Meltdown/Spectre update slow downs?

The initial update i did via MS awhile back didnt seem to slow the machine down. But i also thought i had read it wouldnt be implemented until my PC manufacturer updated their bios. I could be wrong though. I forget.

I noticed that after updating the bios with HP Z400 fix, browsing seems a tad bit slower. I didnt really benchmark anything pre/post, but after using this pc for awhile, i can notice a slight difference. Its more noticeable when browsing but when i launch apps i notice a minute lag that wasnt there before.

using xeon 6 core 12 thread w3690

Anyone else notice this. i had read that there would probably be slowdowns for older machines, i just didn't notice it until i did the bios update.

Anyone on the newer amd platforms notice the difference?

I notice it on all my machines. The only way to not notice it is to be a damn liar, or not apply the patches. My XPS13 has an i7 8550U and it took a hit. Which pisses me off the most since it's #1 fairly new, #2 already a bit restrained because by design it is a low resource "U" class coffee lake, long battery machine and super power efficient. So I'm basically banking on some decent upgrades down the line to bring life back into a new machine that a few months ago was impressively powerful for this class.

My personal benchmarks that I use(because IDGAF about gaming, since maxing gaming for 99% of games is pretty easy mode, so for my tasks, takes a back seat to serious computing for production) consist of doing scientific hash assessments, just a great raw bench that nobody GAF about in the review-world unless they're hardcore nerd as well.
The XPS13 was at 89mh/s before (which is freaking great for a portable device) and 59mh/s after.

I have a bunch of older gens all over the map that are in my direct and extended maintenance. A ton of systems here, family systems, etc. I only have the one Ryzen system right now, and it is by far and away the lowest performance hit. Also helps that the bios allows PSP disabling without an extended workaround that doesn't fully disable to maintenance OS. Now it's not all roses for AMD, my old opteron system still took a good hit as well. The dual opteron server, while nowhere near as exploited, also took some hit. But in all fairness, that's a machine I should just upgrade already. It still holds up and ram is still too pricey. 76mh/s before, 68mh/s after.
The part about the Ryzen system that kind of sucks though, is if you wanted to not patch, take the risk and keep your old performance.. well, the bios updates expand on a lot of memory and m.2 compatibility. So you have to choose. I couldn't make a m.2 drive bootable no matter what unless I used a bios that included new microcode for specre workarounds(WD Black 3DNand, so something that came out long after the mobo, so I'm lucky it was patched to begin with). Before and after for scientific hashing, I did 146mh/s before, and 132mh/s after patching with 0 impact from any GPU or SSD change.

Wife's desktop is a 7600 non-K and she's not computer literate, but she noticed the difference and said it was night and day after it was fully patched. Never benched it, wish I had though. But if she notices these things, it's legit.

Sister has an older laptop(mostly a tablet user, so she's just trying to extend this one uptil we see another cycle of super cheap ram), IIRC, a socket P 8600(which is a core 2 duo platform, not coffee lake). Had to switch her from antix to Vector Light(ultralight linux distros that thrive on older machines) Antix is a 128mb *rated* kernel in memory OS, and VLL is a 64mb *rated* kernel in memory OS. Puppy, which is light, not ultralight, is IIRC, a 512mb *rated* kernel in memory, so you can imagine just how far down the rabbit hole this has gone just to save a perfectly functional machine for the purposes of ... checking facebook, email, and watching youtube/prime/netflix videos. While it would run all those things fine unpatched, patches made it impossible to have more than one window open at a time in a 'mainstream' web browser. I switched her to Brave for media heavy sites and midori for email and text heavy sites, but even Brave isn't a silver bullet.

The amazing thing... it's not just x86. Any reasonable ARM device that has actual security patches possible also took a noticeable hit in performance. Yeah yeah, stob being n00bish and focusing on phones/tablets here, think about ALL the ARM devices you have. From your TV providers set-top DVR and complimentary cable boxes to your routers, modems, automobile CPUs, IoT devices, etc. People that have installed security cameras that stream to a cloud provider that once could handle 5 cameras safely can now only capture data off 3 at the same bitrate and frames. That is a HUGE drop and it's no wonder people leave their hardware unpatched.



So yeah. If you're not seeing much of a hit with recent CPUs, then it's because your system isn't fully patched by your own choice, or the manufacturers have not fully released patches or bios updates.

Crystallas said:

So yeah. If you're not seeing much of a hit with recent CPUs, then it's because your system isn't fully patched by your own choice, or the manufacturers have not fully released patches or bios updates.

Click to expand...

Many companies are choosing to not patch for Meltdown/Spectre until the performance hits and stability are cleaned up.

That alone should tell us about how nasty the performance hit is on this patching.

Meltdown was preventable, that is the main problem. No reason to believe the next archs will learn from this lesson.

Sad thing is, it's been widely agreed upon that these designs are a net negative and will always bare a potential security risk. If they want backdoor maintenance running but think users can remotely shut them down... that means they believe that someone can exploit them to begin with which cancels the opt-in portion from the get go.


Specre works in a completely different manner, and a lot of people don't like that meltdown and spectre are bundled together when explaining the basics of these exploits because users fail to learn from the issues with speculative computing. Therefore consumers will not demand changes. I've been against speculative function for a long time as well, but it doesn't matter because the next generation will simply carry an even clunkier speculative structure to work around the exploits of a number of past architectures. Then that will get exploited again, and we're back to square one. So if anyone believes this issue is going away with new architectures, and it will be 'fixed' , then lol, I have oceanfront property in Rockford for sale.

But spectre is to some degree also preventable. There is a philosophical debate on whether it's better to have or not have. While being proven right along with many other *actual* electronic engineers that develop entire architectures, it's simply sad that those who pushed for this design are software heads who couldn't even build a basic fpga.

So Specre works at like an estate or a hacienda?

Ares said:

Many companies are choosing to not patch for Meltdown/Spectre until the performance hits and stability are cleaned up.

That alone should tell us about how nasty the performance hit is on this patching.

Click to expand...

well i know hp delayed releases for many of their enterprise pc bios. it had been up then was taken down for months for them to work out some bugs according to them.

havent been gaming much but played some cs:go to see if in game was slower for me.. probably not a good game to gauge since its a fairly low resource dependant game.. i did notice that lag when launching, but it seemed about the same fps once it got going. i only picked that game because i roughly remember my old fps and it seemed about the same. im not sure on lows though..

maybe i should go back and play some bf1, witcher 3 or fallout 4. those pushed me system some.. but more gpu than cpu..