion ARM Cortex/Neoverse IP + SoCs (no custom cores) Discussion

Page 37 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
FlameTail said:
www.xda-developers.com

Someone fixed the Windows Subsystem for Android because Microsoft refuses to — here's how you can do it

The Surface Pro 11 is finally good as a tablet
www.xda-developers.com www.xda-developers.com


Will the same issues pop up when the CPUs get upgraded to ARMv9? I guess not- since ARMv9 and ARMv8 aren't too dissimilar.
Click to expand...

No.

I suspect the original article should have said "AArch32" and "AArch64" instead of "ARMv7" and "ARMv8." The latter is kind of nonsensical. AArch32 is still specified under the umbrella of ARMv8 - it's just optional.
 
DZero said:
Wait... as for Android isn't Intel the first to introduce Hyperthreading on phones?
And yeah is x86
Click to expand...
Huh, TIL Atom originally had hyperthreading with Saltwell.

It went out the window pretty quickly with Silvermont µArch tho.
 
Last edited:
Doug S said:
More examples of how Intel really had no clue what mobile needed. HT in mobile is stupid now, it was objectively insane back then.
Click to expand...

IIRC, that wasn't a mobile product but an HPC product with AVX-512.

Edit: Guess there was a desktop/notebook product too apparently as mentioned.
 
Doug S said:
More examples of how Intel really had no clue what mobile needed. HT in mobile is stupid now, it was objectively insane back then.
Click to expand...

Intel original Atom was simple in-order core. With HT it was actually pretty competitive against small out-of-order cores back then. And most importantly, those were times when cpus used still were mainly 1-core. With one core HT actually improves system responsiveness - a lot - compared to same core without HT as there is two hardware threads without need to switch out primary task for backround tasks.
 
FINALLY some more info on AmpereOne. Aurora will be a 512 Core CPU:

www.servethehome.com

Ampere AmpereOne Aurora 512 Core AI CPU Announced

Ampere AmpereOne Aurora will have 512 cores, integrated AI acceleration, and even HBM memory integrated for a monster future chip
www.servethehome.com www.servethehome.com

But still ways away, not sure they can really survive given that Amazon and Google use their own design (and Microsoft is 100% following suit) and the heated up competiton from AMD and Intel's small cores, the market seems really competitive.

jGK3hV5.jpeg
 
Gideon said:
FINALLY some more info on AmpereOne. Aurora will be a 512 Core CPU:

www.servethehome.com

Ampere AmpereOne Aurora 512 Core AI CPU Announced

Ampere AmpereOne Aurora will have 512 cores, integrated AI acceleration, and even HBM memory integrated for a monster future chip
www.servethehome.com www.servethehome.com

But still ways away, not sure they can really survive given that Amazon and Google use their own design (and Microsoft is 100% following suit) and the heated up competiton from AMD and Intel's small cores, the market seems really competitive.

jGK3hV5.jpeg
Click to expand...

Is the initial AmpereOne available yet? They announced it in 2022 and said it would be available later that year but it went MIA. They re-announced it with official specs in May 2023 and said again it would be available by the end of the year in Oracle Cloud, but no one seems to know where it is, lol. At this point, they're running into Turin (D) and have lost their core count advantage. How long will it be before MX actually makes it to market? Don't get me wrong, I am actually pulling for them, but with their major delays thus far, competition increasing, and big cloud vendors rolling their own ARM chips, I just don't know if we will actually see them get to 512 core products. I hope I'm wrong, though.
 
Gideon said:
FINALLY some more info on AmpereOne. Aurora will be a 512 Core CPU:

www.servethehome.com

Ampere AmpereOne Aurora 512 Core AI CPU Announced

Ampere AmpereOne Aurora will have 512 cores, integrated AI acceleration, and even HBM memory integrated for a monster future chip
www.servethehome.com www.servethehome.com

But still ways away, not sure they can really survive given that Amazon and Google use their own design (and Microsoft is 100% following suit) and the heated up competiton from AMD and Intel's small cores, the market seems really competitive.

jGK3hV5.jpeg
Click to expand...
That may be why they shelved their IPO plans. They’re the primary ARM cpu supplier for Oracle Cloud and smaller clouds like Hetzner.

I wouldn’t be surprised if Oracle eventually buys Ampere altogether.

I still expect them to IPO though. I’ll buy a few shares.
 
Apple A13 : 2.65 GHz
Apple A17 : 3.78 GHz

Snapdragon 865 (Cortex A77) : 2.84 GHz
Snapdragon 8 Gen 3 (Cortex X4) : 3.4 GHz

There is a 4-year gap between both. Within that same timespan, Qualcomm/ARM increased their clock speed by 0.46 GHz, while Apple increased it by 1.13 GHz. More than 2x the amount!

The Snapdragon 8 Gen 4 with custom Oryon core is rumoured to clock at 4.37 GHz! The Apple A18 Pro will also certainly clock north of 4 GHz.

Some people are calling for ARM to do the same and increase the Fmax of their Cortex X cores. I disagree. I think ARM should stick to what they have been doing best for the past few years: Steady YoY IPC gains.

Increasing clock speed drastically like that comes with a drastic power increase too. It's the Netburst ideology. Apple is embarked on a dark path. For the first time, they used HP libraries for the CPU in M4, and the ST power increased by 50% compared to M3 (according to Geekerwan). Apple can get away with it, only because they are starting from a very good baseline.
 
FlameTail said:
Some people are calling for ARM to do the same and increase the Fmax of their Cortex X cores. I disagree. I think ARM should stick to what they have been doing best for the past few years: Steady YoY IPC gains.
Click to expand...
Means to an end.
All that matters is end product PPA.
FlameTail said:
It's the Netburst ideology
Click to expand...
?
Netburst means skinny heavily pipelined core with trace cache.
None of the modern OoO machines are that.
 
FlameTail said:
The Snapdragon 8 Gen 4 with custom Oryon core is rumoured to clock at 4.37 GHz! The Apple A18 Pro will also certainly clock north of 4 GHz.
Click to expand...

I don't know if we can assume that. Maybe they don't use HP cells for the P cores in A18? Even if they do I'd say 4 GHz is the absolute max we're likely to see, it won't be north of that.
 
Scree4553_Chrome.jpg
Intel/AMD are not stagnating, but their improvements have slowed to a crawl.

If ARM can deliver 10% CAGR IPC gains every year, their future in the PC world is secured.
 
It's not only AMD, but also Intel.

Arrow Lake/Lunar Lake debuts their "new" core- Lion Cove. The next new core is supposedly for Nova Lake in late 2026/early 2027.

Cougar Cove in Panther Lake is supposedly just Lion Cove ported to 18A (a repeat of the situation in Raptor Lake -> Meteor Lake).

So if ARM can put out new cores with decent IPC gains every year, they will surpass x86.


Is 50% back on the table? 😛
moinmoin said:
@FlameTail why did you start posting screenshots of people's posts on this very forum instead just quoting them using the forum's functionality? That's honestly annoying.
Click to expand...
My apologies.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top