Discussion Intel Nova Lake in H2-2026: Discussion Threads

[adroc_thurston]

Note some of narratives that were going on, at first some would say mainstream RDNA 5 iGPU was to come during/compared with PTL's 12Xe3

that was never, ever a thing.
PTL intercept (back when it was still 2025) was the OG KRK with 16CU RDNA4.5.

now it looks like it would come even later.

It was always like that.
Not a question of GPU perf at all, but who gets to do client GEMM accel.

 

[regen1]

that was never, ever a thing.

It was always like that.
Not a question of GPU perf at all, but who gets to do client GEMM accel.

I was thinking Computex 2027 or later(for RDNA 5 iGPUs) which seems to be the way as it looks CES 2028+
Well, all kinds of weird comparisons were there and kinda unreasonable.

Weren't you one of those who was saying NVL-H's 12Xe3p would be "roadkilled" ?

yeah they are.

don' matter, it's gonna get roadkilled by the gfx13 blob either way.

Seems no competition for a year or so now.
We'll see what comes next. I am hoping both will do real well in whatever way(whether someone wins by big amount or they are equally competitive or whatever, interested to see progress).

 

[adroc_thurston]

Weren't you one of those who was saying NVL-H's 12Xe3p would be "roadkilled" ?

Yeah it's just gonna get mogged.
It's not a part to compete with Intel, it's made to take NV slots away.

as it looks CES 2028+

Nope.
CES28 will be a big gfx refresh for both vendors doe.

We'll see what comes next I am hoping both will do real well in whatever way(whether someone wins by big amount or they are equally competitive or whatever, interested to see progress).

Meh, Intel has no CPUs until Unified Core.

 

[regen1]

Yeah it's just gonna get mogged.
It's not a part to compete with Intel, it's made to take NV slots away.

Yeah, you carry on with your unreasonable 1 year old vs new comparisons.

Nope.
CES28 will be a big gfx refresh for both vendors doe.

CES 2028+ means it includes CES 2028. CES 2028 is expected.

Meh, Intel has no CPUs until Unified Core.

Was talking about iGPU but yeah, again change narrative to some distant future.
Ignore PTL and others since those don't suit your narrative.

 

[adroc_thurston]

Yeah, you carry on with your unreasonable 1 year old vs new comparisons.

Intel doesn't have anything better. RZL is a CPU tile swap.

CES 2028+ means it includes CES 2028. CES 2028 is expected.

uh, nope.

Was talking about iGPU but yeah, again change narrative to some distant future.
Ignore PTL and others since those don't suit your narrative.

PTL sucks balls.
2024 CPU perf on N3-class node is big yuck. iGPs fine at least but it works and performs like a 50% bigger LNL would.

 

[regen1]

Intel doesn't have anything better. RZL is a CPU tile swap.

RZL is decent or not isn't the point. How's that relevant to notebook iGPU discussion that was going on ?

PTL sucks balls.
2024 CPU perf on N3-class node is big yuck. iGPs fine at least but it works and performs like a 50% bigger LNL would.

What matters it is that it's a superior product in x86 world, who cares if the competition didn't show up ?

uh, nope.

Nah that's what it means, notation doesn't become different for you. Anyway semantics.

 

[adroc_thurston]

How's that relevant to notebook iGPU discussion that was going on ?

Intel doesn't update their iGPs much after NVL.

What matters it is that it's a superior product in x86 world, who cares if the competition didn't show up ?

The only superior bit is the BL. But you could've gotten that a year ago with LNL so what are we even talking about?

 

[Fjodor2001]

Meh, Intel has no CPUs until Unified Core.

Just wonder what you expectations are for UC. It’s gonna be an evolvement of the E core, and it’s going to be available in b.L style configs just like P and E cores (or compact or whatever they’ll call it).

 

[adroc_thurston]

Just wonder what you expectations are for UC.

moar IPC at ????? fmax.
The latter is the biggest question.

 

[regen1]

Intel doesn't update their iGPs much after NVL.

That has to be seen and again nothing to do with this year or next. Intel's recent notebook iGPU IP cadence has been pretty decent for now. If it uses Xe3p twice that's fine, enough knobs(memory, clocks, units, etc.) to eek out perf, after that they need something more(Xe4 or some rumored "Xe3p V2"), that remains to be seen. Very far out to speculate. It's not like AMD has done faster iterations always.

The only superior bit is the BL. But you could've gotten that a year ago with LNL so what are we even talking about?

LNL didn't provide those levels of performance, it was good otherwise.

 

[adroc_thurston]

It's not like AMD has done faster iterations always.

Theirs are not tied to that, but to Radeon roadmap.

LNL didn't provide those levels of performance, it was good otherwise.

Same 1t lmao.
nT is like STX1 levels but that's 2024 stuff.

 

[regen1]

Same 1t lmao.
nT is like STX1 levels but that's 2024 stuff.

Does Strix/Gorgon provide those level of battery? or do they offer any tangible perf advantage/ 1T perf ? No.

 

[adroc_thurston]

Does Strix/Gorgon provide those level of battery?

no.

or do they offer any tangible perf advantage/ 1T perf ? No.

They've been on the market for 1+ years. it kind of counts.
For general usecases LNL did this 1t and this BL in late'24 so PTL is almost redundant.

 

[regen1]

They've been on the market for 1+ years. it kind of counts.

ARL-H was too have been out about a year with much more design choices compared to Strix Point, which had launched somewhat earlier to ARL-H.

 

[adroc_thurston]

ARL-H was too have been out about a year with much more design choices compared to Strix Point, which had launched somewhat earlier to ARL-H.

ARL-H sucks donkeyballs so we can stop pretending any of it was worth it.
LNL was a nice part with good designs though (godbless X1c).

 

[regen1]

ARL-H sucks donkeyballs so we can stop pretending any of it was worth it.

Really ? It had lot more design wins and more interesting ones(and more volume) too while launching later than Strix Point. Seemed alright enough, 1T perf was there for x86.

 

[adroc_thurston]

It had lot more design wins and more interesting ones(and more volume) too while launching later than Strix Point

Doesn't matter, the part itself was ass.

Seemed alright enough, 1T perf was there for x86.

In LNL yes. kind of.
Apple was already blasting off with M4.

 

[regen1]

Doesn't matter, the part itself was ass.

In LNL yes. kind of.
Apple was already blasting off with M4.

Well, that's your opinion. ARL-H did alright for parameters that mattered.

 

[adroc_thurston]

ARL-H did alright for parameters that mattered.

slower than stx1 with N3. big yuck.

 

[regen1]

slower than stx1 with N3. big yuck.

Top vs Top, Similar or higher 1T perf, bit lower nT perf.
Users generally don't care what node is used.
More design choices and more volume that's what matters.
HX370 and AI 9 365 were pretty good but not as much volume or choice, simple.

 

[adroc_thurston]

Users generally don't care what node is used.

that's cope man, everyone else on N3 did a nice blast off.
Just be glad that PTL has real nice BL and enjoy the show.

 

[regen1]

that's cope man, everyone else on N3 did a nice blast off.

It delivered by outselling. Behind in eeking out from nodes? Yes, doesn't matter to end user that much when x86 competition on whatever node doesn't give more tangible perf or design choices.

 

[adroc_thurston]

It delivered by outselling

From that POV eternal 666 years of Raptor Lake deliver anything anyone would ever need.

Behind in eeking out from nodes? Yes, doesn't matter to end user that much when x86 competition on whatever node doesn't give more tangible perf or design choices.

this is cope.
They suck and you know it.

 

[regen1]

From that POV eternal 666 years of Raptor Lake deliver anything anyone would ever need.

Don't skew the comparison, again. It was Strix Point vs ARL-H since you have objections with PTL being similar ST while ignoring other stuff/aspects of present.
Raptor didn't have ARL-H's or Strix Points' battery life, iGPU, etc.

this is cope.
They suck and you know it.

Said like one truly hates something.

 

[adroc_thurston]

Don't skew the comparison, again. It was Strix Point vs ARL-H since you have objections with PTL being similar ST while ignoring other stuff/aspects of present.

other aspects do not really compensate for RPL-H 1t perf in 2026.
This is the second Intel N3-class mobile platform with no 1t gains relative to 2023 stuff.

Raptor didn't have ARL-H's or Strix Points' battery life, iGPU, etc.

Raptor was cheap and good, and ARL-H was expensive and Bad. Pretty simple.

Said like one truly hates something.

I can't hate Intel at all, but they're not very competent people.
Especially not DCAI folks (lobotomite farm).