The Intel Lakefield Deep Dive: Everything To Know About the First x86 Hybrid CPU
by Dr. Ian Cutress on July 2, 2020 9:00 AM ESTHow To Treat a 1+4 Hybrid CPU
At the top of the article, I explained that the reason for using two different types of processor core, one big on performance and the other big on efficiency, was that users could get the best of both worlds depending on if a workload could be run efficiently in the background, or needed the high performance for a user experience interaction. You may have caught onto the fact that I also stated that because Intel is using a 1+4 design, it actually makes more sense for multi-threaded workloads to run on the four Atom cores.
Using a similar power/performance graphs, the effect of having a 1+4 design is quite substantial. On the left is the single core power/performance graphs, but on the right is when we compare 1 Sunny Cove to all 4 Tremont cores working together.
Where the previous graph considered a 1+1 design, which is more relevant in those user experience scenarios listed above, on the right is the 1+4 design for when the user demands a heavier workload that might not be latency critical. Because there are four Atom cores, the blue line multiplies by four in both directions.
Now obviously the real world scenario is somewhere between the two, as it is possible to use only one, two, or three of the smaller cores at any given time. The CPU and the OS is expected to know this, so it can govern when workloads that can be split across multiple cores end up on either the big core or the small core.
In this graph from Intel, we have three distinct modes on which threads can operate.
- ‘Sunny Cove/SNC’ is for responsiveness and user experience threads,
- ‘Tremont/TNT Foreground’, for user related tasks that require multiple threads that the user is waiting on.
- ‘Tremont/TNT Background’, for non-user related tasks run in efficiency mode
Even though the example here is web browsing, it might be best to consider something a bit beefier, like video encoding.
If we run video encoding, because it is a user related task that requires multiple threads, it will run on the four Tremont cores (TNT FG). Anything that Windows wants to do alongside that gets scheduled as TNT BG. If we then open up the start menu, because that is a responsiveness task, that gets scheduled on the SNC core.
Is 1+4 the Correct Configuration?
Intel here has implemented a 1+4 core design, however in the smartphone space, things are seen a little differently. The most popular configuration, by far, is a 4+4 design, simply because a lot of smartphone code is written to take advantage of multiple foreground or multiple background threads. There are a number of cost-down designs that reduce die area and power by going for a 2+4 implementation. Everyone seems adamant that 4 is a good number for the smaller cores, partly because they are small and cheap to add, but because Arm’s quad-core implementation is a base unit for its IP.
The smartphone space in recent quarters has also evolved from a two tier system of cores. In some of the more leading edge designs, we now have three types of core: a big, a middle, and a small. Because of the tendency to stay with eight core designs, we now get 1+3+4 or 2+2+4 designs, powered by complex schedulers that manage where to put the threads for the best user experience, the best battery life, or somewhere in the middle. Mediatek has been famously dabbling in 10 core designs, going for a 2+4+4 approach.
One thing missing from all of these implementations is an SoC with one big core and four small cores. Smartphone vendors don’t seem to be interested in 1+4 silicon, and yet Intel has decided on it for Lakefield. This is borne out of decisions made on both sides.
From the smartphone perspective, when hybrid designs came about, the big cores just weren’t powerful enough on their own. In order to offer something more than simply basic, at least two cores were needed, but because of how Arm architected the big and little designs, it almost became standard to look into 4+4 implementations of big and small cores. It was only until this configuration was popularized over a couple of years, and Arm big cores got more powerful, that chip designs started looking at 2+4, or 1+3+4 designs.
On Intel’s side of the fence, the biggest problem it has is the size of the Sunny Cove core. By comparison, it’s really, really big. Because the graphics core is the same as Ice Lake and reuses its design, there simply isn’t enough room within the 82 mm2 compute die to add another core. Not only that, but there is a question of power. Sunny Cove wasn’t built for sub-1W operation, even in the Tremont design. We see big smartphone silicon pulling 4-5W when all eight cores are active – there is no way, based on our understanding of Intel’s designs, that we could see four (or even two) Sunny Cove cores being in the optimal performance per watt range while being that low. Intel’s Lakefield graphics, with 64 EUs, is running at only 500 MHz – a lot lower than the Ice Lake designs. Even if Intel moved that down to a 32 EU design to make space for another Sunny Cove core, I reckon that it would eat the power budget for breakfast and then some.
Intel has made the 1+4 design to act as a 0+4 design that sometimes has access to a higher performance mode. Whereas smartphone chips are designed for all eight cores to power on for sustained periods, Lakefield is built only for 0+4 sustained workloads. And that might ultimately be its downfall. This leads onto a deep discussion about Lakefield’s performance, and what we should expect from it.
Facebook
Twitter
RSS
221 Comments
View All Comments
Quantumz0d - Sunday, July 5, 2020 - link
PC gaming marketcap is supposed to be at $40Bn by 2022, total gaming market is $120Bn including everything, and Consoles are built on AMD x86 technology and now DX12U and you think that is a niche ?ARM is not going to do anything just because Apple did, there are so many trials by so many companies and the best company which is known for it's ROI with R&D, Qualcomm abandoned all of it's Server ARM marketshare dreams with the death of their full custom Centriq. x86 runs blazingly fast and optimized with Linux which is what the world is powered just because ARM is good in thin and light garbage doesn't make it a superstar.
ARM is not going to get into Desktop at all, no one is going to write their programs again to suppor that HW, and no company is going to invest in DIY market before Server/DC market. Supercomputer market is not the DIY or Enterprise, look at the Top Supercomputers, Chinese Tianhe and 2 positions are with Chinese only, AMD CRAY Zen based IF supercomputer is about to come as well.
Wilco1 - Sunday, July 5, 2020 - link
The #1 supercomputer is Arm, and Arm servers beat x86 servers on performance, cost and power, so not a single "fact" in your post is correct.lmcd - Sunday, July 5, 2020 - link
That first statement is hilariously disconnected from the second. Fugaku at 3x the cost per flop of its next competitor hardly backs up your assertion.ARM servers might beat x86 servers on performance, cost, and power but it's not looking that good vs x86_64. The latter arch is commodity hardware, software, and talent hiring.
Wilco1 - Monday, July 6, 2020 - link
Just looking at the peak FLOPS in comparisons is deceiving. Fugaku is a very different design as it does not use GPU accelerators like most supercomputers. That means it is far better than the rest in terms of ease of programming and efficiency. So even if the upfront cost is higher, they expect to get far more out of it than other super computers.I'd say Arm servers are doing really well in 2020, clearly companies want a change from the x86 duopoly. Much of the talent is at companies that do Arm designs. How else do you think Arm CPUs are getting 20-30% faster per year, and mobile phones already outperform the fastest x86 desktops?
Quantumz0d - Tuesday, July 7, 2020 - link
No company wants to develop an in house IP, that R&D and ROI is not easy, Amazon did it because to chop off some costs and set up a plan for the low end AWS instances with Graviton 2, Altera is still yet to show, Centriq abandoned by Qcomm with so much of marketing done around Cloudflare and top class engineering work, the team which made 820's full custom core.AND What the fuck you are babbling on fastest x86 desktops (Like Threadripper 3990X, or 3950X, 10900K) outperformed by mobile phones ? Ooof, you are gulping down the AT's SPEC scores aren't you ?
ARM servers LMAO, like how AMD upped their DC marketshare with EPYC7742, dude stop posting absolute rubbish. ARM marketshare in data centers is in 0.5% area where IBM also resides.
Quantumz0d - Monday, July 6, 2020 - link
Tiahu is fucking Chinese Sunway Processor based Supercomputer and it's top #3 so what did they do ? jack off to Zen with Hygon or did they make all Chinese use Chinese made processors ? Stop that bullshit of Supercomputer nonsense, IBM has been there since ages and they had SMT8 with Power9 uarch which came in 2017 (Summit which is #2, it was first since 2018) what did they do ? x86 is consumer based and DC market is relying only on that. ARM DC market-share is less than fucking 2%, AMD is at 4.5%, Intel is at 95% that is 2019 Q4.I don't know why people hate x86 as if it's like their life is being threatened by them, the fact that x86 machines are able to run vast diverse rich software selection and more freedom based computing, people want ARM based proprietary dogshit more, Apple series trash wich their APIs or the Bootloader locked (much worse like chastity) or Unlocked Android phones, even with GNU GPL v2 and Qcomm's top OSS CAF the godddamned phones do not get latest updates or anything but a Core2Quad from decade ago can run a fucking Linux or Win7 / Win10 without any bullshit issue.
Wait for the SPEC A series iPhone 12 benchmarks and then you be more proud of that garbage device which cannot compute anything outside what Apple deems it.
Wilco1 - Friday, July 3, 2020 - link
It would be good to run benchmarks on the 2 variants of Galaxy Book S. One comparison I found:https://browser.geekbench.com/v5/cpu/compare/25848...
So Lakefield wins by only 21% on single-threaded (that's a bad result given it is Cortex-A76 vs IceLake at similar clocks), and is totally outclassed on multithreaded...
lmcd - Sunday, July 5, 2020 - link
Current scheduler doesn't even guarantee that's the Sunny Cove core.Wilco1 - Monday, July 6, 2020 - link
Given Tremont can't get anywhere near Cortex-A76 performance, we can be sure single-threaded result is the Sunny Cove core.PaulHoule - Friday, July 3, 2020 - link
This is an example of the "Innovator's Dilemma" scenario where it is harder to move upmarket (in terms of performance) than downmarket.Put a phone processor into a box with a fan and people will be blown away by how fast it is -- they've never seen an ARM processor cooled by a fan before.
Put a desktop processor into a thin tablet with little thermal headroom and people will be blown away by how slow it is.
So first it is a situation that Intel can't win, but second it is a disaster that this low performance (downmarket) chip is expensive to produce and has to be sold upmarket. Sure you can stick any number of dies together and "scale up" a package in a way that looks as if you scaled up the chip by reducing the feature size, but when you reduce the feature size the cost per feature goes down in the long term -- when you stick a bunch of cheap chips together you get an expensive chip.