|
SwissArmyDruid posted:It's like "does my ext4/5 need ECC, because btrfs doesn't need ECC" questions all over again. I mean, the ideal scenario is creating for yourself the situation where workstation parts are things you might conceivably need and then just using the computer to push entertaining pixels. Throw ECC in there for extra good renders. I could get by on a 4C/8T machine with my current GPU for my CAD and video games. But now, hardware availability might let me get up to a 32C/64T Threadripper machine, and hopefully that Navi GPU with 1080 perf @ $250 happens, because then I can virtualize my CAD environment in Windows and render in whatever Linux distro I choose with all those cores. And write it off on my taxes. And then play video games. Thank you AMD 
|
#
?
Jan 4, 2019 23:19
|
|
|
|
| # ? Apr 29, 2024 13:26 |
|
|
PC LOAD LETTER posted:Eventually but it'll be years before it does. Disagree: it won't happen until compilers or perhaps the major engines start having really good, mostly-automated methods to create multiprocessing optimizations. Which is kinda like saying it won't happen without a magic wand. Multi-threaded programming is real hard, games are not naturally suited for it, and studios have limited resources especially in the true programmer department. History has shown that just handing some new hardware to game developers does not result in them having the ability or resources to meaningfully take advantage of it. Combat Pretzel posted:Yeah, I don't get the drama home users create about these exploits. Enthusiasts who read the tech news just pick up on the general level of hype and alarm that the industry covers it with. And those exploits got some real "the sky is falling" type coverage because the biggest attack target is The Cloud, which the industry as a whole just spent billions of dollars in advertising alone to sell everyone on. If the Clown were not safe it could cause a massive industry crash. OTOH meltdown really would have been a giant problem if they hadn't had fixes ready to go by the time of publication. A webpage running javascript could scan mamory or keylog your whole system from inside the browser sandbox, that is a huge vulnerability for home users. Anyone running intel systems who intentionally disables the meltdown mitigations is a fool.
|
|
#
?
Jan 4, 2019 23:52
|
|
|
NewFatMike posted:I mean, the ideal scenario is creating for yourself the situation where workstation parts are things you might conceivably need and then just using the computer to push entertaining pixels. Throw ECC in there for extra good renders. That's not the situation that was being asked, though. The situation being asked was "do I need to turn off SMT to be SUPER ULTRA COMPLETELY TOTALY ABSOLUTELY SECURE". I mean, short of putting yourself in an Absolutely Safe Capsule, or literally rolling your own silicon based on RISC-V? No. You will never be safe. You will only ever be safe *enough*. SlayVus posted:I just tried their CorePrio program on my 1950X and was running Indigo myself. Without CorePrio and without /affinity, I was getting 1.6 on Indigo. With just /affinity 0xFFFFFFFE I was getting 1.73 in Indigo(Launching Indigo with Core 0 affinity off). With CorePrio and no /affinity, I shot up to 2.2 in Indigo with my 1950x OC'd to 3.9GHz. So not the same 50% increase in performance he was, but I did see a ~37% increase in performance in just Indigo. Yay, glad I was able to share news that helped. SwissArmyDruid fucked around with this message at 23:59 on Jan 4, 2019 |
|
#
?
Jan 4, 2019 23:55
|
|
|
Klyith posted:OTOH meltdown really would have been a giant problem if they hadn't had fixes ready to go by the time of publication. A webpage running javascript could scan mamory or keylog your whole system from inside the browser sandbox, that is a huge vulnerability for home users. Anyone running intel systems who intentionally disables the meltdown mitigations is a fool. All browsers have mitigated js meltdown attacks fyi.
|
|
#
?
Jan 4, 2019 23:58
|
|
|
SwissArmyDruid posted:Yay, glad I was able to share news that helped. It didn't completely fix my dropped frames in OBS, but I went from having tens of thousands of drop frames in like an hour span to only having 54 dropped frames after an hour of gaming. This poo poo seemed to have fixed my OBS issues.
|
|
#
?
Jan 5, 2019 01:18
|
|
|
BangersInMyKnickers posted:All browsers have mitigated js meltdown attacks fyi. Yeah, but if you're disabling meltdown mitigations you're probably running Brave Browser and followed some idiot's guide on Reddit to make the internet faster
|
|
#
?
Jan 5, 2019 01:37
|
|
|
Klyith posted:Disagree: it won't happen until compilers or perhaps the major engines start having really good, mostly-automated methods to create multiprocessing optimizations. Which is kinda like saying it won't happen without a magic wand. And yeah games that make use of 8 threads is a faaar cry from a game that'd make use of say 16, 24, or 32 threads, like these newer Zen2 model Ryzens are rumored to have, but I don't think its an unreasonable real world example to point to of progress over time on this subject. And yes it took years for that to happen but I said as much already. Klyith posted:History has shown that just handing some new hardware to game developers does not result in them having the ability or resources to meaningfully take advantage of it.
|
|
#
?
Jan 5, 2019 02:00
|
|
|
Games still benefit from single-threaded performance much more (and will continue to in the future) because they don't tend to be well-suited to multithreaded processing. Magic middleware won't be able to change that.
|
|
#
?
Jan 5, 2019 02:03
|
|
|
PC LOAD LETTER posted:there are games that are indeed starting to use 8 threads today Yeah, but the question is how much work those other threads are doing and how limited the game is by the performance or the 1 or 2 most important ones? Looking back at some of the reviews from the original ryzen launch, it's interesting to note that while there are some games that have significant performance loss on a 1300X (4c/4t), they get pretty much all of it back on a 1400 (4c/8t) and marginal benefit from 6 or more real cores. My interpretation of that would be the threads >4 aren't really doing that much work, though they do need the faster scheduling of HT. But only the people that wrote it know if they used more than 4 threads because the consoles have more, and if it could have been equally good with 4 instead. PC LOAD LETTER posted:Also true but I also never said things worked this way either! Sorry, didn't mean to imply that, or that you think multithreading is easy. Just my reasons for why I have a different opinion.
|
|
#
?
Jan 5, 2019 03:40
|
|
|
SwissArmyDruid posted:That's not the situation that was being asked, though. The situation being asked was "do I need to turn off SMT to be SUPER ULTRA COMPLETELY TOTALY ABSOLUTELY SECURE". Ah, yeah, you're right. I think I made a jump in thinking about it that wasn't there.
|
|
#
?
Jan 5, 2019 03:42
|
|
|
Lambert posted:Games still benefit from single-threaded performance much more (and will continue to in the future) because they don't tend to be well-suited to multithreaded processing. It is however worth pointing out on this line of thought that developers have been aware for a long time now that big (ie. 20%+) single thread performance increases were pretty much not going to happen anymore (due to thermal/power limits for CPU's + inherent IPC limits of x86 itself) and all that AMD/Intel would eventually be able to offer in the way of consistent and big performance gains for x86 would be through more multithreading/cores. I'm sure they'd muuuch prefer to have more single thread performance instead but ultimately they're gonna have to make due with what they have to work with which is part of the reason why they've slooowly begun to support and use more than 4 threads in game as CPU's 4+ cores with HT become commonplace. So I see no reason for the long term trend of gradually using more and more threads in game to change even if games are and always will continue to be poorly suited to multithreading. Well at least until Amdahl's Law takes full effect anyways. Lambert posted:Magic middleware won't be able to change that. Klyith posted:Yeah, but the question is how much work those other threads are doing and how limited the game is by the performance or the 1 or 2 most important ones?
|
|
#
?
Jan 5, 2019 05:37
|
|
|
Yeah things definitely are changing, just like 3-4 years ago people here were still recommending dual cores for some budget gamers and now 4/4 CPUs are starting to have issues in some games. There was definitely no concern from most people at the time I got my 6600k that the lack of hyperthreading would ever limit gaming performance, turns out that thinking was wrong.
MaxxBot fucked around with this message at 06:28 on Jan 5, 2019 |
|
#
?
Jan 5, 2019 06:24
|
|
|
MaxxBot posted:Yeah things definitely are changing, just like 3-4 years ago people here were still recommending dual cores for some budget gamers and now 4/4 CPUs are starting to have issues in some games. There was definitely no concern from most people at the time I got my 6600k that the lack of hyperthreading would ever limit gaming performance, turns out that thinking was wrong. Well I was recommending 4790K with a H/B mobo over OCing a Haswell i5 with a Z-board back in 2014, there were already a few games where the stock factory overclocked 4.2/4.4GHz 4790K beat OCed 4690K and yet people thought I was committing CPU heresy because "but-but-but OC IS EASY MAN AND WORTH IT"
|
|
#
?
Jan 5, 2019 09:37
|
|
|
I have the worst luck it seems. My PSU self combusted on the one 40c day we had last week. Fingers crossed my components are ok, because I can't afford a new PSU let alone a new motherboard, CPU or GPU.
|
|
#
?
Jan 5, 2019 11:27
|
|
|
Palladium posted:Well I was recommending 4790K with a H/B mobo over OCing a Haswell i5 with a Z-board back in 2014, there were already a few games where the stock factory overclocked 4.2/4.4GHz 4790K beat OCed 4690K and yet people thought I was committing CPU heresy because "but-but-but OC IS EASY MAN AND WORTH IT" 4790k without bothering to overclock (with h97, for example) became a staple of the build thread, hell, I even built one for a family member
|
|
#
?
Jan 5, 2019 11:33
|
|
|
Most games nowadays have somewhat decent multithreading outside of niche (as the only developer, I'll make my own engine) or games with very specific use cases (I need everything to be fully deterministic), render command submission used to be one of the main bottlenecks but a lot of games nowadays just build command lists on separate threads and then actually submit them on the render thread. The other issue is physics since that's significantly more of a pain in the rear end to multithread and still have them be deterministic and I haven't heard of any recent work done on that (Bullet has had a GPU version as a WIP for about 10 years now for example). An issue with a bunch of games and threading nowadays however is not pinning worker threads to specific cores so the end result is you'd have like 10 worker threads constantly context switching instead of actually doing useful things. 
|
|
#
?
Jan 5, 2019 11:57
|
|
|
Monster Hunter World PC has literally hundreds of threads running at any one time and around 20% CPU usage is spent just switching them around. It's the crappiest port I've ever seen.
|
|
#
?
Jan 5, 2019 13:45
|
|
|
Zedsdeadbaby posted:Monster Hunter World PC has literally hundreds of threads running at any one time and around 20% CPU usage is spent just switching them around. It's the crappiest port I've ever seen. Well the technology has to begin somewhere. At least they seem to be pushing the envelope and utilizing a lot of threads?
|
|
#
?
Jan 5, 2019 14:11
|
|
|
Lube banjo posted:Well the technology has to begin somewhere. At least they seem to be pushing the envelope and utilizing a lot of threads? Ideally you want N threads, where N is the number of logical processors. An arbitrarily high number of threads is bad, each thread switch is a kernel level operation, requiring 2 context switches, from thread A to kernel and back to thread B. This is expensive. In comparison a thread running a loop fetching small jobs will take much less to switch to a new task after finishing the current task and won't be interrupted, killing the cache efficiency, by a different task. It's an anti-pattern where you assign a thread to every little thing. Button animation? New thread. Scrolling a view? New thread. A sewage agent in the latest Sim City? Give it a thread. You might also find it in corporate business java applications, it's not like you expect better from them. Also assign a thread for every single I/O channel you have. Thread per file, thread per network connection, let's benchmark kernel context switching.
|
|
#
?
Jan 5, 2019 14:38
|
|
|
Lube banjo posted:Well the technology has to begin somewhere. At least they seem to be pushing the envelope and utilizing a lot of threads? Not when you get things like this; quote:https://i.imgur.com/VvuZgpX.gifv
|
|
#
?
Jan 5, 2019 15:07
|
|
|
I don't know how the gently caress somethingawful doesn't support properly downsizing gifv embeds yet
|
|
#
?
Jan 5, 2019 15:09
|
|
|
Zedsdeadbaby posted:Not when you get things like this; What am I supposed to be seeing here?
|
|
#
?
Jan 5, 2019 16:21
|
|
|
Sidesaddle Cavalry posted:What am I supposed to be seeing here?
|
|
#
?
Jan 5, 2019 16:34
|
|
|
Llamadeus posted:Every flick of the controller resulting in a frametime spike Not every flick. There's one flick towards the end that causes a stutter from 16.7 ms to 26.6 ms. Jitter within +-1 ms is very normal
|
|
#
?
Jan 5, 2019 16:40
|
|
|
Oh that's a frametime histogram. Ohhh. Gotcha I haven't perused the Monster Hunter thread in a long long time and a lot of the port controversy i only heard of by word of mouth and read secondhand accounts. (i liked the port because it let me play with my friends so i tend to be a profuse apologist for it) Hundreds of threads and not 32? karoshi posted:let's benchmark kernel context switching. *entitled gamer voice* Then it's time for kernel context switching to get faster. Get to it, Microsoft! Intel! Wintel!
|
|
#
?
Jan 5, 2019 16:50
|
|
|
Zedsdeadbaby posted:I don't know how the gently caress somethingawful doesn't support properly downsizing gifv embeds yet SA has one volunteer that does some backend stuff in his spare time. I don't think lowtax has any employees anymore.
|
|
#
?
Jan 5, 2019 18:06
|
|
|
Llamadeus posted:Every flick of the controller resulting in a frametime spike Also that looks like terrible input lag.
|
|
#
?
Jan 5, 2019 18:38
|
|
|
NewFatMike posted:https://twitter.com/KOMACHI_ENSAKA/status/1081174660136353792 They mean “2 = Model Revision”, right?
|
|
#
?
Jan 5, 2019 18:47
|
|
|
Subjunctive posted:They mean “2 = Model Revision”, right? Yeah. An alternate theory is that the 121 part isn't base clock but the core config.
|
|
#
?
Jan 5, 2019 19:06
|
|
|
Lambert posted:Games still benefit from single-threaded performance much more (and will continue to in the future) because they don't tend to be well-suited to multithreaded processing. Magic middleware won't be able to change that. In contrast, I would argue that with each passing console generation that remains under AMD's control, is more time spent acclimatizing devs to multicore programming. Remember that last year, Sony committed a great many Ryzen LLVM improvements to the project, which will almost certainly be involved in the forthcoming PS5, as well as have ripple effects to anything else that AMD touches. Zedsdeadbaby posted:Monster Hunter World PC has literally hundreds of threads running at any one time and around 20% CPU usage is spent just switching them around. It's the crappiest port I've ever seen. You know, I wonder if that CorePrio tool might help MHW. I don't have MHW installed at the moment, can anyone check? SwissArmyDruid fucked around with this message at 19:26 on Jan 5, 2019 |
|
#
?
Jan 5, 2019 19:24
|
|
|
SwissArmyDruid posted:In contrast, I would argue that with each passing console generation that remains under AMD's control, is more time spent acclimatizing devs to multicore programming. Multicore consoles handily pre-date AMD entering the market for their CPUs though? It's not like anyone but Nintendo, maybe, would be stupid enough to revert to single-core or low core count consoles, and they've also never been on AMD CPUs anyway.
|
|
#
?
Jan 5, 2019 19:52
|
|
|
Fishmech gonna fishmech. Fine. Acclimatiling devs to multicore programming in a way that should yield results outside of a console environment, as this is the first generation where the majority of the console market runs off of hardware built on an x86 architecture that is consequently more applicable to the wider industry as a whole. And now I remember why I put you on ignore.
|
|
#
?
Jan 5, 2019 20:31
|
|
|
SwissArmyDruid posted:Fishmech gonna fishmech. What because you're salty your statements are nonsensical? I can assure you that coding systems well for the triple core symmetric PPC core in the 360 was quite applicable to getting good results in PC multicore programming too.
|
|
#
?
Jan 5, 2019 20:36
|
|
|
Easier it is for ports the better imo
|
|
#
?
Jan 5, 2019 22:27
|
|
|
Arzachel posted:Yeah. An alternate theory is that the 121 part isn't base clock but the core config. Khorne fucked around with this message at 01:05 on Jan 6, 2019 |
|
#
?
Jan 5, 2019 22:28
|
|
|
GRINDCORE MEGGIDO posted:Easier it is for ports the better imo Yeah. The X360's CPU and cache structure was very different from a typical x86 CPU of the time. The Xb1's and PS4's CPU and cache structure is nearly identical to a typical x86 desktop CPU in comparison.
|
|
#
?
Jan 5, 2019 23:10
|
|
|
The current gen consoles came after a generation that was born in the recession era and the console developers really didn't want to do strange, expensive bespoke designs again. That's why the current bunch use practically off the shelf pc configurations as opposed to weird hosed up triple core ppc or cell tech with spes and such. There's a lot of expertise overlap for multi thread programming as a result
|
|
#
?
Jan 6, 2019 10:58
|
|
|
fishmech posted:What because you're salty your statements are nonsensical? I can assure you that coding systems well for the triple core symmetric PPC core in the 360 was quite applicable to getting good results in PC multicore programming too. It even had SMT..
|
|
#
?
Jan 6, 2019 12:30
|
|
|
Ryzen Mobile 3000 series got announced ahead of the CES keynote: https://www.anandtech.com/show/13771/amd-ces-2019-ryzen-mobile-3000-series-launched Process improvement over Raven Ridge, fully on Zen+ 12nm architecture. Looks like clock boosts in similarly addressed thermal envelope, except there are now 35W parts. Some Excavator stuff got pulled back from the grave for 1C/2T Chromebook parts. Still caps out at 4C/8T, which is a little disappointing. Hopefully this means the keynote will be much more exciting/focused on desktop parts.
|
|
#
?
Jan 6, 2019 19:47
|
|
|
|
| # ? Apr 29, 2024 13:26 |
|
|
You're forgetting the most important part: DAY ZERO MOBILE APU DRIVER UPDATES (gott in himmel, finally)
|
|
#
?
Jan 6, 2019 20:58
|
|