Meltdown and Specter: Do They Really Impact PC Game Performance? – article

It has been described as one of the biggest PC security breaches we’ve ever seen. The front pages were dominated by news of dangerous security breaches inherent in the same hardware design of Intel’s microprocessors of the past 20 years, thus extending to millions of units in circulation and in use today. If your gaming PC is based on an Intel processor, then it is vulnerable to the newly discovered ‘Meltdown’ exploit. Side by side with Meltdown is another major security problem, dubbed Specter, which unlike Meltdown can plague your PC, tablet or smartphone regardless of whether its processor is Intel-branded or not.

So what’s the risk in layman’s terms? Essentially, Meltdown allows malware code to gain access to your CPU’s protected memory, an area that should be inaccessible. Sensitive data of any kind are therefore potentially accessible. Specter offers another avenue of access to your sensitive data, to the point that, although it is more difficult to update, it could create headaches for months or even years from now.

Fixes are already coming in a very short time, and many more will arrive soon, but the concern is that the performance of the CPUs can take a hard impact. Epic Games recently revealed the extent of the performance impact on its Fortnite gaming servers, via a graph showing a huge increase in peak CPU usage once Meltdown patches have been applied. The question is, what kind of medium extension will these patches have, as they are now, on PC gaming?

Covering the security holes generated by the two threats requires separate updates. First of all, Windows Update patches Microsoft operating systems by offering protection for Intel CPUs vulnerable to Meltdown by moving the CPU kernel to a separate virtual address space. In conjunction, a BIOS update on your motherboard aims to make the parallelized CPU prediction instructions less aggressive, making Specter less effective. Both of these updates can slow down your PC but the good news is that the overall impact measured so far in many benchmarks is on the order of 2 or 3 percent, while only the I / O operations of the fastest SSDs (such as NVMe) they seem to suffer considerably from slowdowns. In the enterprise sector the situation is completely different, with CPU virtualization being significantly compromised, as seen on Fortnite servers.

Win10 Not patched Win10 Meltdown Patch Win10 Meltdown + Microcode Patch
The Witcher 3, Ultra, No HairWorks 139.8fps 128.3fps 126.6fps
Rise of the Tomb Raider, Very High, DX12 121.6fps 117.2fps 121.6fps
Far Cry Primal, Ultra 128.4fps 127.0fps 126.2fps
Crysis 3, Very High 129.3fps 129.2fps 126.8fps
Ashes of the Singularity, CPU Test 35.3fps 35.5fps 35.6fps
Assassin’s Creed Unity, Ultra High 131.1fps 131.2fps 130.3fps

But we still wanted to measure the impact of these patches ourselves, so we chose the Core i5 8400 as the subject of our tests, comparing our previously performed benchmarks with the same titles running on the same configuration, first with only the Meltdown patch for Windows applied, and then also with Specter-specific firmware active. Asus was surprisingly quick to deliver updates and we used a Z370 chipset-based Maximus X Hero for our test system.

The table above confirms the universally accepted view that gaming is not compromised by more than a negligible percentage in most situations. However, we have gradually experimented over the last few years how gaming CPUs should be tested and we are still not totally satisfied with the methodology. Discovering the areas of a game where a CPU bound occurs is already no easy feat, even with the current technique of coupling the processor to an overclocked Titan X Pascal and setting 1080p as the resolution. The basic idea is to push CPU performance and memory bandwidth to the maximum, ahead of the GPU which is usually the main performance limiting factor in games. But despite everything, even here we have found that different areas and in-game rendering scenarios can reduce or widen the difference between the two processors.

That said, five of the six tests here can push Intel processors with four cores and four threads to 100 percent utilization (with the exception of Far Cry Primal, which relies heavily on single core IPC), but c ‘is much more headroom for the hexa-core i5 8400. What we found is that, even with both patches in place, some games make no difference: Ashes of the Singularity’s very heavy CPU benchmark remains consistent across all. and three test runs, as happens in our run at Notre Dame in Assassin’s Creed Unity, where the differences fall within the margin of error. CPU-hungry Crysis 3 shows no signs of abating with Meltdown while with the Specter patch applied only 1 percent performance is lost. Far Cry Primal? It loses 1fps with every security patch you install.

The Meltdown patch caused a 3.6 percent reduction in performance when running our test in the Rise of the Tomb Raider Geothermal Valley. However, our test of The Witcher 3, which impacts heavily on storage and memory bandwidth, raised the deficit to 9.4 percent with the BIOS update to cover the Specter flaw. Interestingly, by running the same test again on a (less powerful) Core i7 4790K, we only had a 3 percent impact (but at that time there was no security firmware update available for our Z97 motherboard, which it means we could only test the impact caused by the Meltdown patch.

What we see at the moment is just the tip of the iceberg that could anticipate a long series of patches, but to date there is both good and bad news.

The good news is that performance is being maintained: our tests artificially push CPU performance to the limit in a world where the GPU is the main limiting factor in games. And even with this artificiality, only one game is the protagonist of a palpable impact on performance but only in a part of what in practice is a really well optimized game, which we have chosen for our stress tests for the CPU. Much of The Witcher 3 runs much smoother.

The bad news is related to Specter. First of all, the full extent of the problem and its possible exploitation could mean that it will be a problem for years to come, but the real concern lies in the micro-code that manufacturers will release for older systems, and consequently the firmware. manufacturers may or may not release for older generation CPUs. In a world where Sandy Bridge 2011 generation processors such as the Core i5 2500K are still widely used in gaming PCs, should users take the opportunity to upgrade? Alternatively, with so many of these CPUs still in use, maybe Intel and motherboard makers shouldn’t make sure these systems are as secure as possible? We will continue to monitor the situation closely over the coming weeks and months until the full extent of the problem is clear, and consequently the remedies adopted to resolve this regrettable incident.