The Silvermont Module and Caches

Like AMD’s Bobcat and Jaguar designs, Silvermont is modular. The default Silvermont building block is a two-core/two-thread design. Each core is equally capable and there’s no shared execution hardware. Silvermont supports up to 8-core configurations by placing multiple modules in an SoC.

 

Each module features a shared 1MB L2 cache, a 2x increase over the core:cache ratio of existing Atom based processors. Despite the larger L2, access latency is reduced by 2 clocks. The default module size gives you clear indication as to where Intel saw Silvermont being most useful. At the time of its inception, I doubt Intel anticipated such a quick shift to quad-core smartphones otherwise it might’ve considered a larger default module size.

L1 cache sizes/latencies haven’t changed. Each Silvermont core features a 32KB L1 data cache and 24KB L1 instruction cache.

Silvermont Supports Independent Core Frequencies: Vindication for Qualcomm?

In all Intel Core based microprocessors, all cores are tied to the same frequency - those that aren’t in use are simply shut off (power gated) to save power. Qualcomm’s multi-core architecture has always supported independent frequency planes for all CPUs in the SoC, something that Intel has always insisted was a bad idea. In a strange turn of events, Intel joins Qualcomm in offering the ability to run each core in a Silvermont module at its own independent frequency. You could have one Silvermont core running at 2.4GHz and another one running at 1.2GHz. Unlike Qualcomm’s implementation, Silvermont’s independent frequency planes are optional. In a split frequency case, the shared L2 cache always runs at the higher of the two frequencies. Intel believes the flexibility might be useful in some low cost Silvermont implementations where the OS actively uses core pinning to keep threads parked on specific cores. I doubt we’ll see this on most tablet or smartphone implementations of the design.

From FSB to IDI

Atom and all of its derivatives have a nasty secret: they never really got any latency benefits from integrating a memory controller on die. The first implementation of Atom was a 3-chip solution, with the memory controller contained within the North Bridge. The CPU talked to the North Bridge via a low power Front Side Bus implementation. This setup should sound familiar to anyone who remembers Intel architectures from the late 90s up to the mid 2000s. In pursuit of integration, Intel eventually brought the memory controller and graphics onto a single die. Historically, bringing the memory controller onto the same die as the CPU came with a nice reduction in access latency - unfortunately Atom never enjoyed this. The reasoning? Atom never ditched the FSB interface.

Even though Atom integrated a memory controller, the design logically looked like it did before. Integration only saved Intel space and power, it never granted it any performance. I suspect Intel did this to keep costs down. I noticed the problem years ago but completely forgot about it since it’s been so long. Thankfully, with Silvermont the FSB interface is completely gone.

Silvermont instead integrates the same in-die interconnect (IDI) that is used in the big Core based processors. Intel’s IDI is a lightweight point to point interface that’s far lower overhead than the old FSB architecture. The move to IDI and the changes to the system fabric are enough to improve single threaded performance by low double digits. The gains are even bigger in heavily threaded scenarios.

Another benefit of moving away from a very old FSB to IDI is increased flexibility in how Silvermont can clock up/down. Previously there were fixed FSB:CPU ratios that had to be maintained at all times, which meant the FSB had to be lowered significantly when the CPU was running at very low frequencies. In Silvermont, the IDI and CPU frequencies are largely decoupled - enabling good bandwidth out of the cores even at low frequency levels.

The System Agent

Silvermont gains an updated system agent (read: North Bridge) that’s much better at allowing access to main memory. In all previous generation Atom architectures, virtually all memory accesses had to happen in-order (Clover Trail had some minor OoO improvements here). Silvermont’s system agent now allows reordering of memory requests coming in from all consumers/producers (e.g. CPU cores, GPU, etc...) to optimize for performance and quality of service (e.g. ensuring graphics demands on memory can regularly pre-empt CPU requests when necessary).

ISA, IPC & Frequency SoCs and Graphics, Penryn-Class Performance
Comments Locked

174 Comments

View All Comments

  • t.s. - Tuesday, May 7, 2013 - link

    If Intel play fair few years back, maybe now we have competitive offerings from AMD. That practice Intel's doing hurt AMD alot. Until now.
  • Homeles - Monday, May 6, 2013 - link

    I'm sure Anand would be drawing plenty of comparisons if he had a Temash tablet in hand.
  • Bob Todd - Monday, May 6, 2013 - link

    As an owner of two Bobcat systems (laptop/mini-itx), I don't think a 25% boost from Jaguar is going to get us into the realm of "good enough" cpu performance for general computing in Windows. The same goes for Intel unless Silvermont is significantly faster than Jaguar. I'm excited that Intel is finally bringing something interesting to the table, even if we end up two to three generations away from a good experience in Windows with their (and AMD's) mobile offerings. This sounds like it will make for a beastly dual core Android phone though, even at lower clocks.
  • jjj - Monday, May 6, 2013 - link

    Hilarious difference in attitude when it comes to Intel.
    Tegra 4 gets into phones by "aggressively limiting frequency." while Intel " Max clock speeds should be lower than what’s possible in a tablet, but not by all that much thanks to good power management. "
    Objectivity at it's best.
  • Homeles - Monday, May 6, 2013 - link

    Your scenario is a false equivalency.
  • Krysto - Monday, May 6, 2013 - link

    Is it? I wouldn't accuse Anand of "objectivity" when it comes to Intel, whether it's on purpose, or involuntary.
  • nunomoreira10 - Monday, May 6, 2013 - link

    The point is tegra 4 was not exactly made for phones while Intel was, for that you have tegra4i

    its not exacly nvidia fault, everybody complained that tegra 3 was lacking, now tegra 4 which is competitive consumes to much, atleast there is a choice.
  • Homeles - Monday, May 6, 2013 - link

    A15s are big cores in relation to its relatives. The only way to fit not 2, not 4, but *5* of them in a phone on 28nm is to downclock them agressively. Just like the only way to fit Ivy Bridge in a tablet is to downclock it agressively.

    Anand did point out that the "the only A15 SoCs we've seen have been very leaky designs optimized for high frequency," and that if power consumption were prioritized instead (which I believe Tegra 4i is supposed to be), it would be less of a blowout.

    It's silly getting defensive about stock ARM cores anyways. It's not an attack on Nvidia by saying their stock ARM cores aren't all too spectacular -- it's not like they poured blood, sweat and tears into making their A15s the best thing ever.

    Finally, Tegra 4 is on a process that is rather significantly inferior to Intel's 22nm process. You think Nvidia would have to downclock agressively if they were on a level playing field and using Intel's 22nm process? I sure don't. But jjj and others here feel the need to get defensive whenever songs of praise are being sung about Intel, even when it's well deserved.
  • extide - Tuesday, May 7, 2013 - link

    I am in agreeance with what you said, but I do believe Tegra 4i is Cortex A9, not A15 like Tegra 4.
  • Wilco1 - Tuesday, May 7, 2013 - link

    The Korean Galaxy S4 has a 1.8GHz Exynos Octa, Tegra 4 does 1.9GHz. In what way are these "aggressively downclocked"? They run at their maximum frequency!

Log in

Don't have an account? Sign up now