Here’s how much the Meltdown and Spectre fix hurt my Surface Book performance - vankirkbutiournin

Every bit a performance junkie, I'm less afraid all but the security measures risks of the Spectre and Meltdown vulnerabilities—later altogether, there are no known exploits busy today—than I am nearly a execution hit from the fixes.

And from what I'm seeing, my concerns are bonded.

My sole experience with a fully updated platform then far is with Microsoft's unconventional Surface Book. It's based on an Intel "Skylake" Core i7-6700U and has 16GB of LPDDR3 and a 512GB Samsung 950 Pro NVMe drive. The Surface Holy Scripture is running the 64-fleck Windows 10 Pro Fall Creator's Update.

I basically horde the machine all week at CES, and happening Friday cockcro when I fired IT up at home, I constitute that Microsoft had pushed outer two pairs of firmware updates that address the Spectre and Meltdown vulnerabilities.

I'll admit, I'm still trying to play catch informed just what the hell on earth is passing on with Spectre and Meltdown, but this was a great chance to run before and later benchmarks on a output machine.

I know from reading Steve William Walton's write-up at TechSpot that the performance of games and most Central processor-intensive apps doesn't modify. But Walton found memory board read/write performance to personify an issue, so this was my first tip of investigation on the Surface Book. I started my tests with a not-yet-black-and-white machine.

Before the Spectre and Meltdown fix

Spectre and Meltdown performance hit — CrystalDiskMark 5.5.0 performance results connected a Microsoft Skin-deep Book before being patched for Meltdown and Apparition.

I started aside running the default mental testing for CrystalDiskMark 5.5.0. It's the slightly older version, but the results are still sound for two areas: puny 4K reads and writes as swell as 4K reads and writes using a queue depth of 32. To the left you'll find the consequence before Surface Firmware 91.1926.768.0 and 90.1837.256.0 were installed.

I made three repeated runs with two to five minutes of rest time in between to let the SSD return to normal temperature. SSDs, as you may know, canful slow when heated. This is 1 particular result, but representative of the results I saw.

Afterwards the Spectre and Meltdown fix

Erst I completed my tests, I was able to reboot the Surface Book and let the firmware patches install.

As you can view in the chart to the right, the sequential read and write performance doesn't interchange much (in point of fact, CrystalDiskMark no more uses a dispirited-line up-depth sequential test). But just A TechSpot found, 4K operation ain't pretty. While 4K read performance was same, the write carrying out dropped by 26 percent.

Far worse, though, 4K take and spell with high line up deepness engage a functioning murder of 42 percent and 39 percent, respectively. Ouch.

As with the pre-patch state, I ran the test three times with similar results.

A more relatable scenario

Computer storage performance hits in polysynthetic read/write tests, of course, can be difficult to relate to. Thus, to see how the patches manifested in operation that's relaxed to relate to, I also ran High-principled Technologies WebXPRT 2015 on the Surface Bible, exploitation the latest version of Microsoft Edge. WebXPRT 2015 is a web browser-based benchmark that measures performance in various scenarios in HTML5 and JavaScript.

Unpatched, I saw an boilers suit score of 450. Once patched, the score born, though non substantially, to 433. That's about 4 percent slower. Intel's own findings thereupon same run show about a 10 percent reduction.

Sol is the state of affairs not as dire atomic number 3 the celluloid computer storage benchmarks give it bent be? Yes and no. We'atomic number 75 still very early in testing the patches, but it's safe to put on that operation drops leave equal conditional on what you really do with your machine.

As TechSpot found, most standard tests (such as pure 3D rendering) and most games won't see a variety. But the greater-than-20-percent storage penalty that both TechSpot and I observed will rear its head occasionally.

In point of fact, matchless test that Intel presents is BAPCo's SYSMark 2014 SE. One of the about advanced benchmarks around, SYSMark uses real-world applications much as Word and Photoshop, and then runs them through tasks that mirror serious-world activity.

screenshot 2018 01 12 12.35.30 — Two sets of patches for the Surface Book address the Apparition and Meltdown exploits.

Even amend, BAPCo has developed a method acting for testing where only the response time of an action is metrical. In Word, for instance, olde worlde benchmarks would type and perform actions in Word fountainhead beyond what even the fastest typist could ever bump off.

But SYSMark 2014 Southeast measures the things that can truly annoy you, like how long it takes to first the application, or have IT perform a search and supplant, or import photos.

As you can imagine, SYSMark testament lean more heavily on storage responsiveness.

Intel's tests happening SYSMark 2014 Selenium mostly indicate nigh an 8 percent overall hit, just in the details for the System Responsiveness test (where you'd carry the entrepot operation to affair more), Intel says its seeing a 21 percent drop. Ouch.

We'll try to independently verify Intel's results on our own builds, but everything I'm visual perception then far says the performance penalty will belik run the gamut from "no openhanded deal" to "this is really testing my patience." Once more: It'll semen down to what you're doing, you bet you'Ra doing IT.

If we're talking an supernumerary 500ms to found an practical application that takes 1,500ms to launch, no big shell out. But if we're talking 34 seconds to import or copy photos instead of 27 seconds, it's going to suffer vexatious really fast and that's what scares me.