meltdown Meltdown and Spectre patch shows significant impact on SSD performance

[Seithon]

As reported by heise.de apparently the impact of the windows patches upon the performance of SSD's is even greater then the impact upon CPU operations. 

 

https://m.heise.de/newsticker/meldung/Intel-Benchmarks-zu-Meltdown-Spectre-Performance-sackt-um-bis-zu-10-Prozent-ab-SSD-I-O-deutlich-mehr-3938747.html

 

One relevant (google translated) section from the article reads

 

c't: Significantly lower SSD performance after security updates

c't has also seen significant drops in SSD performance after installing BIOS and security updates (Core i7-8700K, Asus Maximus X Hero, Samsung 960 Pro, Random Access with 4K blocks and 32 I / O-targets). The Samsung SSD only reached 105,986 I / Ops for reading and 79,313 I / Ops for writing compared to 197,525 / 185,620 I / OPs (previous BIOS, Windows with KB4054517). Further analysis, benchmarks on the effects of patches, and reviews of updates can be found in the upcoming c't issue 3/18.

[Doobeedoo]

Yeah it definitely hampered the IO indeed. 

[samcool55]

That's like, 50% gone? That's brutal...

 

[Ryujin2003]

I've got to updatemy BIOS and OS this weekend. I'll bench my driver's prior to the update.... MSI doesn't even have the old BIOS available, so no flashing back if it kills performance.....

[vorticalbox]

1 hour ago, Doobeedoo said:

Yeah it definitely hampered the IO indeed. 

All the reports I have read shows hits to i/o on servers I assumed the same on desktop. 

 

My station at work is definitely feeling slower than it used to

[Zodiark1593]

I'll definitely be waiting on patching until the performance hits are mitigated. Neither of my systems are connected to the internet much at all, soo...

[ItsMitch]

I guess I'll be putting of my SSD purchase

[DrMacintosh]

Still faster than a HDD so idc

[HarryNyquist]

"I'll put off patching a critical security flaw in the design of my CPU because I don't want to lose performance on my SSD."

 

I dunno about any of you, but I patched as soon as I saw MS had the package up. Yes, I'll admit things feel slightly more sluggish on my SSD than they used to. But I'd rather do that, than risk exposing a critical security flaw in the design of my CPU to an internet-connected machine. Perhaps with time this will be mitigated, but stalling on updating on something this big is just idiotic.

[thexzenon]

1 hour ago, SC2Mitch said:

I guess I'll be putting of my SSD purchase

Waiting for what if I may ask?

[Jito463]

6 minutes ago, crisro996 said:

Waiting for what if I may ask?

New and improved SSD 2.0.

[Zodiark1593]

57 minutes ago, HarryNyquist said:

"I'll put off patching a critical security flaw in the design of my CPU because I don't want to lose performance on my SSD."

 

I dunno about any of you, but I patched as soon as I saw MS had the package up. Yes, I'll admit things feel slightly more sluggish on my SSD than they used to. But I'd rather do that, than risk exposing a critical security flaw in the design of my CPU to an internet-connected machine. Perhaps with time this will be mitigated, but stalling on updating on something this big is just idiotic.

I'll let the lot of you beta test the patches, thank you very much. Only thing that regularly touches the internet at home is my phone, so I'm in no rush whatsoever.

 

[vorticalbox]

3 hours ago, HarryNyquist said:

"I'll put off patching a critical security flaw in the design of my CPU because I don't want to lose performance on my SSD."

 

I dunno about any of you, but I patched as soon as I saw MS had the package up. Yes, I'll admit things feel slightly more sluggish on my SSD than they used to. But I'd rather do that, than risk exposing a critical security flaw in the design of my CPU to an internet-connected machine. Perhaps with time this will be mitigated, but stalling on updating on something this big is just idiotic.

It is such a massive flaw that end uses won't be a target for a long time. Servers will be the big targets. 

[Crunchy Dragon]

4 hours ago, SC2Mitch said:

I guess I'll be putting of my SSD purchase

I've had to put it off for so long haha. I never really had the budget for one to begin with and prices for a good quality SSD these days are insanely high.

 

[Shadestones]

My SSD performance measured with CrystalDiskMark is nearly half what it used to be in Seq. Writes and 4k Writes. I don't notice any difference in speed, but if it really is that much slower, I'm disappointed.

[RyanEsau]

That's rather brutal... Does this affect SATA SSDs as well? M.2 NVMe which uses PCIe lanes which levers on the CPU doesn't it? So seems natural (but unfortunate) that it's hit but SATA is... SATA. Sucks but not a big deal to me I guess long as I still get my speedy boot times. I only put the OS/Programs on my SSD, just a Samsung 850 Pro 128GB, everything else goes on my 2TB Seagate Firecuda SSHD.

 

I'm still waiting though. I had went ahead and did the Windows Updates followed by the BIOS update for my Motherboard last night but then I couldn't get into Windows anymore...

 

Tried system restore but didn't work for some reason so I ended up having to wipe my SSD (Which is luckily just the OS by itself), reflash previous bios, and reinstall Windows. 

 

Also I remember after my reboot for the windows updates, that split second before I just restarted again to do the bios update I remember seeing a program (Asus GPU Tweak II I think? For my GPU) had crashed/failed to start. Does the patch affect more than just anti-virus/malware programs?... That makes it seem like a bit of a wait is in store to seamlessly move on over.

 

Seems like Avast is up-to-date (as of Dec Update) from what I read, but I also use MalwareBytes and I got a bit confused on that front.

 

(as of 1/04/18): Since the Malwarebytes Database Update 1.0.3624, all Malwarebytes users are able to receive the Microsoft patch to mitigate Meltdown.

https://blog.malwarebytes.com/security-world/2018/01/meltdown-and-spectre-what-you-need-to-know/

 

Database Update? Do they mean the 'Update Package'?

 

[MageTank]

Just tested this on my 960 evo.

 

Before patch KB4056892:

 

After patch KB4056892:

 

 

Not really seeing a difference at all here. If there is a performance difference, it's only a slight one. 

-2% in Seq Read, no difference in 4K Q8T8 read, -3.5% in 4K Q32T1 read, -12% in 4K Q32T1 write, and -9% in 4K Q1T1 write. While 12% and 9% seem like big numbers, should also note that these tests vary quite heavily from pass to pass, and I've seen it sway both ways while running the tests. You can also see improved writes for both Seq and 4k Q8T8.

 

If anyone has more accurate benchmarks, I can test them as well, with and without the patch, to see if any difference can be seen. One thing is for certain, I don't feel any difference in performance with the way I use my system. 

[leadeater]

10 hours ago, MageTank said:

If anyone has more accurate benchmarks, I can test them as well, with and without the patch, to see if any difference can be seen. One thing is for certain, I don't feel any difference in performance with the way I use my system. 

IOmeter is the only one that is going to be able to show the difference. You'll need to run through a large array of tests and multiple runs etc, which can all be setup with IOmeter and left to run. Doing it with 1 worker, 5 workers and 10 workers would also be required with a run time of 10 minutes and 60 minutes. Burst performance may be effected more than sustained, maybe sustained is impacted more.

[SilkyDistress]

16 hours ago, SC2Mitch said:

I guess I'll be putting of my SSD purchase

It think its only hit on NVMe SSD's cause they are directly connected with processor with PCI-E interface, SATA SSD's shouldn't be hit.

[Sir_Loin_Ofsteak]

Yaaay! I get to do another BIOS update which will do unknown things to my system.

 

*joy*

[Sickshooter0]

12 hours ago, vorticalbox said:

It is such a massive flaw that end uses won't be a target for a long time. Servers will be the big targets. 

Agreed; initially. They are the juiciest targets. However server managers have more resources to lock down the exploit(s). The mom & pop computer user who don't realize, that in order to benefit from the OS security updates, they have to also update their BIOS, will eventually be the targets.

[jagdtigger]

14 minutes ago, Sickshooter0 said:

Agreed; initially. They are the juiciest targets. However server managers have more resources to lock down the exploit(s). The mom & pop computer user who don't realize, that in order to benefit from the OS security updates, they have to also update their BIOS, will eventually be the targets.

Many wont do that out of fear that something will go wrong and they can kiss their MB goodby...

 

I just upgraded the BIOS on my home server, but it come out in 2017-09-29. As for my gaming PC the last update come out in 2014-07-22... (How fast the time goes by, its already 4 years old...)

Edited January 12, 2018 by jagdtigger

[Sir_Loin_Ofsteak]

I'm actually interested how they're going to push these updates to people. I am a hobbyist when it comes to PCs, so I know how to update it and the potential pitfalls of it going wrong, but your average user doesn't know what a BIOS is, let alone how to update it. 

 

I think that if this update really is vital to security, we're going to see users vulnerable for decades. And possibly a lot of dead hardware from people giving it a go and screwing it up...

[vorticalbox]

9 minutes ago, Sir_Loin_Ofsteak said:

I'm actually interested how they're going to push these updates to people. I am a hobbyist when it comes to PCs, so I know how to update it and the potential pitfalls of it going wrong, but your average user doesn't know what a BIOS is, let alone how to update it. 

 

I think that if this update really is vital to security, we're going to see users vulnerable for decades. And possibly a lot of dead hardware from people giving it a go and screwing it up...

why the exploit is getting blocked at a kernel level a BIOS upgrade probably isn't needed. Could be wrong though.

[-TesseracT-]

Let me get this straight... This only applies to nvme ssds, and not to my 850 evo?