Well this took an interesting turn.
https://old.reddit.com/r/hardware/comments/bos7y5/intel_tried_to_bribe_vu_university_amsterdam_into/
Security has always been an afterthought when designing CPUs
Free promotion for the new Ryzen CPUs most likely releasing in two weeks.
I used to have that CPU with terrible thermals (due to intel toothpaste) and now have a 2700x and I couldn't be happier with this beast.
Current computer uses an Intel CPU and was built back in 2015 which was before AMD announced Ryzen.
Will definately be going AMD on my next computer build, and probably won't be switching back to Intel afterwards.
You can bet that Intel is diverting a deluge of dollars into researching AMD uarch vulnerabilities.
SMT is a generally leaky technique -- for example PortSmash measures timing differences in arbitration of the execution units between two simultaneous threads, to sniff private data. Their proof-of-concept was on Skylake + Kaby Lake, but they explicitly call out AMD Ryzen as an open question, and a topic for future research.
A comment by the author of that paper, from a 2nd November 2018 article:
"Security and SMT are mutually exclusive concepts," he added. "I hope our work encourages users to disable SMT in the BIOS or choose to spend their money on architectures not featuring SMT."
Note he says SMT in general, not HyperThreading specifically.
The Microarchitectural Data Sampling (MDS) issue is a speculative execution side-channel attack that may allow malicious actors to locally execute code in order to extract sensitive data that would otherwise by protected by Intel processors’ architectural mechanisms.
This had me mystified since the beginning of the Spectre/Meltdown debacle. How exactly does a processor architecture protect data in the first place? Can someone explain in more detail?
It's a 2ghz core duo (or Dual core) 1st gen with 2gb DDR2 ram. It's not amazing. but its specs also put it just outside the useless zone. It could conceivably be used for light programming or a word processor for a retired person.
A few short months back I decided to buy a big dick computer because fuck suddenly earning some good money felt amazing and dangit imma be one of those money before reason people just this once.
it wasn't long after the i9 9900 was announced so I wait a little while and the reviews be like "yeah this shit is expensive but owns fukken bones" so I decide to splurge, I highly doubt I'll see another point in my life when I can drop £600 on a cpu without jeopardizing other financial obligations.
And here we are, even after holding off a little to let the hype die down suddenly I get a fucking surprise wedgie for my efforts.
i've not used amd since the other side of 2010 but fuckin' hell, time to go back next upgrade methinks.
Apple just dropped their patch, and it causes up to FORTY PERCENT perf drops
https://support.apple.com/en-us/HT210107
Sounds like we know how much money Intel owns us in a class action now.
I wish more laptops had AMD processors.
AFAIK, that is only if you follow their guide to actually disable HT. Otherwise, they don't disable it in the patch.
Glad I went with Ryzen. Then again when you get a deal like a Ryzen 5 2600 for $116, you take it
Aaaaa my laptop is going to be a true dual core now. No good for running tests on, I'm already spoiled by my 16-thread R 1700. Time to look at those new Ryzen ThinkPads!
One mechanism is memory protection with paged virtual memory. Each process has a virtual address space which it keeps all of its data in; the hardware maps this onto actual physical memory and system resources, and the OS manages those mappings. The mapping is transparent: user processes only see the virtual space, and can't see details of the physical machine. By default the physical regions mapped to are non-overlapping, so if one process browses through its address space, it will never see data belonging to another process.
The Spectre attacks partly revolve around an attacker process tricking another process into accessing some of its own secret data, in such a way that the data "leaks" out. That is, the machine's state is subtly changed in a way that depends on the secret data. The attacking process can detect this subtle change (e.g. a certain line was evicted from the cache, changing the attacker's program execution timing) and then reason about what the secret data must have been.
Properly designed software will check its inputs and make sure nothing nefarious is going on. Unfortunately, for performance reasons, processors will speculate as to which way the check will go, and run ahead and start executing its best guess as to what the next code will be. The attacker passes some data into the victim process that causes a speculated access to the secret data. In some cases the attacker can load the dice and make the access more likely to be mis-speculated.
Once the CPU figures out it guessed wrong, it discards data it shouldn't have accessed. Unfortunately, if the secret data is (for example) used to calculate the address of another access, this access will have side effects on the caches, with the exact side effect depending on the value of the data.
https://files.facepunch.com/forum/upload/238785/8a6cc6fa-dcb7-4be6-a27d-5ea8ece2d9e9/image.png
BUY HIGH SELL LOW
Good thing I'm considering shifting to Zen2 when it launches later in the year to replace my 3570k, though I suppose this is a natural consequence of Intel's stagnant position on the market. While I was a bit young to be into the processor scene at the time I've an uncle who swore by his Athlon X2 and I'd like to think AMD can claw its way back to the position it had before and smash Intel on both a personal and professional market level, bringing back some solid competition and giving them a reason to innovate and massively improve instead of sitting on the same architecture like they have for the last decade
9th generation probably doesn't have HyperThreading because they've known for a while now that everyone knows how broken it is.
Sorry, you need to Log In to post a reply to this thread.
