Storage Tests and Determination

The terminal set of tests I'll be looking at are storage benchmarks, because it'south here where we saw the biggest impact on the desktop platforms. Information technology'southward a similar story on both the PCIe and SATA SSD equipped laptops, with massive reductions in performance seen in well-nigh every state of affairs.

The 8thursday gen XPS 13 with a PCIe NVMe SSD suffers significantly in the sequential write test, with a performance decrease of 24 percent, whereas the fiveth gen XPS 13 with a SATA SSD is largely unaffected.

With queue-2 1-thread random performance, though, the 8thursday gen system takes a 5 and 15 percent hitting in reads and writes respectively, whereas the 5th gen system gets punished with eighteen and 39 percent drops.

It'south interesting to see how other queue depths and thread counts affect operation. With 8-thread, 8 queue depth random transfers, the eightth gen system takes a fourteen percent hitting in both reads and writes, while the Broadwell machine takes a larger 34 percent striking in reads merely the aforementioned 14 per centum in writes.

And and so switching to a single-thread random transfer with a queue depth of 32, and the SATA drive gets punished with a more than 40 percent reduction in operation, compared to 35 and 12 percent drops in reads and writes respectively on the PCIe drive.

In the cease, the performance impact of the Meltdown and Spectre patches is a bit more astringent on laptops with an Intel U-series CPU within compared to the desktop platforms we looked at before.

On the desktop, most productivity workloads saw performance drops that were within the margin of error, or a few percent at worst. However on both the Kaby Lake Refresh and Broadwell laptops we looked at today, near every workload was impacted by an amount that's also large to aspect to a simple margin of mistake.

Workloads like Cinebench, x264 encoding, Excel, Adobe Premiere, Photoshop, and MATLAB all experienced noticeable performance declines typically around the v percent mark, though the worst case was a 19 percent drop in Photoshop on the i5-5200U. Other tests like PCMark, Handbrake, 3DMark and both compression and decompression were largely unaffected.

1 of the more interesting things to notation from these benchmarks is there doesn't seem to exist a significant deviation in the amount of slowdown between the iii-year-erstwhile Broadwell laptop and the modern Kaby Lake Refresh laptop in productivity workloads. It'due south certainly not a situation where the older and slower Broadwell is affected more than than newer parts, in some situations the eightth gen parts are hit harder, while in others vthursday gen suffers more.

Of course, storage performance declines significantly on both systems, specially the write performance, though that isn't surprising because what we saw on the desktop side of things.

Nosotros'll go along to monitor the situation with the Meltdown and Spectre patches, specially every bit an already worrisome and complicated thing is turning into a big disaster for Intel -- security issues aside -- in rushing a ready for the Spectre flaws, the company has discovered and now identified that the current microcode update is causing random crashes and other instability problems on a variety of hardware, sometime and new.

Based on this information, Intel is now advising OEMs to pull their Spectre BIOS updates until a newer release is tested and distributed. There take also been reports that the updates don't even gear up the vulnerabilities. Meanwhile, Google is coming to the rescue by offering a software solution that they claim not only patches the security holes simply come up at negligible impact in performance.

No dubiety we'll be seeing even more updates to endeavor and lock downwards these problems in the coming weeks and months, and nosotros'll be letting y'all know if any of them further bear upon (or fix) functioning.