@bozologic yes! the other rare good example. 👍

@JSON_Objects yeah I do a ton of urban recycling, just not totally close to my house. Usually a well trafficked (foot traffic) corner nearby

@MrKWatkins well yeah we're swimming in "free, but not really" ala Facebook and (waves hands around) everything else

@lizardbill I read it originally, was very mixed on it, tried re-reading it recently and gave up midway

@lizardbill that's good because I was kinda underwhelmed by the comic. "The Boys" is *DEEEEEPLY* flawed as a comic, to the point of being nearly unreadable. I mean it has the right themes but.. ugh.

"what if super hero, but deeply flawed" is (surprisingly) an underrepresented genre in comics. But you know if we *did* have super heros, this is exactly what we'd end up with.

@lizardbill He is so so sooo good in that role

@johnolilly Yeah the GaN stuff is worth the $. It's nice.

@Pinboard @tqbf but does it have an auto-renewing Let's Encrypt cert

@betaporter as I said, these are not subtle differences. Believe me, if they were within 20% I wouldn't care. That's a 3.6x time difference.

Double nerdiness today, just for you 🤓🤓 https://t.co/ZrRQ0ZIs7L

@kevinhooke @MaximilianoKorp 🤷‍♀️ I have a _lot_ of devices here and the scores are quite repeatable for me (assuming same OS version, same browser version). But I run squeaky clean defaults almost everywhere.

@MaximilianoKorp @kevinhooke make absolutely sure you're running a clean browser, no extensions. Ideally the "native" browser for your platform (Safari on iOS, Chrome on Android, even Edge on Windows since it's just Chrome anyways) Adblock extensions in particular hurt perf quite a bit.

@kwiens fuck yeah, sell *allll* the Apple things. Instead, cancel all the android things. kthxbye https://t.co/AtmnvIOdCO

@JoannaStern @MacRumors Monster Cables are back, baby!!

@seesharp I cannot wait to read this man‘s obituary

@MissQuickstep @DanielOCnr @ID_AA_Carmack @danielbowen our Discourse specific benchmarks match these results exactly, so that seems unlikely. But we are built on Ember.js, which is a part of the Speedometer 2.0 benchmark as well.

@DanielOCnr @MissQuickstep @ID_AA_Carmack @danielbowen geekbench is probably an optimization target for them already anyway since it's been around a looong time and it gets quoted a bunch of places. But per Intel, the smart money is on handing most of this logic over to the CPU since it can react faster and has all the data

@DanielOCnr @MissQuickstep @ID_AA_Carmack @danielbowen the general philosophy here is "don't try to outsmart the compiler", or in this case "don't try to outsmart the CPU designers". But when the CPU designers are as incompetent as Qualcomm, who knows, maybe

@MissQuickstep @DanielOCnr @ID_AA_Carmack @danielbowen that's usually decided by the scheduler automatically, all the "scale up under load" and "scale down when idle" magic. Much of this logic has been moved into the CPU so it can decide faster (it's closest to the data), e.g. Intel SpeedShift https://t.co/YbQebaEbjI

@betsyphd @seesharp 🤣🤣🤣🤣 https://t.co/H2noSUwnv0

@DanielOCnr @ID_AA_Carmack @danielbowen My theory is they're in a holding pattern while their custom silicon is being formulated. Hopefully sooner rather than later.

@kurko I want this on my tombstone

@DanielOCnr @ID_AA_Carmack @danielbowen that or Apple has hired CPU design geniuses, whose work nobody else can replicate. 🤷‍♂️

@DanielOCnr @ID_AA_Carmack @danielbowen I have been looking at this for *years* and as much as I have tried, the only conclusion I can inevitably reach is that Qualcomm really, really, *really* sucks at their job

@DanielOCnr @ID_AA_Carmack @danielbowen "Snapdragon 865 .. fabricated on TSMC's advanced 7nm (N7P) process, the 865 features four Kryo 585 Silver high-efficiency cores operating at 1.8 GHz along with three high-performance Kryo 585 Gold operating at 2.42 GHz and another prime Kryo 585 Gold core operating at 2.84 GHz"

@ID_AA_Carmack @bobbydigitales @danielbowen if there were other viable SoCs in android devices (which btw would be great for the marketplace) this theory would be easier to test! It's like an alternate WinTel universe where the "Tel" part is *very very bad*

@billisdoing @danielbowen pretty sure they are. My advice to them is this: build faster.

@ID_AA_Carmack @danielbowen the best you can really say is that, based on Geekbench at least, the 865 should be closer in JS perf to the iPhone 8 (895 vs 907), but somehow isn't. The 855 tracks exactly with the iPhone 7 (750 vs 728). Who knows, maybe the 875 will do.. something..

@ID_AA_Carmack @danielbowen Geekbench is another example, and it's not JS-specific at all. The differences in single-threaded perf are .. dire https://t.co/abq2FcGVkX

@ID_AA_Carmack @danielbowen in our experience, JavaScript performance scales almost linearly with single-threaded CPU perf in every scenario we've tested. There are some minor browser differences, e.g. Safari is indeed a bit faster, but you can't overcome massive single-threaded perf discrepancies.

Basically I'm not gonna be happy until every Qualcomm facility is razed to the ground, and the earth underneath plowed with salt so nothing can ever grow there again. But that's just me.

@sherifhanna What Qualcomm does do, is FUCKING SUCK AT THEIR JOB

@sherifhanna Qualcomm does not use stock ARM cores. "Qualcomm Kryo is a series of custom or semi-custom ARM-based CPUs included in the Snapdragon line of SoCs." https://t.co/5lHjD819Of

@1stdescents congrats, your 2016 device is literally faster than every other android device in existence 🥇

@danielbowen to be fair this is 100% qualcomm's fault, not Android. But it is an endemic problem, because what other SoCs are you gonna pick? There effectively aren't any.

@sherifhanna so is Apple. Research it.

and just because I love beating dead horses... on the left, Google Pixel 2XL (Snapdragon 835), on the right, Google Pixel 5 (Snapdragon 765G) both on Android 11, latest updates 🤦‍♂️ https://t.co/pZamJgbnUi

@jeffehobbs It'd be more "fair" if they bought them, like Instagram. Still bad for the world, just like the Instagram acquisition.

@jongalloway damn now I need to go get my ballot back from city hall so I can change it

@AllWeLikeWorms @Pinboard is there a difference?

@Pinboard you are just pissed because Pinboard wasn't in the list

@jlericson time to move

@amyjokim agree, with one clarification: the only insight that can be trusted is recording what they actually do.. https://t.co/ZWWcRAULTE

@mathowie @ZachWeiner https://t.co/Iz9gtd1qWL

@rlarno the rules vary but the success of "buying" a win is very much in question https://t.co/PXm8nKEkq2

@rlarno the candidates running the campaigns need money to pay staffers, place signs, make ads, run ads, etc. This money goes to the candidates in that US state.

@textfiles sure, I donate monthly, but here's another $100. Consider using @discourse as your long term memory alongside chat. Careful though, it's kind of expensive at $0/month.

@johnolilly https://t.co/gudjiIsUvB

@Pinboard @emilsit https://t.co/TURL4HBrO7

@Kip_Ricker @eknkc What's ESPECIALLY bonkers is the TDP. That's a 6w (prev-gen!) device versus a 105w device, and the 6w device is producing comparable single-threaded perf 🤯 .. Apple has had true magic on their hands for a while now. I can only imagine what A15 and A14X/Z are gonna be like

@nicksahler 😀

@refi_64 🙏

@mathowie well, that's it: you can never work anywhere but at home again

@slashequip TL;DR everyone's using the same Qualcomm SoC dreck in the WinTel (sorry, AndQual) ecosystem, so for a given model, simply reference the Qualcomm model number and you'll know exactly how bad its performance will be on the chart

@jwz @carlfish Show me on this doll where BillG touched you

@jwz @carlfish no snarky response to this one, @jwz? I am disappoint.

@slashequip I believe this is a question you can solve with a judicious application of Google ;)

@jwz @carlfish and damn I forgot to even bring up how evil Facebook is. Microsoft didn't destroy democracy, did they? I mean that wasn't even *possible* back then.

@jwz @carlfish disagree. The level of market dominance that Google currently has is easily an order of magnitude over what Microsoft had. It's not even close.

@sankara @carlfish Google has levels of market dominance that Microsoft could only *dream* of. Like an order of magnitude more.

@codekoan the (mildly) good news is that iPhone 7 performance is decent, at least

@furan geekbench single core scores scale *exactly* with Speedometer 2.0 in my experience

@jeteve @marceloid statistically, people use the default browser on the device. So it tracks. Turns out the default browsers are also *really fast* on their respective devices.. but gated by how crappy the SoC perf is (or isn't)

@sklivvz @codesgood *especially* on mobile! It's a bit more credible on desktop / laptop

@sklivvz @codesgood Regardless, "8 CPUs with really fast single-threaded perf" is always going to be a bigger win than "8 CPUs with really bad single-threaded perf". Arbitrarily adding more slow cores is basically a smokescreen, except for verrrry limited and specific workloads.

@sklivvz @codesgood It is true that once you get over 140, it's kind of a pointless metric. Looking at the existing data, Qualcomm should get there in .. approximately 2030. Once that happens, I promise on a stack of Bibles to shut up about this topic forever.

@sklivvz @codesgood it's not a "very specific workload", it iterates through several major, *extremely* popular frameworks for building JavaScript apps. Which are.. the future. And have been for a while.

@carlfish https://t.co/JRE6MZFWM8

@sklivvz @codesgood depends on your workload. Are you compiling code 24/7 on your android phone? Are you encoding movies 24/7 on your android phone? Are you raytracing 24/7 on your android phone? Then by all means, lots of slow cores is your huckleberry. For everyone else.. notsomuch

@carlfish well, wake me up when someone writes a book like this about Bill Gates https://t.co/iDPw4HaiAm

@marceloid I would avoid weirdo third party browsers. Test with the default browsers -- Chrome on Android, Safari on iOS. Edge on Windows 10 produces (shock!) nearly identical results to Chrome these days..

@arclight @Pinboard eh, at this scale, it's a game of inches. Little nudges like this do work, but it's also all too easy to create unintended consequences, so it has to be monitored closely

@carlfish I guess my observation is that it seems awfully quaint in comparison to what Uber / Oracle / Intel are doing *today* and *right now*. I'd also argue that the Google kind of anti-trust is way more significant in terms of actual effect on the world. Like people *really* use Bing?

@sklivvz @codesgood single threaded perf *IS* javascript perf, it's also 95% of perf for most daily computing tasks (and gaming, too). Literally nobody is encoding movies 24/7 on their Androids

@MaximilianoKorp @mortenmertner as I've already said, once qualcomm gets to ~140 I promise you on a stack of Bibles that I will shut up about this *forever*. Based on current trends that should happen around 2030

@taviso @migueldeicaza @Pinboard My first reaction when someone reports a serious sec vuln in Discourse is "how *dare* you!" but then the rational mind prevails and I'm like "thank you sooo much". Definitely organizational immaturity if you are getting the former and not the latter.

@lyda @Pinboard @migueldeicaza it's a fair criticism! But I loved GUIs and hated commandlines, so..

@Pinboard @lyda @jwz @migueldeicaza Build more Ubers and Oracles to compensate!

@CrazyWeeMonkey @vaultboy900 it was evident by the iPhone 6

@Pinboard @migueldeicaza to be fair prior to PowerShell the windows commandline wasn't *great*

@Pinboard @migueldeicaza honestly 99% of OSCON criticism can be summarized as "bah, the commandline sucks and we hate GUIs"

@Pinboard @migueldeicaza Even at drunk-Ballmer peak, Microsoft was never particularly evil. Compare with Oracle (true, pure evil), Uber, Intel, gosh, I could go on and on..

@henbas yes! this is the low-power form factor A14, not the tablet/laptop A14X, which will be even faster still!

@immrroj @dpa99c this is the "640kb should be enough for anyone" argument and it's a sucky argument. On top of that... https://t.co/gLbD9kzKxE

@RaghuRajput make sure you're running squeaky clean, no plugins, adblock, etc. My android device is one I use for testing so it is basically at complete and total factory defaults all the time.

@henbas a little of both, Safari is quite good here, but the SoCs apple is producing are *monstrously* fast in single-core perf, and getting faster every year

@eknkc it's true that numbers beyond ~140 don't really matter. The true sadness comes from realizing that the Qualcomm 865 pulls a whopping *65*. Believe me, once Qualcomm gets to 140 I'll shut up about this forever. Based on current trends that should happen in.. 2030?

@ktejnora It's mostly the weak CPU performance in the Qualcomm SoCs. To their credit, Google did do a big round of Android / V8 JS perf optimizations circa 2016-2017, but when your CPUs *SUUUUCK*, there's only so much you can do

@OleenaNatiras it's 100% consistent with our Discourse performance tests. Have a read.. https://t.co/ZWZdqmb2my

@__alter123 it's a 100% accurate depiction of JS perf on mobile. Source: we've been building one since 2012, and are experts on the topic. Have a read https://t.co/ZWZdqlTqXY

@slashequip The one in the chart is literally the Pixel 1, the first model. Basically just look up what Qualcomm dreck is in your device and you can predict the (poor) performance exactly

@k4shif 🙏

@KlausGradinger if you can't get more perf on mobile SoCs, when it's just about the only area of computing left that offers clear year-on-year performance wins, let me say this in the gentlest possible way I can: YOU FUCKING SUCK AT YOUR JOB (QUALCOMM)

Also, data confirmed: just ran this on Xiaomi Mi 9 (Snapdragon 855) Android device, updated to latest everything https://t.co/QPqRsXBJfv

@MaximilianoKorp (note that browser plugin based adblockers will definitely cost you browser perf as well. I suggest disabling them before running the benchmark.)

@CrazyWeeMonkey @vaultboy900 Google did not go with the 865 this round of "new" phones for some inexplicable reason. It's almost like there's a monopoly on Android SoCs and maybe that's not the best idea

@MaximilianoKorp We build a JavaScript app that runs in the browser, and have since 2012, and I can *assure* you this is 100% accurate depiction of JS perf on mobile. Have a read https://t.co/ZWZdqlTqXY

@jd87 even lower, since they didn't use the 865 or 855 this round 😱