How concerned should I be about meltdown and spectre exploits?

[UsandThem]

So? Most of the benchmarks were in the +/- 3%.

And even that "worse performance" on storage related items is only related to prior Intel performance.

If you compare those storage related benchmarks between AMD and Intel. Intel is not showing any kind of serious deficit.

As much as some partisans like to use it as FUD against Intel, Specter/Meltdown is not a determinant in choosing between Intel/AMD.

this is the 2nd time you used "FUD", but storage bandwidth has definitely been reduced (especially in NVME drives). So while you disagree with him, instead of saying "FUD", link to another review that shows otherwise.

I'm not even arguing the fact that normal PC users (non-enthusiasts) would see any real - world difference. But running things like F@H, I've noticed a small difference since I patched everything.

 

[PeterScott]

I'm not even arguing the fact that normal PC users (non-enthusiasts) would see any real - world difference. But running things like F@H, I've noticed a small difference since I patched everything.

So you notice a small difference in niche case. I am not discounting that. I am discounting this as FUD:

Well, if the BIOS already has the fix, and windows update auto-installs the patches, and you have a SSD or NVME SSD, it will drastically slow down the disc access.

It's a small difference in niche case, not a drastic general one for disk access. The latter is FUD.

Plus I have yet to see anything that addresses the point that Intel hasn't actually fallen behind AMD in storage access. There was this earlier test posted thread, that shows an array of storage benchmarks, where if Anything, Intels still has faster access, post meltdown/specter.

https://www.techspot.com/review/1646-storage-performance-intel-z370-vs-amd-x470/

 

[Markfw]

So you notice a small difference in niche case. I am not discounting that. I am discounting this as FUD:



It's a small difference in niche case, not a drastic general one for disk access. The latter is FUD.

Plus I have yet to see anything that addresses the point that Intel hasn't actually fallen behind AMD in storage access. There was this earlier test posted thread, that shows an array of storage benchmarks, where if Anything, Intels still has faster access, post meltdown/specter.

https://www.techspot.com/review/1646-storage-performance-intel-z370-vs-amd-x470/

You don't call 20 percent drastic ? from this: https://www.theregister.co.uk/2018/01/09/meltdown_spectre_slowdown/

"Red Hat has clocked the patch performance impact as ranging from one to 20 per cent."

I can keep looking and posting, but I am sick of you accusing me of FUD, because its not. For high IO situations, it can be drastic.

 

[PeterScott]

You don't call 20 percent drastic ? from this: https://www.theregister.co.uk/2018/01/09/meltdown_spectre_slowdown/

"Red Hat has clocked the patch performance impact as ranging from one to 20 per cent."

I can keep looking and posting, but I am sick of you accusing me of FUD, because its not. For high IO situations, it can be drastic.

The Register is just reporting on the Hysteria at the time.

The Redhat link you didn't read in enough detail. Here is the worse case on the Redhat link:
"Measurable: previously 8-19% - updated w/ retpoline to be 4-8% - Highly cached random memory, with buffered I/O, OLTP database workloads, HPC (High Performance Computing) scale-out environments using MPI and benchmarks with high kernel-to-user space transitions were measured to be impacted the most. Examples include OLTP Workloads (tpc), sysbench, pgbench."

So niche case, and the worse case is now 8%.

So no, I don't consider 8% in a niche case Drastic.

 

[DaveSimmons]

Do many home users run Red Hat Enterprise Linux as their desktop OS? That's news to me.

If not, the performance hit to RHEL seems off-topic for this thread.

 

[Markfw]

Do many home users run Red Hat Enterprise Linux as their desktop OS? That's news to me.

If not, the performance hit to RHEL seems off-topic for this thread.

It was I guess a bad example. Its all in the meltdown thread, I just didn't want to spend hours re-reading thousands of posts to find the links where it was up to 30% for home users. But some here will not accept that, so fine, buy what you want, I don't care to spend more time here.

I am done with this thread.

 

[samboy]

I've deliberately disabled (InSpectre) all of these patches for performance reasons; 5-20% doesn't sound like much but it's effectively giving away all the CPU progress we have made in the last 4-5 years!

I run with administrator privilege on my system (as many people do) and this really makes any of these patches pointless....... anything that gets in has full access anyway (which is why running with admin privilege is not recommended!).

In a cloud environment I can understand the concern; but on a personal home system I think this has been blown all out of proportion. It does make a very interesting read and has launched a few careers for some folks..........

I don't think general home users should be concerned with any of this at all.........

 

[epsilon84]

It was I guess a bad example. Its all in the meltdown thread, I just didn't want to spend hours re-reading thousands of posts to find the links where it was up to 30% for home users. But some here will not accept that, so fine, buy what you want, I don't care to spend more time here.

I am done with this thread.

Again, no one is saying there is no performance impact, at least measurably, but the real world impact for the vast majority of desktop users is negligible.

Let's put it this way - if someone actually has to ASK about how they will be affected by Spectre/Meltdown, it ironically actually means they won't be affected much at all. People that need the high amount of IO volume to actually see a difference from these patches would already know how it affects their workload.

Blanket statement about 'drastically slower disc access' really doesn't help the OP because he will think 'oh god my SSD will turn into a HDD' when in fact, Z370 still matches or beats X470 in the majority of storage tests as I posted earlier.

 

[DaveSimmons]

^ and the same for single-core performance. Is there an effect? Yes. Does intel still offer the fastest single-code performance? Also yes. Context matters.

If Spectre affected nvidia GPUs but not AMD, the 1080ti would still be the best single-card gaming GPU even if it was now 3-10% slower. (And 3-30% slower for AI CUDA apps, to complete the analogy.)

 

[moinmoin]

The whole Spectre range of issues is still not resolved. We still regularly get new forms of Spectre related attacks (the most recent ones being publicized ret2spec, SpectreRSB, and NetSpectre), many of which are reportedly still not yet publicized nor patched. AMD *is* better off than Intel with current chips as they do respect the separation of kernel and user space data in caches which prevents the more egregious attacks like Meltdown from working. Also the option of disabling fixes is not a real one if you intend to use new hardware (which likely already has a BIOS version installed containing fixes) and intend to use an up-to-date OS (which you prolly should). So while desktop users may be able to ignore the attacks, they may not want to ignore the performance impacts which come along with preventing them from working.

 

[DaveSimmons]

The whole Spectre range of issues is still not resolved. We still regularly get new forms of Spectre related attacks (the most recent ones being publicized ret2spec, SpectreRSB, and NetSpectre), many of which are reportedly still not yet publicized nor patched. AMD *is* better off than Intel with current chips as they do respect the separation of kernel and user space data in caches which prevents the more egregious attacks like Meltdown from working. Also the option of disabling fixes is not a real one if you intend to use new hardware (which likely already has a BIOS version installed containing fixes) and intend to use an up-to-date OS (which you prolly should). So while desktop users may be able to ignore the attacks, they may not want to ignore the performance impacts which come along with preventing them from working.

Which doesn't change the fact that intel CPU cores are still faster post-fix and for normal users SSD speed isn't noticeably slower than AMD.

Also, NetSpectre is not a concern at all for normal desktop users since it takes far to long to learn any significant amount of information.

I'm glad that AMD is finally competing with intel, but for normal desktop users raising Spectre as a major concern really is FUD. If you are a big fan of AMD then buy AMD, but it's silly for most people to do so "because Spectre."

 

[moinmoin]

Which doesn't change the fact that intel CPU cores are still faster post-fix (...)

Which "post-fix"? And will it still be true whenever (if ever) the whole Spectre attack vector is truly resolved? All things considered it's still early days and so far we don't actually see fixes but software/firmware/microcode workarounds that prevent attacks from working. We may get the final answer once all the chip manufacturers have their hardware design fixes baked into the chips.

 

[DaveSimmons]

Which "post-fix"? And will it still be true whenever (if ever) the whole Spectre attack vector is truly resolved? All things considered it's still early days and so far we don't actually see fixes but software/firmware/microcode workarounds that prevent attacks from working. We may get the final answer once all the chip manufacturers have their hardware design fixes baked into the chips.

FUD = Fear, Uncertainty and Doubt.

If you're playing worst-case what-if, then some unrelated new side channel exploit that has a greater impact on AMD could be discovered next week. Or some general execution flaw could be found in either AMD or intel chips.

 

[Vattila]

I would go with an Intel processor because I care mostly about emulation

By emulation, I presume you are talking about Dolphin or such for emulating gaming consoles (and not Vice for legacy 8-bit hardware, like Commodore 64, etc.). Is emulation performance really still a problem for Ryzen?

"Already mentioned elsewhere, but where Ryzen shined when it released, Raven Ridge 2200g/2400G are the real deal. for 169$ and 99$, you can actually play most recents emulators, at full speed without spending a penny on graphic card."

https://www.reddit.com/r/emulation/...gamecube_and/dxg1csp/?st=jkeur7dq&sh=c77b8ad7