SPEC - Single-Threaded Performance

Starting off with SPECint2017, we’re using the single-instance runs of the rate variants of the benchmarks.

As usual, because there are not officially submitted scores to SPEC, we’re labelling the results as “estimates” as per the SPEC rules and license.

We compile the binaries with GCC 10.2 on their respective platforms, with simple -Ofast optimisation flags and relevant architecture and machine tuning flags (-march/-mtune=Neoverse-n1 ; -march/-mtune=skylake-avx512 ; -march/-mtune=znver2).

While single-threaded performance in such large enterprise systems isn’t a very meaningful or relevant measure, given that the sockets will rarely ever be used with just 1 thread being loaded on them, it’s still an interesting figure academically, and for the few use-cases which would have such performance bottlenecks. It’s to be remembered that the EPYC and Xeon systems will clock up to respectively 3.4GHz and 4GHz under such situations, while the Ampere Altra still maintains its 3.3GHz maximum speed.

SPECint2017 Rate-1 Estimated Scores

In SPECint2017, the Altra system is performing admirably and is able to generally match the performance of its counterparts, winning some workloads, while losing some others.

SPECfp2017 Rate-1 Estimated Scores

In SPECfp2017 the Neoverse-N1 cores seem to more generally fall behind their x86 counterparts. Particularly what’s odd to see is the vast discrepancy in 507.cactuBSSN_r where the Altra posts less than half the performance of the x86 cores. This is actually quite odd as the Graviton2 had scored 3.81 in the test. The workload has the highest L1D miss rate amongst the SPEC suite, so it’s possible that the neutered prefetchers on the Altra system might in some way play a more substantial role in this workload.

SPEC2017 Rate-1 Estimated Total

The Altra Q80-33 ends up performing extremely well and competitively against the AMD EPYC 7742 and Intel Xeon 8280, actually beating the EPYC in SPECint, although it loses by a larger margin in SPECfp. The Xeon 8280 still holds the crown here in this test due to its ability to boost up to 4GHz across two cores, clocking down to 3.8, 3.7, 3.5 and 3.3GHz beyond 2, 4, 8 and 20 cores active.

The Altra showcases a massive 52% performance lead over the Graviton2 in SPECint, which is actually beyond the expected 32% difference due to clock frequencies being at 3.3GHz versus 2.5GHz. On the other hand, the SPECfp figures are only ahead of 15% for the Altra. The prefetchers are really amongst the only thing that come to mind in regards to these differences, as the only other difference being that the Graviton2 figures were from earlier in the year on GCC 9.3. The Altra figures are definitely more reliable as we actually have our hands on the system here.

While on the AMD system the move from NPS1 to NPS4 hardly changes performance, limiting the Altra Q80-33 from a monolithic setup to a quadrant setup does incur a small performance penalty, which is unsurprising as we’re cutting the L3 into a quarter of its size for single-threaded workloads. That in itself is actually a very interesting experiment as we haven’t been able to do such a change on any prior system before.

Test Bed and Setup - Compiler Options SPEC - Multi-Threaded Performance
Comments Locked

148 Comments

View All Comments

  • Wilco1 - Monday, December 21, 2020 - link

    Using Zen 2 is not correct since it uses much larger transistors. Using Kirin 990 5G density gives an estimate of 330mm^2 for Graviton 2. The size of N1 cores has been published for 7nm, so we know it is 1.4mm^2. You're right that PCIe lanes would add to it as well - assuming the PHYs have the same size as DDR PHYs at the same speed, 64 lanes would be about 12-15mm^2. That would increase it to about 365mm^2.
  • milli - Monday, December 21, 2020 - link

    Kirin 990 5G uses N7+. Altra uses N7.
    Not only is the process different but they're also totally different categories of products concerning transistor density. A mobile SOC can be very dense. It barely has any IO (which is not transistor dense). Also GPU, DPU, IMG, ... all are extremely dense.
    Kirin 990 5G is 90MTr/mm^2.
    No way a server class SOC is going to be more than 60MTr/mm2.
    Renoir = 62, Navi 21 = 52, Zen2 = 54, Vega 20 = 40, Navi 10 = 41.
    Ampere isn't going to magically break 60.

    "The size of N1 cores has been published for 7nm, so we know it is 1.4mm^2"
    Those are ARM numbers and that is only if you use high density libraries.
  • Wilco1 - Monday, December 21, 2020 - link

    Arm servers don't need high performance libraries - even mobile phones clock over 3 GHz using high density libraries. See https://images.anandtech.com/doci/13959/03_Infra%2... (note 3.1GHz and 1.4mm^2 with 1MB L2 on 7nm is ~100MT/mm^2)

    Using ~90MT/mm^2 for 7nm is reasonable since that is the reported density of recent 7nm chips (Kirin 990 5G is 91, 4G is 88 - the older 980 gets 93). Mobile SoCs already have a large amount of IO and analog logic and we are multiplying that amount by 3x.

    This shows how stupid it is to use high performance libraries in server chips - they don't need to run at 5GHz!
  • milli - Monday, December 21, 2020 - link

    We have different opinions but there's only one true fact: the die size is not disclosed. So anything anyone says is just a pure guess. You can't throw it around as fact.
  • milli - Monday, December 21, 2020 - link

    Navi 10/20 chips run at < 2Ghz and are 40MTr/mm. Just because Altra runs at 3.3Ghz, doesn't mean that it doesn't use HPL.
  • Josh128 - Friday, December 18, 2020 - link

    Exactly-- no way in hell this thing is just 350mm^2. The package is huge. Why would a 350mm^2 die need such a giant package?
  • Wilco1 - Friday, December 18, 2020 - link

    The package is only 16% larger than EPYC. Do you see any opportunity to reduce the huge number of pins? There are 8 memory channels plus full 128 PCIe lanes.
  • mode_13h - Sunday, December 20, 2020 - link

    Yes, the problem Altra Max will likely face is more memory bottlenecks. Also, I wonder if they'll have to dial clocks down, a little, to keep the power-efficiency numbers attractive.
  • Wilco1 - Monday, December 21, 2020 - link

    Altra Max drops max frequency to 3GHz, but it's not clear whether the TDP stays the same.
  • Gondalf - Friday, December 18, 2020 - link

    Are you sure :). Come on

Log in

Don't have an account? Sign up now