|
chocolateTHUNDER posted:Yeah, I upgraded to a 5600X3D this past summer (the microcenter exclusive one with 6 cores) and the leap from my 2700x to this is insane. Totally feel you on the 5600X3D upgrade. Everything runs so much smoother and faster, especially for gaming or when you're juggling a bunch of tasks. It's like getting a brand-new PC without actually getting one, lol. I popped in a 1TB NVMe SSD recently, and wow, does it make a difference. We're talking about your PC waking up in a snap and games loading before you can even blink. Pairing this with the 5600X3D, my computer is on fire. Found out about this combination on this software blog and not only. It's full of useful tips. (USER WAS PERMABANNED FOR THIS POST)
|
#
?
Apr 3, 2024 12:43
|
|
|
|
| # ? May 2, 2024 23:46 |
|
|
Lorentz Factor posted:Totally feel you on the 5600X3D upgrade. Everything runs so much smoother and faster, especially for gaming or when you're juggling a bunch of tasks. It's like getting a brand-new PC without actually getting one, lol. I’m still rocking a SATA SSD  idk, I’ve looked at benchmarks for game loading between SATA SSD & NVMe and most of them only shave off a second or two? Meh
|
|
#
?
Apr 5, 2024 01:51
|
|
|
chocolateTHUNDER posted:I’m still rocking a SATA SSD Pretty much. So far the only game I've seen where a super-fast NMVe drive made a big difference was Starfield, at launch, because the game had some hideously unoptimized small-block unqueued file reads that would choke even a gen5 drive and cause frame stutters. Even the PC port of Ratchet & Clank works fine on a sata SSD. So yeah keep rocking that sata cable, anyways now is not the time to buy SSD storage.
|
|
#
?
Apr 5, 2024 02:56
|
|
|
chocolateTHUNDER posted:I’m still rocking a SATA SSD Queues are what the reads, writes, trims, and other operations go into when dealing with any storage. Since the bulk data in video games consists of loading big sequences of random I/O (both of which are things SSDs are better at than spinning rust), it doesn't meaningfully benefit from all of the extra command queues found on NVMe storage, and most games don't implement reads to take advantage of the bigger queues. Klyith posted:Pretty much. So far the only game I've seen where a super-fast NMVe drive made a big difference was Starfield, at launch, because the game had some hideously unoptimized small-block unqueued file reads that would choke even a gen5 drive and cause frame stutters. Even the PC port of Ratchet & Clank works fine on a sata SSD.
|
|
#
?
Apr 5, 2024 13:01
|
|
|
oops
|
|
#
?
Apr 5, 2024 13:33
|
|
|
Do we have any expectation on Zen 5 availability? Do the subsequent x3D parts lag by a good bit usually?
|
|
#
?
Apr 5, 2024 19:23
|
|
|
v1ld posted:Do we have any expectation on Zen 5 availability? Do the subsequent x3D parts lag by a good bit usually? The original 5800X3D was a "gimmick" toward the (anticipated) end of Zen 3's lifespan. The 7800, 7900 and 7950 X3D models were announced right at the start of Zen 4's lifespan. Given their extreme popularity, I would absolutely expect that the (higher-end) X3D parts for Zen 5 will be announced in the initial lineup. EDIT: VVV Though, what Cygni said. The first X3D parts will probably be available a few months after the non-X3D ones; I would expect the 9800X3D (or a part like it, the 1CCD 8+ processor core model) to be available comparatively early, given how well the 7800X3D has done and how the response to the 7900/7950 was lukewarm. SpaceDrake fucked around with this message at 19:45 on Apr 5, 2024 |
|
#
?
Apr 5, 2024 19:31
|
|
|
Current rumors are that they will be announced at Computex that opens June 3rd, as they have the keynote there this year. Availability is sorta unknown yet, but all the leaks point to fall this year. August - September may be a good guess? There were some rumors that the X3D chips may arrive sooner to better compete with Intel's next launch, but the latest rumors suggest that plan is off. Wouldn't expect X3D until Spring if thats the case. For comparison, Zen 4 launched in September. X3D then launched in February technically (with the parts that underperformed vs expectations), and the 7800X3D didnt land until April. Cygni fucked around with this message at 19:34 on Apr 5, 2024 |
|
#
?
Apr 5, 2024 19:31
|
|
|
Seems like Zen 5 Bios's are rolling out, which implies a somewhat imminent release, Computex seems like a good bet. https://www.pcgamer.com/hardware/pr...next-gen-chips/ Earlier leaks suggested later half of the year but if they announce in Computex I would think they'd be looking at releasing.....something soon, though it depends on what the rollout actually looks like. I'd agree, X3D is probably 6ish months later, if they do release Zen 5 in June then I'd guess x3D would come out with the next Intel launch.
|
|
#
?
Apr 5, 2024 20:02
|
|
|
Thanks, folks. No absolute need to upgrade at the moment, think I can wait that long. I saw the bios news, but then searching for more news on Zen 5 availability had a leak by MLID as the most prominent response on Google. Made a hasty retreat to the thread immediately.
|
|
#
?
Apr 5, 2024 20:32
|
|
|
I thought the BIOS update for zen5 was due to the mobile chips just starting to hit the market?
|
|
#
?
Apr 5, 2024 21:31
|
|
|
The mobile chips are not anywhere near hitting the market and haven't even been announced yet, so no.
|
|
#
?
Apr 5, 2024 21:55
|
|
|
AMD has not yet announced when they are going to announce the new products, but they are doing the opening keynote at Computex and given the bios leaks everyone just assumes that that's when they do the announcement. For the past few generations, AMD announcement has predated the actual availability of chips by about 1-2 months, so June 4th announcement means chips probably on shelves in July.
|
|
#
?
Apr 5, 2024 22:13
|
|
|
Tuna-Fish posted:AMD has not yet announced when they are going to announce the new products, but they are doing the opening keynote at Computex and given the bios leaks everyone just assumes that that's when they do the announcement. For the past few generations, AMD announcement has predated the actual availability of chips by about 1-2 months, so June 4th announcement means chips probably on shelves in July. If that's the case I'd guess X3D chips by November/December.
|
|
#
?
Apr 5, 2024 22:34
|
|
|
CES
|
|
#
?
Apr 6, 2024 00:06
|
|
|
Koskun posted:I thought the BIOS update for zen5 was due to the mobile chips just starting to hit the market? I think you're thinking of the recently released 8000 series APU's which are pretty much the same as the 7000 series APU's (so same Zen4 CPU cores, same RDNA3 iGPU) but now they have a "NPU" built in for AI acceleration. Anyways, its a rumor but supposedly Zen5 will have a full 512 bit wide FPU (Zen4/c is 2x 256 bit) and Zen5c will be the same which will be interesting if that turns out to be true. If they can keep the chip from throttling when the FPU is in use it could be a nice boost for some work loads. If nothing else the PS3 emulation guys will be super happy.
|
|
#
?
Apr 6, 2024 05:52
|
|
|
PC LOAD LETTER posted:I think you're thinking of the recently released 8000 series APU's which are pretty much the same as the 7000 series APU's (so same Zen4 CPU cores, same RDNA3 iGPU) but now they have a "NPU" built in for AI acceleration.
|
|
#
?
Apr 6, 2024 06:10
|
|
|
some scuttlebutt about npu's displacing other features on upcoming cpu's in microsoft's rush to add ai crap to their os, but i i don't know how much ms get to dictate those things. ideally we can have something without an npu, but obviously that's not gonna happen for laptops https://old.reddit.com/r/GamingLeaksAndRumours/comments/1bxe90e/amds_next_generation_mobile_chips_were_supposed/ quote:Source: https://forums.anandtech.com/threads/zen-5-speculation-epyc-turin-and-strix-point-granite-ridge-ryzen-9000.2607350/page-350#post-41186460 i do wonder what all the trade-offs are gonna be for this ai rush on the consumer end of cpu's and apu's
|
|
#
?
Apr 6, 2024 17:09
|
|
|
Doesn't Apple's M series chips have (what we could call ) a NPU that is used to speed up a bunch of tasks?
|
|
#
?
Apr 6, 2024 17:14
|
|
|
Rawrbomb posted:Doesn't Apple's M series chips have (what we could call ) a NPU that is used to speed up a bunch of tasks? It has the Neural Engine NPU (16 cores/18T op/s in M3) but it’s not clear exactly what it’s being used for. There are some PyTorch conversion tools, but not much direct API for accessing it AIUI.
|
|
#
?
Apr 6, 2024 22:30
|
|
|
kliras posted:some scuttlebutt about npu's displacing other features on upcoming cpu's in microsoft's rush to add ai crap to their os, but i i don't know how much ms get to dictate those things. ideally we can have something without an npu, but obviously that's not gonna happen for laptops Microsoft gets to dictate what kind of PCs (laptops and prebuilts) ship with an official Windows certification, and hardware manufacturers will design around those certification/license requirements. If Microsoft says an NPU that can do 40 TOPS required for future Windows PCs, then you can expect AMD and Intel to do what they can to put those NPUs into all of their products. There could conceivably be DIY-only desktop parts that lack an NPU because AMD and Intel won't have to worry about those requirements then (assuming this won't become a hard-and-fast requirement for installing windows at all), but that would require a lot of extra investment. 40 TOPS is the number Microsoft gave as the requirement they're currently looking at: https://www.tomshardware.com/pc-com...npu-performance Intel's Meteor Lake chips currently have a 10-TOPS NPU, while AMD's Hawk Point chips have a 16-TOPS NPU. Both are very far from this requirement, so AMD and Intel are probably having to redesign some things to meet that requirement. And that will probably mean a bigger NPU that will require sacrificing something else on the die (like cache) All so you can run a local chatbot a little faster.
|
|
#
?
Apr 6, 2024 23:18
|
|
|
Is PCIe too slow to just throw an NPU on that and avoid hitching a design anchor to CPU design?
|
|
#
?
Apr 6, 2024 23:51
|
|
|
i would guess the problem with that is the card would need its own memory, so its cheaper to bundle it into the CPU where it can use shared RAM
|
|
#
?
Apr 6, 2024 23:56
|
|
|
CaptainSarcastic posted:Is PCIe too slow to just throw an NPU on that and avoid hitching a design anchor to CPU design? Intel is putting it on the SoC die, and i bet AMD also splits it off to a secondary die in their future designs after Zen5. Still silicon that we will all have to pay for even if it never gets used, though.
|
|
#
?
Apr 7, 2024 00:08
|
|
|
I do so not loving care about all this AI assistant poo poo. Also, there's rumors going around that there's going to be an AM5+ soon.
|
|
#
?
Apr 7, 2024 09:18
|
|
|
Combat Pretzel posted:Also, there's rumors going around that there's going to be an AM5+ soon. AMD has already publicly stated AM5 will be in use until very late 2025 or early 2026 so we've got at least 1 more major CPU core update coming after Zen5 for it. There is talk that the mobile socket might change to support a new DIMM standard (CAMM) but that is something entirely different than the desktop AM5 socket and is due to the memory standard possibly changing. AMD tends to lag Intel on changing sockets for new memory so I doubt this will happen "soon" as well.
|
|
#
?
Apr 7, 2024 15:59
|
|
|
CaptainSarcastic posted:Is PCIe too slow to just throw an NPU on that and avoid hitching a design anchor to CPU design? At the end of the day it’s all vector ops whether it’s called AVX on the CPU or tensor cores/vector units or whatever on a GPU or NPU. Integrated NPU silicon is all about providing efficient silicon to do inferencing (ie, executing an AI model) without fully powering up the entirety of the GPU’s tensor cores or CPU’s AVX.
|
|
#
?
Apr 7, 2024 17:34
|
|
|
Yeah, microsoft has said that they specifically don't want the GPU to be involved (in laptops, at least) because of how inefficient it is to power that compared to just using an NPU. Intel's own benchmarking on Meteor Lake shows this, where the fastest result is a mix of GPU and NPU, but pure NPU consumes the least amount of energy (207 joules vs 339): Not sure how this would compare to a dedicated nvidia GPU using its tensor cores, but that's probably going to be less efficient too since you'll be powering a lot of extra silicon. What Microsoft wants is for AMD and Intel to just use big-rear end NPUs that can do the above test in 5 seconds or less on their own without using crazy amounts of power to do it. It would be nice if Microsoft had a different set of requirements for desktop chips since power efficiency isn't as big of a concern, but so far they haven't said much about that. PC LOAD LETTER posted:Unless the memory standards change (not happening soon for desktop) or they want to suddenly support a iGPU that'll burn another 50W+ on the same package with the CPU's burning 150W+ (also not happening) AM5 isn't changing for a good while. https://www.tomshardware.com/pc-components/cpus/amd-may-have-a-new-platform-for-upcoming-ryzen-cpus The existence of an AM5+ socket is alluded to in the change notes for this tool. You shouldn't assume that just because AMD said they'll support AM5 for X length of time that all new architectures released in that timeframe will utilize AM5. AMD never made any such promise. "Support" could just come in the form of new CPUs released with old architectures, similar to how the 5700X3D released this year on AM4. edit: https://twitter.com/platomaniac/status/1776982918977532393 The reference to "AM5+" has nothing to do with a new socket, but I still think there's a good chance we see Zen 6 end up on a new socket. I am distrustful of AMD's marketing and promises, and I don't think we should blindly assume that "support" means new architectures. Dr. Video Games 0031 fucked around with this message at 21:10 on Apr 7, 2024 |
|
#
?
Apr 7, 2024 19:30
|
|
|
How far off are these new Windows standards?
|
|
#
?
Apr 7, 2024 19:42
|
|
|
They haven't said. There's a 50/50 chance that Microsoft waffles long enough that the AI boom wanes before any of this becomes reality, they only do a half-hearted attempt at pushing "AI PCs," and then drop the matter and pretend it never happened within 5 years.
Dr. Video Games 0031 fucked around with this message at 19:51 on Apr 7, 2024 |
|
#
?
Apr 7, 2024 19:48
|
|
|
it wouldn't be microsoft if they didn't introduce something either too early or too late for it to take off
|
|
#
?
Apr 7, 2024 19:49
|
|
|
kliras posted:it wouldn't be microsoft if they didn't introduce something either too early or too late for it to take off i'm looking at you, "3d objects" top level entry in windows explorer
|
|
#
?
Apr 7, 2024 20:04
|
|
|
Dr. Video Games 0031 posted:They haven't said. There's a 50/50 chance that Microsoft waffles long enough that the AI boom wanes before any of this becomes reality, they only do a half-hearted attempt at pushing "AI PCs," and then drop the matter and pretend it never happened within 5 years. inshallah
|
|
#
?
Apr 7, 2024 20:19
|
|
|
Dr. Video Games 0031 posted:The existence of an AM5+ socket is alluded to in the change notes for this tool. Vague allusions to all sorts of things that never come to light are going to be found in these tools, firmware, and software updates so getting worked about them as if they're solid evidence is pointless. Dr. Video Games 0031 posted:You shouldn't assume that just because AMD said they'll support AM5 for X length of time that all new architectures released in that timeframe will utilize AM5.
|
|
#
?
Apr 8, 2024 15:57
|
|
|
There were 4 Socket A CPU revisions, each requiring a new motherboard. Older CPUs could be used in newer boards, but new CPUs can’t run at full speed in older boards Socket 939 was not compatible with Socket 754 despite using the same chipsets and same architectures. Some later 939 even occasionally used the 754 bus speeds. AM2 CPUs could be used in AM2+ boards and vice versa, but some features were only active if an AM2+ cpu was in an AM2+ board FM2 CPUs worked in FM2+ boards, but FM2+ CPUs did not work in FM2 boards AM3 CPUs worked in AM3+ boards (and AM2/AM2+), but AM3+ CPUs did not work in AM3 boards What can we learn from this? I dunno man the past be like that
|
|
#
?
Apr 8, 2024 16:21
|
|
|
Apparently DDR6 is going to be coming out in 2026. Maybe AMD is gonna try to do the 2 memory controllers thing? It worked well for Intel in the DDR4-5 transition. So AM5+ might be the initial DDR6 platform that's compatible with AM5, but you need the new CPU to use DDR6.
|
|
#
?
Apr 8, 2024 16:49
|
|
|
Well the guy who put AM5+ on github and caused this rumor simply forgot AM3+ existed and is sorry about the confusion https://twitter.com/platomaniac/status/1776982918977532393
|
|
#
?
Apr 8, 2024 17:32
|
|
|
Cygni posted:What can we learn from this? It should be noted that many of the "new" sockets were minor revisions of the existing socket, which was why there was significant compatability across chips from the """old""" sockets, and that most of these revisions were addressing relatively minor issues. There isn't a hint of issues requiring a socket change coming down the pipe for AM5, it didn't play out that way for AM4, and AM3+ was around for quite a while too without issue. Klyith posted:So AM5+ might be the initial DDR6 platform that's compatible with AM5, but you need the new CPU to use DDR6. I'm no EE but it sounds like its going to need a whole new socket. I wouldn't be surprised if AM5 and AM6(?) exist along side each other for a while the same way AM4 and AM5 do with AM5 taking 4's place of being the cheaper option. PC LOAD LETTER fucked around with this message at 17:57 on Apr 8, 2024 |
|
#
?
Apr 8, 2024 17:46
|
|
|
What’s the motivation for moving power management onto the DIMM for DDR6? Is there some limit that’s being hit when pushing it from the motherboard, or do they just want to avoid motherboards loving with voltages badly?
|
|
#
?
Apr 8, 2024 18:04
|
|
|
|
| # ? May 2, 2024 23:46 |
|
|
PC LOAD LETTER posted:I can understand being cynical about any companies' comments about future products but truly new socket is expensive enough that launching one for a specific architecture is prohibitive for AMD and they're not going to do it without good reason. I think either the AMD CEO or execs have stated that Zen 6 is coming for AM5 as well so its not unreasonable to interpret "support" as supporting new arch and not just BIOS updates or continued production. I don't recall ever seeing any news stories about anyone at AMD saying that Zen 6 will be on AM5. I just did a search and haven't found anything. There's no reason to assume it will be on AM5 until AMD says so. I also don't think it will be a terrible crime if it isn't. AMD architectures are slower to roll out and last longer than Intel ones, so switching after two architectural revisions isn't as big of a gently caress you as when Intel does it. Dr. Video Games 0031 fucked around with this message at 18:17 on Apr 8, 2024 |
|
#
?
Apr 8, 2024 18:09
|
|