Discussion Intel current and future Lakes & Rapids thread

[Thala]

Cannonlake, IceLake, and TigerLake support it. Cannonlake was largely irrelevant; nevertheless, Intel has made an effort to support AVX512 on client cores at the design level. Sadly, they're struggling to produce and sell any of those designs. Golden Cove represents the fourth generation of Intel cores to support AVX512 across their entire product lineup.

Cannonlake, IceLake, TigerLake and Cascade/CooperLake do not support the same subset of AVX512 extensions - it is a big mess in AVX512-land. I do not see any general software will be picking up anything else than the common denominator, which is AVX512 F/CD.
And even then, with AMD not supporting AVX512 it is just a waste of die space - making it even harder for Intel to compete with AMD.

 

[Cardyak]

Interesting chart here to demonstrate just how much of a mess Intel has made with the fragmentation of AVX512

 

[DrMrLordX]

Cannonlake, IceLake, TigerLake and Cascade/CooperLake do not support the same subset of AVX512 extensions - it is a big mess in AVX512-land.

True, but largely academic. The point is that as of Cannonlake, Intel was actually supporting AVX512 in general, even on small mobile dual-core chips. It's not like the Skylake generation and previous when some low-level chips might not support AVX/AVX2 at all. TigerLake supports the most AVX512 extensions, and Golden Cove/Alder Lake might support even more.

 

[jpiniero]

No. As far as I know, Emmitsburg = Eagle Stream PCH. So These are same one.
Both names are shown because the names shown are different depending on the document being referenced.

Emmitsburg is apparently what the Alder Lake platform is called. I guess yeah, maybe it's just a cost cutting move where server and client share the same PCH.

 

[Exist50]

Emmitsburg is apparently what the Alder Lake platform is called. I guess yeah, maybe it's just a cost cutting move where server and client share the same PCH.

No, Komachi is correct. Emmitsburg = Eagle Stream PCH.

 

[IntelUser2000]

By the way I don't think 7nm CPUs will be delayed. The biggest hit will be to their discrete GPU ambitions, specifically the Xe-HPC, and very likely Ponte Vecchio.

Intel claimed the 2021 7nm lead product is a dGPU for datacenters. Sapphire Rapids is early 2022, so that puts Granite Rapids in early 2023. Alderlake is late 2021, so successor to that is late 2022. They were talking about 6 months delay to a product but 12 months behind their goals in process so its possible that post-Alderlake might be 2-3 months behind. That puts the 7nm high volume client chip at Dec/Jan?

Originally 7nm would have been ramped up at the same time as their 10nm++ process.

Personally I was expecting post-post-Lakefield to be one of the early CPU launches on 7nm. If my guesses are true, that could be affected.

This explains the other report that said TSMC doesn't expect big orders from Intel. Little doubt in mind that its the Aurora contract that's affected and its very possible that they ripped out the GPU portion to use AMD/Nvidia. Intel might still try to sell Ponte Vecchio as a general product with the compute die sourced from TSMC.

It's starting to be a big gap now, but if they don't suffer any more setbacks it might not be too terrible for them.

 

[jpiniero]

This explains the other report that said TSMC doesn't expect big orders from Intel.

180k is a pretty big order.

 

[IntelUser2000]

180k is a pretty big order.

Yes, and there was another report saying that TSMC doesn't expect Intel deal to be a long term thing. You also know that certain press were sure of Intel using Samsung for production woes, and nothing of that sort came true?

Intel already orders quite a lot from TSMC, and even if that number is true, it could be just that.

 

[maddie]

180k is a pretty big order.

Over what period? I for one, haven't seen that mentioned anywhere.

 

[jpiniero]

Yes, and there was another report saying that TSMC doesn't expect large additional orders.

Intel already orders quite a lot from TSMC, and even if that number is true, it could be just that.

180k pretty much has to be a Core or Xeon product.

 

[jpiniero]

Over what period? I for one, haven't seen that mentioned anywhere.

Good question, I would think they would want most of it up front but spread over say a year.

 

[mohit9206]

Will Tiger lake i3 also have the new higher core count Gen12 iGPU? And will it launch later this year on budget laptops? Will it compete with Vega 3 graphics in 3200U?

 

[btarlinian]

I'm not sure if the "reported" N6 order is meaningful with regards to outsourcing of processors. Intel has historically made plenty of products at TSMC (e.g., Ethernet controllers). And they've also bought plenty of companies that make their chips at TSMC (e.g., Mobileye, Altera, Habana, Barefoot, etc.) Outside of Altera, none of these companies have moved their products to internal Intel fabs (and definitely won't in the current situation). I could imagine that the combined volume of all of these SKUs along with a discrete Xe GPU would add up to the reported volumes.

 

[Bouowmx]

Will Tiger lake i3 also have the new higher core count Gen12 iGPU? And will it launch later this year on budget laptops? Will it compete with Vega 3 graphics in 3200U?

The Core i3 is the i3-1115G4, so not the G7 tier. This info is based on the leaked SKU list.

Previous thread posts say G4 has 48 EUs, half of 96 EUs in G7. Not too sure how concrete the G4's specs are.

 

[jpiniero]

I'm not sure if the "reported" N6 order is meaningful with regards to outsourcing of processors. Intel has historically made plenty of products at TSMC (e.g., Ethernet controllers). And they've also bought plenty of companies that make their chips at TSMC (e.g., Mobileye, Altera, Habana, Barefoot, etc.) Outside of Altera, none of these companies have moved their products to internal Intel fabs (and definitely won't in the current situation). I could imagine that the combined volume of all of these SKUs along with a discrete Xe GPU would add up to the reported volumes.

180k is a ton of wafers. Not to mention that it would be well over a billion dollars. I'm sure Altera will be going back to TSMC but I don't even think they would need that many wafers.

 

[mohit9206]

"The 3DMark Fire Strike results that we've received show that even the entry-level Core i3 Tiger Lake with 48 EUs is about 20% faster than the Ice Lake Iris Plus Graphics G7 with 64 EUs."
Wow this is exciting. This means that cheap Tigerlake i3 laptops will most definitely beat Vega 3 laptops and maybe match Vega 6 laptops of the 3200u and 3500u respectively. I don't like Intel but it seems AMD is simply allowing Intel to gain the lead even in iGPU department which is honestly crazy because AMD had begun with a huge lead when it comes to integrated graphics.

 

[IntelUser2000]

"The 3DMark Fire Strike results that we've received show that even the entry-level Core i3 Tiger Lake with 48 EUs is about 20% faster than the Ice Lake Iris Plus Graphics G7 with 64 EUs."

3500U has Vega 8 graphics, not Vega 6. The Iris Plus G7 of the Icelake generation is pretty much on par with the Picasso generation so 20% faster than that will definitely beat it.

 

[NTMBK]

"The 3DMark Fire Strike results that we've received show that even the entry-level Core i3 Tiger Lake with 48 EUs is about 20% faster than the Ice Lake Iris Plus Graphics G7 with 64 EUs."
Wow this is exciting. This means that cheap Tigerlake i3 laptops will most definitely beat Vega 3 laptops and maybe match Vega 6 laptops of the 3200u and 3500u respectively. I don't like Intel but it seems AMD is simply allowing Intel to gain the lead even in iGPU department which is honestly crazy because AMD had begun with a huge lead when it comes to integrated graphics.

I don't think it's that crazy. The huge lead in IGP performance won AMD almost no laptop sales, so now they're copying Intel's strategy- great CPU performance, and "good enough" GPU.

 

[uzzi38]

The Core i3 is the i3-1115G4, so not the G7 tier. This info is based on the leaked SKU list.

Previous thread posts say G4 has 48 EUs, half of 96 EUs in G7. Not too sure how concrete the G4's specs are.

G7 tier means nothing either to be fair.

But in anycase, yeah should be 48EUs.

Updated: Intel Tiger Lake Gen12 Xe iGPU offers significantly better performance than Ice Lake Gen11 at same TDP, finally surpasses Vega 8 in AMD Renoir

We have received exclusive 3DMark Fire Strike results pertaining to Tiger Lake-U Gen12 Xe iGPUs, which are slated to be available as 48, 80, and 96 EU parts. The results seem to indicate that Xe can offer nearly twice the performance of corresponding Ice Lake Gen11 at the same TDP while also...

www.notebookcheck.net

 

[mikk]

I don't think it's that crazy. The huge lead in IGP performance won AMD almost no laptop sales, so now they're copying Intel's strategy- great CPU performance, and "good enough" GPU.

It's different though. Intel had better much better ST,+MT peformance and battery life time at this time, so it was no surprise AMDs vastly better GPU didn't help much and their volume was really low anyways. Intel won't outperform 8C Renoir in MT but they are more than competitive in ST and probably battery life alongside a likely better GPU. Intels big issue is the H segment, it's not looking good there.

 

[Bouowmx]

Huh, what is this?

Intel Tiger Lake features 10nm SuperFin architecture - VideoCardz.com

More information on Intel’s next-generation mobile architecture. Intel Tiger Lake features SuperFin We have more details on Intel’s upcoming Tiger Lake architecture. The series will be announced on September 2nd during a special virtual event. In the meantime, we have gathered new information...

videocardz.com

Intel is calling it the 10nm SuperFin architecture because it features a redesigned transistor (SuperFin) and capacitor design (Super MIM). According to people familiar with the matter, this intranode architecture will provide a performance uplift comparable to a full-node transition. A redefined FinFET will provide additional gate pitch (higher drive current), improved gate process (higher channel mobility), and enhanced expiation source/drain (lower resistance). Additionally, 10nm SuperFin architecture will benefit from the introduction of Super MIM capacitor, delivering 5x the increase in MIM (metal-insulator-metal) capacitance and novel thing barrier reducing the resistance by 30%.

The company is developing a 10nm Enhanced SuperFin architecture, promising additional performance, interconnect innovations, and optimization for data centers

Feels like marketing talk out of no where, but I'm some retard, so I wouldn't know.

Also in the article, ring bandwidth got doubled for Tiger Lake; I assume from 32 to 64 bytes/cycle.

 

[IntelUser2000]

Feels like marketing talk out of no where, but I'm some retard, so I wouldn't know.

They may have attached fancy words to it but the gains are real. The MIM capacitance increase over Icelake's process is greater than Icelake versus Cannonlake.

Product also speaks for itself. Base frequency increases by nearly 2.5x and peak Turbo by 20%.

The Willow Cove core will deliver higher frequency at lower voltages than Sunny Cove.

That's apparently in addition to the process.

 

[NTMBK]

Huh, what is this?

Intel Tiger Lake features 10nm SuperFin architecture - VideoCardz.com

More information on Intel’s next-generation mobile architecture. Intel Tiger Lake features SuperFin We have more details on Intel’s upcoming Tiger Lake architecture. The series will be announced on September 2nd during a special virtual event. In the meantime, we have gathered new information...

videocardz.com

Feels like marketing talk out of no where, but I'm some retard, so I wouldn't know.

Also in the article, ring bandwidth got doubled for Tiger Lake; I assume from 32 to 64 bytes/cycle.

I guess Intel ran out of +s?

 

[IntelUser2000]

So they'll have two process versions for Tigerlake. The following is for Celerons and Pentiums.

Intel SuperFin transistor and Super MIN capacitor (left), Source: Raja Koduri

 

[ikjadoon]

reddit found a Tiger Lake listing already up. Of course, it's HP.

Click here and scroll to the last laptop: https://store.hp.com/us/en/vwa/z---categories-for-promotions/form=Standard-laptop

PROCESSOR
Intel® Core™ i7-1165G7 (2.8 GHz base frequency, up to 4.7 GHz with Intel® Turbo Boost Technology, 12 MB L3 cache, 4 cores) [20,21]