Did anyone hear about this news? A number of sites have begun reporting about a very big security vulnerability which has been found in all Intel CPUs produced in the past 10 years. It’s a hardware, not a software level problem (pertaining to the architecture of the CPU itself) which if not patched can allow hackers to attack the PC. What’s worse is that if the vulnerability is patched (and it most certainly will or the PC will be vulnerable to hacks) this will cause anywhere from 5% to up to 30% performance decrease after patching depending on the specific model of your Intel CPU.
If true, doesn’t sound good at all. We’ve got some Intel machines used as render servers. Any ideas if this patch will also decrease Intel CPU performance in software such as Vray, 3ds Max, Photoshop or any other widely used ones? It is also said in the news that AMD processors don’t have this security problem so thankfully no patch for them is needed. I guess good thing we already replaced most of our machines with AMD ones a few months ago?
Not good at all, be interesting to see what happens. I think 90% of the world uses Intel! (total guess there) .. Anyway I’m hoping it’s an elaborate prank
OK but if this slowdown happens, dont we have the right to RMA this?
Weve spent almost 5.000 dollars on 2 XEONS and if they slow down by 30% they become quite useless for what we do with them. So in reality since they re under warranty we should either be eligible to get some money back OR return and buy the ones withtou such bug. Afterall we bought processors with certain speed parameters so if they dont have them, its the same issue as with all the VW cars etc.
Curious how this will work out, i will have to contact my reseller and perhaps also the lawyer to be prepared.
According to an article I’m reading “Intel said the actual impact would depend on the design of software programs, and that the worst affected ? those that need to make frequent calls on the operating system ? could be slowed by as much as 30 per cent”.
I would imagine this isn’t the case for V-Ray, but what about Max, other DCCs, Affinity, Photoshop…
Some benchmarks with KPTI-Patch (Windows 10 Insider Preview - Build 17063)
V-Ray seems not to be affected (at least when running the benchmark) - the newer build is even 1 sec. faster than the Fall Creators Update (1709)
(last application benchmark, above the CristalDiskMark benches)
That’s good to hear. But it’s best to see more tests with a wide variety of CPU models. They say the severity of the slowdown depends on the CPU model.
What about IPR/interactive rendering though? That’s a bit more different than plain old rendering which allocates the memory it needs once before it proceeds. I read that the slowdowns are much more pronounced when there is heavy constant communication between CPU and memory which I think is what happens in IPR/interactive rendering mode?
The problems don’t stop here, unforunately. If you are using an Nvidia card, it needs a security patch as well. A new driver update will be available tomorrow. Let’s hope GPU performance doesn’t degrade as a result of it.
And to make matters worse, the Windows update was just a small part of blocking the security holes. To be fully secured from the vulnerabilities, people with Intel based systems will also have to install special emergency BIOS updates for their motherboards after which, surprise-surprise, SSD performance may decrease by up to 41% Corona Benchmark also shows 3% slower results:
This image is also interesting and shows bad consequences after fully securing with both the Windows update and the BIOS update (chart taken from here):
I did install the BIOS update for my laptop, as well as the relevant Windows updates; I haven’t measured if SSD speed was affected, but rendering speed was not (once everything that’s needed was loaded in RAM).
That’s reassuring. I guess time will tell what applications get f***** the most by this. Well, this stuff happens. It’s not nice but it happens. But that SSD speed decrease gave me goose bumps.
Yes, but I still think in the majority of cases for desktop use and render farms without virtualization there won’t be significant issues. Some disk operations might be somewhat slower, but I don’t know if the difference will be that obvious. There is also this:
where the performance impact in actual use cases is quite small.
We ran some tests and the pre-fix render time was 43:19 and the post-fix render time was 45:35. On a mega long render the pre-fix time was 6:40:38 and the post-fix time was 6:43:02. So even with a long running render, the time difference wasn’t huge or increasing. Though, I’ve heard rumblings that DR will be more affected than just single machine rendering. Still haven’t tested that yet.