So this is this decade's Y2K.
So this is this decade's Y2K.
(\__/)
(='.'=)
(")_(")
Been helped or just 'Like' a post? Use the Thanks button!
My broadband speed - 750 Meganibbles/minute
I'd be careful saying that mate, an engineer will pop out of a socket and knife you.
The only reason Y2K didn't come to be a problem was because of the millions of hours of work that went on, around the clock, by dedicated engineers, coders, programmers, and various other folks. It was an absolute massive effort that was pretty much a complete success
This is nothing like Y2K: Y2K was a very real issue, that could be fixed. This is being mitigated. Only way it'd be like Y2K would be if at the stroke of midnight all clocks started running slightly slower and there'd be feck all we could do about it
1. Most of the Internet runs on LAMP stacks. That's Linux (The OS Devs that found this), Apache, MySQL, PHP. All of those need to do two things: Read and Write to Memory and Hard Drive. When you take all of the little slow downs, they stack up.
2. The fix has been found to kill scientific/mathematical calculation speed. While the calculations still occur, they are slow, and if software was specifically designed or engineered to get a response in a certain amount of time, then you need to completely redesign the software, and any time dependent observations may be affected up to 30% or more (especially if virtualized).
3) The fields most affected by math and science calculations off the top of my head are most publicly funded science research, medical imaging, financial databases, etc.
Fortunately, you will only experience a partial slowdown if you visit websites that have AMD hardware. Unfortunately, most of the enterprise level world runs on Intel.
Couple this with the insider trading and the known security flaws of Intel, and you have Intel getting BTFO'd, if you will.
If Intel rallies from this breech of trust, I will be astounded.
I haven't chimed in on this yet though I've been following it closely.
I would just warn you that there are certain performance regressions from the OS patches, but there are some other (possibly larger) performance regressions for 'Intel' platforms from the also required firmware patches, most of which are missing right now, except if you have certain Asus boards as far as I know. So I would expect the real impact of this to come through only as the firmware / BIOS patches appear alongside the OS patches, whether that be Linux Windows or Mac/others.
Keep in mind the change numbers you see now may not be the actual outcome once vendor bios fixes hit.
edit: it is still likely to mainly hit certain server workloads though
Last edited by Millennium; 07-01-2018 at 04:30 PM.
hexus trust : n(baby):n(lover):n(sky)|>P(Name)>>nopes
Be Careful on the Internet! I ran and tackled a drive by mining attack today. It's not designed to do anything than provide fake texts (say!)
Last edited by the_avatar; 08-01-2018 at 02:19 AM. Reason: typo
Oh no, the world will end!!! Much like Y2K was over blown. And no, that's not to suggest this was/is a non issue, just that it got over blow a bit, and allowed all the Intel HATERS another reason to freak out over. Now, the nit picking begins.
The 30% performance hit figure was not a prediction, it was a measurement of a real workload. Compiling code in Linux matters to some of us as that is basically our job.
I know quite a few people who are running around dealing with this. Just because it doesn't impact gamers (much) doesn't make it unimportant.
They've been designed to prevent, or rather mitigate against it, on a hardware level for over six years, unfortunately because the original core design has been around a lot longer than that it contains legacy "hardware", that's to say Intel introduced x86 instructions and the ability for the hardware to process those instructions many years ago but because there wasn't a known way to exfiltrate data from the processors local memory during speculative execution, and because writing extra code that excluded certain types of data from speculative execution takes time, i suspect, most developers didn't bother using the new instructions.
MFENCE, LFENCE, INVPCID and the associated PCID (Process-Context Identifiers) have been around for ages.
There are currently 1 users browsing this thread. (0 members and 1 guests)