AMD’s New CPU Drivers are Supposed to Improve Execution, Yet Our Tests Show Generally

AMD's new CPU drivers
Written by Faiza Iftikhar

Recently, AMD released new chipset drivers for its Ryzen CPU platforms. In the wake of hearing about updates to the Zen 2 architecture and discuss what’s in store from the Upcoming Ryzen 3000 CPUs, I was extremely curious to see whether the new drivers would influence execution in a positive way. To understand what’s supposed to be going on, however, there’s some background information to cover.

AMD's new CPU drivers

The Windows 10 scheduler recently was to some degree unaware of the lower level details of your CPU cores. It had some essential information on physical versus logical cores (eg, Hyper-Threading and SMT make a moment ‘logical’ center that shares assets with a second core), yet that was about it. The issue with this approach is that as CPUs turned out to be increasingly complex, not all cores are created an equivalent.

Specifically for AMD, it’s Ryzen CPUs have two CCXes (Core Complexes), each with half of the L3 cache. Conceivably, if an application is assigned to a center on the first CCX and after that spawns a thread that gets allotted to a core on the second CCX, there will be higher latencies for getting data across the CCX boundary. Things get potentially even worse with Threadripper, where there are two separate chips in a bundle, each with two CCXes—addition two separate memory controllers from each chip. Pulling data from the other chip’s L3 cache, or from RAM associated with the other Chip’s memory controller, will be slower than getting access to local data.

AMD, it's Ryzen CPUs

The Windows 10 May 2019 update, joined with the new chipset driver from AMD, is supposed to improve the procedure scheduling so that strings from the similar application are bound to keep running on the same CCX, in this manner decreasing the opportunity for higher latency memory and cache accesses. And that brings us to the testing and outcomes.

I’ve been refreshing all my CPU benchmarks for the pending launch of AMD’s Ryzen 3000. May 2019 update in hand, I prepared all my testbeds and began benchmarking. But this new chipset driver could potentially refute those earlier outcomes! So yesterday I downloaded the drivers and retested the Ryzen 7 2700X. Here’s the summary of outcomes, appearing the “new” vs. the “old” ( or only a week or so prior), alongside the percent tweak in execution. What I don’t have is a full set of outcomes from before the May 2019 update, and it’s possible a large portion of the fixes for thread scheduling are as of now handled okay by Microsoft’s built-in drivers. Enough discussion, let’s let give a chance to benchmarks to tell the story:

CPU benchmarks

Wow. Um, that’s not by any stretch of the imagination exciting. There’s always a touch of wiggle room in benchmarks, which is the reason each test is run numerous times. Anything that shows in excess of a 1% drop in performance was kept running at least 4 times, and the above table demonstrates the best results achieved. With the new chipset driver installed, the execution was somewhat lower in most by far of CPU benchmarks, with PCMark 10 faring particularly inadequately. Power use was additionally down slightly (a good thing, kind of), not that the deltas truly matter that much.

Looking at the gaming outcomes, the majority of these are inside of margin of error (1%), yet Far Cry 5 shows a slightly bigger dip while Assassin’s Creed Odyssey demonstrates a relatively noteworthy 6% improvement.

But generally? The progressions are to a great extent simply noise. The multithreaded execution is unchanged (0.04%), while the total set of benchmarks demonstrates a miniscule 0.6% drop. Again, this could be on the ground that the Windows 10 May 2019 update previously took care of most of these changes, or it might just be that the string scheduling simply doesn’t make a difference that much. Or maybe this will have a greater amount of an effect on Ryzen 3000, and not make any difference such a great amount with existing Ryzen CPUs.

I’m really happy there were no massive changes, as it implies I don’t need to retest all the CPUs yet again. Be that as it may, it sure would have been nice to see a palpable tweak in performance. I’d recommend you just keep whatever chipset drivers you’re as of now running in the meantime—or upgrade for any bug fixes, however, don’t anticipate a suddenly faster PC.

About the author

Faiza Iftikhar

Leave a Comment