Question Alder Lake - Official Thread

[Hulk]

With the release of Alder Lake less than a week away and the "Lakes" thread having turned into a nightmare to navigate I thought it might be a good time to start a discussion thread solely for Alder Lake.

 

[jpiniero]

Do you have a source for that? It seems reasonable that they only are running the 8+8 die right now. But, they theoretically could have used the 6+8 die for the 12600K and 12600KF. So, is there evidence that they are not doing so?

There is no 6+8 die. Only 8+8 or 6+0.

 

[dullard]

There is no 6+8 die. Only 8+8 or 6+0.

Right in the middle:

https://images.anandtech.com/doci/16881/69.jpg

 

[jpiniero]

Right in the middle:

https://images.anandtech.com/doci/16881/69.jpg

That's for mobile not desktop.

 

[dullard]

That's for mobile not desktop.

Yes, but what is to stop them from using it?

 

[jpiniero]

Yes, but what is to stop them from using it?

Doubt it's fully compatible with the socket. Even if it was they wouldn't do that since there's plenty of mobile products to dump defective dies.

 

[Saylick]

I wouldn't have guessed that there's a smaller 6+0 die, which I interpret as being only 6 GC cores and zero Gracemont cores. What's the point of such a die if Intel is going all in with the hybrid approach? It could have been a 4+4 or 4+8 product to offer a smaller die for the lower end desktop SKUs, but they went with 6 big cores instead? Seems like a headscratcher.

 

[jpiniero]

I wouldn't have guessed that there's a smaller 6+0 die, which I interpret as being only 6 GC cores and zero Gracemont cores. What's the point of such a die if Intel is going all in with the hybrid approach? It could have been a 4+4 or 4+8 product to offer a smaller die for the lower end desktop SKUs, but they went with 6 big cores instead? Seems like a headscratcher.

Marketing. Didn't want lower end products to have more cores than higher end products yet didn't want to regress on the number of big cores on i5.

 

[naukkis]

I wouldn't have guessed that there's a smaller 6+0 die, which I interpret as being only 6 GC cores and zero Gracemont cores. What's the point of such a die if Intel is going all in with the hybrid approach? It could have been a 4+4 or 4+8 product to offer a smaller die for the lower end desktop SKUs, but they went with 6 big cores instead? Seems like a headscratcher.

For desktop 6 big cores is way better approach than less big cores plus small cores. Intel's choice is good, that 4+8 would lose to 5600x at games that scale beyond 4 threads. This time 6 big cores is sweet spot, most of users won't need more but many workloads prefer 6 cores over 4.

 

[blckgrffn]

Marketing. Didn't want lower end products to have more cores than higher end products yet didn't want to regress on the number of big cores on i5.

Appears to be lots of marketing going on here. Plus E cores only bring 1T, so to surpass/match the old 6C/12T model you'd need to go at least 4P+4E to get to core and thread counts that are "higher".

6P+4E would get you to 16T total which seems like a forward looking sweet spot (to be redundant).

Perhaps the I3's will be even more interesting? I don't know that I've seen any leaks on that front, but it's possible we could get 4P+2E or even 2P+4E - if 4C/8T was the old spec even a 6C/10T would appear to be a solid upgrade. You know, on paper.

 

[tomatosummit]

Marketing. Didn't want lower end products to have more cores than higher end products yet didn't want to regress on the number of big cores on i5.

I think it's silicon space and performance.
Largest mobile cpu would be a simimilar die size to the desktop cpu assuming 64eu are a similar size to two Pcores.
So they want a smaller die for the mainstream part and the volumes allow them to create a unique part for the performance required, the ultra mobile really wouldn't fit in where the 12400/500/600 cpus are going to be marketed, be it by offered cores or performance (games) given, although it's odd that they'd chose a small igpu cpu for the skus that will go into alll business and budget desktop computers for the next couple of years.

 

[blckgrffn]

I think it's silicon space and performance.
Largest mobile cpu would be a simimilar die size to the desktop cpu assuming 64eu are a similar size to two Pcores.
So they want a smaller die for the mainstream part and the volumes allow them to create a unique part for the performance required, the ultra mobile really wouldn't fit in where the 12400/500/600 cpus are going to be marketed, be it by offered cores or performance (games) given, although it's odd that they'd chose a small igpu cpu for the skus that will go into alll business and budget desktop computers for the next couple of years.

Is the i5 really a different die, though? If we already ascertained that, I missed it

Wait, I see that discussion right above. In that screen shot though... it looks like a 6+ some amount of little cores though doesn't it? What are the two colors of blue supposed to be for?

 

[tomatosummit]

Is the i5 really a different die, though? If we already ascertained that, I missed it

I didn't belive it initially but it appears to be real with the msi presentation. I presume intel leaves it out of marketing materials because it doesn't show off hybrid computing which is the big new thing.

I expect it to have other cheaper choices taken in manufacturing too. Probably has no solder and is tim on a thicker die with thinner ihs, need to eek out those margins for cpus that will cost ~$100.

 

[blckgrffn]

I didn't belive it initially but it appears to be real with the msi presentation. I presume intel leaves it out of marketing materials because it doesn't show off hybrid computing which is the big new thing.

I expect it to have other cheaper choices taken in manufacturing too. Probably has no solder and is tim on a thicker die with thinner ihs, need to eek out those margins for cpus that will cost ~$100.

I mean it makes some sense to have P only dies, but with the E cores being so small I would have thought they would keep them right down to the new pentium line which could be one P core and 4 E cores for some really whack use of a small die but "moar cores".

Like the octa core A55 phones, there is absolutely value in those small cores being marketed as upgrades.

Plus it seems like if Intel was serious about it the would have the heterogenous model be full stack.

 

[jpiniero]

Perhaps the I3's will be even more interesting? I don't know that I've seen any leaks on that front, but it's possible we could get 4P+2E or even 2P+4E - if 4C/8T was the old spec even a 6C/10T would appear to be a solid upgrade. You know, on paper.

i3 is 4+0, Pentium/Celeron is 2+0.

 

[blckgrffn]

i3 is 4+0, Pentium/Celeron is 2+0.

:/

Thanks for the info, sorry I didn't know.

I guess they gotta save something for the refresh. It seems like the number of PC's using the new P & E model is going to be a lot lower than I was expecting it to be.

 

[tomatosummit]

:/

Thanks for the info, sorry I didn't know.

I guess they gotta save something for the refresh. It seems like the number of PC's using the new P & E model is going to be a lot lower than I was expecting it to be.

laptops are all hybrids, bigger in the consumer market compared to desktops.
One of the leaked roadmaps did show 1P+4e for the lowest end, sub 10watts tier.
Laptops are where I hope alderlake shines, the 2+8 version anyway, noway 6 golden cove cores aren't going to need some juice.

 

[blckgrffn]

laptops are all hybrids, bigger in the consumer market compared to desktops.
One of the leaked roadmaps did show 1P+4e for the lowest end, sub 10watts tier.
Laptops are where I hope alderlake shines, the 2+8 version anyway, noway 6 golden cove cores aren't going to need some juice.

Ah yes, I had completely overlooked this in my haste to judge the lineup. I am sure that compared to the number of i3/Pentium/Celeron 12th gen desktops there might be a full order of magnitude more 12th gen laptops and hybrid devices. Lots of SFF OEM systems could use mobile parts as well.

Thanks for the reminder!

 

[Hougy]

Do you have a source for that? It seems reasonable that they only are running the 8+8 die right now and will be using the best 6-core chips for high-end mobile. But, they theoretically could have used the 6+8 die for the 12600K and 12600KF. So, is there evidence that they are not doing so?

Yes it's for lower clocked non OC models, the fastest 6+0 seems to be the i5-12600. However this model comes with reasonable turbo speeds of 4.8 Ghz ST and 4.4 Ghz MT. On wccftech there is a SKU overview.

I have no evidence for anything I said, I wanted to speculate together with this forum.

I don't think it's likely that 12600K uses the 6+8 die also. The 6+0 could have K models core i3s, but these won't be sold with very high frequencies, so they don't need good bins. If they are not used for mobile chips, like you suggested, it seems the golden samples would be wasted.

What will the worst 8+8 dies become? 12600Ks?

 

[jpiniero]

What will the worst 8+8 dies become? 12600Ks?

If it's busted that badly it could be just about anything in the stack, all the way down to a 2+0 Celeron.

 

[tomatosummit]

If it's busted that badly it could be just about anything in the stack, all the way down to a 2+0 Celeron.

There are probably limitations. Depending where the faults are located, only certain parts have redundancy.
igpu fails can become K series
Ecore fails can become i7
Pcore fails can become i5
Uncore/imc/io fails can become paperweights
I doubt there are a significant amount of further failure types to require chopping an i9 even further down, 6/4/2+0 on desktop is possible but a die that broken is probably on borrowed time.

 

[igor_kavinski]

i3 is 4+0, Pentium/Celeron is 2+0.

Your source please, if you don't mind?

 

[jpiniero]

I doubt there are a significant amount of further failure types to require chopping an i9 even further down, 6/4/2+0 on desktop is possible but a die that broken is probably on borrowed time.

IIRC Intel did this with Comet Lake - selling the big die as Pentiums if it was busted that far. If Intel's binning suggested that it wouldn't last the 3 years they wouldn't sell it.

Your source please, if you don't mind?

That's the rumor. Since Intel hasn't officially announced the locked parts, it's conceivable that they could change their mind and all i5 could end up being 6+4 and the i3 being 6+0. But the i3 is definitely going to be coming from the 6+0 die exclusively.

Having the locked i5 be able to use the 6+0 die would cut Intel's costs quite a bit.

 

[Exist50]

That's the rumor. Since Intel hasn't officially announced the locked parts, it's conceivable that they could change their mind and all i5 could end up being 6+4 and the i3 being 6+0. But the i3 is definitely going to be coming from the 6+0 die exclusively.

Mainstream i5 will definitely be 6c as well. That's where much of the volume is, and no way Intel will want to shift all of that to the 8+8 die. But I think igor was asking about i3 and below spec rumor.

Anyway, with Raptor Lake going 8+16 for the big die, it would make sense if the mainstream die also went to 6+8.

 

[jpiniero]

IIRC historically the i3 has been the full smaller die though.

 

[igor_kavinski]

That's the rumor. Since Intel hasn't officially announced the locked parts, it's conceivable that they could change their mind and all i5 could end up being 6+4 and the i3 being 6+0. But the i3 is definitely going to be coming from the 6+0 die exclusively.

That's quite a jump if i3 ends up being 6+0. Most games can't put more than 6 cores to good use so it would make a really decent budget gaming CPU. I also just found out that there is no i3 in the RocketLake line-up. WTH? Is their yield that good that they dont have defective chips for i3?