Intel/Windows 30% performance reduction

[vortex681 806]

1 hour ago, Rob Ainscough said:

But a 30% reduction on a virtualized AWS (cloud servers, especially render farms) or even bitcoin miners then that number can have real financial implications.

You, me, others represent a tiny subset of the testing data sample.  Just because it has NOT impacted you as much, doesn't mean it hasn't impacted others.

My comment wasn't only about the latest Intel problems but about many issues that arise and are discussed here. Some of the online articles I've read recently make it seem as if the end of the world is nigh! The very title of this thread ("Intel/Windows 30% performance reduction") is just scaremongering as far as the average home user is concerned. If the title had said "Intel issue may cause performance reductions in certain circumstances" that would have been much more sensible and realistic.

How many of us here run cloud servers or server farms for a living? I'd guess very few, if any. I'm not for a moment saying that no systems will experience significant slowdowns (as, clearly, some will), just that for most of us with general-use/gaming desktop systems it's unlikely to be significant. There's an interesting video here from Hardware Unboxed which compares performance before patch, with patch and  finally with patch and BIOS updates: https://www.youtube.com/watch?v=JbhKUjPRk5Q. The only real performance reduction seems to come from the storage benchmarks. What's more significant is that NVMe storage devices seem to be affected much more than SATA devices. Most of the non-storage results have a 3-4% performance reduction at most. This seems to be consistent across the sites which have tested after patching. Most games look to have little or no change and even then, probably within the margin of error for the test.

[TheFamilyMan 279]

My two cents on all this mess:  A huge hardware design paradigm shift is about to take place that was just aching to happen.  Welcome to the new computing world of security first, and performance relegated to second consideration.

[jabloomf1230 2,445]

I knew that when The Register had the scoop on this story, that the extent of the problem for the average PC user was well overblown.

[lownslo 592]

14 minutes ago, jabloomf1230 said:

I knew that when The Register had the scoop on this story, that the extent of the problem for the average PC user was well overblown.

It's not only The Register.  There's an entire cottage industry that exists just waiting to pounce on the sky-is-falling incidents like this.  From the reporting all the way to the "solutions" (that sometimes involve your money), there's plenty of folks who thrive on this stuff.

Greg

[Uteman 51]

On 1/9/2018 at 11:41 AM, jabloomf1230 said:

I knew that when The Register had the scoop on this story, that the extent of the problem for the average PC user was well overblown.

I am nowhere near as confident as you are, all the statements from Intel and Microsoft have said that 1. it won't affect the average user much and 2. the affect on performance is entirely based on CPU workload. That puts flight sim users who regularly see 70% plus workloads well outside the realms of the average home user who is lucky to see their CPU at 20% for more than a nanosecond. The first lot of Intel microcode is out for Linux and that should give some benchmarks for the next code release that hopefully will tackle the Microsoft issues then we will know for sure on our own machines.

I hope you are right  in your prediction.

[G-RFRY 1,708]

Feel sorry for the AMD users some have had to rollback of reinstall there OS due to the spectre patch. 

Also spotted this.

http://www.tomshardware.co.uk/microsoft-intel-slowdown-old-chips,news-57684.html.

[simmerhead 885]

Yep, AMD users be careful: https://www.theverge.com/2018/1/9/16867068/microsoft-meltdown-spectre-security-updates-amd-pcs-issues

 

[vortex681 806]

7 hours ago, Uteman said:

I am nowhere near as confident as you are, all the statements from Intel and Microsoft have said that 1. it won't affect the average user much and 2. the affect on performance is entirely based on CPU workload. That puts flight sim users who regularly see 70% plus workloads well outside the realms of the average home user

It's not just about CPU load, it's also about what the CPU is actually doing. This from another sim site about the potential performance impact:

"the specific slowdown is during the user/kernel mode switch. FS spends the vast majority of its time in user mode. The only time you do lots of these transitions are when you're doing heavy network and/or disk I/O".

If you don't believe that (or Intel and Microsoft), believe those like Mathias here who've actually tried it and see very little difference.

[Uteman 51]

I have read a fair bit of the reported Microsoft PR and it is general in context and detail, for example:

~~~~~~~~~~~~~~

"Windows 10 machines running older processors like Haswell “show more significant slowdowns, and we expect that some users will notice a decrease in system performance,” says Myerson. The same older Haswell machines running Windows 7 or Windows 8 will also experience slowdowns that Myerson says “most users” will notice.

Windows 7 and Windows 8 will be the worst hit simply because these older operating systems have features like kernel-level font rendering that will be impacted by the Spectre and Meltdown mitigations even further than Windows 10. Regardless, Microsoft says on Skylake or newer chips “Intel has refined the instructions used to disable branch speculation to be more specific to indirect branches, reducing the overall performance penalty of the Spectre mitigation.”

Windows 10 running on Skylake, Kaby Lake or newer CPU show benchmarks show “single-digit slowdowns”, but most users shouldn’t expect to see noticeable slowdowns

Windows 10 running on Haswell or older CPUs “show more significant slowdowns” and “some users will notice a decrease in system performance”

Windows 7 or Windows 8 running on Haswell or older CPUs means “most users will notice a decrease in system performance"

~~~~~~~~~~~~~~~~ taken from the Verge article linked above.

You can draw your own conclusions from the released PR info, I haven't made any. Until end users see and test the final/declared Intel/Microsoft fix in their own machine and I see it on my own machine I will stand by my comment - which was and I repeat - I hope Jays prediction is correct.

[jabloomf1230 2,445]

I have no prediction except that I said the same thing here when The Register initially leaked the story. They are the definition of Tabloid Tech. 

[flyinion 280]

Well, guess I'll never see the full performance impact.  I have the MS patches and patched Nvidia drivers but my Asus motherboard is NOT receiving a new BIOS (Z97 series ROG board) so I'll never really have the fix that might truly impact my 4790K Haswell.  With just the software patches installed I've seen zero changes in 3DMark11 and Firestrike benchmarks.

[jabloomf1230 2,445]

The only 30% reduction has been in the amount of legitimate flight sim discussions sacrificed because of this non-issue.

[G-RFRY 1,708]

More updates heading our way this time Motherboards.

http://www.tomshardware.co.uk/motherboard-vendors-release-bios-updates-spectre,news-57708.html

[TheFamilyMan 279]

On 1/10/2018 at 2:34 PM, flyinion said:

Well, guess I'll never see the full performance impact.  I have the MS patches and patched Nvidia drivers but my Asus motherboard is NOT receiving a new BIOS (Z97 series ROG board) so I'll never really have the fix that might truly impact my 4790K Haswell.  With just the software patches installed I've seen zero changes in 3DMark11 and Firestrike benchmarks.

Exactly my experience with my 4770K Z87 build.

This situation almost feels as if it's a backhanded attempt by Intel to force hardware updates upon us while they've provided us with so little performance-wise incentive to do so over the last few years.  If anything, since it's solely their fault, Intel should fully compensate vendors and insist that they generate bios updates going all the way back to at least Sandy Bridge, or offer update rebates for "unsupported" CPUs.

[addman 33]

Good to know that I won't be upgrading my whole AMD platform to an Intel platform in the foreseeable future. I know AMD is affected too, to a certain degree but it's Intel who's playing a shady game right now. I have an Asus B350-F board so I don't have to upgrade that part, or any other, when I will upgrade to the next generation (Ryzen 2) of Zen-based CPU's. Feel sorry for Win7 users still sitting on Haswell, which are still relevant, or older platforms, they'll most likely have some sort of noticeable performance degradation.