-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Hyper threading vs SMT performance comparison
- Thread starter neo3703
- Start date
Interesting to see gaming benchmarks on low-end SKUs (I had only seen productivity ones on higher-end CPUs), but not exactly a recent test (uses 10th Gen Intel and AMD Zen 2) so 2-3 generations behind.
SMT is actually the generic technical term for independent thread execution on each CPU core to enhance utilisation of available resources. Hyper-threading is Intel's marketing term for their SMT implementation, AMD doesn't use a trademark. Generally effectiveness of an SMT implementation is a function of how well the underlying core is designed for ST applications, and if they use aggressive out-of-order execution. A design that benefits greatly from SMT usually indicates a wasteful design in terms of ST efficiency (space and power) and one that doesn't benefit from it often means a more efficient design. There are also CPUs with 4-way or even 8-way SMT, although generally they are out of fashion in modern microarchitectures. These are all design trade-offs and double-edged swords, i.e. bigger SMT gains don't mean better at all. It's possible they could have extracted more performance from the same die space if they optimised their microarchitecture differently.
The alternative to SMT for higher resource utilisation in a CPU core is a better and more aggressive out-of-order execution that allows better instruction-level parallelism, which enhances ST performance as well as MT (SMT only enables thread-level parallelism and therefore only improves MT performance). This is generally the more modern approach in microarchitecture and is used in Apple/ARM designs, as well as Intel's new efficiency cores in 12th and 13th gen products.
SMT is actually the generic technical term for independent thread execution on each CPU core to enhance utilisation of available resources. Hyper-threading is Intel's marketing term for their SMT implementation, AMD doesn't use a trademark. Generally effectiveness of an SMT implementation is a function of how well the underlying core is designed for ST applications, and if they use aggressive out-of-order execution. A design that benefits greatly from SMT usually indicates a wasteful design in terms of ST efficiency (space and power) and one that doesn't benefit from it often means a more efficient design. There are also CPUs with 4-way or even 8-way SMT, although generally they are out of fashion in modern microarchitectures. These are all design trade-offs and double-edged swords, i.e. bigger SMT gains don't mean better at all. It's possible they could have extracted more performance from the same die space if they optimised their microarchitecture differently.
The alternative to SMT for higher resource utilisation in a CPU core is a better and more aggressive out-of-order execution that allows better instruction-level parallelism, which enhances ST performance as well as MT (SMT only enables thread-level parallelism and therefore only improves MT performance). This is generally the more modern approach in microarchitecture and is used in Apple/ARM designs, as well as Intel's new efficiency cores in 12th and 13th gen products.
Last edited: