Non-K 13th Gen Core i5 (such as i5-13400) Based on Older "Alder Lake" Architecture, Hints Intel Slide

[SL2]

Except lower end parts apparently will be Alder Lake refreshed, as new story here suggests.

He said Raptor tho.

 

[TheoneandonlyMrK]

I hope the guys and girls calling AMD's naming scheme see the irony today.
They're both shit at naming schemes.
And they'll both try and pull the wool a bit to shift stock.
Not good but to be expected.

But to me this is a direct indication of they're yields, clearly for every 12900K they got they created too many 12400's and 12600's.

Again not unexpected given the monolithic nature but these things are not in they're daily PR spiel.

 

[SL2]

But to me this is a direct indication of they're yields, clearly for every 12900K they got they created too many 12400's and 12600's.

That, or a bit related, they have a mountain of unsold chips, just like in the GPU world. Although it could have happened even with good yields given the current demand.

You are wrong here. You are comparing CPUs with different number of cores.

Actually I'm not wrong. The 13900K is the successor to the 12900K. I don't know the price, but I'd be surprised if they don't end end up having the same launch price, give or take.

What I am saying is that, if Intel was coming out with CPUs that where keeping the number of cores, but just switching from Alder to Raptor Lake architecture, the performance difference would have been small, considering the non K models have a TDP ceiling and Raptor's performance advantage comes also from higher frequencies, not juster better IPC.

You are wrong here, twice, as both the 12900K and the 13900K were running at 5.2 GHz in my example, and the power consumption was about the same (yeah I did not share that one before).

 

[bug]

I must be stupid, but I have to ask: the slide says increased L2 and L3 cache, the article says "based on Alder Lake". Where the link from the former to the latter?

 

[PCL]

I thought a 13400F was a lock to replace my venerable 6600k, but looks like the race between it and a 7600X will be closer than I thought.

 

[thestryker6]

I thought a 13400F was a lock to replace my venerable 6600k, but looks like the race between it and a 7600X will be closer than I thought.

Did you mean 13600KF? because otherwise you're comparing a product in the $175-200 range (basing this on ADL + a guess on RPL price increases) to a $300 part...

I must be stupid, but I have to ask: the slide says increased L2 and L3 cache, the article says "based on Alder Lake". Where the link from the former to the latter?

ADL e-cores have 2MB/cluster and the comparative ADL parts didn't come with e-cores (ex: 12400 was 6p/0e but 13400 is 6p/4e)

 

[PCL]

Did you mean 13600KF? because otherwise you're comparing a product in the $175-200 range (basing this on ADL + a guess on RPL price increases) to a $300 part...

No, I did not mean that. Price is one factor in my comparison, but far from the only one. My point is that Intel will allegedly be bringing less to the fight than I thought besides "cheaper".

 

[bug]

ADL e-cores have 2MB/cluster and the comparative ADL parts didn't come with e-cores (ex: 12400 was 6p/0e but 13400 is 6p/4e)

I still don't see how infer from that that 13400 will be using ADL cores.

 

[thestryker6]

I still don't see how infer from that that 13400 will be using ADL cores.

RPL p/e cores both have more cache so the numbers would be higher and not match ADL.

 

[Wirko]

"Competition just has too much momentum, and we haven't executed well enough." - Pat Gelsinger

Yeah, no shit.


I disagree, a 28 % uplift in R23 is nothing to sneeze at, and nothing like several pointless refreshes in the past.

In case anyone forgot what a refresh looks like:

Spoiler

Yeah, that's an astonishing 7 % improvement at 4 GHz..
View attachment 261154

Refresh was 4790/4690/4590 vs. 4770/4670/4570 (K and non-K), not Haswell vs. Ivy Bridge.

 

[ModEl4]

If the slide is authentic, indeed it suggests that i5 non-K it will be based on Alder Lake but I seems strange.
Raptor Lake will be monolithic, and in this P+E core era it would make sense 2 dies like what we had with Alder Lake (8P+16E & 6P+8E instead of 8P+8E & 6P+0E that we had on Alder Lake case)
So essentially this slide suggests only one die (8P+16E) that will support 13600K also (according to this 13600 will be Alder, so for Intel financially it makes more sense to just sell 8P+8E Alder Lake dies to i5 non-K prices instead of make a lower end Raptor Lake die?

 

[SL2]

Refresh was 4790/4690/4590 vs. 4770/4670/4570 (K and non-K), not Haswell vs. Ivy Bridge.

You're right, but I picket that one out of laziness. Your example would show even less improvement than 7 %, which would have been expected from a refresh.

 

[R0H1T]

It wasn't as much a refresh as a stopgap/contingency for the failure that was Broadwell. Intel's actually been slipping since 22nm, it was at least one (two?) quarter behind their original schedule & then the first 14nm chips were practically MIA. They launched this under the garb of better TIM & of course higher clocks & that tradition has continued for 8 "generations" now!

 

[Wirko]

If the slide is authentic, indeed it suggests that i5 non-K it will be based on Alder Lake but I seems strange.
Raptor Lake will be monolithic, and in this P+E core era it would make sense 2 dies like what we had with Alder Lake (8P+16E & 6P+8E instead of 8P+8E & 6P+0E that we had on Alder Lake case)
So essentially this slide suggests only one die (8P+16E) that will support 13600K also (according to this 13600 will be Alder, so for Intel financially it makes more sense to just sell 8P+8E Alder Lake dies to i5 non-K prices instead of make a lower end Raptor Lake die?

It surprises me too, I'd expect that developing a smaller variant of silicon is a fairly low-cost operation with high enough financial gain. For example, a small 4+0 die for the i3, which I'm sure has massive sales.

I understand that a new set of photomasks has to be made, costing many millions of $. On the other hand, masks have a limited lifetime, so they have to be re-made (several times?) during the life of a chip anyway.

 

[docnorth]

I must be stupid, but I have to ask: the slide says increased L2 and L3 cache, the article says "based on Alder Lake". Where the link from the former to the latter?

It’s based on AL, but with increased cache. @kane nas posted an enlightening table (if it’s correct).

 

[ModEl4]

It surprises me too, I'd expect that developing a smaller variant of silicon is a fairly low-cost operation with high enough financial gain. For example, a small 4+0 die for the i3, which I'm sure has massive sales.

I understand that a new set of photomasks has to be made, costing many millions of $. On the other hand, masks have a limited lifetime, so they have to be re-made (several times?) during the life of a chip anyway.

And a hypothetical scenario that Intel had excess stock of 8P+16E Alder Lake die doesn't make sense at all imo, if slide's true it was 100% planned from the start.

--------------------------------------------

Total i5 L2 sum has increased even if this is Alder Lake, but since L2 is private (1P core/1 group of E cores) slide's wording is the correct way to express it (essentially private L2 didn't increase for 13600 and below)
Again If the slide is true, It's really disappointing and i guess i just don't want to believe it.

 

[thestryker6]

And a hypothetical scenario that Intel had excess stock of 8P+16E Alder Lake die doesn't make sense at all imo, if slide's true it was 100% planned from the start.

Very much this any plans for RPL were set in stone a while ago Intel wouldn't have had their engineers design a core then just not use it. They likely knew that AMD wouldn't be bringing anything Zen 4 to the sub $300 market, chose to focus on the high end, and save money where there wasn't going to be new competition.

There are many reasons for Intel to do this (these are just off the top of my head):

• They don't have to run new production lines while keeping ADL open to meet OEM contracts
• They may be trying to maximize Intel 7 capacity for SPR due to having to run a new revision
• They're losing market share and while it's not a "oh no we're going to die" situation this should mean more profit on those chips than if they'd been a new design

This obviously isn't a pro-consumer move, but neither is AMD neglecting that entire market with new products. I do wish it wasn't happening because anyone in that market would be better served by the new cache capacity and more efficient clockspeed scaling.

 

[Gica]

Raptor is based on a new architecture, it's not just a refresh. They are those architectures that have been lying in the drawers while Intel cut the Gordian knot of 10nm (announced, I think, for 2017, in mass production with a 4-year delay, maybe more).
From my own garden, jump multicore from generation to generation.
As an idea:
10500: max 4.2GHz for all cores
12500: max 4.1GHz for all cores
Even compared to the Rocket 11600 (max 4.4 GHz for all cores), the performance increase of the Alder 12500 processor is colossal.

 

[RandallFlagg]

This is turning out to be yet another false rumor based on weak speculation and bad assumptions. 13400 GB is showing up with Raptor Lake cache config.

These tech sites and tech tubers have had a horrifically bad record on this kind of stuff this year, from rumors of Nvidia 4000 being delayed to 2023 (obv a fail) to ARC being cancelled (which was just debunked a couple of days ago).

People need to apply a little bit more critical thinking skills to this stuff. Too many of these people just want to make money from the clicks. They are wasting a lot of people's time, and at some level probably taking on a lot of liability if they propagate false rumors that defame/damage reputations.

This is a raptor lake cache configuration :

 

[Gica]

And "10 Cores, 16 Threads"

 

[thestryker6]

This is turning out to be yet another false rumor based on weak speculation and bad assumptions. 13400 GB is showing up with Raptor Lake cache config.

These tech sites and tech tubers have had a horrifically bad record on this kind of stuff this year, from rumors of Nvidia 4000 being delayed to 2023 (obv a fail) to ARC being cancelled (which was just debunked a couple of days ago).

People need to apply a little bit more critical thinking skills to this stuff. Too many of these people just want to make money from the clicks. They are wasting a lot of people's time, and at some level probably taking on a lot of liability if they propagate false rumors that defame/damage reputations.

This is a raptor lake cache configuration :

View attachment 262483

I literally just searched for gb scores on the 12600K/KF and it kinda looks familiar wouldn't you say?

 

[Gica]

I literally just searched for gb scores on the 12600K/KF and it kinda looks familiar wouldn't you say?

There are many differences in specifications (system, maximum frequency, etc.).
It is possible that it is fake, but the leak we are discussing is not confirmed either.