Task vs ValueTask performance

ackermann.cs

using System;
using System.Threading.Tasks;
using BenchmarkDotNet.Attributes;
using BenchmarkDotNet.Running;

namespace ValueTaskBenchmarks
{
    // Uses the Ackermann function to measure performance of task and friends
    // c.f. https://en.wikipedia.org/wiki/Ackermann_function

    [MemoryDiagnoser]
    public class Ackermann
    {
        // Ackermann(3, 3) = 61, total recursive calls = 2432
        [Params(1, 2, 3)]
        public int n;
        [Params(1, 2, 3)]
        public int m;

        [Benchmark(Baseline = true)]
        public int Baseline()
        {
            return Ackermann(m, n);

            int Ackermann(int m, int n) => (m, n) switch
            {
                (0, _) => n + 1,
                (_, 0) => Ackermann(m - 1, 1),
                _ => Ackermann(m - 1, Ackermann(m, n - 1)),
            };
        }

        [Benchmark]
        public ValueTask<int> ValueTask()
        {
            return Ackermann(m, n);

            async ValueTask<int> Ackermann(int m, int n) => (m, n) switch
            {
                (0, _) => n + 1,
                (_, 0) => await Ackermann(m - 1, 1),
                _ => await Ackermann(m - 1, await Ackermann(m, n - 1)),
            };
        }

        [Benchmark]
        public Task<int> Task()
        {
            return Ackermann(m, n);

            async Task<int> Ackermann(int m, int n) => (m, n) switch
            {
                (0, _) => n + 1,
                (_, 0) => await Ackermann(m - 1, 1),
                _ => await Ackermann(m - 1, await Ackermann(m, n - 1)),
            };
        }
    }

    class Program
    {
        static void Main()
        {
            BenchmarkRunner.Run<Ackermann>();
        }
    }
}

results.md

BenchmarkDotNet=v0.12.0, OS=Windows 10.0.18363
Intel Core i7-8665U CPU 1.90GHz (Coffee Lake), 1 CPU, 8 logical and 4 physical cores
.NET Core SDK=3.1.100-preview3-014645
  [Host]     : .NET Core 3.1.0 (CoreCLR 4.700.19.53102, CoreFX 4.700.19.55104), X64 RyuJIT
  DefaultJob : .NET Core 3.1.0 (CoreCLR 4.700.19.53102, CoreFX 4.700.19.55104), X64 RyuJIT

Method	n	m	Mean	Error	StdDev	Median	Ratio	RatioSD	Gen 0	Gen 1	Gen 2	Allocated
Baseline	1	1	8.857 ns	0.2074 ns	0.2388 ns	8.866 ns	1.00	0.00	-	-	-	-
ValueTask	1	1	125.333 ns	3.9694 ns	10.6635 ns	121.942 ns	15.05	1.69	-	-	-	-
Task	1	1	63.987 ns	1.3532 ns	1.3897 ns	63.966 ns	7.22	0.23	-	-	-	-
Baseline	1	2	26.763 ns	0.4669 ns	0.4139 ns	26.648 ns	1.00	0.00	-	-	-	-
ValueTask	1	2	476.952 ns	9.3495 ns	8.7456 ns	477.926 ns	17.84	0.35	-	-	-	-
Task	1	2	236.332 ns	4.6040 ns	5.1173 ns	235.215 ns	8.81	0.20	-	-	-	-
Baseline	1	3	189.247 ns	3.7806 ns	7.0076 ns	189.080 ns	1.00	0.00	-	-	-	-
ValueTask	1	3	3,812.292 ns	75.6200 ns	80.9126 ns	3,804.229 ns	19.82	0.86	-	-	-	-
Task	1	3	2,307.302 ns	45.5397 ns	52.4436 ns	2,291.761 ns	12.02	0.61	0.3777	-	-	1584 B
Baseline	2	1	11.248 ns	0.2124 ns	0.1986 ns	11.321 ns	1.00	0.00	-	-	-	-
ValueTask	2	1	183.435 ns	1.7635 ns	1.4726 ns	183.324 ns	16.29	0.30	-	-	-	-
Task	2	1	89.209 ns	1.7744 ns	1.6598 ns	89.951 ns	7.93	0.17	-	-	-	-
Baseline	2	2	49.115 ns	0.6867 ns	0.5734 ns	48.889 ns	1.00	0.00	-	-	-	-
ValueTask	2	2	904.934 ns	17.1505 ns	17.6123 ns	907.298 ns	18.48	0.49	-	-	-	-
Task	2	2	491.162 ns	9.7270 ns	10.4077 ns	493.575 ns	10.02	0.21	-	-	-	-
Baseline	2	3	1,238.368 ns	24.3141 ns	32.4586 ns	1,229.467 ns	1.00	0.00	-	-	-	-
ValueTask	2	3	20,072.025 ns	393.1447 ns	551.1340 ns	20,180.771 ns	16.22	0.71	-	-	-	-
Task	2	3	13,368.605 ns	264.7312 ns	247.6297 ns	13,256.192 ns	10.78	0.34	4.7455	-	-	19872 B
Baseline	3	1	14.554 ns	0.2493 ns	0.2332 ns	14.621 ns	1.00	0.00	-	-	-	-
ValueTask	3	1	254.058 ns	5.0642 ns	7.8843 ns	255.125 ns	17.80	0.58	-	-	-	-
Task	3	1	127.259 ns	2.4915 ns	3.0598 ns	128.881 ns	8.69	0.23	-	-	-	-
Baseline	3	2	77.015 ns	0.9136 ns	0.8099 ns	76.799 ns	1.00	0.00	-	-	-	-
ValueTask	3	2	1,510.296 ns	28.8578 ns	28.3422 ns	1,521.415 ns	19.59	0.45	-	-	-	-
Task	3	2	860.356 ns	16.3538 ns	16.0616 ns	862.778 ns	11.19	0.26	0.0515	-	-	216 B
Baseline	3	3	7,396.098 ns	145.2481 ns	155.4139 ns	7,453.833 ns	1.00	0.00	-	-	-	-
ValueTask	3	3	92,352.103 ns	1,839.9734 ns	1,721.1122 ns	92,585.132 ns	12.50	0.40	-	-	-	-
Task	3	3	65,442.848 ns	1,305.4800 ns	1,503.3927 ns	65,409.296 ns	8.84	0.27	30.2734	-	-	126864 B