|
Trimson Grondag 3 posted:not too much apparently: tools people need to do actual work and everything which touches a database (i.e. every cloud and web thing) is going to be slower, but never fear, the video games are fine
|
#
?
Jan 3, 2018 07:55
|
|
|
|
| # ? Apr 24, 2024 02:28 |
|
|
will this affect my blast processing
|
|
#
?
Jan 3, 2018 07:55
|
|
|
H.P. Hovercraft posted:will this affect my blast processing why don't you process this blast https://www.youtube.com/watch?v=jLKnCeeAW48
|
|
#
?
Jan 3, 2018 08:00
|
|
|
The_Franz posted:tools people need to do actual work and everything which touches a database (i.e. every cloud and web thing) is going to be slower, but never fear, the video games are fine good, cloud and web apps are dumb
|
|
#
?
Jan 3, 2018 08:13
|
|
|
as an end user is it another thing you defeat by not clicking on random attachments and disabling all js by default ?
|
|
#
?
Jan 3, 2018 08:32
|
|
|
this intel thing has been known about for quite some time, it just finally hit the news cycle, or someone has successfully leveraged it in the wild. Breaking Kernel Address Space Layout Randomization with Intel TSX, Blackhat USA, 2016
|
|
#
?
Jan 3, 2018 08:33
|
|
|
orange juche posted:this intel thing has been known about for quite some time, it just finally hit the news cycle, or someone has successfully leveraged it in the wild. bingo on the second one
|
|
#
?
Jan 3, 2018 08:39
|
|
|
Great, another perf hit was just what my poor little ultrabook needed when compiling C++ projects 
|
|
#
?
Jan 3, 2018 09:46
|
|
|
fortunately, i found a dec alpha multia at work last saturday. i'm secure, and with no performance impact
|
|
#
?
Jan 3, 2018 10:42
|
|
|
does anyone know if this affects newer xeons as well, or do they not have the same issue?
|
|
#
?
Jan 3, 2018 13:27
|
|
|
|
|
#
?
Jan 3, 2018 15:51
|
|
|
tk posted:Oh, good, so just literally everything that I work on. 
|
|
#
?
Jan 3, 2018 15:51
|
|
|
|
|
#
?
Jan 3, 2018 15:53
|
|
|
until you save it to disk 
|
|
#
?
Jan 3, 2018 17:14
|
|
|
orange juche posted:this intel thing has been known about for quite some time, it just finally hit the news cycle, or someone has successfully leveraged it in the wild. they disabled tsx a long time ago
|
|
#
?
Jan 3, 2018 17:15
|
|
|
i can't wait for my $20 check in the mail
|
|
#
?
Jan 3, 2018 17:15
|
|
|
hifi posted:i can't wait for my $20 check in the mail
|
|
#
?
Jan 3, 2018 17:20
|
|
|
i bought an intel processor and a loaf of bread so my mailbox is gonna be a fuckin payday also lomarf i got a new board & chip this weekend and i took the 8700k instead of a ryzen 1800x 
|
|
#
?
Jan 3, 2018 17:30
|
|
|
Raere posted:im betting on risc v sounds risky
|
|
#
?
Jan 3, 2018 17:42
|
|
|
lol I bet this is going to turbofuck a bunch of IPS platforms as well
|
|
#
?
Jan 3, 2018 18:10
|
|
|
https://www.computerbase.de/2018-01/intel-cpu-pti-sicherheitsluecke/  der plural auf anekdoten nicht ist data, etc etc 
|
|
#
?
Jan 3, 2018 18:13
|
|
|
yes, things that spend 99.9% of their time doing computations in user space are going to be unaffected. great job everyone.
|
|
#
?
Jan 3, 2018 18:39
|
|
|
|
|
#
?
Jan 3, 2018 18:40
|
|
|
https://www.fool.com/investing/2017/12/19/intels-ceo-just-sold-a-lot-of-stock.aspx Intel CEO sold a poo poo ton of stock on11/29.... JAIL TIME!
|
|
#
?
Jan 3, 2018 18:47
|
|
|
Rivethead posted:https://www.fool.com/investing/2017/12/19/intels-ceo-just-sold-a-lot-of-stock.aspx lol
|
|
#
?
Jan 3, 2018 18:48
|
|
|
The Management posted:yes, things that spend 99.9% of their time doing computations in user space are going to be unaffected. great job everyone. i thought file io needed a syscall quote:But there is still a bigger rash to be seen. On a second SSD in the system in the form of a Samsung 960 Pro in M.2 storage space, the CrystalDiskMark shows a difference of almost seven percent in two out of eight calculated values (screenshot), with another two to four percent behind. This is consistent with the database benchmarks linked above. legitimately asking, what sorts of things other than databasey stuff would beat us to death with context switches
|
|
#
?
Jan 3, 2018 18:50
|
|
|
https://www.phoronix.com/scan.php?page=article&item=linux-415-x86pti&num=2 owning some amd stock feels great right now
|
|
#
?
Jan 3, 2018 18:52
|
|
|
flakeloaf posted:i thought file io needed a syscall it does, but things like games and file compression utilities tend to read and write large chunks of data with fewer io calls and spend most of their time doing operations on that data the other big area which takes a hit is heavy compilation workloads which spawn lots of processes, each of which do lots io operations to resolve dependencies and whatnot.
|
|
#
?
Jan 3, 2018 19:15
|
|
|
flakeloaf posted:i thought file io needed a syscall one thing to note is that while the kernel has to get unmapped from the process the reverse is not true, so the base cost of a syscall is getting cranked up (changing page tables is expensive since it invalidates tlb) making a *large* syscall gets no more expensive, since the kernel can still scoop a huge write out of the process memory, or a huge read into it, so as long as requests are bulky it'll work out fine beyond high loads of small pieces of io from databases a lot of random software will no doubt suffer if it relies on a bunch of small kernel services like synchronization primitives or monitoring file system events or tiny ipc that used to be a lot cheaper
|
|
#
?
Jan 3, 2018 19:31
|
|
|
that would explain why fsmark's "create 5000 1mb files really fast" now takes twice as long on an nvme drive  but really how often do i expect to be doing that yay for generous return policies i guess; the last amd cpu i had was.... uhh..... an athlon xp decisions decisions
|
|
#
?
Jan 3, 2018 19:34
|
|
|
so does this affect torrentings
|
|
#
?
Jan 3, 2018 19:37
|
|
|
also whats the energy impact of branch prediction and on like phones is it dignificantly
|
|
#
?
Jan 3, 2018 19:39
|
|
|
flakeloaf posted:i thought file io needed a syscall yes, file IO is a syscall but when you’re encoding video you tend to read large chunks (or mmap them) so the number of times you enter the kernel is very small compared to the time you spend doing math in a loop in user space. as said before databases do a ton of IO, but so do servers of just about every kind. other things that will hurt are things that are heavily IPC-based. lots of middleware stuff. any time you have two processes communicating with a high throughput that’s going to take a hit. more generally you can say that all software that isn’t compute intensive will probably suffer. the degree to which it will varies.
|
|
#
?
Jan 3, 2018 19:43
|
|
|
echinopsis posted:also whats the energy impact of branch prediction and on like phones is it dignificantly without branch prediction and speculative execution your phone is useless garbage.
|
|
#
?
Jan 3, 2018 19:46
|
|
|
Rivethead posted:https://www.fool.com/investing/2017/12/19/intels-ceo-just-sold-a-lot-of-stock.aspx executives need to file a declaration of their sales schedule in advance and are legally required to stick to it. this would only be a problem if he didn’t declare these sales in advance.
|
|
#
?
Jan 3, 2018 19:50
|
|
|
i see intel has found a subtler way to stave off antitrust investigations than cash payments to amd. good on them
|
|
#
?
Jan 3, 2018 20:00
|
|
|
it looks like the patch that disables page table isolation for amd is making it into the kernel. it might not be merged until the 4.16 release though
|
|
#
?
Jan 3, 2018 20:01
|
|
|
where intel is going to get hosed is their data center customers. amazon, google, and Microsoft are not going to like getting significantly less performance from the chips that they bought, and they buy the highest end most expensive chips. they are going to demand some big discounts to make up for this.
|
|
#
?
Jan 3, 2018 20:10
|
|
|
and then consumers are going to say "well what about us" and hopefully intel has something more intelligent to offer than "well if you can prove you need to divide 4005FB by 2FFFFF..."
|
|
#
?
Jan 3, 2018 20:13
|
|
|
|
| # ? Apr 24, 2024 02:28 |
|
|
flakeloaf posted:legitimately asking, what sorts of things other than databasey stuff would beat us to death with context switches OpenCL and CUDA?
|
|
#
?
Jan 3, 2018 20:33
|
|
