The Cortex-X2: More Performance, Deeper OoO

We first start off with the Cortex-X2, successor to last year’s Cortex-X1. The X1 marked the first in a new IP line-up from Arm which diverged its “big” core offering into two different IP lines, with the Cortex-A sibling continuing Arm’s original design philosophy of PPA, while the X-cores are allowed to grow in size and power in order to achieve much higher performance points.

The Cortex-X2 continues this philosophy, and further grows the performance and power gap between it and its “middle” sibling, the Cortex-A710. I also noticed that throughout Arm’s presentation there were a lot more mentions of having the Cortex-X2 being used in larger-screen compute devices and form-factors such as laptops, so it might very well be an indication of the company that some of its customers will be using the X2 more predominantly in such designs for this generation.

From an architectural standpoint the X2 is naturally different from the X1, thanks in large part to its support for Armv9 and all of the security and related ISA platform advancements that come with the new re-baselining of the architecture.

As noted in the introduction, the Cortex-X2 is also a 64-bit only core which only supports AArch64 execution, even in PL0 user mode applications. From a microarchitectural standpoint this is interesting as it means Arm will have been able to kick out some cruft in the design. However as the design is a continuation of the Austin family of processors, I do wonder if we’ll see more benefits of this deprecation in future “clean-sheet” big cores designs, where AArch64-only was designed from the get-go. This, in fact, is something that's already happening in other members of Arm's CPU cores, as the new little core Cortex-A510 was designed sans-AArch32.

Starting off with the front-end, in general, Arm has continued to try to improve what it considers the most important aspect of the microarchitecture: branch prediction. This includes continuing to run the branch resolution in a decoupled way from the fetch stages in order to being able to have these functional blocks be able to run ahead of the rest of the core in case of mispredicts and minimize branch bubbles. Arm generally doesn’t like to talk too much details about what exactly they’ve changed here in terms of their predictors, but promises a notable improvement in terms of branch prediction accuracy for the new X2 and A710 cores, effectively reducing the MPKI (Misses per kilo instructions) metric for a very wide range of workloads.

The new core overall reduces its pipeline length from 11 cycles to 10 cycles as Arm has been able to reduce the dispatch stages from 2-cycles to 1-cycle. It’s to be noted that we have to differentiate the pipeline cycles from the mispredict penalties, the latter had already been reduced to 10 cycles in most circumstances in the Cortex-A77 design. Removing a pipeline stage is generally a rather large change, particularly given Arm’s target of maintaining frequency capabilities of the core. This design change did incur some more complex engineering and had area and power costs; but despite that, as Arm explains in, cutting a pipeline stage still offered a larger return-on-investment when it came to the performance benefits, and was thus very much worth it.

The core also increases its out-of-order capabilities, increasing the ROB (reorder buffer) by 30% from 224 entries to 288 entries this generation. The effective figure is actually a little bit higher still, as in cases of compression and instruction bundling there are essentially more than 288 entries being stored. Arm says there’s also more instruction fusion cases being facilitated this generation.

On the back-end of the core, the big new change is on the part of the FP/ASIMD pipelines which are now SVE2-capable. In the mobile space, the SVE vector length will continue to be 128b and essentially the new X2 core features similar throughput characteristics to the X1’s 4x FP/NEON pipelines. The choice of 128b vectors instead of something higher is due to the requirement to have homogenous architectural feature-sets amongst big.LITTLE designs as you cannot mix different vector length microarchitectures in the same SoC in a seamless fashion.

On the back-end, the Cortex-X2 continues to focus on increasing MLP (memory level parallelism) by increasing the load-store windows and structure sizes by 33%. Arm here employs several structures and generally doesn’t go into detail about exactly which queues have been extended, but once we get our hands on X2 systems we’ll be likely be able to measure this. The L1 dTLB has grown from 40 entries to 48 entries, and as with every generation, Arm has also improved their prefetchers, increasing accuracies and coverage.

One prefetcher that surprised us in the Cortex-X1 and A78 earlier this year when we first tested new generation devices was a temporal prefetcher – the first of its kind that we’re aware of in the industry. This is able to latch onto arbitrary repeated memory patterns and recognize new iterations in memory accesses, being able to smartly prefetch the whole pattern up to a certain depth (we estimate a 32-64MB window). Arm states that this coverage is now further increased, as well as the accuracy – though again the details we’ll only able to see once we get our hands on silicon.

In terms of IPC improvements, this year’s figures are quoted to reach +16% in SPECint2006 at ISO frequency. The issue with this metric (and which applies to all of Arm’s figures today) is that Arm is comparing an 8MB L3 cache design to a 4MB L3 design, so I expect a larger chunk of that +16% figure to be due to the larger cache rather than the core IPC improvements themselves.

For their part, Arm is reiterating that they're expecting 8MB L3 designs for next year’s X2 SoCs – and thus this +16% figure is realistic and is what users should see in actual implementations. But with that said, we had the same discussion last year in regards to Arm expecting 8MB L3 caches for X1 SoCs, which didn't happen for either the Exynos 2100 nor the Snapdragon 888. So we'll just have to wait and see what cache sizes the flagship commercial SoCs end up going with.

In terms of the performance and power curve, the new X2 core extends itself ahead of the X1 curve in both metrics. The +16% performance figure in terms of the peak performance points, though it does come at a cost of higher power consumption.

Generally, this is a bit worrying in context of what we’re seeing in the market right now when it comes to process node choices from vendors. We’ve seen that Samsung’s 5LPE node used by Qualcomm and S.LSI in the Snapdragon 888 and Exynos 2100 has under-delivered in terms of performance and power efficiency, and I generally consider both big cores' power consumption to be at a higher bound limit when it comes to thermals. I expect Qualcomm to stick with Samsung foundry in the next generation, so I am admittedly pessimistic in regards to power improvements in whichever node the next flagship SoCs come in (be it 5LPP or 4LPP). It could well be plausible that we wouldn’t see the full +16% improvement in actual SoCs next year.

2022 Generation: Moving Towards Armv9 The Cortex-A710: More Performance with More Efficiency


View All Comments

  • RSAUser - Wednesday, May 26, 2021 - link

    Basically interesting for cases when you don't want to add an A73, e.g. It's pretty big news in the watch space where it's been the same 4/5yo architecture for a very long time. Reply
  • mode_13h - Thursday, May 27, 2021 - link

    > It's pretty big news in the watch space

    I'm actually surprised people are even using A55s in smartwatches, or that ARM is targeting the A510 at them. I'd figured the most they could get away with would be the A35.

    I guess pairing a couple A55s with some A35s might be a way to get responsiveness *and* battery life. Is that something people do?
  • mode_13h - Wednesday, May 26, 2021 - link

    It'd be interesting to see how efficient the A73 would be, if you dropped its clock to match the A510's performance. Reply
  • AntonErtl - Thursday, May 27, 2021 - link

    Yes. ARM gives some flowery wordings for the lower performance of the A510 compared to A73 (and Andrei reworded in the way ARM wants us to think: "very similar IPC and frequency capabilities whilst consuming a lot less power"; looking at the numbers given by ARM, the A510 has >20% less performance than the A73, at 35% less power. The DVFS stuff I have seen makes me expect that the A73 has the same or lower power at the same performance, if you lower the clock by 20% (or whatever the slowness factor of the A510 is).

    Andrei already showed us in his Exynos 9820 review that the A75 has better Perf and Perf/W for nearly all of the performance range of the A55. So I find it surprising that ARM went for another in-order design for the little core of ARMv9, instead of something like an ARMv9-enabled A75. For me it will certainly be an interesting microarchitecture to study, but I guess it will take some time until it appears in some Odroid or Raspi board.
  • mode_13h - Saturday, May 29, 2021 - link

    > the A75 has better Perf and Perf/W for nearly all of the performance range of the A55.
    > So I find it surprising that ARM went for another in-order design for the little core of ARMv9

    You're forgetting about PPA, though. The A510 is probably a lot smaller (ISO-process) than the A75.

    > I guess it will take some time until it appears in some Odroid or Raspi board.

    Look for A76-enabled SBCs late this year or early next. Rockchip's RK3588 will have 4x A76.

    Raspberry Pi will probably be stuck on A72 or A73 for a couple more generations, since they plan to stay on 28 nm, for a while. Meanwhile, the Allwinner SoC in ODROID's N2 is made on 12 nm.
  • AntonErtl - Sunday, May 30, 2021 - link

    Looking at the Exynos 9820 die shot, te A55 is ~3.4 times smaller than the A75, but it also has ~3.4 times lower top performance and a similar factor at the lowest common perf/W point, and from the looks of the line, in between. I doubt that the A510 is better in perf/area. But maybe it's the difference that ARM is claiming between the workloads Andrei used for evaluating performance (SPEC CPU2006) and what the A55 and A510 are doing in practice; if they mainly wait for peripherals, I can believe that their performance does not matter much.

    Thanks for the info on SBCs to be expected.
  • Wereweeb - Wednesday, May 26, 2021 - link

    I'll ignore all the warfare in the comments, and just say this: imagine a 16-'core' A510 SoC. Sorry. Reply
  • mode_13h - Wednesday, May 26, 2021 - link

    So, if you built a HPC CPU with A510 @ one core per complex, 2x 128-bit SVE2, and max L2 cache, how would area-efficiency (PPA) and power-efficiency (PPW) compare with a V1-based chip on the same node?

    Let's assume the workload has enough concurrency to scale up to all the A510 cores, and that there's enough ILP that the A510's lack of OoO isn't a significant impediment.
  • Shakal - Thursday, May 27, 2021 - link

    Pardon my ignorance but what exactly is an "Alternate path predictor"? They mention that for the X2 core but I've not found any reference to what it is. I've heard of path based predictors but how does the alternate come into play? Reply
  • - Thursday, May 27, 2021 - link best website for techy terms. best for fitness and weight. is a place for weight loss, weight gain, diet, body building.

Log in

Don't have an account? Sign up now