Q What is the effective CPU speed index?

A A measure of CPU speed geared towards typical users. Intel i9-9900K ≈ 100%.

---

Gaming is by far the most demanding CPU activity most users undertake. CPU Effective Speed (average bench) is calibrated to estimate differences in EFps between PCs. We publish EFps data, with video footage, for hundreds of PC configurations using replicable gameplay in the world’s most popular games. We use testable, verifiable data to ensure that users can easily compare their own results against ours

The first few threads

Desktop tasks such as surfing the web with multiple tabs, watching videos and listening to music rarely use more than four threads. Very few of today's popular games benefit from more than six threads. There is not much difference in fps between a 4 thread i3-9100F and an overclocked 16 thread Ryzen 2700X, in fact, the 9100F is 10% faster. CPU latency has more impact than core count. Gaming performance is primarily influenced by the GPU rather than the CPU.

More threads

Higher thread counts are useful for workstation tasks such as cryptography and virtual machine hosting. If dedicated GPU hardware (NVENC/QuickSync) is not an option, streamers and video producers can, suboptimally, use additional CPU cores for encoding. On the 27th of March, 2020 UserBenchmark's six core database server averaged 10,000 queries per second with a CPU load of just under 10%. High data throughput is more sensitive to latency than core count.

CPU memory latency

Lower latency results in quicker data retrieval and faster computations. CPU architectures exhibit different latency characteristics. The Zen CPU architecture has significantly higher latency (60 ns) than Skylake (45 ns) which is partly why Skylake delivers superior gaming (fewer frame drops) and higher database throughput despite having comparable processing cores.

Updates

Our indices are based on today’s performance requirements rather than "Moar Core" marketing visions.

July 2019

We reduced the contribution of thread counts higher than eight. The 32-core AMD 2990WX moved from first position to 48th. Meanwhile the 8-core Intel 9900K moved from 7th to first position.

Smear campaign

Within hours of the July 2019 changes, an army of anonymous call center shills, posing as AMD "fans", accused UserBenchmark of impartiality. We are not affiliated with any brands. We act solely in the interest of our users.

November 2020 (pending)

During the notable Ryzen 5000 release event, as well as discussing the importance of single core performance and CPU latency, AMD provided benchmarks for 10 games of their choice. The upcoming 5900X averaged 26% faster than the 3900XT and 7% faster than the Intel 10900K. These figures imply that the 10900K is around 19% faster than the 3900XT. Our effective speed only puts the Intel 10900K 14% ahead of the 3900XT, so according to AMD's figures, UserBenchmark overestimates Ryzen 3000 by ≈ 5%. Meanwhile, AMD "fans" continue to smear UserBenchmark via an army of anonymous accounts on reddit, youtube and forums.