Hybrid CPUs: Sunny Cove and Tremont

Now that we’ve gone over the concept of the heterogeneous core design, it’s time to dig into each of the cores separately and some of the tradeoffs that Intel has had to do in order to get this to work.

Big Sunny Cove

As mentioned previously, the big core in Lakefield is known as Sunny Cove, and stands as the same core we currently see in Intel’s Ice Lake mobile processors today. It is officially Intel’s second 10nm-class core (the first one being the DOA Cannon Lake / Palm Cove), but the first one in mass production.

We have covered the Sunny Cove core microarchitecture in great detail, and you can read about it here:

Examining Intel's Ice Lake Processors: Taking a Bite of the Sunny Cove Microarchitecture

The quick recap is as follows.

Very similar to a Skylake design, except that:

  • Better prefetchers and branch predictors
  • +50% L1 Data Cache
  • +100% L1 Store Bandwidth
  • +100% L2 Cache w/improved L2 TLB
  • +50% Micro-op Cache
  • +25% uops/cycle into reorder buffer
  • +57% reorder buffer size
  • +25% execution ports
  • AVX-512 with VNNI

The side effect of increasing the L1 Data cache size was a decrease in latency, with the L1-D moving to a 5-cycle rather than a 4-cycle. Normally that would sound like a 25% automatic speed drop, however the increased L1 size, L1 bandwidth, and L2 cache all help for an overall improvement.

Intel claimed that Sunny Cove should perform ~18% better clock-for-clock compared to a Skylake core design. In our initial review of Ice Lake, we compared the i7-1065G7 processor (Ice Lake) to the Core i9-9900K processor (Coffee Lake, a Skylake derivative), and saw a 19% increase in performance per clock, essentially matching Intel’s advertised numbers.

(However it should be noted that overall we didn’t see that much of an improvement at the overall chip and product level, because the Ice Lake ran at a lower frequency, which removed any raw clock speed gain.)

Small Tremont Atom

Arguably the Tremont core is the more interesting of the two in the Lakefield design. Lakefield will be the first consumer product built with a Tremont core inside, and as a result we have not had a chance to test it yet. But we have gone over the microarchitecture extensively in a previous article.

Intel's new Atom Microarchitecture: The Tremont Core in Lakefield

The reason why Tremont is more exciting is because updates to Intel’s Atom line of processor cores happen at a much slower pace. Traditionally Atom has been a core that focuses on the low cost part of the market, so there isn’t that much of a need to make it right at the bleeding edge as it commands lower margins for the company. It still plays a vital role, but for context, here is what year we’ve seen new Atom designs come into the market:

  • 2008: Bonnell
  • 2011: Saltwell
  • 2013: Silvermont
  • 2015: Airmont
  • 2016: Goldmont
  • 2017: Goldmont Plus
  • 2020: Tremont

Tremont is the first new Atom microarchitecture design for three years, and technically only the third Atom design to be an out-of-order design. However, Tremont is a big jump in a lot of under-the-hood changes compared to Goldmont Plus.

  • Can be in a 1-core, 2-core, or 4-core cluster
  • +33% L1-Data Cache over Goldmont+, no performance penalty
  • Configurable L2 cache per cluster, from 1.5 MB to 4.5 MB
  • +50% L2 TLB (1024-entry, up from 512)
  • New 2x3-wide decoder, rather than single 3-wide decoder
  • +119% re-order buffer (208, up from 92)
  • 8 execution ports, 7 reservation stations
  • 3 ALUs, 2 AGUs
  • Dual 128-bit AES units
  • New Instructions*

What made the most noise is the new dual 3-wide decoder. On Intel’s primary Core line, we haven’t seen much change in the decoder in recent generations – it still uses a 5-wide decoder, split between 1 complex decoder and 4 simple decoders, backed with a micro-op cache. Tremont’s new dual 3-wide decoder can manage dual data streams in order to keep the buffers further down the core fed. Intel stated that for the design targets of Tremont, this was more area and power efficient than a 6-wide decoder, or having a large micro-op cache in the processor design (Atom cores have not have micro-op caches to date). Intel states that the decoder design helps shape the back-end of the core and the balance of resources.

Also worthy of note in Tremont is the L1-Data cache. Intel moved up from a 24 KiB design to a 32 KiB design, an increase of 33%. This is mostly due to using the latest manufacturing node. However, an increase in cache size is typically accompanied with an increase in latency – as we saw on Sunny Cove, we moved from a 4-cycle to a 5-cycle. However in Tremont’s case, the L1-Data cache stays at 3-cycle for an 8-way 32 KiB design. Even Skylake’s L1-D cache, at an 8-way 32 KiB design, is a 4-cycle, which means that Tremont’s L1-D is tuned to surpass even Skylake here.

The final point, Tremont’s new instructions, requires a section all on its own, specifically because none of the new instructions are supported in Lakefield.

What’s Missing in Lakefield

One of the biggest issues with a heterogeneous processor design is software. Even if we go beyond the issues that come with scheduling a workload on such a device, the problem is that most programs are designed to work on whatever microarchitecture they were written for. Generic programs are meant to work everywhere, while big publishers will write custom code for specific optimizations, such as if AVX-512 is detected, it will write AVX-512.

The hair-pulling out moment occurs when a processor has two different types of CPU core involved, and there is the potential for each of them to support different instructions or commands. Typically the scheduler makes no guarantee that software will run on any given core, so for example if you had some code written for AVX-512, it would happily run on an AVX-512 enabled core, but cause a critical fault on a core that doesn’t have AVX-512. The core won’t even know it’s an AVX-512 instruction until it comes time to decode it, and just throw an error when that happens. Not only this, but the scheduler has the right to move a thread when it needs to – if it moves a thread in the middle of an instruction stream, that can cause errors too. The processor could also move a thread to prevent thermal hotspots occurring, which will then cause a fault.

There could be a situation where the programmer can flag that their code has specific instructions. In a program with unique instructions, there’s very often a check that tries to detect support, in order to say to itself something like ‘AVX512 will work here!’. However, all modern software assumes a homogeneous processor – that all cores will support all of the same instructions.

It becomes a very chicken and egg problem, to a certain degree.

The only way out of this is that both processors in a hybrid CPU have to support the same instructions completely. This means that we end up with the worst of both worlds – only instructions supported by both can be enabled. This is the lowest common denominator of the two, and means that in Lakefield we lose support for AVX-512 on Sunny Cove, but also things like GFNI, ENCLV, and CLDEMOTE in Tremont (Tremont is actually rather progressive in its instruction support).

Knowing that Lakefield was going to have to take the lowest common denominator from the two core designs, Intel probably should physically removed the very bulky AVX-512 unit from the Sunny Cove core. Looking at the die shot, it's still there - there was some question going into the recent disclosures as to whether it would still be there, but Intel has stated on the record repeatedly that they removed it. The die shot of the compute silicon shows that not to be the case.

For x86 programmers doing instruction detection by code name or core family, this might have to change. In the smartphone world, where 4+4 processor designs are somewhat the norm, this lowest common denominator issue has essentially been universally adopted. There was some slight issue with a Samsung processor that had a non-unified cache setup, which ended up being rectified in firmware. But both sets of CPUs had to rely on lowest common denominator instructions.

Thermal Management on Stacked Silicon How To Treat a 1+4 Hybrid CPU
Comments Locked

221 Comments

View All Comments

  • PaulHoule - Saturday, July 4, 2020 - link

    @DrK,

    the engineering on this part is like what you'd get if you contracted out to Rockwell or Litton Industries for a brain for a Stinger missile. Compact, brilliantly packaged, with adequate performance, but no concern at all about thermal dissipation because the missile is going to hit or miss its target before the CPU fries.

    Foveros is an expensive technology for a mass market device (cheap tablet) because the fabrication cost depends on the total area and there is an expensive step of stitching the chips together at the end. If you could avoid fabricating "glue" components and just snap together chips from a library this might be an amazing technology to build 500 of something at low development cost and time (e.g. weeks) If you have to make a new mask for the chip, however, it is a lot less fun.

    So far as AVX the problem is as you say: "who cares about AVX?" Intel has shipped a backlog of features that people don't use because of overhead and complexity. As a software dev I get paid to work on certain aspect of my products, and maximizing performance with the latest instructions may or may not be on my agenda. If it is easy to do I will push for it but it means debugging compatibility problems it is a tough ask. "Optimal" performance for a range of users can mean shipping many versions of a function; the performance of loading, installing, updating, those libraries will be not in the least optimal.

    Intel is like that Fatboy Slim album, 'We're #1, Why Try Harder?' The world has changed and Intel is not the #1 CPU firm any more. Intel has to get more Paranoid or it might not Survive.
  • Spunjji - Monday, July 6, 2020 - link

    Why start with "I'm not one to criticise" and then do it? Clearly you are, and as a rhetorical flourish it's tedious in the extreme.

    1 - It's a first-gen product and it shows, but they're putting it in premium products.
    2 - No deep-dive, for sure, but Intel's own figures are not very encouraging.
    3 - Citation needed here. There's no sign of it being used outside of low-power premium devices.
    4 - Who cares about AVX indeed! Tell that to the Intel fanboys pissing all over the AMD threads?

    I'm entirely in favour of your final conclusion, but it's not really supported by the previous statements. 🤷‍♂️
  • Oxford Guy - Friday, July 3, 2020 - link

    Bricklake or bust.
  • Meteor2 - Friday, July 3, 2020 - link

    Ultimately this is another attempt by Intel to stay relevant in a space where it's always struggled: mobile. With the progress being made by Apple, Microsoft, and Qualcomm using ARM, Intel is looking at losing an ever-growing chunk of what was the laptop market.

    But whatever Intel tries, bottom line is that ARM is more efficient than x86.
  • Beaver M. - Friday, July 3, 2020 - link

    Thats not the issue. The issue is that theres not much software in that sector for x86.
  • Valantar - Sunday, July 5, 2020 - link

    A few errors in the article: 2 16-bit channels of LPDDR4X should be 2 32-bit channels of LPDDR4X, given that Renoir (with 4 32-bit LP4X channels at the same clock speed) delivers exactly 2x the bandwidth. Right?

    You should also proofread the pasted-in laptop descriptions; a lot of stuff in them clashes with the previous text.

    Beyond that though: great article! Part of the reason why I love AT is for these technical yet understandable deep-dives. Looking forward to the next one.
  • Pixelpusher6 - Sunday, July 5, 2020 - link

    Interesting choice to place the DRAM right over the core, seems like it would make more sense to move it next to the chip but on package. I guess my question is was it worth the complexity to implement this Foveros design to save a little space? It seems like they could have gotten the same benefit by using a traditional packaging i.e. with a little large package. Can you imagine paying $2500 like the price of that Lenovo and having Atom-esque performance?
  • Pixelpusher6 - Sunday, July 5, 2020 - link

    *larger
  • Farfolomew - Monday, July 6, 2020 - link

    Agreed on the DRAM placement. It seems really out of place. Another "dime size" piece of silicon right next to the Lakefield CPU doesn't seem like it would take up much more board space, and would alleviate a ton of the heat dissipation problems by allowing the compute-layer die to be directly connected to a heatsink
  • serendip - Monday, July 6, 2020 - link

    It seems to be an interesting technical answer to a question nobody asked. Board space is a lot cheaper than what Lakefield would cost. It could also cost more for Intel to produce and they'd be stuck carrying multiple RAM SKUs.

    Heat dissipation could be a major issue. The slow chip could become even slower if it has to constantly throttle down because of thermal loads. Intel is sadly mistaken if this is supposed to be an ARM competitor.

Log in

Don't have an account? Sign up now