|
njsykora posted:Well that sucks, feels super weird to do a single region product launch in this day and age. I'm just glad that AMD is selling these rather than throwing them in the garbage. I love weirdo runt products from the clippings.
|
# ¿ Jun 30, 2023 15:06 |
|
|
# ¿ May 16, 2024 23:09 |
|
Kerbtree posted:Hah, looked at the 5800x3d userbenchmark page for a giggle and they're certainly saying the quiet parts out loud: God I wish userbenchmark had been around during the Pentium 4 days.
|
# ¿ Jul 3, 2023 15:05 |
|
BlankSystemDaemon posted:RT is great for screenshots, but I sure as poo poo don't notice it when I'm playing a game where it's enabled. You're not alone that most things driving ultra graphics over the last few years feel like marginal improvements at huge performance costs. Then occasionally something comes out like TLOU where the only settings options are "looks amazing but needs 12GB VRAM" or "PS2 textures".
|
# ¿ Jul 3, 2023 17:49 |
|
CodFishBalls posted:To chirp in, Micro Center is likely going to do a 5600x3d bundle, bundling it with a B550 board (my suspicion is the Tuf B-550 since they got quite a bit in stock at my local store ) and 16 gb of 3200 cl16 g.skill ram (probably ripjaw again, what they have a lot of) drat, hell of a bundle if that's it, maybe enough to make me road trip.
|
# ¿ Jul 7, 2023 02:42 |
|
Cygni posted:isn’t there like 1 thing that actually uses avx512 that anyone cares about, and it’s a ps3 emulator? Why do all the work to use hardware features that aren't common? In the one place where AVX512 should have mattered, Intel Xeons, they've hosed that up too. Low end Xeons only have 1AVX 512 FMA port per core, so they have half the throughput of the more expensive Xeons. If you do pony up for the more expensive Xeons, those reduce clocks on significantly if you actually use AVX512 instructions: https://travisdowns.github.io/blog/2020/08/19/icl-avx512-freq.html By the time Intel finally got their act together, with Ice Lake, AMD had better performing processors around that don't have AVX512, and is starting to make huge gains in server market share. If we start actually getting AVX512 adoption in software, it's going to be because of AMD. Hell, I've hand written some vectorized code with intrinsics recently, and AVX 512 was way, way nicer to work with than AVX2. My worse, less efficient AVX2 code path ended up being faster on Skylake, and we have a lot of Skylake. It was slightly faster on Ice Lake, but not enough to keep around. I wish they would have called it AVX3 and made the 512 bit part optional. That seems like most of the source of problems but there's a lot of nice instructions in there that are perfectly usable with 256 bit vectors.
|
# ¿ Jul 11, 2023 02:57 |
|
AARP LARPer posted:Hi everyone, You can use Ryzen Master to set power targets, and that's going to be more effective at reducing wattage than messing around with offsets: https://support.punchtechnology.co.uk/hc/en-us/articles/6438162546461-How-to-set-a-TDP-PPT-Power-Limit-on-AMD-Ryzen-CPU-Processors. Is your machine on a UPS or any other device that measures power usage? How low are you looking to get? What power supply are you using, and is it a correctly sized, efficient one?
|
# ¿ Jul 15, 2023 21:46 |
|
hobbesmaster posted:That’d be cpu migration or cluster switching and not an HMP implementation like Intel, right? The scheduling problems might not be as bad, but any situation where you've got fast cores and slow cores creates scheduling problems. The small Zen cores might have the same instruction set but they don't clock or behave quite the same.
|
# ¿ Jul 30, 2023 18:50 |
|
Josh Lyman posted:I remember when someone, maybe from HardOCP, sold a baybus where you had to drill holes in your own front plate to install it. I think I got one back in like 2001? A few years after that I had an off-the-shelf cooler master cooler that had a rheostat that slotted into a PCI slot cover, so you could reach behind your PC to crank it up for game time.
|
# ¿ Aug 7, 2023 21:09 |
|
Paul MaudDib posted:pre-Haswell/pre-Ryzen is coincidentally also when NVMe support started to be rolled out, and you can't play the game without NVMe storage either. it's easy to justify not writing that pre-AVX2 path when the users mostly can't play anyway for other reasons. Wait what, games actually require NVMe storage now? TLOU at that? Wow, I assumed that SATA SSDs would just make for slightly slower load times.
|
# ¿ Aug 9, 2023 00:45 |
|
ConanTheLibrarian posted:Epycs only allow one DIMM per channel and the number of people who need 4 DIMMs is a drop in the ocean so AMD don't have a lot of motivation to make a memory controller that can handle 4 DIMMs well. All Epycs have been designed to support 2 DIMMs per channel, and AMD says that they're able to fix their latest generations memory problem with just a BIOS update. https://www.tomshardware.com/news/amd-responds-to-claims-of-epyc-genoa-memory-bug-says-update-on-track 2DPC gets Epyc to 48 DIMMs in a 2 socket server.
|
# ¿ Aug 10, 2023 14:55 |
|
VorpalFish posted:Call me crazy, but I suspect now that they're putting igpus just good enough for office pcs on the io dies of chiplet ryzens maybe there won't be any more socketed monolithic APUs. Given that every console and laptop part is still monolithic and will continue to be, laptops need big iGPUs more than desktops, and the current fastest iGPUs are all laptop only, AMD could pretty easily bring them to desktop if they wanted but they may not choose to. If they don't sell them socketed they'll still be available in mini-PCs.
|
# ¿ Sep 10, 2023 16:10 |
|
BlankSystemDaemon posted:Modern mainframes have over a GB of distributed L3 cache. https://www.anandtech.com/show/17323/amd-releases-milan-x-cpus-with-3d-vcache-epyc-7003 You can get 1.5GB of L3 cache in a 2 socket commodity server as of last year.
|
# ¿ Sep 30, 2023 21:08 |
|
Shaocaholica posted:So X3D cache is accessible by all cores with the same penalty and its slower than L3 but orders of magnitude faster than main memory? And X3D is not available on Epic/Threadripper...yet but there are plans? X3D has been available on Epyc for a whilez released March 2022: https://www.anandtech.com/show/17323/amd-releases-milan-x-cpus-with-3d-vcache-epyc-7003 The 3D cache is part of L3 and it makes the whole L3 a little bit slower than a non-3D part, but usually having all that extra cache helps way more than the tiny hurt you get from it being slower.
|
# ¿ Oct 19, 2023 17:38 |
|
Dr. Video Games 0031 posted:Not sure where else to post this, but details on Qualcomm's upcoming Snapdragon X Elite have leaked: https://videocardz.com/newz/snapdragon-x-elite-leak-12-core-oryon-cpu-4-6tf-adreno-gpu-lpddr5x-av1-and-wifi7 Is there a native port of MS Office yet? I'd bet that for a whole lot of usage cases, if you've got an ARM browser and MS Office, then the substandard emulation for the long tail of other applications is fine. Come to think of it, anything that's .NET and JIT-ed anyway will be completely fine already too. I know that ARM is making inroads and gaining territory in the server segment, and clearly Apple has had enormous success. The last time that x86 and ARM went head to head was the Atom Phone days, but that was a weird market and time and Intel may well have sabotaged themselves. For Windows laptops, even if ARM doesn't gain market share quickly it'll limit AMD and Intel's profit margins.
|
# ¿ Oct 23, 2023 00:58 |
|
Farmer Crack-rear end posted:shsc just loves segmentation, there's like three separate work chat threads. see klyith mentioning overclock chat when, hey, there's already an overclocking megathread! I'd love to see a unified x86 thread, ideally with a OP that mentions Centaur and VIA.
|
# ¿ Oct 31, 2023 00:56 |
|
VorpalFish posted:To play devil's advocate, that's kind of how max turbo speeds end up working on everything, especially low power mobile chips. You only ever see it with 1c or 2c load, and that gets dispatched to a preferred core. Friend, have you taken a look at the Intel thread and their "14th generation" desktop chips?
|
# ¿ Nov 3, 2023 15:16 |
|
Mental Hospitality posted:The laptop Phoenix APU's seem quite performative in those mini-pcs I keep seeing on YouTube. I wonder if you'll be able to go hog wild with the power limits to get even more performance from the desktop parts. I'm sure you'd reach a point of diminishing returns but if someone already has a beefy air cooler it would be neat to see just how much performance you could extract out of the chip. I'd posit "not much" given that we already saw the 5800H laptop vs 5700G desktop comparison and it looks broadly like this: https://www.cpubenchmark.net/compare/3907vs4323/AMD-Ryzen-7-5800H-vs-AMD-Ryzen-7-5700G If you're able to throw way more power at the GPU, you also start running smack into memory bandwidth limitations. 2 channel DDR5 just isn't anywhere near GPU memory bandwidth.
|
# ¿ Nov 9, 2023 17:49 |
|
Rinkles posted:speaking of, the Steam Deck refresh (OLED + die shrink) apparently performs slightly better thanks to having faster memory. With how modern CPU clock & power management works, a die shrink alone will probably result in a very small performance boost by itself. They added a bigger heatsink & fan also, it's just a general tweak
|
# ¿ Nov 9, 2023 19:17 |
|
Cross-Section posted:With all the tech sales coming up, should I buy a good pair of DDR5 sticks in preparation for the eventual Ryzen 9000 CPUs or is it likely that RAM prices in late 2024 (or whenever these CPUs drop) will be as low as they currently are? I wouldn't do this for DDR5. Memory speeds, voltages, and support is still improving and if you want a bunch of memory or really fast memory I'd wait and see what next-gen processors want. I do think this is the bottom for DDR4 prices though!
|
# ¿ Nov 12, 2023 16:28 |
|
BurritoJustice posted:In terms of factorio, with large bases it is completely memory bottlenecked. This means that the VCache still helps, as cache increases effective memory bandwidth, but Intel's much faster memory subsystem gives them the win in general. Amusingly, I do actually have the world record on the FactorioBox 50K benchmark (61UPS), despite using a 7950X3D, but I cheated a little by using Linux and a bunch of kernel tweaks that help massively (custom malloc with huge pages and much more). You can get similar performance using Intel on bone stock windows. Is Factorio more memory latency sensitive or bandwidth sensitive? Does it do any better on 4-8 channel DDR5 setups, or does the higher latency that bigger memory controllers and potential RDIMMs bring hurt performance?
|
# ¿ Nov 23, 2023 06:51 |
|
Anime Schoolgirl posted:Z1 NUClike in development What's the appeal? 7020U is Zen 2, right? Wouldn't you be better off with the 5800H, Zen 3, which you can still get with 16GB of RAM and 500GB of SSD for under $300?
|
# ¿ Nov 27, 2023 23:24 |
|
SpaceDrake posted:So I had the itch to play around today, and I was a little surprised to see that 2001SE actually booted and ran on my 7800X3D/4070 Win11 system... but 3DMark 03, 05 and 06 wouldn't. (I was less surprised to see that 2000 didn't start correctly, and 99 refused to even install. ) Sadly, they give extremely nonspecific errors when starting up, so it's hard to troubleshoot what's going on. I'm tempted to download Vantage and see how well the system crunches through that, but 2001SE was funny enough as it was, and the 11 result was impressive enough for my blood. Maybe it wanted Creative EAX? Get yourself an Audigy, friend. poo poo, you just reminded me that I still have a dedicated sound card in my desktop because the onboard realtek made me mad. Asus Xonar to the rescue!
|
# ¿ Dec 28, 2023 16:17 |
|
gradenko_2000 posted:is AMD really going to be shipping product where only the left-most digit is moved up from a 7 to an 8, but all the rest of the digits (meaning the actual technical specifications) are the same? Its just traditional, every vendor does it. Remember the Nvidia 8800 GT / 9800 GT?
|
# ¿ Jan 3, 2024 03:56 |
|
OhFunny posted:Apple did this too. iPhone numbering jumped from 8 to 10, so not to be a number behind Samsung. I thought it was because naive software checks for "iPhone 9x" would think it was running on iPhone '95.
|
# ¿ Jan 5, 2024 00:11 |
|
Canned Sunshine posted:Zen 5 is also moving from 4 to 6 ALUs, so I'm wondering now if Zen 5c will see that, or will remain at 4 if it's based off of Zen 4. Ostensibly the 4c/5c cores are competing against Intel E-cores that lack not just clocks but only have 128-bit wide vector units, right? They have no need to worry about getting outperformed.
|
# ¿ Jan 13, 2024 17:28 |
|
Jeff Fatwood posted:Where are the 5700X3D reviews, that drat thing released today and there's no coverage. I think I'm going to drop one in to replace a 1700X. Does it come with a cooler? I'm using a stock Wraith, I wonder if box coolers have gotten any better in the last 7 years. Intel's new ones look pretty sick.
|
# ¿ Jan 31, 2024 19:23 |
|
Anime Schoolgirl posted:all X3D chips don't have coolers bundled. That's a tightwad move for some expensive chips.
|
# ¿ Feb 1, 2024 01:17 |
|
Shipon posted:absolutely no one buying an X3D was going to use the stock cooler anyway, it would have been a waste of money Should I not? I was planning on it, after doing a hand-me-down / leftovers build where the only other new parts were $50 worth of RAM and a $40 case, only to discover that a 1700X is slower than I thought and get tempted to blow the budget and upgrade. My kids want to play Planet Zoo and other stuff that won't run on the Intel iGPU they started gaming on. $249 feels like a pretty decent value, I guess I could cheap out and get a 5600X for half that but the X3D and 8 cores sure is neat. Edit: I have an AMD branded cooler with an RGB LED ring around it that looks a hell of a lot like the Phenom II X4 965 box cooler I had in 2009. I thought this was a wraith and what came with the 1700X, but I got this CPU, cooler, and MB given to me by my brother in law and don't know the whole history. Rather than taking a photo and showing y'all, I googled and apparently this is a Wraith Prism. It looks like it should do OK with a 5700X3D? Twerk from Home fucked around with this message at 01:36 on Feb 1, 2024 |
# ¿ Feb 1, 2024 01:32 |
|
repiv posted:RE Engine calls checkerboard rendering interlacing for some reason That feels accurate enough, neat. I thought checkerboarding had fallen by the wayside now that we have better upscaling tech?
|
# ¿ Feb 4, 2024 21:26 |
|
The same thing happened with Intel, the 12700K is faster than and has been cheaper than the 12700 non-K.
|
# ¿ Feb 13, 2024 21:21 |
|
Josh Lyman posted:Would it make sense to think of my 7900X as 2x 6 core CPUs rather than 1x 12 core? If so, is Intel’s P+E architecture better about this type of thing? Intel's P+E architecture in client chips has a single L3 cache, yes. However, Intel's setup has its own big problems with caching, coherency, and inter-core bandwidth. When the E cores wake up it slows down the ring bus, slowing down how fast the P-cores are able to access L3 or memory: https://chipsandcheese.com/2021/12/16/alder-lake-e-cores-ring-clock-and-hybrid-teething-troubles/. Also, the E-cores share an L2 cache but have to synchronize through the L3 cache, meaning that core-to-core latency is worse for E cores that share an L2 cache than P-cores with each have independent L2 caches.
|
# ¿ Feb 26, 2024 01:35 |
|
forbidden dialectics posted:...is it possible to learn this power? Look at your mobos QVL RAM list, pick the highest validated speed you see on the list, crank up the voltage one notch past that and run tight timings. Every board's QVL list has some insane poo poo on there like running a 2400 kit at DDRR-4400 or faster at 1.6V or something.
|
# ¿ Feb 27, 2024 17:27 |
|
Klyith posted:I can think of lots of good reasons. Macs hold their own at games as long as you buy some $75 software to run Windows games that may or may not continue to work in the future: https://www.codeweavers.com/crossover/ It's incredibly annoying that a bunch of games that used to work natively on Mac no longer do. We lost a lot when Apple removed the ability to run 32 bit applications, and are losing way more as OpenGL keeps rotting. I do wonder if Apple is going to go all the way and just kill OpenGL in the next few years. My experience has been that tons of games that I expected to run on Mac OS do not. CS: GO, TF2, Guild Wars 2, none of them run on Macs at this point. Twerk from Home fucked around with this message at 05:25 on Apr 9, 2024 |
# ¿ Apr 9, 2024 03:54 |
|
VorpalFish posted:It's more than that - the die with vcache on it is much harder to cool so they dial clocks and voltages way back - the 7950x3d has a die without vcache and that's almost certainly how it hits those clocks. I didn't realize that the X3D chips were exposed to the OS as heterogenous, I thought that part of the issue is that the OS and thus the scheduler have no idea that there's 2 different kinds of cores there?
|
# ¿ Apr 12, 2024 16:54 |
|
|
# ¿ May 16, 2024 23:09 |
|
Subjunctive posted:I actually have a use-case at work for this sort of setup right now (run many tiny web assembly functions, as parallel as we can get, over small data) but I doubt I can get GCP to deploy them for us! Isn't that usage case of "tons of cores without significant I/O or memory bandwidth" about to be completely devoured by both ARM64, whether Ampere or in-house, or E-core Intel? I haven't looked at the GCP pricing comparison, but I do feel like we're heading for a future where for those kinds of workloads it's not cost effective to run them on mainstream Intel or AMD. I guess we'll see how Zen 5c SKUs and pricing shake out, or if the big cloud guys buy from Ampere or just roll their own ARM.
|
# ¿ Apr 13, 2024 16:39 |