First came a lake, then a dinosaur, then a meteor - Intel Raptor Lake is rumored to launch by the end of September

[Lightwreather]

Summary

According to Moore’s Law is Dead, First 13th Gen desktop parts are supposed to launch by the end of the third quarter this year, which means the end of September at most.

 

Quotes

Quote

MLID has an update on the possible Raptor Lake release date. First 13th Gen desktop parts are supposed to launch by the end of the third quarter this year, which means the end of September at most. Interestingly, the mobile H/HX/U series should launch already in the fourth quarter. The mobile launch date rumors are a bit more interesting considering that the most recent laptop launches were usually attached to new graphics architectures. That said, Raptor Lake Mobile arriving in the fourth quarter would indicate that we are also getting new NVIDIA or Intel GPUs.

At Investor Meeting 2022 Intel confirmed Raptor Lake is to feature up to 24 cores and 32 threads. This configuration is achieved by combining 8 Performance cores and 16 Efficient cores. The company also confirmed Raptor Lake CPUs will be socket-compatible with LGA1700 motherboards already available for Alder Lake systems.

 

My thoughts

Anyone else just noticing that after Raptorlake comes MeteorLake?

Tangent aside, it seems Intel's Raptor lake will arrive on time to fight against AMD's Zen4 offerings, only giving about (likely) 2 months of glory if a July release is to be believed. It also seems Intel is doubling down on and doubling E-core core counts. I'm actually feeling a bit excited about the fact that there might be an actual tussle for the performance crown (even though I'm unlikely to upgrade in that time). Well, whether MLID is right about a september launch, we're gonna have to wait and see.

 

Sources

Videocardz

MLID

[CTR640]

Screw this news, I want to know what comes after Raptor!

[IkeaGnome]

Just now, CTR640 said:

Screw this news, I want to know what comes after Raptor!

Meteor, then Arrow. Seems it's the end of times Horizon Zero Dawn style.

[thewelshbrummie]

59 minutes ago, J-from-Nucleon said:

It also seems Intel is doubling down on and doubling E-core core counts. I'm actually feeling a bit excited about the fact that there might be an actual tussle for the performance crown (even though I'm unlikely to upgrade in that time). Well, whether MLID is right about a september launch, we're gonna have to wait and see.

Pretty much on the expected schedule then, since Intel has released a new generation give or take every year (whether they're a new arch is a different question, but they have without fail done near annual updates to their CPU SKUs).

 

The rumour I saw in November in terms of core count seems to be holding to be accurate, mainly that the -e cores will be doubled across the board including adding them to the whole i5 range (rather than only the xx600K(F) SKU). Potentially makes the 13400 a very interesting CPU if that turns out to be accurate.

[Alcarin]

Very nice news, anticipating upgrading to this for my rig and potentially my gf's rig (though she may get my current platform instead).

 

Is it still 10nm or going to be 10nm+? Also is it still going to retain DDR4 compatibility or just be DDR5 only?

[elfensky]

2 hours ago, CTR640 said:

Screw this news, I want to know what comes after Raptor!

According to wikipedia, Osprey. 

[porina]

3 hours ago, J-from-Nucleon said:

According to Moore’s Law is Dead, First 13th Gen desktop parts are supposed to launch by the end of the third quarter this year, which means the end of September at most.

This isn't much of a prediction. An approximately annual cadence has been historic. If he ends up wrong the blame will be shifted to Intel delaying. With that in mind, I'd say Q3 would be aggressive given Alder Lake was a November launch, but we are seeing a different Intel now and we have AMD's platform catch up possibly in a similar time.

 

53 minutes ago, Alcarin said:

Is it still 10nm or going to be 10nm+?

It will remain on Intel 7 same as Alder Lake. Even Intel wasn't consistent on processes formerly known as 10nm depending if you count Canon Lake or not.

Cannon Lake: First Intel 10nm that Intel rather forget

Ice Lake: First 10nm Intel wants to talk about

Tiger Lake: First 10nm that could clock well, named 10SF

Alder Lake: generally comparable to TSMC 7nm, named 10SFE then later Intel 7.

 

53 minutes ago, Alcarin said:

Also is it still going to retain DDR4 compatibility or just be DDR5 only?

I'd guess so but don't have an official reference. It shares the same platform as Alder Lake, and given the widespread DDR4 boards for such, it would seem logical to continue for that generation. I wouldn't expect DDR4 support beyond that though.

[LAwLz]

1 minute ago, porina said:

This isn't much of a prediction. An approximately annual cadence has been historic. If he ends up wrong the blame will be shifted to Intel delaying. 

That's exactly what I was thinking too. MLID likes to make shit up for clicks and then blame it on the manufacturer when his rumors turn out to be wrong. 

 

 

Anyway, let's hope the doubling of E cores gets applied across the board. That would make some of the mid range chips absolutely amazing. They are already really good, but it would be a big bump to mutincore scores. 

[Arika]

Up next, MLID predicts the sun will rise tomorrow

[thewelshbrummie]

2 hours ago, Alcarin said:

 Also is it still going to retain DDR4 compatibility or just be DDR5 only?

Raptor Lake CPUs should work on 600 series chipsets with a BIOS update. No idea if 700 series will support DDR4 but there won't be any issue using a 13th gen CPU on Z690 based motherboard.

 

Put simply, every Intel release since 2nd gen has allowed CPU and chipset compatibility can be mixed across 2 consecutive generations (e.g. 2nd/3rd, 6/7th, 10/11th). The only exception I know if is that OG 4th gen Z87 boards can't take either of the 2 Broadwell based 5th desktop CPUs that Intel launched. I see no reason for Intel to change that.

[dizmo]

MLID is almost as (if not as) bad as WCCFTech for sources. I'd really recommend against using him in the future. 

[Lightwreather]

3 hours ago, dizmo said:

MLID is almost as (if not as) bad as WCCFTech for sources. I'd really recommend against using him in the future. 

Huh, that is something I did not know, I'll keep that in mind. Thanks

[RejZoR]

Predicting launches is like making prediction announcements that new iPhone will be launched for sure in fall 2022 and it'll have the best chipset yet! It's not rocket science to look back and make predictions because companies tend to stick with them if they work, Intel is no exception. Only difference is that Apple doesn't quite have competition in smartphones segment, but Intel does with CPU's. It now depends how much on fire AMD will be with Zen4/AM5 which will dictate if Intel will push things forward to steal the attention, assuming they have things ready enough anyways. If not they'll probably launch some petty "we're still here" press release until they actually deliver something. Then again AMD did that too in the past...

[mr moose]

7 hours ago, Arika S said:

Up next,p MLID predicts the sun will rise tomorrow

Turns out it's just the earth spinning, the sun stayed in the same place.  MLID blamed the universe in general and God specifically for delaying the sunrise.

[starsmine]

7 hours ago, dizmo said:

MLID is almost as (if not as) bad as WCCFTech for sources. I'd really recommend against using him in the future. 

Yet people in the industry regularly speaks with him with respect... so idk about that. I would not put him at the same level as WCCFTech. WCCFTech do make stuff up, or rather get baited by bad info regularly. 
Leaks always need skepticism because things change, regularly. MLID is always upfront with how confident he is in his information and leaks, and what information is likely to change, and by how much. 

I feel like people need to change their thinking when it comes to intelligence. Its not about being right 100%, its about collecting information from reliable sources, and converting and combing that information into something meaningful, with reasonable confidence intervals. 

Like yes a review is going to be able to have a margin of error sub 1%, but because things CHANGE leaks can not. and to disparage reporting because it doesn't have hindsight is weird.

Things inside businesses change to. Like its not incorrect to say there was a Quinten Tarentino Star trek film in some form of pre production, just because it never left that stage. 


Anyways. point is, take the news with healthy amount of salt, and by news, I mean specifically that it will launch at end of september, because there is always the possibility they find an issue in an engineering sample that delays it a month. Or some shift in market makes they rush it out a month earlier because they dont need new mobos or the like. (aka why intel has not released a specific release date yet, so they are not held to it, even they are not 100% sure they can hit it this far out, its just a goal of launch at this date)

[WolframaticAlpha]

2 hours ago, RejZoR said:

it'll have the best chipset yet

Well, looking at Rocket Lake, Bulldozer and SDM810, I have pretty much given up getting shocked when the previous generation is better

[leadeater]

3 hours ago, starsmine said:

Yet people in the industry regularly speaks with him with respect... so idk about that

Who speaks to him is what matters, any random person "in the industry" could talk to him but that doesn't mean that person is in way actually involved in what they are talking about or has any real details at all and it just simply rumor spreading.

 

I am "in the industry" and I do not talk to MLID however I don't know a damn thing more than anyone else typically any earlier than anyone else so talking to me would be entirely fruitless, however if MLID did talk to me I'd be real tempted to feed him BS just to mess with him. I could do it in a very believable and convincing way.

 

3 hours ago, starsmine said:

I feel like people need to change their thinking when it comes to intelligence. Its not about being right 100%, its about collecting information from reliable sources, and converting and combing that information into something meaningful, with reasonable confidence intervals. 

Well I think people need to apply more critical thinking and not be so believing of "people on the internet".

 

People like MLID are the Palm Readers or Cold Readers of the internet.

 

And FYI people don't like MLID not because he's talking about rumors etc and get things wrong it's how it's done, the elevation of self importance, that constant hiding behind "protecting sources" and most importantly never admitting getting things wrong. And getting things wrong, admitting it or not, isn't even the issue it's the why it was wrong that is the problem, because there is typically very strong suspicion that is due to just making things up based on next to nothing. No matter how much credible things you say and do these are always negated by "lies".

[starsmine]

12 minutes ago, leadeater said:

And FYI people don't like MLID not because he's talking about rumors etc and get things wrong it's how it's done, the elevation of self importance, that constant hiding behind "protecting sources" and most importantly never admitting getting things wrong.

I do find that beyond obnoxious. While he does have some of the best source pools and accurate leaks in the industry... his attitude drains me often. Like he may be correct in terms of that he may have leaked something in 90 min into a 2 hour video a month ago, But then the man acts like you should have already seen it. And while I appreciate not having to lay the ground work every single video, he presents it in a way like you are dumb for not watching all of his videos/reading all of his articles all the way through and not remembering every detail 3 weeks later. I think have seen him admit to having bad analysis like twice, but not often enough. One does not even really have to admit much fault with that depending on the circumstances. 

On topic of Raptor Lake, I want to know more about how they will fit 8 more cores without changing pin outs. I wonder where I can find a white paper on the pin out. And since intel isnt doing tik tok anymore, how significant the changes will. a 10% ipc increase on the big cores sounds nice for not doing a node shrink. 

[porina]

27 minutes ago, starsmine said:

On topic of Raptor Lake, I want to know more about how they will fit 8 more cores without changing pin outs. 

The socket pins are not directly related to core count. Consider what signals do go through the socket: Ram, IO (e.g. PCIe), power. All the cores will talk through the same interface to the rest of the world.

[KaitouX]

Good to see that another increase in core count, even if it's only E-cores. And if this actually manages a 10% IPC increase, I feel like it will be hard for AMD to match it in single core performance.

 

It will be interesting to see how AMD reacts to this, they can't compete at all with their current prices and core counts when it comes to multicore, they will need to drop the current core counts at least one tier(Ryzen 5 as a 8 core part, Ryzen 7 as 12 core) to be able to compete, or lower the prices to Ryzen 2000 levels(R5 for $200, R7 for $300), possibly both depending on the single core and gaming performance they get from Zen 4.

A price drop might be necessary anyway if DDR5 prices don't drop soon, as the extra cost from DDR5 with Zen 4 would make Intel options with DDR4 a no-brainer. And the core count increase at the same price needs to happen, as it will be soon the Ryzen 5th anniversary with no core count increases for the same price below $500, if anything price per core only increased since the 1st gen.

[tim0901]

1 hour ago, leadeater said:

And FYI people don't like MLID not because he's talking about rumors etc and get things wrong it's how it's done, the elevation of self importance, that constant hiding behind "protecting sources" and most importantly never admitting getting things wrong.

Don't forget the obvious bias towards a certain red company. You just have to look at his video titles to notice that his language surrounding Nvidia is always super negative, while AMD's is for the most part more neutral to positive.

 

Also he's quite happy to use other creator's content without credit. Like this chart from LTT for example: https://youtu.be/8nAwu6CnmgY?t=70 or perhaps this chart from Techspot: https://youtu.be/8nAwu6CnmgY?t=151 . While, yes, fair use laws do probably make this behaviour legal, that's just not on.

[porina]

53 minutes ago, KaitouX said:

if this actually manages a 10% IPC increase, I feel like it will be hard for AMD to match it in single core performance.

AMD wont be standing still either. Zen 4 will have some IPC increase, and if they can include a clock increase also that would be quite a potent combination. AMD will also be moving to TSMC 5nm so they'll have a node generation advantage against Intel.

 

53 minutes ago, KaitouX said:

It will be interesting to see how AMD reacts to this, they can't compete at all with their current prices and core counts when it comes to multicore

AMD are taking a different approach. If we look at the current full desktop CPUs, they're all high X models at each core count. There are no low tier versions that were popular with value seekers in the past. Instead, the lower end of each core count is served by the APUs. I suspect this is the operation model for AMD at least for the near future. You want the top models, you pay for it. You pay less, you get less (cache, PCIe lanes). I get why they're doing this. The chiplet approach is going to be more costly and makes more sense on the higher end where pricing can match. For lower costs, the monolithic APUs will reduce manufacturing cost. I'd go as far as to wonder if they might even kill off the full 6 core version, since it would be lacking for those seeking a higher end solution and paying for it. 

 

53 minutes ago, KaitouX said:

A price drop might be necessary anyway if DDR5 prices don't drop soon, as the extra cost from DDR5 with Zen 4 would make Intel options with DDR4 a no-brainer.

If as I suspect Zen 4 will only appear on the higher end initially, they'll be fighting for the same high end system builders who are also looking at Intel DDR5 systems. The value buyers I don't feel will be catered for initially and that will continue to be served by older offerings.

 

53 minutes ago, KaitouX said:

And the core count increase at the same price needs to happen, as it will be soon the Ryzen 5th anniversary with no core count increases for the same price below $500, if anything price per core only increased since the 1st gen.

Core count isn't everything. We had the endless arguments and will likely continue to have the same arguments for eternity. All a sensible buyer can do is look at performance for what is important to them and take that into consideration as one of many data points.

[LAwLz]

3 hours ago, leadeater said:

Who speaks to him is what matters, any random person "in the industry" could talk to him but that doesn't mean that person is in way actually involved in what they are talking about or has any real details at all and it just simply rumor spreading.

 

I am "in the industry" and I do not talk to MLID however I don't know a damn thing more than anyone else typically any earlier than anyone else so talking to me would be entirely fruitless, however if MLID did talk to me I'd be real tempted to feed him BS just to mess with him. I could do it in a very believable and convincing way.

Another issue I have with him is that he and his viewers constantly shifts the blame to the manufacturer whenever he gets things wrong, even when it doesn't make any sense.

He said that the RTX 3080 would have 3840 CUDA cores. It turned out it has 8704. That's not something Nvidia decided to change a couple of months before it launched.

 

He also likes the cold reading technique of shotgunning guesses, and then just points to the one he got right.

In the span of like two months he said the RTX 30 series would be made on:

• TSMC 7N
• Samsung 10nm
• Samsung 8nm

It's not hard to get one right answer when you are guessing 3 times on something that essentially only has 3 outcomes.

 

 

Has anyone done a follow-up on the RDNA2 stuff he leaked? Would be interesting to see how much he got right.

[LAwLz]

20 minutes ago, porina said:

Core count isn't everything. We had the endless arguments and will likely continue to have the same arguments for eternity. All a sensible buyer can do is look at performance for what is important to them and take that into consideration as one of many data points.

Core count isn't everything, but Intel has the single core performance crown, and if they can use a bunch of E cores to get a big lead in multi-core performance, then I think that's a fair strategy.

 

Intel's processors already has very good, if not the best, performance when it comes to these common workloads that only use up to 6 threads. Once you go past that into these very highly threaded workloads then it doesn't really matter if it's running on a bunch of small cores or a few big cores since it will scale well anyway. 

[KaitouX]

7 hours ago, porina said:

AMD wont be standing still either. Zen 4 will have some IPC increase, and if they can include a clock increase also that would be quite a potent combination. AMD will also be moving to TSMC 5nm so they'll have a node generation advantage against Intel.

I understand that, I just am not sure if Zen 4 is able to get a 30%+ increase in single core performance, which would be about what is necessary to keep up with Raptor Lake. If they want to pull the Intel and compete solely on the single core performance and gaming, they will need to be significantly ahead of Intel on that front to make any sense.

 

My current guess is that Zen 4 will be close to Raptor Lake in single core performance but slightly slower, significantly below at multicore, with power consumption likely being a win to AMD when compared to Intel's K SKUs, but a loss against the non-K ones. Gaming performance is a weird one but probably neither will be significantly ahead of the other overall, considering that both are planning big cache increases, and I'm assuming similar single core performance.

8 hours ago, porina said:

AMD are taking a different approach. If we look at the current full desktop CPUs, they're all high X models at each core count. There are no low tier versions that were popular with value seekers in the past. Instead, the lower end of each core count is served by the APUs. I suspect this is the operation model for AMD at least for the near future. You want the top models, you pay for it. You pay less, you get less (cache, PCIe lanes). I get why they're doing this. The chiplet approach is going to be more costly and makes more sense on the higher end where pricing can match. For lower costs, the monolithic APUs will reduce manufacturing cost. I'd go as far as to wonder if they might even kill off the full 6 core version, since it would be lacking for those seeking a higher end solution and paying for it.

The issue with this strategy is Intel non-K SKUs, AMD needs to beat those soundly if they want to ask for premium prices, but that might be pretty hard to do.

AMD, just like in the GPU front, is trying to position itself as if they are the one with the majority of the market share, even though that isn't the case.

8 hours ago, porina said:

If as I suspect Zen 4 will only appear on the higher end initially, they'll be fighting for the same high end system builders who are also looking at Intel DDR5 systems. The value buyers I don't feel will be catered for initially and that will continue to be served by older offerings.

Please correct me if I'm wrong, but isn't high-end DDR4 faster than current DDR5 at most tasks? I thought that DDR5 only was better at some extremely niche tasks currently. While I understand that there are people that will go DDR5 regardless, if you want the best performance in the high-end DDR4 might still be the better option.

8 hours ago, porina said:

Core count isn't everything. We had the endless arguments and will likely continue to have the same arguments for eternity. All a sensible buyer can do is look at performance for what is important to them and take that into consideration as one of many data points.

Core count isn't everything, Intel did prove that with the 8700K and 9900K, but it's one of the important factors of CPU performance in many tasks. But this argument only goes through if the lower core count part have significant advantages on the single-core/cache side, which might not be true when comparing Raptor Lake to Zen 4.

7 hours ago, LAwLz said:

Intel's processors already has very good, if not the best, performance when it comes to these common workloads that only use up to 6 threads. Once you go past that into these very highly threaded workloads then it doesn't really matter if it's running on a bunch of small cores or a few big cores since it will scale well anyway. 

Some core heavy applications do have a sweet spot where the higher core counts starts to scale not as well, so it might matter in those.