geekhack
geekhack Community => Off Topic => Topic started by: tp4tissue on Thu, 18 January 2018, 18:00:52
-
Slow your computer down by up to 7% in Day to day applications,
Thanks Intel, love your CPUs
http://www.catalog.update.microsoft.com/Search.aspx?q=KB4056894
http://www.catalog.update.microsoft.com/Search.aspx?q=KB4056897
-
jokes on you guys i'm still on XP
-
jokes on you guys i'm still on XP
XP affected
-
what does this fix?
-
what does this fix?
So... cpu makes prediction, this process can be intercepted and they can potentially keylog you from javascript banners.
As long as you're a noscript guy, the risk is minimal, BUHHHHhhhhh it's still recommended because this type of attack is bound to become ubiquitous...
-
jokes on you guys i'm still on XP
XP affected
oh
guess i'll downgrade to ME then
-
jokes on you guys i'm still on XP
XP affected
oh
guess i'll downgrade to ME then
If xp is affected, then everything with NT4 kernal is affected..
You gotta go down to 95 probably..
-
ooff.. already feeling the sluggishness in PC response..
dat 2% hits hard... [attachimg=1]
-
jokes on you guys i'm still on XP
XP affected
oh
guess i'll downgrade to ME then
If xp is affected, then everything with NT4 kernal is affected..
You gotta go down to 95 probably..
95 it is then
i hope i can still install Docker and vscode
-
ZOMG.. 20% reduced 4K random read on my SSD..
WTF intel, wtf.....
-
ZOMG.. 20% reduced 4K random read on my SSD..
WTF intel, wtf.....
Keep in mind this is just the first patch. Intel put it out as a quick CYA thing due to the lawsuits coming out. Intel likely won't shoot themselves in the foot by not addressing these performance hits, so expect them to do some microcode updates in the future.
The question is how far back these updates will go, though. The initial sets of patches are going to go back as far as the x99 platform, but this is Intel. They may try to force older platform users to update to a new one.
-
ZOMG.. 20% reduced 4K random read on my SSD..
WTF intel, wtf.....
Keep in mind this is just the first patch. Intel put it out as a quick CYA thing due to the lawsuits coming out. Intel likely won't shoot themselves in the foot by not addressing these performance hits, so expect them to do some microcode updates in the future.
The question is how far back these updates will go, though. The initial sets of patches are going to go back as far as the x99 platform, but this is Intel. They may try to force older platform users to update to a new one.
I doubt it..
I'm actually considering shorting their stock, hahahaha
If you think about it, given how wide this is, and how a HUGE drain on productivity this may turn out to be,
Enough class action lawsuits in the right places, Intel could have trouble making money for a year or two..
But that's assuming they don't release their Super secret ultra fast CPU w/ 200% IPC improvement..
Because we know they've got it..
-
OVERHEARD AT INTEL, PROBABLY
Engineer 1: Hey, this processor speed is amazing!
Engineer 2: Yeah, but there is a big security hole, due to the prediction threading.
Engineer 1: Does anyone else know about the security flaw?
Engineer 2: No, they wouldn't even expect where to look for it.
Engineer 1: So, nobody would find out, at least for a few years?
Engineer 2: I guess not...
Engineer 1: And we get a bonus based on performance specs?
Engineer 2: That's right, but...
Engineer 1: We have ourselves a winner!
-
Meanwhile Coreda wonders the significance of the color teal
-
the real joke is, it seems likely that if you have anything older than Haswell-E you are ****ed for a real spectre bios patch... pretty much ever..
https://www.hardocp.com/news/2018/01/17/uefi_bios_updates_for_spectre
according to kyle at [H] motherboard makers are saying not to expect updates for older platforms.. so desktop haswell ,ivy sandy..etc all pretty much screwed..possibly.
-
the real joke is, it seems likely that if you have anything older than Haswell-E you are ****ed for a real spectre bios patch... pretty much ever..
https://www.hardocp.com/news/2018/01/17/uefi_bios_updates_for_spectre
according to kyle at [H] motherboard makers are saying not to expect updates for older platforms.. so desktop haswell ,ivy sandy..etc all pretty much screwed..possibly.
I was afraid of this. I have a 3770K (Ivy Bridge) on a Z77 motherboard. The last BIOS update is from 2014. I wasn't planning on upgrading anything but GPU for several more years, but now I'm definitely not upgrading until this design flaw is fixed in hardware.
-
the real joke is, it seems likely that if you have anything older than Haswell-E you are ****ed for a real spectre bios patch... pretty much ever..
Yeah, X99 is the oldest platform getting patches, from what I've read. That includes both Spectre (which affects all brands that use branch prediction) and Meltdown (which is Intel and ARM only).
The good news is most major browsers are working to mitigate the Javascript issues Spectre presents, which is the easiest way to use the exploit. And at least the Spectre patch issues are understandable--there are far too many patches needed to fix it, so they have to draw the line somewhere.
-
jokes on you guys i'm still on XP
Inb4 microsoft windows.
You should never have upgraded from Windows 98.
-
the real joke is, it seems likely that if you have anything older than Haswell-E you are ****ed for a real spectre bios patch... pretty much ever..
https://www.hardocp.com/news/2018/01/17/uefi_bios_updates_for_spectre
according to kyle at [H] motherboard makers are saying not to expect updates for older platforms.. so desktop haswell ,ivy sandy..etc all pretty much screwed..possibly.
I was afraid of this. I have a 3770K (Ivy Bridge) on a Z77 motherboard. The last BIOS update is from 2014. I wasn't planning on upgrading anything but GPU for several more years, but now I'm definitely not upgrading until this design flaw is fixed in hardware.
A FELLOW 3770Ker **** YEAH
-
using x86 in the current year come on guys step it up
-
There should not be a reason to use an Intel CPU, but there are so many issues.
I use an Intel instead of a AMD because of AMD's unwillingness to acknowledge a bug in Ryzen (https://www.phoronix.com/forums/forum/hardware/processors-memory/955368-some-ryzen-linux-users-are-facing-issues-with-heavy-compilation-loads/page7) that affects mostly Unix/Linux users, and because of lack of good mATX and mITX motherboards for the AM4 socket. (That was the case when it was time to buy, anyway. AMD looks better now)
I use x86-64 because of hardware and software support. Not just Linux distributions but I would also like to play some games for which almost all require x86 and many require Windows (which I would only run virtualised, of course).
I would have liked to build a modern MIPS or SPARC system into a vintage SGI or Sun case (respectively) that I have, just for the principle, but there are also no modern high-performance MIPS or SPARC systems with good motherboards, etc. so it wouldn't be more useful than at best a Raspberry Pi but at many times the cost.
A 64-bit ARM-based PC would be cool. It is the most capable, modern CPU architecture out there today ... but I want a proper PC running a free Unix variant, not a locked-in throttled iPad or Android device that runs on battery, with deficient passive cooling, with planned obsolescence and no ports.
In the long-term I would like to use The Mill (https://millcomputing.com/) but it is still years from silicon. Maybe a prototype in FPGA this year though. It should run faster and cooler than any x86-64 or AArch64 and it is impervious to Meltdown, Spectre and return-oriented programming.
I follow this closely and if I had the money, I would invest in it.
I (and many others) have long had a dream of CPU instruction architecture begin irrelevant, that operating systems would instead use a architecture-neutral binary format that would be compiled to native code first when a program is installed. The goal would be that programs could be developed and tested on one architecture and then be deployed on any CPU out there with no compatibility issues.
I have revisited the problem now and then over the years. Fifteen years ago, GCC was unfortunately the best framework to start with and it was a tangled mess... Now, there is LLVM (which is not a VM), but while it could be useful, it is far from a solution. The problem of making software safe, secure and bug-compatible over CPU architectures is Hard. The more you dwell into the problem, the more and more esoteric technical issues you will find, and the more you will add to your TODO-list of software in the tool-chain needed to support this.
-
COME ON Ryzen... beat dat intel-bastard back..
Ya'll remember overclocking dem 2.4ghz pentium 4s on a $60 motherboard to 3.2ghz ?
Dos' were the days.......