|
Llamadeus posted:This has come up a few times before, Paul's just implying that some of those who chose Zen/Zen+ (maybe Zen 2 as well?) processors over Coffee Lake chose wrong. Specifically gamers who could afford an 8700K but somehow convinced themselves that the better multicore per dollar of Ryzen mattered more to them than just gaming. He's made this case several times in the past and never bothers to calculate system cost. Paul MaudDib posted:anyway, I think a lot of people's primary "heavy workload" is gaming. And I also think a lot of people were telling themselves "I buy a 1600 now, then I drop in a $200 5600 in a couple years". But the price increases have really screwed that up, now you're dropping in a $300 5600X instead, and was buying a 1600 and a 5600X and taking lower performance for those intervening 3 years really worth it over just paying a little more for the 8700K? The price increases really shifted the equilibrium of the "buy flagship now vs upgrade later" equation. Once again ignoring motherboard prices, which were a significant cost difference up until the last year. Also ignoring the likelihood that a non-X 5600 will be available in a few months. If we're gonna make purchasing decisions from 3 years ago with 20/20 hindsight, then let's apply some basic foresight that a Zen 3 CPU will be available for less than $300 at some point, no? Dude you buy fairly high-end systems and buy them often, which is totally fine. But your theories and opinions on how to optimize a low budget are consistently bad or wrong. That's not the life you live, cheap bang-for-buck components are not the parts you research for yourself. I don't give people advice on RGB stuff beyond "buy from a single brand/type; asus has the best built-in mobo support if you don't get stuff with a dedicated controller" because I don't have any experience with RGB. It's not my thing so I don't know how best to make it work. That's you and budget buying. And when talking about gaming, the truth is that there's zero difference for most people between a 8700K and 1600 unless you have just upgraded to a 3080 in the last month. GPU is king, all you need from the CPU is enough to not bottleneck. (For some people with an old 1600 system, the smartest upgrade might actually be the memory -- ram was peak price back then and is super-cheap now, and with zen's love of fast memory ditching the 2666 for 3600 might take care of the 1% low frames.)
|
#
?
Oct 13, 2020 23:41
|
|
|
|
| # ? Jun 6, 2024 01:00 |
|
|
It's hilarious to me that no matter how objectively correct Paul is about this people will still try to deny it. Coffee Lake was a loving great buy and everyone who argued for buying AMD over an 8600k for the average user who primarily needs peak performance for gaming was about as wrong as possible.
|
|
#
?
Oct 13, 2020 23:43
|
|
|
i guess the lesson is don't have a history of melting down about AMD if you want to have an objective discussion about the pros and cons of various chips (the fishmech rule of internet argumenting)
|
|
#
?
Oct 13, 2020 23:49
|
|
|
K8.0 posted:It's hilarious to me that no matter how objectively correct Paul is about this people will still try to deny it. Coffee Lake was a loving great buy and everyone who argued for buying AMD over an 8600k for the average user who primarily needs peak performance for gaming was about as wrong as possible. Especially when you factor in overclocking. 8600ks running at 5ghz aren't even thinking about shopping for a CPU today if all they do is game on their own without streaming.
|
|
#
?
Oct 13, 2020 23:51
|
|
|
I went to Zen over Coffee Lake despite knowing the performance differences because I thought the idea of being on a single platform for several years was quaint and appealed to me.
|
|
#
?
Oct 13, 2020 23:55
|
|
|
Llamadeus posted:This has come up a few times before, Paul's just implying that some of those who chose Zen/Zen+ (maybe Zen 2 as well?) processors over Coffee Lake chose wrong. Specifically gamers who could afford an 8700K but somehow convinced themselves that the better multicore per dollar of Ryzen mattered more to them than just gaming. I haven't ever seen 8700k's going for 250 dollars. Maybe they existed but it was unlikely to be available en-masse at 250 for a significant period of time. I doubled checked PCPartpicker 2 year price history and Amazon's price history to see if I was just an amnesiac but I could not find them selling below 340 USD with *any regularity* prior to July 2020. I am talking about a one time dip when Amazon had it 300 bucks for like a day. When the 8700k was relevant, the choice for a budget gamer was to spend for a 1600X @ 209 dollars or a 2600X @ 229 at their respective launch MSRPs. That is well over 100 dollars in savings for marginal performance gains (~15% when CPU bound in most gaming loads) when most of the time gamers are GPU bound which renders the performance gains over the AMD parts even less relevant. The mistake would have come in if you overspent on an AMD CPU at around 330 dollars (1700X or 2700X) for more cores for a gaming workload but that is a very specific mistake that people who know what gaming workloads look like would never have recommended. In that specific corner case then yea, the 8700k was the way to go. If you manage to find Intel CPUs for the prices Paul found them at, then sure buying AMD was wrong but unfortunately outside short sale windows, those kinds of prices simply didn't exist. When people come looking for budget systems, every dollar counts when trying to give them the best bang for the buck. Zen and Zen+ absolutely could have and would have been the correct choice given many people's budget envelopes as that 100+ USD in savings almost equates to upgrading up an entire tier worth of GPU class. Also many do not have the know-how, patience, or wherewithal to engage in CPU overclocking to get the most out of an Intel processor when simply asking them to assemble a PC on their own is a daunting challenge.
|
|
#
?
Oct 14, 2020 00:15
|
|
|
Paul MaudDib posted:8700K has been as low as $250 
|
|
#
?
Oct 14, 2020 00:46
|
|
|
Some Goon posted:I do wonder how the move from 4 core ccxs to 8 core has impacted yields and by extension margin on the new parts. TSMC has stated that their 7nm parts have a shockingly low error rate. Like 'we normally see this 3+ years into the node' low error rate. N5 is looking to be even better, due in large part to using mature EUV processes for the masking vs. quadruple exposures using DUV.  Yields are about as good as 16nm, at something like 0.09 defects per cm^2. Wafer yields probably dropped 5% or something, with a bit more binning 6 core parts out of the 8 core CCXs, but moving to an 8 core 32MB L3 unified CCX was definitely the right move. 8 core consumer parts behave basically like a monolithic die, and I know cloud companies are super horny for 256 threads of cloud compute goodness in one socket.
|
|
#
?
Oct 14, 2020 01:21
|
|
|
Ż\_(ツ)_/Ż here's the $250 deal to which I'm referring: https://www.reddit.com/r/buildapcsales/comments/cv96js/cpu_i78700k_microcenter_in_store_only_250_no_mobo/ Microcenter also regularly ran the retail boxed version at $300, as they usually do. https://www.reddit.com/r/buildapcsales/comments/9yq3zn/cpu_intel_core_i78700k_micro_center_29999_36060/ but it's been pretty low for a long time https://www.reddit.com/r/buildapcsales/comments/83744t/cpu_intel_core_i78700k_potentially_26399_32999/ https://www.reddit.com/r/buildapcsales/comments/8x3bsm/cpu_i78700k_37_ghz_6core_processor_26241_34989/ https://www.reddit.com/r/buildapcsales/comments/90z84n/cpu_intel_8th_gen_core_i78700k_processor_coffee/ https://www.reddit.com/r/buildapcsales/comments/7zixf0/cpu_i78700k_31305_am_i_reading_this_right/ https://www.reddit.com/r/buildapcsales/comments/8cgnjk/cpu_i7_8700k_31999_free_shipping_walmart_rollback/ etc etc. Wasn't hard to get it below $300 in 2018 if you wanted, thanks to regular eBay promos (20-25% off sitewide) and much better supply back then (the constant shortages were really a 10-series thing, they affected 9 series to a much smaller degree and didn't really affect 8-series at all). And it regularly dipped around $300. Buying 2000-series over the 8700K was p. dumb. Like, if $260 was too much to pay for a hexacore in 2018 what does it say about AMD's $300 hexacore in 2021? Paul MaudDib fucked around with this message at 01:52 on Oct 14, 2020 |
|
#
?
Oct 14, 2020 01:42
|
|
|
lamo at posting a bunch of random reddit links where the comments are largely "I couldn't get it at the price" and double lol at the $250 deal being a Black Friday sale price
|
|
#
?
Oct 14, 2020 01:45
|
|
|
shrike82 posted:lamo at posting a bunch of random reddit links where the comments are largely "I couldn't get it at the price" gotta be quick if you want the dealz, but it wasn't really that hard if you watch buildapcsales. $250 wasn't a black friday sale price, unless they moved black friday to august (and that deal was running for months) lamo at not really reading the post edit: had a duplicated link instead of the $250 deal, fixed Paul MaudDib fucked around with this message at 01:55 on Oct 14, 2020 |
|
#
?
Oct 14, 2020 01:50
|
|
|
this is the real issue with paul's historical stanning for the 8700K - his arguments went beyond the merits of the chip itself and into weird meltdown territory of having to outright lie about stuff like prices
|
|
#
?
Oct 14, 2020 01:51
|
|
|
Paul MaudDib posted:gotta be quick if you want the dealz, but it wasn't really that hard if you watch buildapcsales. This is a weird hill to die on
|
|
#
?
Oct 14, 2020 01:55
|
|
|
nerdrum posted:This is a weird hill to die on just citing my sources, why was it brought up at all then?
|
|
#
?
Oct 14, 2020 01:57
|
|
|
Paul MaudDib posted:just citing my sources, why was it brought up at all then? Lets be real, it was hard, not easy, to get the 8700k at the prices you mentioned. Not impossible but from the reddit responses, it mirrors 3080 F5'ing hard. Your own reddit thread links indicate these were very short windows to buy.
|
|
#
?
Oct 14, 2020 02:02
|
|
|
Are these new Ryzen CPUs almost guaranteed to have better perf-per-watt than my 8700k for gaming? I'm building a backpack VR mITX PC, so keeping the wattage as reasonable as possible is paramount.
|
|
#
?
Oct 14, 2020 02:16
|
|
|
Paul MaudDib posted:just citing my sources, why was it brought up at all then? You brought it up, and your sources are poo poo. RT though, in 2018, if you were looking to build a nice-rear end rig coffeelake was the way to go, Zen+ topped out at 80-90 fps. It was great for 60fps builds since it's a 25% savings over a 8600, more if you could live with the stock cooler, on the cpu alone, and had plenty of headroom. If your were building for 144 then then you wanted Intel, and plenty of people bought Intel for that reason. But Zen2 pretty much killed that gap and coffee lake refresh and coffee lake refresh refresh haven't had much to offer that contests it. Fantastic Foreskin fucked around with this message at 02:32 on Oct 14, 2020 |
|
#
?
Oct 14, 2020 02:26
|
|
|
Zero VGS posted:Are these new Ryzen CPUs almost guaranteed to have better perf-per-watt than my 8700k for gaming? This is cool as heck, but you probably want a top-end laptop CPU for this, right?
|
|
#
?
Oct 14, 2020 02:30
|
|
|
Zero VGS posted:Are these new Ryzen CPUs almost guaranteed to have better perf-per-watt than my 8700k for gaming? AMDs been murdering Intel on perf/watt for ages now, they're on a smaller process and Intel has been brute forcing speed to compensate for their fab problems. No Zen3 benchmarks yet but there's no reason to believe it won't continue to do so. How are you planning on assembling it? All the talk about computers in desks in the part picking thread has me idly thinking about computers without traditional cases. Twerk from Home posted:This is cool as heck, but you probably want a top-end laptop CPU for this, right? Also this. Dunno where you can get an embedded board with a high-end cpu but laptop processors are so much more efficient than desktop ones.
|
|
#
?
Oct 14, 2020 02:30
|
|
|
I had to look up what a stan was.  Despite how much money people may have spent on their choice of CPU, lets face that most people who bought a midrange or better piece of silicon from Intel or AMD in recent years are practically doing just fine. Maybe not getting biggest number, but whatever, face it — the billion dollar machine you fed currency into doesn't give a gently caress about defending its honor. Addressing this generally. Different folks have different figures of merit. Objective discussion in good faith, please.
|
|
#
?
Oct 14, 2020 02:42
|
|
|
Twerk from Home posted:This is cool as heck, but you probably want a top-end laptop CPU for this, right? No, I already have a Skyreach Mini case, which I can strap to a backpack or a military ALICE frame. I also have some eBike batteries which weigh about 8 pounds and can each put out 550 watts for an hour (or a bit more than twice that for two hours). The main challenge is that my PC runs on a 12v input so I have to make sure that after reducing the batteries from their native 36v down to 12v, that I'm conditioning the power as much as possible and minimizing voltage ripple. So far this DC-DC converter is looking like one of my better options but I'm still being a perfectionist about it and seeing if I can do better: https://www.amazon.com/720W-Converter-Voltage-Regulator-Waterproof/dp/B081GMD5HZ A few other people have done them and there's writeups: https://randomfoo.net/2016/03/28/the-most-portable-vr-workstation https://smallformfactor.net/forum/threads/ono3-pelican-1430-to-s4-portable-vr-workstation.423/
|
|
#
?
Oct 14, 2020 02:46
|
|
|
Zero VGS posted:Are these new Ryzen CPUs almost guaranteed to have better perf-per-watt than my 8700k for gaming? As an engineer I'm curious if there's any structural stuff you're doing inside the case for when you're flailing about in VR? Big honkin' GPUs, cantilevered CPU coolers, etc. come to mind. I imagine things not designed for dynamic loads like that may not like it much but I know absolutely nothing about backpack PCs...no idea if this is even something generally considered or generally it's fine? edit: thanks for more info before I even posted.  certifiably, neat
|
|
#
?
Oct 14, 2020 02:47
|
|
|
Zero VGS posted:No, I already have a Skyreach Mini case, which I can strap to a backpack or a military ALICE frame. I also have some eBike batteries which weigh about 8 pounds and can each put out 550 watts for an hour (or a bit more than twice that for two hours). The main challenge is that my PC runs on a 12v input so I have to make sure that after reducing the batteries from their native 36v down to 12v, that I'm conditioning the power as much as possible and minimizing voltage ripple. So far this DC-DC converter is looking like one of my better options but I'm still being a perfectionist about it and seeing if I can do better: https://www.amazon.com/720W-Converter-Voltage-Regulator-Waterproof/dp/B081GMD5HZ My dude, Digi-Key time. Gonna edit this in a few mins when I dig up something. Won't win on cost, but also won't be a fireball on your back. e: I don't what your exact voltage range on the batteries is (is that 36 V at 100% SOC?), but here's my search query and some interesting options like DIN-rail mount: https://www.digikey.com/en/products...dsBd%2B5gkK9XkA https://www.digikey.com/en/products/detail/mean-well-usa-inc/SD-200B-12/7706494 https://www.digikey.com/en/products/detail/mean-well-usa-inc/DDR-120C-12/8591545 CUI's got some finned ones, but $$$$. Either way, get some capacitance on the output and you'll be in decent shape. movax fucked around with this message at 02:51 on Oct 14, 2020 |
|
#
?
Oct 14, 2020 02:48
|
|
|
movax posted:I had to look up what a stan was. Its 20 years old 
|
|
#
?
Oct 14, 2020 02:48
|
|
|
hobbesmaster posted:Its 20 years old I think I missed the original definition of it / was too much of a shut-in, and then it started showing up all over Twitter and I was confused. But I learned something new, hooray.
|
|
#
?
Oct 14, 2020 02:52
|
|
|
hobbesmaster posted:Its 20 years old I haven’t really seen it being thrown around regularly until this year.
|
|
#
?
Oct 14, 2020 03:02
|
|
|
hobbesmaster posted:Its 20 years old I guessed it when I first started seeing it, but it was well after it was relevant that it started showing up for me.
|
|
#
?
Oct 14, 2020 03:03
|
|
|
Is the 5900x just two 8 core chips that got binned down? Is that worse or better than the 5800x that has a single "perfect" 8 core for single threaded stuff or does it not matter. Does 12 cores mean it'll suck up literally twice as much power as a 6 core if both are running at 100%?
|
|
#
?
Oct 14, 2020 03:32
|
|
|
MeruFM posted:Is the 5900x just two 8 core chips that got binned down? It's two 8 core chip complexes with two cores disabled on each. It's not the same as two 6 core chips independent on your motherboard but it's not wholley different either, but AMDs been doing it for a couple years so software has adapted. For single core it shouldn't matter, but no benchmarks yet. On previous Zen chips the interconnect was a substantial portion of the power draw so this would not be so. It still shouldnt scale linearly but how it performs vs the single ccx SKUs on power is still unknown.
|
|
#
?
Oct 14, 2020 03:38
|
|
|
Methylethylaldehyde posted:TSMC has stated that their 7nm parts have a shockingly low error rate. Like 'we normally see this 3+ years into the node' low error rate. N5 is looking to be even better, due in large part to using mature EUV processes for the masking vs. quadruple exposures using DUV. i am also extremely horny for 256 or, better yet, 512 threads in a single socket
|
|
#
?
Oct 14, 2020 04:01
|
|
|
MeruFM posted:Does 12 cores mean it'll suck up literally twice as much power as a 6 core if both are running at 100%? This is generally not how it works. It's more like it'll consume more power, but not twice as much. This means each core will be running at a reduced clock speed, which actually makes the fully loaded 12 core significantly more power efficient than the fully loaded 6 core. The 3900X and 3950X are efficiency monsters at a lot of highly threaded workloads, getting work done in not just less time but for fewer watt-hours.
|
|
#
?
Oct 14, 2020 04:14
|
|
|
MeruFM posted:Is the 5900x just two 8 core chips that got binned down? Not exactly. Note that the 5900X has a target boost clock of 4.8 Ghz, which is 100 Mhz higher than the 5800X, and 200 Mhz higher than the 5600X. There's still some binning involved because you still need the part to hit those advertised clocks, though having to work with six cores per CCX does give you more flexibility. MeruFM posted:Is that worse or better than the 5800x that has a single "perfect" 8 core for single threaded stuff or does it not matter. if a workload is only single-threaded, then it's going to run on a single thread, and Windows scheduler has gotten pretty good at making sure it doesn't keep sending workloads across different CCX's, so it probably doesn't matter between the 5800X and the 5900X (besides the 5900X clocked just a little bit higher) if a workload can take advantage of multiple threads, then it can be done just as fast without running the CPU at full clip relative to a CPU with fewer threads, so it's not a given that a CPU with more cores/threads will consume more energy
|
|
#
?
Oct 14, 2020 04:31
|
|
|
Stan is a masterpiece of a song and it's weird seeing it become a somewhat commonplace term on internet discussion and social media only very recently
|
|
#
?
Oct 14, 2020 04:43
|
|
|
https://theoutline.com/post/2425/when-stan-became-a-verb?zd=1&zi=ng5jrmxvquote:But according to the folks at the Oxford English Dictionary, which tracks the historical usage of words, the first recorded use of “stan” as a verb that they have found so far was in a tweet from April 2008: “I stan for santogold. I may even like her more than MIA.” It got no replies, no retweets, and no likes, but could nevertheless prove to be an important historical artifact should “stan” make it into the OED.
|
|
#
?
Oct 14, 2020 04:46
|
|
|
Malcolm XML posted:i am also extremely horny for 256 or, better yet, 512 threads in a single socket the sound of a million code monkeys crying out in horror when they realize they really do have to multithread ALL of their code
|
|
#
?
Oct 14, 2020 04:46
|
|
|
What are you talking about, 2008 was only a couple of years ag- oh my god
|
|
#
?
Oct 14, 2020 04:48
|
|
|
Cygni posted:the sound of a million code monkeys crying out in horror when they realize they really do have to multithread ALL of their code Who will win in a fight, ever more lovely web programmers or continually more powerful hardware?
|
|
#
?
Oct 14, 2020 04:53
|
|
|
movax posted:My dude, Digi-Key time. Gonna edit this in a few mins when I dig up something. Won't win on cost, but also won't be a fireball on your back. lol the only one that fits the specs is $450... Though searching around online I did find an HDPlex DC-DC ATX that can do 800w and has the full voltage range that i need for $278: https://hdplex.com/hdplex-800w-dc-atx-with-16v-63vdc-input.html
|
|
#
?
Oct 14, 2020 05:51
|
|
|
Cygni posted:the sound of a million code monkeys crying out in horror when they realize they really do have to multithread ALL of their code I've had this opinion in other places (and been shouted down because obviously i'm lying, for some reason), but the complexity will be absorbed into the engine/framework/toolchain. Parallelism is already very easy, thanks to the worker pool model becoming standardized. You don't have to care about the threads, or what type of threads they are, or if they're even actually threads or might be full processes. Simple (non-interdependent) concurrency is only marginally more complex. Concurrency with dependencies is more complicated (and in my experience requires some wrapping your head around, but so did OOP when you grew up with purely procedural languages like us olds did). But it's very doable if you give a poo poo, and there's good support for it in at least one real, non-academic language. I'm sure someone will eventually work out a generalized approach for it that handles all the common use cases -- whatever that is at the time. So yes, the programmers who don't really care, who are working on stuff where nothing is truly critical, will continue to not care about quality or resource usage. But the people who enable those programmers will. And language designers will.
|
|
#
?
Oct 14, 2020 07:14
|
|
|
|
| # ? Jun 6, 2024 01:00 |
|
|
 Asus?!? This along with the fact that they're refreshing their B450 boards with zen 3 specifically in mind makes me worry that they'll drop support on my old B450.
|
|
#
?
Oct 14, 2020 07:54
|
|
