The current high-performance low power 5W Intel processors are part of Amber Lake family, which will be followed by Comet Lake in 2020, and Tiger Lake in 2021 according to a roadmap leaked a few months ago.
It looks a little too early for this, but NotebookCheck.net reports an unnamed Intel Tiger Lake Y-series processor has showed up in UserBenchmarks website.
The system is said to feature an “Intel TigerLake Y LPDDR4x T4 Crb” motherboard, with an “Intel Tiger Lake Client Platform” coupled with 12GB DDR4 memory. The processor is only named “Intel 000” and comes with 4 cores, 8 threads having a base clock of 1.2 GHz, and an (average) Turbo clock of 2.9 GHz. The SoC also comes with a Gen12 Intel UHD Graphics, but the computer also features a much faster AMD RX 580 graphics cards so bear this in mind if you’re looking at graphics benchmarks.
If true and accurate, the CPU performance is really impressive as it matches the 45W Intel Core i9-8950HK in single-core benchmarks and the 45W Intel Core i7-8850H in quad-core tests based on results from UserBenchmark website. This is all the more surprising that the CPU was throttled at 83% by Windows according to the benchmark.
Another way to look at the results is to check them out against Intel Core i7-8500Y dual-core/quad-thread Amber Lake-Y processor. That’s tricky because there are massive variations between systems with the same processor, even for single-threaded tests. So I took an average system.
Viewed that way it looks reasonable with around 20% single thread performance increase, and twice the speed in quad-core benchmarks with the Tiger Lake processor having four actual cores instead of four threads for the Amber Lake processor.
Intel Tiger Lake-Y processor will be manufactured using a 10 nm process, and according to the aforelinked roadmap really have a 5W TDP. It’s unclear what Intel may have done to achieve that kind of performance with such a low power budget.
Jean-Luc started CNX Software in 2010 as a part-time endeavor, before quitting his job as a software engineering manager, and starting to write daily news, and reviews full time later in 2011.
Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress
2.9 appears to be the avg turbo — max single-core turbo is 3.7.
The single-core aggregate score for the Tiger Lake is 121. The corresponding for 8950HK is 142 (peak bench), for a max turbo of 4.8.
121 / 142 = .85
37 / 48 = .77
85 / 77 = 1.1, or IOW a 10% IPC advancement.
If the GPU supports tensorflow to run NN will inferences, then this would be a candidate to compete with Jetson nano. They seem quite similar.
It is interesting to note on the leaked roadmap that even in 2021 most chips will still be etched at 14nm, like the skylake in 2015, thus no major lithography update in 6 years. This could be the definitive end of the tick/tock schedule. Maybe this will force semiconductor makers to try to switch to other compounds even if it means 20 years of research before reaching the same level of integration we have today.
The tick/tock/tock/tock (as last known) cycle is done for. Intel have finally figured it out that their future lies in chiplets, which is not at all ironical..