> A weeks ago, my friends and I were talking about the inner workings of Zen 5. We were talking about how the CPUID instruction works, and how AMD MSRs are technically editable if you ask the processor nicely.
I don't know about your friends, but as we're on HN, I'm sure others here have friends like mine, who absolutely have conversations about how the low level shit that facilitates our world works.
1. Start building stuff that is hard to build that requires touching these niche topics. Especially stuff you don't know how to build
2. As you encounter problems, you'll have to scour for solutions (AI doesn't know these things due to lack of training data). In the process you will find people who are also working on these problems. Ask these people well-formed, intelligent questions.
I am skeptical of your second claim here… if you can “scour for solutions”, and you find something about it on the internet, then AI could find it the same way.
Just cause the AI could find the info definitely does not mean it will find and apply that knowledge correctly to solve a problem.
I find AI shockingly bad ad searching the web, as SEO blogspam sites heavily pollute AI context windows, while relevant and important resources are typically very densely presented reference material which must be constantly revisited.
It doesn’t need to. It has already all the fundamental knowledge it needs. Just set it up on a system with an editable proc file system and it would be able to figure it out.
By 'talk' I suspect he means discord and by friends he means display names. Not that there's anything wrong with that. I catch myself saying 'talk' when I'm talking about something a friend told me over chat.
>> A weeks ago, my friends and I were talking about the inner workings of Zen 5. We were talking about how the CPUID instruction works, and how AMD MSRs are technically editable if you ask the processor nicely.
> As do we all.
I think they interpreted “as do we all” as pointing out humorously that this is an unusual friend group. So, speculation that it might have formed online makes sense, because online spaces can sometimes facilitate that sort of thing.
They are two different situations but why is the distinction meaningful here? I rarely even remember the venue of most conversations, just that it happened.
Because for a lot of us it’s hard to imagine finding a half dozen or so people who can talk in person like that outside of a conference or workplace. Discord facilitates that because it assembles people based on interests and such. You’re just going to more easily have that kind of conversation than you would “out in the real world.”
My guess is they are functionally saying “this probably happened on discord if anybody is wondering how this is even possible and not made up for effect” but I might be interpreting too much
My friends and I share interests but they can’t all talk about the relative pros/cons of full frame vs. cropped sensors in digital cinema with me. That’s kind of the framing here if that makes sense. We share a lot of interest and can talk in depth about certain topics, but there are plenty of topics that I am interested in or just know a lot more about that none of them can really discuss with me, so I have to find those communities elsewhere
Presumably because they all traveled there for a temporary part of their lives. And after university, they presumably scattered to the places where they built their careers and families.
> so, to test, one of us took a heavily PBO'd 9700X and changed /proc/cpuinfo to be a "9700X3D" and ran a Passmark run to see if the software would be fooled...
The two articles I saw about this both emphasized that the high clock speed (from the PBO) was inconsistent with the name of the CPU that implied it would be lower performance than the 9800X3D.
Most of the sites I check regularly have been pretty good about calling out inconsistent leaks or rumors, contrary to the “all journalism is trash” comments down below. On the other hand, if you were following someone who presented this singular benchmark result as proof of something without looking at the details, it might be a good time to reconsider the quality of your sources. I did see some lazy Twitter personalities parroting the result without any actual thought.
This is all extra confusing (as to why people republished this) because a 9850X3D was already rumored a couple weeks ago as a higher binned 9800X3D, which would actually make sense, as well as a 9950X3D2 with dual X3D CCDs.
> I feel badly for all of the people who may have held off on a 9800X3D purchase because of this Passmark that we thought wouldn't work.
I'm considering a new build soon, but RAM prices are out of control, like they've more than doubled since June! (Damn AI bubble...) I guess I'll have to get by with my Ryzen 1800X a bit longer.
I’m rocking an AM4 build still and very happy with the bump I got from going to a 5800X and maxing out the RAM (primarily for productivity use rather than gaming).
A major takeaway from this is that the news media can easily be misled and report false information. Everyone sees this whenever there's a news article in a field they are an expert in, but then they trust all of the other articles in fields they are not.
> The Gell-Mann amnesia effect is a cognitive bias describing the tendency of individuals to critically assess media reports in a domain they are knowledgeable about, yet continue to trust reporting in other areas despite recognizing similar potential inaccuracies.
I was hoping for a slightly budgetier X3D chip but I went and got a plain 9700 a few months ago. I realized I probably don't need the performance and the extra power budget/efficiency of using a 65W chip was nice.
If a tech journalist actually understood their subject matter at a competent level, they wouldn't be a journo and go out and get a job in that subject matter dramatically increasing their salary. Those that can't do, teach. Those that can't teach, become journalists???
I sometimes use lwn.net as an exemplary showcase of things non-tech journalists should learn (e.g.: add references whenever paraphrasing material some or all readers might have direct access to)
As do we all.
1. Start building stuff that is hard to build that requires touching these niche topics. Especially stuff you don't know how to build
2. As you encounter problems, you'll have to scour for solutions (AI doesn't know these things due to lack of training data). In the process you will find people who are also working on these problems. Ask these people well-formed, intelligent questions.
I find AI shockingly bad ad searching the web, as SEO blogspam sites heavily pollute AI context windows, while relevant and important resources are typically very densely presented reference material which must be constantly revisited.
[1] https://github.com/AngryUEFI/ZenUtils
Other examples include "let's build a submarine" https://media.ccc.de/v/37c3-11828-how_to_build_a_submarine_a..., creating your own 2000s style phone ringtone/wallpaper subscription service https://blamba.de/ or running toslink audio over regular long-distance fiber links https://blog.benjojo.co.uk/post/sfp-experiment-ultra-long-ra...
> As do we all.
I think they interpreted “as do we all” as pointing out humorously that this is an unusual friend group. So, speculation that it might have formed online makes sense, because online spaces can sometimes facilitate that sort of thing.
My guess is they are functionally saying “this probably happened on discord if anybody is wondering how this is even possible and not made up for effect” but I might be interpreting too much
I have good friends that love to discuss highly technical topics over a beer or whiskey.
Ahhh, I see, I see...
Clearly proximity is involved
The two articles I saw about this both emphasized that the high clock speed (from the PBO) was inconsistent with the name of the CPU that implied it would be lower performance than the 9800X3D.
Most of the sites I check regularly have been pretty good about calling out inconsistent leaks or rumors, contrary to the “all journalism is trash” comments down below. On the other hand, if you were following someone who presented this singular benchmark result as proof of something without looking at the details, it might be a good time to reconsider the quality of your sources. I did see some lazy Twitter personalities parroting the result without any actual thought.
"New CPU in Passmark" news has become so regular, I've long since assumed that they are not leaks at all, but intentional product hype.
EXIF metadata is editable, too. Similar that it could be useful intelligence, but it is very easy to deceive others with it.
The mainstream journalism about this was actually pretty good
No
I'm considering a new build soon, but RAM prices are out of control, like they've more than doubled since June! (Damn AI bubble...) I guess I'll have to get by with my Ryzen 1800X a bit longer.
I hope I'm never working on a project where -j32 on a 5900XT is noticeably faster than than -j16 on a 5700X3D.
Then again, -j1 is nice, when you need to time a break. (https://xkcd.com/303/)
https://en.wikipedia.org/wiki/Gell-Mann_amnesia_effect
Clearly there's a market for a 9700X3D though!
Or do you think journalists are going to wait for ‘peer review’ for their breaking news?