Intel Random Restart Bug & Data Center Performance After Patches

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,510
Back on January 11, Intel started seeing reports of random system reboots after applying Meltdown and Spectre patches on Intel Broadwell and Haswell CPUs. Seems as if this was not an isolated occurrence and this has now been verified with other Intel partners. And this looks to be hammering a larger base of CPUs than first reported. Ivy Bridge, Sandy Bridge, Skylake, and Kaby Lake processors are also affected. Microcode is in the works to address this and being worked on now.


We have now issued firmware updates for 90 percent of Intel CPUs introduced in the past five years, but we have more work to do. As I noted in my blog post last week, while the firmware updates are effective at mitigating exposure to the security issues, customers have reported more frequent reboots on firmware updated systems.

As part of this, we have determined that similar behavior occurs on other products in some configurations, including Ivy Bridge-, Sandy Bridge-, Skylake-, and Kaby Lake-based platforms. We have reproduced these issues internally and are making progress toward identifying the root cause. In parallel, we will be providing beta microcode to vendors for validation by next week.


Intel has also followed up its initial desktop performance metrics with benchmarks for data center. All looks well except for data IO, which of course is no concern to data centers. /s


As usual, take these with a grain of salt, and hopefully you admins have some pre-patch metrics to use for comparison on your own systems.

The benchmark results reported above may need to be revised as additional testing is conducted. The results depend on the
specific platform configurations and workloads utilized in the testing, and may not be applicable to any particular user’s
components, computer system or workloads. The results are not necessarily representative of other benchmarks and other
benchmark results may show greater or lesser impact from mitigations.
 
  • Like
Reactions: Jza
like this
Here is something I have measured:


I use VS2015 at work to build a fairly large project, mostly C code, some VB.NET, some C#. Used to be able to do a full rebuild in 43 seconds max, sometimes 41. It's now at 54 seconds steady, so I am looking at a good 25% reduction in performance. And no, Intel, it's not just artificial SSD benchmarks shitting the bed. This is real work.

This is a 5600-U CPU on a HP laptop using SSD, which is SATA. The model is a EliteBook 840 Gen2. It's running Win7 64-bit.

I will note that the Windows updates came in yesterday and I didn't see a measurable regression. Today I put on the HP BIOS and that's when the above was measured.

Intel can rub everybody's nose into those benchmarks they did, but as an end-user, I have to say this security fix really, really blows.
 
I get Intel is trying to minimize the impact of the patches, I just wish they would also put out numbers from slightly older Xeons as well. To me that's the more pertinent information for most folks.


On another note I shudder inside when I think of all the average users out there who Intel is expecting to have to run not only a Windows Update but also updating firmware. I wonder what the % of functionally unpatched systems are going to be (I'd guess 80%?). There are going to be some fun times out there security wise once exploits get weaponized.
 
Seriously thinking of uninstalling Meltdown patch not upgrading ANY bios...I have 5 Intel laptops from Core 2 duo to Kabylake...Intel obviously does not have their act together anymore....I'll take my chances instead of killing performance or machines. This should never have happened Thanks Intel....
 
Seriously thinking of uninstalling Meltdown patch not upgrading ANY bios...I have 5 Intel laptops from Core 2 duo to Kabylake...Intel obviously does not have their act together anymore....I'll take my chances instead of killing performance or machines. This should never have happened Thanks Intel....
That InSpectre tool posted yesterday has the ability to enable/disable the patches (Windows-side, obviously not BIOS-side), so you may be able to use that for your needs.
 
Intel 4770K + Z87th Gryphon here. ASUS: "Buy our latest products or go fuck yourself."
I'm currently unemployed on savings right now trying to get a job in Silicon Valley. Honestly, there is no compelling reason for me to upgrade from this. (Maxwell Titan in this system as well.)

Not the end of the world but it still sucks. I'll be lucky if I get 10$ coupon off a 200+ CPU + 200+ MB + 170ish for 16GB of ram from the class action lawsuit.
 
i downloaded the tool ran as administrator on windows 8.1 machine I disabled protection but tool DID NOT disable it. Does not allow enabling or disabling .....
 
i downloaded the tool ran as administrator on windows 8.1 machine I disabled protection but tool DID NOT disable it. Does not allow enabling or disabling .....
I wish I could offer some insite there, alas, being on a slow (metered) cellular connection and running AMD hardware in all my computers, I've not bothered patching my systems. Therefore, there's nothing for it to disable anything. However, even when ran as Administrator, the two buttons do not show up. Whether or not that is because there are no updates installed to even tinker with, I don't know...

Best bet would be to try and report it to the author, provided you read the pages relating to the program and weren't able to glean any solution...
Sorry I couldn't be of further help :\
 
Here is something I have measured:


I use VS2015 at work to build a fairly large project, mostly C code, some VB.NET, some C#. Used to be able to do a full rebuild in 43 seconds max, sometimes 41. It's now at 54 seconds steady, so I am looking at a good 25% reduction in performance. And no, Intel, it's not just artificial SSD benchmarks shitting the bed. This is real work.

This is a 5600-U CPU on a HP laptop using SSD, which is SATA. The model is a EliteBook 840 Gen2. It's running Win7 64-bit.

I will note that the Windows updates came in yesterday and I didn't see a measurable regression. Today I put on the HP BIOS and that's when the above was measured.

Intel can rub everybody's nose into those benchmarks they did, but as an end-user, I have to say this security fix really, really blows.
The tests they did most likely were cherry picked.
 
Listening to some of the people in the Intel subforum, their response has been...
9cc.gif

I think one of them even went so far as to say there was "zero effect."
 
Listening to some of the people in the Intel subforum, their response has been...
9cc.gif

I think one of them even went so far as to say there was "zero effect."
They could be catching fire and they would still say the same thing.
 
Here is something I have measured:


I use VS2015 at work to build a fairly large project, mostly C code, some VB.NET, some C#. Used to be able to do a full rebuild in 43 seconds max, sometimes 41. It's now at 54 seconds steady, so I am looking at a good 25% reduction in performance. And no, Intel, it's not just artificial SSD benchmarks shitting the bed. This is real work.

This is a 5600-U CPU on a HP laptop using SSD, which is SATA. The model is a EliteBook 840 Gen2. It's running Win7 64-bit.

I will note that the Windows updates came in yesterday and I didn't see a measurable regression. Today I put on the HP BIOS and that's when the above was measured.

Intel can rub everybody's nose into those benchmarks they did, but as an end-user, I have to say this security fix really, really blows.

I understand your concerns. I'll be upfront here, I've been complacent in being vocal on this issue because I'm just dealing with a lot right now. ( Unemployment and unfair bias. ) On my 2015 Macbook Pro 13" with a i7 dual core and 3.1 Ghz base clock, I recently recompiled my gcc 7.1 set up. The compile time went from just under two hours to about two and half hours. Though in all honestly, I normally leave these kind of rare compilation jobs to run over night where I wouldn't notice or switch to the desktop rig. (In this case, trouble sleeping meant I would check on it every other time I fought to go to sleep.)

A bit of shame on me, because there is a vocal group running around the net (and here in Silicon Valley) calling bullshit on the performance impact and reciting Intel's PR.
 
Last edited:
I'm hoping some clarity comes soon. Some of these patches has hosed one of my boxes. I just want to move forward and continue to get some use out of this system.
 
Back
Top