|
Rinkles posted:How do the numbers work with AMD, why were (iinm) the 4000 and 6000 Ryzen chips mobile only? I wouldn’t try to detect a pattern in computer naming, generally. “A bunch of San Jose State marketing grads thought it would sell more units for some reason“* is generally the rationale. *As a fellow CSU undergrad, I am legally allowed to rip on em.
|
#
?
Jul 30, 2022 05:06
|
|
|
|
| # ? Apr 25, 2024 12:10 |
|
|
Dr. Video Games 0031 posted:I really don't know what "the same as the last generation, but with new architecture" is supposed to mean. Isn't the new architecture the whole point? If the combination of new architecture and a node shrink gives a good boost to overall performance and perf per watt, then what else are you looking for? The Ryzen 7000 series will also support AVX-512, if that's important to you. Taking AMD's word at face value, they said they're expecting an IPC gain of 11% iirc. Memory will give you at least a little bit of an IPC boost too, for Alder Lake, even in pure ST load, anandtech found about 3.5% higher IPC for using DDR5 over DDR4, and [url=https://www.anandtech.com/show/17047/the-intel-12th-gen-core-i912900k-review-hybrid-performance-brings-hybrid-complexity/8]the advantage in multithreaded was much larger. So you do get some small gains for using DDR5 in general, not huge but it's a big chunk of an 11% gain. You also do gain latency even just from a "naive" shrink iirc, things are closer together. If it's 11% IPC gain that's probably pretty minor architectural changes apart from AVX-512 (but that's not minor at all). AMD backpedaled after that talk got started and said that's more like a minimum, the rumor mill hopes more like a 20-30% gain in general, that would obviously be a lot more significant than the 11% number. And it also supposedly clocks higher iirc. Kinda funny, finfets have finally caught up to planar transistor frequencies and then... babe, it's time for your GAAFETs! (yes honey) I'm reasonably hopeful, they can probably do like another zen2->zen3 sized core tweak or two. TSMC 5nm has good logic shrink and almost no SRAM/cache shrink - so you might see more architectural focus this gen rather than just "maek moar caches!!". Or maybe it's just a low-risk straight shrink with some minor tweaks and AVX-512, while they work on a deeper architectural changes for a future gen. Who knows. AMD is milking it for a separate Zen4 and Zen4D launch and I think that's probably a sign of confidence too. Like, everyone knows they're coming at this point, c'mon. I personally wonder if the IPC gain for ddr5 would be higher in AVX-512 tasks too. Even with half-rate I think the bandwidth utilization might well go upwards. V-Cache might help as well - bandwidth might not go up (not sure how that's implemented) but it would let you keep a larger working set in cache. Like, at this point I'd really like to wait for Zen5 and a second-gen memory controller but if they salami slice it, it could be forever before they finally get around to releasing 8900X3D lol. But yeah I really do think AVX-512+vcache is the next big check-box on my list, I'd love a 5900X3D on paper but unless you've already bought into AM4 it's kind of hard to justify doing a whole build on "old" chips with something that should be quite a bit better coming out real soon. 7900X3D is what I'm looking for really, even if it just ends up being "zen3 but shrunk and with avx-512". Sadly, the early boards always suck compared to the later ones too, the fun ones happen at the end of the gen where everything's worked out and polished and they have a gazillion cool features ("gee timmy, how come X570 ProArt gets two thunderbolt ports!?"). Cool niche stuff like Asrock Rack X670D4U etc will probably take longer to actually become available (even if the designs come out quick) as well. Paul MaudDib fucked around with this message at 05:44 on Jul 30, 2022 |
|
#
?
Jul 30, 2022 05:39
|
|
|
Rinkles posted:How do the numbers work with AMD, why were (iinm) the 4000 and 6000 Ryzen chips mobile only? they release new a mobile chip series annually, but their desktop chips don't keep to as regular of a release schedule, and there isn't always direct correspondence between architectures. it breaks down like this: the desktop 1000 series is Zen 1 the desktop 2000 series is Zen+ (a Zen 1 refresh) and the mobile 2000 series is Zen 1 the desktop 3000 series is Zen 2, but the mobile 3000 series is Zen+ the mobile 4000 series is Zen 2 (and there were actually some 4000 series desktop parts that were APUs with the GPU disabled, they were not very good budget chips) the desktop 5000 series is Zen 3 but the mobile 5000 series is a mix of Zen 2 and Zen 3 the mobile 6000 series is Zen 3+ which is a mobile-only Zen 3 refresh but i think they're trying to keep it a bit more consistent now as it seems like the 7000 series will just be Zen 4 on both? but their marketing team could also totally abandon that before too long anyway, it's all arbitrary helpfully they're releasing 7000 series CPUs and GPUs around the same time, to make things more confusing
|
|
#
?
Jul 30, 2022 06:45
|
|
|
Rinkles posted:How do the numbers work with AMD, why were (iinm) the 4000 and 6000 Ryzen chips mobile only? Zen 1 was Ryzen 1000 But the Zen 1 APUs, that came later, were already named Ryzen 2000 Zen+ was Ryzen 2000 But the Zen 2 APUs, that came later, were already named Ryzen 3000 Zen 2 was Ryzen 3000 But then Zen 2 APUs, that came later, were already named Ryzen 4000 AMD then took Zen 2 Mobile and also named it Ryzen 4000 This way, when they moved on to Zen 3, the desktop CPUs, the APUs, and the mobile chips would all be Ryzen 5000 (except this still wasn't a completely clean naming scheme) Now, they've released mobile chips, and they're different enough from Zen 3 that it's justifiable to name them Ryzen 6000 But the next release is going to be later enough that it HAS to be Ryzen 7000, because bigger number = better than (which is also why the first two series misaligned anyway)
|
|
#
?
Jul 30, 2022 07:43
|
|
|
I vaguely seem to remember that it was going to be a "mobile chips are an optimized version of last gen's architecture", but that very quickly fell by the wayside.
|
|
#
?
Jul 30, 2022 09:41
|
|
|
Paul MaudDib posted:AMD backpedaled after that talk got started and said that's more like a minimum, the rumor mill hopes more like a 20-30% gain in general, that would obviously be a lot more significant than the 11% number. And it also supposedly clocks higher iirc. Kinda funny, finfets have finally caught up to planar transistor frequencies and then... babe, it's time for your GAAFETs! (yes honey) They said ">15% single thread performance increase" in the announcement event while also touting 5.5 boost clocks so the napkin math worked out to gently caress all ipc improvements despite the doubled L2.
|
|
#
?
Jul 30, 2022 10:01
|
|
|
Cygni posted:Ok well at least the dude at Asrock has a sense of humor. This is the same leaked 7600X result as before, with the name changed: CoolCab posted:lmao, at least they're owning it 90% chance this was done by the Userbenchmarks admin. That guy is an extreme intel nutjob. When an intel ES does well, it's normal. When an AMD one does well, it's marketing BS. He already altered the algorithm behind the scenes so AMD's previous #1 result here is now behind Intel
|
|
#
?
Jul 30, 2022 12:12
|
|
|
userbenchmark is banned on a lot of other forums and social media aggregates I don't feel like naming, for good reason The guy is pants-on-head nuts and extremely biased
|
|
#
?
Jul 30, 2022 12:19
|
|
|
Cygni posted:I wouldn’t try to detect a pattern in computer naming, generally. “A bunch of San Jose State marketing grads thought it would sell more units for some reason“* is generally the rationale. some reasons: "Our competitor released a model with a higher number. Quick, put out a model with a bigger number than theirs!" "Welp, sales are trending down for the current models and our new tech is delayed. Quick, put out a model that's juiced a tiny bit and has a bigger number, we need those Q3 numbers to be good!" (Also, it works.)
|
|
#
?
Jul 30, 2022 13:15
|
|
|
It's marketing fact that a K or X in the name draws the eye to it. And more digits = better, your brain thinks 6950 is better than 6900 is better than 7000.
|
|
#
?
Jul 30, 2022 16:24
|
|
|
the arms race between the i9-12900KS and the r7 5800X3D clearly demonstrates that within our lifetimes amazon will require two lines just for the product code
|
|
#
?
Jul 30, 2022 16:30
|
|
|
 mr president, we must not allow a number a product code length gap!
|
|
#
?
Jul 30, 2022 16:35
|
|
|
NewFatMike posted:If you’re just disappointed that each market segment has the same amount of cores as the previous 5 iterations, just say that. I'm disappointed that each market segment has the same number of cores for several generations now, it sucked when Intel did it from i7-860 through i7-7700K, and it sucks that it's happening again with 6 core mainstream parts and 8 core upper-middle parts. Intel looks about to deliver us a core-splosion though, this year's parts are getting 16 E-cores and they should be able to keep shipping more E-cores cost and space effectively.
|
|
#
?
Jul 30, 2022 16:41
|
|
|
Ok then just say that next time instead of making a bunch of poo poo up whole cloth lmao
|
|
#
?
Jul 30, 2022 17:59
|
|
|
that was a different person
|
|
#
?
Jul 30, 2022 18:22
|
|
|
I don't know that 16 more cores would make a big difference to most people, even enthusiasts.
|
|
#
?
Jul 30, 2022 18:49
|
|
|
Look, it's 16 better
|
|
#
?
Jul 30, 2022 18:51
|
|
|
Oh my B didn’t check the chain. Sorry friend poster.
|
|
#
?
Jul 30, 2022 19:15
|
|
|
Twerk from Home posted:I'm disappointed that each market segment has the same number of cores for several generations now, it sucked when Intel did it from i7-860 through i7-7700K, and it sucks that it's happening again with 6 core mainstream parts and 8 core upper-middle parts. Doesn't matter how many cores you see in task manager if the performance sucks
|
|
#
?
Jul 30, 2022 19:54
|
|
|
If you have a 100 thread workload, those CPUs exist and you probably have the money to get one (or just rent instances). For desktop and gaming use, core count beyond the new 6 core mainstream still doesn’t really matter. https://youtu.be/Nd9-OtzzFxs
|
|
#
?
Jul 30, 2022 20:28
|
|
|
Zedsdeadbaby posted:userbenchmark is banned on a lot of other forums and social media aggregates I don't feel like naming, for good reason If only we could get Google to stop listing them so high in search results about CPUs - they still have a much bigger platform than they should when they are a blatantly insane Intel shill. A lot of people who don't know better are getting misinformation thanks to what seems like a lot of SEO done by userbenchmark.
|
|
#
?
Jul 30, 2022 23:39
|
|
|
Cygni posted:I personally don't see a way for AMD to stay competitive in the non-gaming laptop world without investing in small cores. But I'm not an expert! It's possible that AMD stays the all big core course, but it sure seems like "the ARM problem" is especially difficult in laptop world. You can design a full fat core that can power gate itself down to a slim configuration, when full fatness isn't needed. I think amd is eventually going to do big.LITTLE anyway, if only because datacenter customers keep asking for it.
|
|
#
?
Jul 31, 2022 05:50
|
|
|
lDDQD posted:You can design a full fat core that can power gate itself down to a slim configuration, when full fatness isn't needed. I think amd is eventually going to do big.LITTLE anyway, if only because datacenter customers keep asking for it. But turning big cores into little cores is hugely inefficient when it comes to die area. Why run a big core in a slim configuration (leaving most of the silicon unused), when you could put four slim cores in the same die area? It's less versatile, but Intel gives up some of that versatility to just have a shitload of e-cores (raptor lake is getting so many) I don't like the design for desktop, but I can see AMD going that route in other spaces. edit: lol 
Dr. Video Games 0031 fucked around with this message at 08:09 on Jul 31, 2022 |
|
#
?
Jul 31, 2022 07:23
|
|
|
I do wonder if the userbenchmarks guy is paid or if he's just insane
|
|
#
?
Jul 31, 2022 09:15
|
|
|
gradenko_2000 posted:I do wonder if the userbenchmarks guy is paid or if he's just insane
|
|
#
?
Jul 31, 2022 10:00
|
|
|
CaptainSarcastic posted:If only we could get Google to stop listing them so high in search results about CPUs - they still have a much bigger platform than they should when they are a blatantly insane Intel shill. A lot of people who don't know better are getting misinformation thanks to what seems like a lot of SEO done by userbenchmark. Dollars to donuts this has everything to do with Google's ad revenue
|
|
#
?
Jul 31, 2022 12:12
|
|
|
Rinkles posted:How do the numbers work with AMD, why were (iinm) the 4000 and 6000 Ryzen chips mobile only? ]They planned to get everything to be the same number with 4000-5000, which is why desktop had no 4000 parts, but then a push from sales/vendors/marketing/reality of production capacity/above screwed it up. They know it's messed up and there's no rhyme or reason to the scheme other than the number must go up or else there's lower sales for things like laptops, etc. You pretty much have to look up the part for the 5000/6000 series CPUs. Khorne fucked around with this message at 19:07 on Jul 31, 2022 |
|
#
?
Jul 31, 2022 18:57
|
|
|
Dr. Video Games 0031 posted:But turning big cores into little cores is hugely inefficient when it comes to die area. Why run a big core in a slim configuration (leaving most of the silicon unused), when you could put four slim cores in the same die area? It's less versatile, but Intel gives up some of that versatility to just have a shitload of e-cores (raptor lake is getting so many) I don't want to give him the page hit, but lol this makes me want to see the 5800X3D page.
|
|
#
?
Jul 31, 2022 19:46
|
|
|
K8.0 posted:I don't want to give him the page hit, but lol this makes me want to see the 5800X3D page. Oh, its impressive: quote:The 5800X3D has the same core architecture / IPC as the 5800X but it runs at lower clock speeds and has an extra 64MB of cache (96MB up from 32MB). This results in relatively low latency at 128MB because those transfers have a higher chance of remaining in cache. An unusually high proportion of early 5800X3D samples appear unable to boost above base clock, upcoming BIOS updates may fix this. Either way, for most real-world tasks performance is comparable to the significantly cheaper 5800X. Some very specific cache sensitive scenarios such as low res. canned game benchmarks with a 3090-Ti will benefit. Be wary of sponsored reviews with cherry picked games that showcase the wins and gloss over the losses. Also watch out for AMD’s army of Neanderthal social media accounts on reddit, forums and youtube, they will be singing their own praises as usual. AMD’s marketers continue to show more interest in this year’s bonuses than the longevity of the brand. Instead of focusing on real-world performance, they aim to dupe consumers with bankrolled headlines. The same tactics were used with the Radeon 5000 series GPUs. Zen4 needs to bring substantial IPC improvements, rather than overpriced "3D" marketing gimmicks. New high end PC gaming builders have little reason to look further than the $260 12600K. Users with an existing AM4 build should wait just a few more months for better performance at lower prices with Raptor Lake or even Zen4. The marketers selling expensive “3D” upgrades today will quickly move onto Zen4 leaving unfortunate buyers stuck on a 6 year old, dead-end, platform. [Mar '22 CPUPro]
|
|
#
?
Jul 31, 2022 20:06
|
|
|
Very inspirational that even an insanely weird Brand Warrior forum poster could start a successful website.
|
|
#
?
Jul 31, 2022 20:46
|
|
|
i love how "6 year old dead end platform" is somehow a bad thing compared to intel, where every motherboard is obsolete as soon as the next chip is out lmao. wonder if intel chips have "dead end platform on launch" disclaimer
|
|
#
?
Jul 31, 2022 20:58
|
|
|
i've always found it as kind of a curious argument because i suspect i'm not alone in never doing an in socket upgrade on a motherboard until AM4. i expanded pretty much everything else but until very recently if you were buying new and intel you're max getting one more generation out of your motherboard and you have to futz around with the very intimidating paste etc. upgradability is a really neat marketing feature and hey maybe you're weird like me and wind up finding cpu upgrading a fun facebook market thing. but for a very long time people correctly identified it as just that in 90+ percent of cases, but now at end of life it suddenly matters
|
|
#
?
Jul 31, 2022 21:06
|
|
|
AM4 is the first time I ever did a same board CPU upgrade as well, even going back to the AthlonXP or P3 days.
|
|
#
?
Jul 31, 2022 21:22
|
|
|
I think I'm a bit of an outlier because I got kind of used to doing CPU upgrades when I worked at a place that did computer refurbishing and repair. One of my personal socket 939 machines had at least three different CPUs as I upgraded over time, bumping up to better processors as they came in. I also tend to hang onto old machines longer than I should (I think I donated my last socket 939 box last year), so have a bit of a hand-me-down process going. Upgrading my current 3600X to something like the 5800X3D would be fairly normal for how I've tended to do things over the years, just more expensively than I used to be able to manage.
|
|
#
?
Jul 31, 2022 21:28
|
|
|
CaptainSarcastic posted:I think I'm a bit of an outlier because I got kind of used to doing CPU upgrades when I worked at a place that did computer refurbishing and repair. One of my personal socket 939 machines had at least three different CPUs as I upgraded over time, bumping up to better processors as they came in. I also tend to hang onto old machines longer than I should (I think I donated my last socket 939 box last year), so have a bit of a hand-me-down process going. Upgrading my current 3600X to something like the 5800X3D would be fairly normal for how I've tended to do things over the years, just more expensively than I used to be able to manage. My god what possible purpose can a 939 machine have to anyone other than as a museum piece? Gwaihir fucked around with this message at 21:55 on Jul 31, 2022 |
|
#
?
Jul 31, 2022 21:48
|
|
|
AM4 is my first CPU upgrade board. Only for the three reasons that by the time I felt the need to upgrade previously the CPU would've be at least 5-7 years old, so there was no point of upgrading to the best in slot in such an old socket. Second, the 5800x3d was perfect upgrade bait for what I wanted (games and VR) and the prevailing thought at the time was that it might be a limited run (such a fool to think that). Third, I previously had the plan to upgrade to the best CPU at the end of AM4 since I opted for a 3600 to hold over until then.
|
|
#
?
Jul 31, 2022 21:52
|
|
|
Gwaihir posted:My god what possible purpose can a 939 machine have to anyone other than as a museum piece? It was my secondary desktop when my AM3 machine was my primary. That old Athlon 4800+ held its own for quite a while, but then it got demoted to part of my hoard in the garage until I finally parted it out and donated it. With SSDs and 4GB of DDR1 it kept up okay with Windows 7, although I think I had to disable the joystick port in BIOS or it would always think there was an unidentifiable piece of hardware in Device Manager.
|
|
#
?
Jul 31, 2022 22:01
|
|
|
The User Benchmark dude is like some weird alternative universe Alex Jones that decided to get triggered by AMD instead of gay frogs or whatever. That dude is clinically insane.
|
|
#
?
Jul 31, 2022 22:33
|
|
|
Broose posted:Second, the 5800x3d was perfect upgrade bait for what I wanted (games and VR) and the prevailing thought at the time was that it might be a limited run (such a fool to think that). tbf, i found the arguments that people were putting together that it represented a poor allocation of resources compared to using that die space on their much much more profitable server/prosumer chips compelling, i'm suprised too. there were tons of rumours that it was hitting production problems or bottlenecks and it didn't seem to make any kind of sense to continue supporting a last gen product that makes your current gen ones look bad, so i expected it to be a 3300X equivalent. tbf it is STUBBORNLY resistant to any kind of promo in my observation, typically the only time you see it on my deals website is when there's a sitewide -x% or something. on there (it doesn't cover all retailers in this region, scan isn't allowed on there for instance, far from scientific) can count it having gone on promo being worthy of getting posted only three times, and all in late july. all time low is CCL right now for 384 pounds which is not much lower than RRP. i think it must have sold better and had higher demand than they expected, particularly in the face of 12th gen. gotta be tons of people who want the best in socket chip to keep their system going as long as it can.
|
|
#
?
Jul 31, 2022 22:35
|
|
|
|
| # ? Apr 25, 2024 12:10 |
|
|
Cygni posted:Very inspirational that even an insanely weird Brand Warrior forum poster could start a successful website. They went full psycho when zen2 dropped and was ahead by most reasonable metrics. They went as far as to revise their site to include warnings that the numbers their own benchmarks, including actual gameplay benchmarks their own site was collecting, were not representative of real world performance. They skewed the algorithm so a few year old i3 was beating even intel's latest CPUs nevermind AMD. They added metrics to fork multicore into multiple sections to specifically exclude >4c from their rating system. And a bunch of other weird stuff, like suddenly attacking AMD's GPUs which they had no problem with in the past. They didn't even attack AMD's gpus for actual flaws, it was the same crazy little green men stuff they started doing on CPUs. AMD wishes they had marketing on the level that person is claiming. It's fanboys, stock shills, people wanting competition in the market, and people enjoying a reasonable product that Intel still isn't really competing with in some market segments. The same stuff comes out during intel & nvidia release/leak season and it has the same wave of people being excited for it. they're so off-base that even Intel communities banned the site when they revised the algorithm to rank some few gen old i3 as the highest performing gaming cpu, which was even before they added all the latency stuff and "AMD is fake" disclaimers It's a real shame because they have a unique dataset and the site is still useful for comparing old hardware to new hardware. It's useful for benchmarking a new or old system and seeing where each component stands vs people with identical hardware. These are really useful things that none of the review/etc sites really do well. Khorne fucked around with this message at 23:23 on Jul 31, 2022 |
|
#
?
Jul 31, 2022 23:16
|
|