AMD has claimed that reports suggesting Windows 10's thread scheduler is responsible for lower-than-expected highly-threaded performance scores on its latest Ryzen processor family are inaccurate, pouring cold water on hopes a patch from Microsoft could further boost the chips' speed.
AMD has denied that there's anything wrong with the Windows 10 thread scheduler, dashing hopes that a patch from Microsoft could unlock further performance gains from Ryzen chips.
While reviews of the AMD Ryzen 1800X, 1700X, and 1700 Zen-architecture eight-core 16-thread processors have shown a considerable improvement in performance over their last-generation predecessors, there have been some statistical anomalies that reviewers have been scrabbling around to explain. One of these is a strange drop in performance when running Windows 10 compared to Windows 7, something some reports have claimed is evidence of an issue with the thread scheduler in Windows 10 failing to properly prioritise the physical cores over the virtual cores and spreading threaded tasks over the two four-core blocks that make up the chip - either or both of which would indeed harm performance. AMD, though, says that it has found no evidence to suggest that this is the case, and that any difference between Windows 10 and Windows 7 performance is likely to be down to the operating systems' internal architectures and not evidence of a scheduling issue.
'We have investigated reports alleging incorrect thread scheduling on the AMD Ryzen processor,
' AMD's Robert Hallock explains in a community update
published late last night. 'Based on our findings, AMD believes that the Windows 10 thread scheduler is operating properly for “Zen,” and we do not presently believe there is an issue with the scheduler adversely utilising the logical and physical configurations of the architecture.
'We have reviewed the limited available evidence concerning performance deltas between Windows 7 and Windows 10 on the AMD Ryzen CPU. We do not believe there is an issue with scheduling differences between the two versions of Windows. Any differences in performance can be more likely attributed to software architecture differences between these OSes.
In other words: according to AMD the Windows 10 thread scheduler is operating correctly and there's nothing for Microsoft to patch, leaving those who had hoped the Ryzen performance would be lifted further by a scheduler update - possibly even one which was originally due to be released in the cancelled February Patch Tuesday - disappointed. Elsewhere in the post, Hallock reports a flaw in the Sysinternals Coreinfo utility which generates incorrect topology data for Zen-architecture chips, data which have been used by some technology news outlets to support the theory of the Windows 10 scheduling issue. Upgrading to Coreinfo v3.31 or later, Hallock claims, will fix the bug.
Hallock has also detailed an interesting quirk of the X-family processors, the Ryzen 7 1800X and Ryzen 7 1700X: their temperature reporting features a fixed +20°C offset not present on non-X chips, meaning that a 1700X at the same temperature as a 1700 will report its temperature as 20°C higher. According to Hallock's explanation, the offset allows all Ryzen processors to share a single fan speed policy - though whether the offset will affect features like thermal throttling is not revealed.
Finally, Hallock addresses the issue of simultaneous multithreading (SMT), the technology which allows the eight-core processors to run 16 threads simultaneously. While some reviews have discovered that enabling SMT results in a drop in game frame rates, AMD's internal research is claimed to suggest the opposite. 'Based on our characterisation of game workloads, it is our expectation that gaming applications should generally see a neutral/positive benefit from SMT,
' Hallock claims, detailing games including Arma 3, Battlefield 1, Mafia III, Watch Dogs 2, Civilization VI, Hitman, Mirror's Edge Catalyst, and The Division as either having no performance change at all or benefiting from the presence of SMT.
Hallock's post can be read in full on the AMD community site