Intel's 13th Gen Core CPU Lineup Seemingly Leaks a Month Ahead of the Launch

[ir_cow]

11th and 12th Gen sure didn't last long eh?

About the nornal cycle time for Intel. A new CPU every year and a new socket every 2.

 

[sepheronx]

About the nornal cycle time for Intel. A new CPU every year and a new socket every 2.

You're right. After going back and looking up the previous chips for the last decade+, it's roughly every year a new chip.

 

[MxPhenom 216]

But weren't they saying that their 10nm is better than others 7nm???

Intels 10nm is on par with TSMC 7nm in regard to density and area. Performance and power likely not.

 

[Vayra86]

Rumors are also saying that those C0 revision i5 are actually a refresh of alder lake

Fresh lakes? More like continued drought..

Intels 10nm is on par with TSMC 7nm in regard to density and area. Performance and power likely not.

Density is variable and depends on architecture and what part of the die youre looking at.

It barely says anything really other than being a ballpark number for what is possible on that node.

 

[RandallFlagg]

You're right. After going back and looking up the previous chips for the last decade+, it's roughly every year a new chip.

Will be four in the past 30 months with Raptor Lake. Meteor Lake late Q2 / early Q3 would imply about every 9 months. Before 2020, it was every year :

Desktop Comet Lake-S CPUs April 30, 2020
Rocket Lake desktop family on March 16, 2021
12th Gen Intel Core CPUs on October 27, 2021
Raptor Lake - Sept/Oct 2022

 

[mplayerMuPDF]

Lowering base clock in favor of E cores, even iGPU still the same 32EU and I bet it's same ol' UHD Graphics.

That's more efficient though than pushing crazy clocks. Sure, it sucks for gaming but some people really need to understand that (high-performance) computing is not just about gaming.

Also, base clocks do not mean what a lot of people seem to think they mean. The base clock is the *highest* frequency that can be sustained by all cores indefinitely. At idle the cores run *below* the base clock. My Llano laptop idles at 800 MHz, while my Ryzen desktop has at least some cores running at 2.7 GHz or lower when idling despite having a 3.2 GHz base clock. "Turbo" clock is also a really misleading term. "Nitro" would be a much better analogy, actually.

 

[Vayra86]

That's more efficient though than pushing crazy clocks. Sure, it sucks for gaming but some people really need to understand that (high-performance) computing is not just about gaming.

Also, base clocks do not mean what a lot of people seem to think they mean. The base clock is the *highest* frequency that can be sustained by all cores indefinitely. At idle the cores run *below* the base clock. My Llano laptop idles at 800 MHz, while my Ryzen desktop has at least some cores running at 2.7 GHz or lower when idling despite having a 3.2 GHz base clock. "Turbo" clock is also a really misleading term. "Nitro" would be a much better analogy, actually.

Gaming wont even meet base clock anyway, not a single game is threaded that well. The vast majority still runs one~four bigger threads and the rest is low usage. Thats why you get the peak clocks out of these CPUs and still wont run into tdp limits. E cores can increase power budget headroom on those tasks so the P cores can max boost.

HOWEVER a lower base clock in parallel and threaded loads is definitely harming your sustained perf in favor of higher peak clocks. Thats how it simply works. We used to have that in mobile CPUs, now its everywhere and its not desktop/workstation oriented even in 2022. Its just Intel marketing interfering with more efficient clocking. Thats their sell. Its not in our favor, its our energy bill that does suffer here.

 

[thegnome]

Rumors are also saying that those C0 revision i5 are actually a refresh of alder lake

Yeah, I heard anything not B-0 is just the old i7/i9 Alder lake chips meant for i5 use.

 

[MxPhenom 216]

Fresh lakes? More like continued drought..

Density is variable and depends on architecture and what part of the die youre looking at.

It barely says anything really other than being a ballpark number for what is possible on that node.

Sort of. Theres still rules that have to be followed for any design/architecture. And those rules contribute to the nodes density possibilities. I do place and route at Seagate so nodes is a constant topic amongst what we do.

 

[RandallFlagg]

Yeah, I heard anything not B-0 is just the old i7/i9 Alder lake chips meant for i5 use.

That's purely speculation though.

For example, the 13600 is a 6+8 part and was leaked as a C0 stepping.

Which Alder Lake part is 6+8? None of the desktop parts are.

Or do we assume it is a cut down 8+8? Or perhaps a reworked Alder Lake mobile part (12700H was 6+8) to work in a different socket and chipset bus?

I find a lot of that to be highly unlikely.

They are assuming that if the Raptor Lake part stepping is at or above Alder Lake stepping it is Alder Lake, which may not be true at all.

 

[Jimmy_]

Lowering base clock in favor of E cores, even iGPU still the same 32EU and I bet it's same ol' UHD Graphics.

Yeah for Desktop they are still in UHD - their laptop SKU has Xe core it seems!

 

[AlwaysHope]

Node names have long been decoupled from actual measurement methods. That is why Intel went from 10nm to calling it Intel 7. TSMC Samsung and others long ago stopped following any kind of standard, so Intel has merely followed suit.

Anyone not aware of that at this point, is just speaking out of ignorance.

If you look at the MT/mm2 (million transistors per square mm) and compare TSMC N7 with Intel 7, you get an aspect that makes them roughly equivalent.


"But TSMC’s vice president of corporate research, Dr. Philip Wong, was keen to point out that after introducing his company’s latest node, despite a history of the node naming scheme actually having some relevance to the silicon features etched into the wafer, the node names are now effectively meaningless."


"Today, these numbers are just numbers. They’re like models in a car – it’s like BMW 5-series or Mazda 6. It doesn’t matter what the number is, it’s just a destination of the next technology, the name for it.
So, let’s not confuse ourselves with the name of the node with what the technology actually offers. "
-Philip Wong, TSMC

TSMC’s 7nm, 5nm, and 3nm “are just numbers… it doesn’t matter what the number is”

"These numbers are just numbers. They're like models in a car - it’s like BMW 5-series or Mazda 6."

www.pcgamesn.com

I get your gist, but doesn't matter what these companies call their technologies, 10nm is 10nm, calling it Intel 7 is just marketing hype.
The point I was highlighting in my 1st post of this thread is how they can stack more cache with 10nm in to raptor lake with the same die space as alder lake. But apparently according to thesmokingman, I'm a "fool" cause' I ain't no expert in node technology with cpu manufacturing.

 

[HenrySomeone]

That 20 thread i5 13500 is going to be the undisputed mainstream champion!

But weren't they saying that their 10nm is better than others 7nm???

It is. The transistor density is higher than TSMC's "7"nm.

Intels 10nm is on par with TSMC 7nm in regard to density and area. Performance and power likely not.

Both of those are superior as well, core-for-core, of course, but especially performance.

 

[RandallFlagg]

I get your gist, but doesn't matter what these companies call their technologies, 10nm is 10nm, calling it Intel 7 is just marketing hype.
The point I was highlighting in my 1st post of this thread is how they can stack more cache with 10nm in to raptor lake with the same die space as alder lake. But apparently according to thesmokingman, I'm a "fool" cause' I ain't no expert in node technology with cpu manufacturing.

That's true to a degree, but the thing is 10nm is based on using a logic gate and a flip flop for measurement (Intel's old way).

What TSMC Samsung and others did, was when they had 20nm and added FinFet they called it 16nm. There was no node shrink. Intel on the other hand, went from 22nm down an entire node and added FinFet. 14nm was accurate, but everyone else was on a 14/16nm node that was really 20nm + FinFet. They did not actually catch Intel's 14nm node until their 12/10nm nodes came out.

This is why Intel's 14nm node was about 30%-50% more dense than anyone elses 14/16nm node.

So using TSMC/Samsung standards, Intel 14nm = their 12/10nm. Imagine that.

That's when 'nanometers' became meaningless. It isn't something that Intel did, but all you have to do is look around and see how every analyst and 90% of 'enthusiasts' are completely taken in by the idea that N7 means 7nm. It doesn't, and it never has. TSMC N7 is about 10nm if you use the old logic gate + flip flop method of measurement.

Intel 4 is also more dense than TSMC N4 HP (high power, not for phones) node. It lands between TSMC N4 and N3.

 

[MxPhenom 216]

I get your gist, but doesn't matter what these companies call their technologies, 10nm is 10nm, calling it Intel 7 is just marketing hype.
The point I was highlighting in my 1st post of this thread is how they can stack more cache with 10nm in to raptor lake with the same die space as alder lake. But apparently according to thesmokingman, I'm a "fool" cause' I ain't no expert in node technology with cpu manufacturing.

Foundaries have been labelling their nodes in such a way that does not follow what it has always meant before. When you say 16nm, 28nm, 7nm etc that is suppose to tell you the length of the channel of the FETs of that process. Ever since about 28nm that has not been the case and its more marketting than anything. This is not unique to Intel either. Every foundary does it now. The length of the FETs usually vary by a few nanometer (high speed high leakage, low speed low leakage FETs/cells), and are larger than the node is rated at. And this metric is something layout engineers cannot alter. They can however adjust the width (transistor sizing 1x, 2x 4x, etc.)which effects drive strength of the FET with the caveat of increased power and area

 

[LemmingOverlord]

These aren't leaks. These are speculation. 2GHz is *not* the core clock on the 13900.

 

[Vader]

These aren't leaks. These are speculation. 2GHz is *not* the core clock on the 13900.

Why not? These are base clocks. 8 cores (+ little cores) running at 2.0 ghz in 65w sustained seems to be inline with previous cpus

 

[LemmingOverlord]

Why not? These are base clocks. 8 cores (+ little cores) running at 2.0 ghz in 65w sustained seems to be inline with previous cpus

A few things, related to the the table data being incomplete:

• there are two base frequencies... one for the P-cores and another for the E-cores. The table doesn't specify. Unless the 2GHz is the E-core base frequency, it's a downgrade from AL (2.4GHz base). This could compromise OS responsiveness.
• it's missing the max boost clock. Why would you opt to leave out max boost? That will probably be the single most important spec for many users. I don't see a scenario where a leaker would only have access to base clocks, TDP, cache values but no max boost clocks?
• there is no total power rating for the CPUs, something Intel would've advised on, because early slideware targets system integrators...

That's why I think this is just a mix of speculation and facts taken from other posts, rather than an actual leak.

 

[phanbuey]

A few things, related to the the table data being incomplete:

• there are two base frequencies... one for the P-cores and another for the E-cores. The table doesn't specify. Unless the 2GHz is the E-core base frequency, it's a downgrade from AL (2.4GHz base). This could compromise OS responsiveness.
• it's missing the max boost clock. Why would you opt to leave out max boost? That will probably be the single most important spec for many users. I don't see a scenario where a leaker would only have access to base clocks, TDP, cache values but no max boost clocks?
• there is no total power rating for the CPUs, something Intel would've advised on, because early slideware targets system integrators...

That's why I think this is just a mix of speculation and facts taken from other posts, rather than an actual leak.

Also the cache is much lower than already confirmed leaks.... this is an old speculative chart.

 

[Rakhmaninov3]

Who even needs more than 64k of memory

 

[tfdsaf]

Base frequency of 2GHz that's refreshing

Its a well known TRICK Intel does, they measure their power draw at base frequency, so they can report lower TDP, when in fact realistically the CPU is likely going to drain 150% more than the advertised TDP. At 2GHz they are likely to have 65W, when in reality that CPU is likely to pull up to 150W.

 

[Solid State Brain]

The Intel base frequency is the minimum guaranteed frequency below which the CPU will not downclock in order to maintain its TDP. Typically, even in a worse-case workload, frequencies will be higher than base frequency.

It requires the motherboard to have been set with proper limits of course, which is almost never the case for "gaming" ones.