SiSoftware Sandra 20/20/11 (2020 R11) Update – Cache Bandwidth & Latency measurements

We are pleased to release R11 (version 30.80) update for Sandra 20/20 (2020) with the following changes:

Sandra 20/20 (2020) Press Release

Benchmark: CPU Cache & Memory Latency

• Increased entropy for in-page & full random access patterns.
• Worst-case latency reported not average (as used by memory bandwidth).
• TLB optimisations for in-page random access pattern. We recommend using large (2MB) pages in order to reduce TLB effects on cache latencies.
• Clock timing optimisations when power state fluctuates during test
• See images (below) that illustrate how the changes affect the reported data/code cache latencies

CPU & Memory Latency – Data – In-Page Random Access

CPU & Memory Latency – Code – In-Page Random Access

Benchmark: GPGPU Cache & Memory Latency (CUDA, OpenCL, DirectX Compute)

• Worst-case latency reported not average (as used by memory bandwidth).
• Uses page size when reported by the GPU (4kB-4MB page size)
• Uses 64-bit timers where available.
• Uses cache-line size when reported by the GPU (32-256-byte line size)
• We recommend using CUDA for most accurate (native clock timing) results.
• See images (below) that illustrate how the changes affect the reported data cache bandwidths

CPU & Memory Bandwidth – Data

Known Issues:

• CPU Crypto Benchmark:
  • AES HWA acceleration not used when available. [20/20 R12 fix]
  • VAES / AVX2 HWA acceleration may not be used on AMD Zen3. [20/20 R12 fix]
  • VIA PadLock AES/SHA HWA acceleration no longer supported. [use older 20/20 R10 or earlier]
• GP-GPU Processing Benchmark:
  • Lower than expected aggregate score when GPU does not support FP16. [20/21 fix]
• CPU Support:
  • No native support for AMD Zen3 due to lack of documentation. [20/21 fix]

Reviews using Sandra 20/20:

• SiSoftware
  • AMD CPUs/GP-GPUs
    • AMD Radeon RX 6000 (RDNA2, Navi2) Review & Benchmarks – GPGPU Performance
    • AMD Ryzen 9 5950X (Zen3) – CPU 16-core/32-thread Performance
    • AMD Ryzen 5 5600X (Zen3) – CPU 6-core/12-thread Performance
    • AMD Ryzen 7 5800X (Zen3)  – CPU 8-core/16-thread Performance
  • nVidia GP-GPUs
    • nVidia 3090, 3080 RTX: Ampere GPGPU performance in CUDA and OpenCL
    • nVidia RTX 2080Ti: Turing GPGPU performance in CUDA and OpenCL
  • Microsoft
    • Performance Impact of Hyper-V virtualisation (Windows 10 Pro) – Compute
  • Intel CPUs / APUs
    • Intel DG1 Discrete (Iris Xe Max Gen12) Review & Benchmarks – GPGPU Performance
    • AVX512-IFMA(52) Improvement for IceLake and TigerLake
    • Intel Core Gen11 TigerLake ULV (i7-1165G7) Review & Benchmarks – CPU AVX512 Performance
    • Intel Iris Plus G7 Gen12 XE TigerLake ULV (i7-1165G7) Review & Benchmarks – GPGPU Performance
• BabelTechReviews
  • The RTX 3070 FE Arrives at $499 to Match the RTX 2080 Ti’s Performance
  • The RTX 3090 Founders Edition Performance Revealed – 35+ Games, SPEC & Workstation & GPGPU Benchmarked
  • The RTX 3080 Arrives – Ampere Performance Revealed – 35+ Games & GPGPU Benchmarked
  • The EVGA Z490 FTW Motherboard Review – the Long Road to a Stable Overclock
  • The T-FORCE 128GB SPARK RGB USB 3.2 Flash(y) Drive Review

Update & Download

Commercial version customers can download the free updates from their software distributor; Lite users please download from your favourite download site.

Download Sandra Commercial (Pro/Biz/Eng/Ent)

Download Sandra Lite