Earlier this week Intel had said that it expects the performance impact security updates designed to protect a computer against the Spectre and Meltdown vulnerabilities to be minimal on devices with recent Intel processors, but more noticeable on systems with older chips.
Now the company is sharing some numbers. We still don’t have any details about the performance impact on systems with chips that are more than a few years old, and Microsoft is only providing details for a few of its recent chips. But the company’s internal testing shows less than a 10-percent performance decline for at least some systems with 6th, 7th, and 8th-gen Core processors.
The company ran SYSmark 2014 SE, PCMark 10, 3DMark Sky Diver, and WebXPRT 2015 on systems with multiple configurations including models with Core i7-8700K (desktop), Core i7-8650U (mobile), Core i7-7920HQ (mobile), and Core i7-6700K (desktop) chips.
As expected, the performance impact of the security update varies depending on the workload.
Most computers scored at least 10 percent lower in the SYSMark 2014 SE responsiveness test, although oddly a Core i7-6700K model with Windows 7 and a hard drive actually saw a slight performance increase in that test, along with a few others.
For other tests, such as the 3DMark Sky Diver graphics benchmark, most systems Intel tested saw no difference in before and after scores at all.
Keep in mind that Intel is under pressure to show that the update will have a minimal impact on performance since the company doesn’t have any PC and server chips to sell that aren’t affected by the Spectre and Meltdown vulnerabilities at a hardware level. So software and microcode updates that change the way existing CPUs work (and which may slow down some operations) are the only way to go for now… and if Intel wants to keep selling its latest processors, the company needs to convince customers that they’re just as usable after the security patches as they were before.
So while there’s no reason to doubt the benchmarks posted by Intel, it’s interesting to note that the company has only shown results for a handful of chips so far and hasn’t posted actual benchmark numbers, just the percentage of the performance impact.
It’s also interesting to note that Microsoft says computers that are more than a few years old suffer the most from recent security updates… and Intel hasn’t shared any benchmarks for chips released before 2015.
I suspect we’ll see plenty of third-party benchmarks in the coming months that paint a more clear picture of the performance impact on Intel-powered computers, as well as the impact on those that use chips from rival AMD.
Looks like I am going to need to upgrade my i7-4790K, since it is 4th generation. This sucks. Maybe time to consider going Ryzen.
I would wait for hardware fixes for spectre and then buy Ryzen. I’m waiting for WPA3 before buying any wifi devices.
Don’t forget power. Intel and others are trying to ignore it but, we as users, need to figure that one out too.
Power isn’t in a position to take advantage of this in any major way. The only question is how much AMD benefits from the fallout.
Also, who’s to say there aren’t other security flaws in the Power architecture? All the focus has been on x86 for the same reason Windows systems are targeted by hackers — it’s by far the most widely used platform.
I’m interpreting jjj’s comment as power consumption.
In geeks we trust. Let’s wait for the users to show findings. The community as it were. Apple anyone? The point is, trust in tech giants might just be a bit low these days so who here is taking Intel’s word on this?
Okay, but if Intel is trivially caught with cooked benchmarking, it gets monumentally worse for them. There is no incentive for them to lie here. These are not massaged benchmarks for as yet unreleased chips to show them in the best light. If they claim 10% slowdown and the real figure is 20% and affects millions of processors in customer data centers around the world, they’ve opened themselves up for a world of hurt, both financially and through bad publicity.
There’s not a chance the lawyers would have let them do that.
Sorry, can’t trust Intel anymore. Waiting for someone else to benchmark.
Also beware benchmarks from hosting companies like Amazon Cloud, Microsoft Azure, and Google whatever. Their business is based on server X costing $Y, not $Y plus (ten, twenty, fifty?) percent.
Comments are closed.