|
I turn my computer off every night because I don't trust windows sleep, and why would I just let it idle all night doing nothing? Still, an extra 20 seconds or so isn't a big deal for me since it's just once or twice a day typically. I'm avoiding turning on fast boot or any other similar features because I heard that some AM5 systems are running into memory stability issues with that on.
|
#
?
Apr 20, 2023 16:21
|
|
|
|
| # ? Jun 6, 2024 05:08 |
|
|
Windows Modern Standby is some of the dumbest poo poo Microsoft has ever done, but if your board supports actual proper S3 sleep instead (and every DIY board I've ever seen does, at least for now) then there's no issues, at least if you also turn off wake timers in power options so Windows can't wake up at 1AM to try and fail to install updates and then never go back to sleep thus defeating the entire pont. Now if we can just convince motherboard makers to implement "keyboard event wakes, but mouse event does not".
|
|
#
?
Apr 20, 2023 16:28
|
|
|
power crystals posted:Now if we can just convince motherboard makers to implement "keyboard event wakes, but mouse event does not". I have personally found you can do exactly that by adjusting settings in device manager, miraculously enough
|
|
#
?
Apr 20, 2023 16:29
|
|
|
Speaking of these fancy X3D CPUs is there a realistic upgrade path to using either the 5000 or 7000 series for me or should I be looking at building with a new MB/Ram? I'm currently running a B450 Aorus Pro Wifi MB, and currently a R7 3700x. Looking to upgrade/rebuild to get better performance in total war warhams 3. (Cooler is some dark rock tower monstrosity with two fans on it)
|
|
#
?
Apr 20, 2023 17:06
|
|
|
You can update your BIOS to support the 5800X3D (or verify your version already supports it) and basically drop that in for an instant $300 upgrade. Upgrading to a 7800X3D will speed up your warhams even more, but it also requires new RAM and a new mainboard either way (not sure if you can get a mounting kit for your CPU cooler, if not you also need a new one for AM5).
|
|
#
?
Apr 20, 2023 17:26
|
|
|
Had basically exactly the same question. Thinking of upgrading my 3600 to a 5800x3d. I've got an MSI Tomahawk B450 (non max) board. Latest bios seems to support it from everything I've read. Anything I should know or be worried about?
|
|
#
?
Apr 20, 2023 17:42
|
|
|
some boards don't have enough storage for support of both older and the most recent cpu's, but if you're on a 3000 cpu, both should probably work fine
|
|
#
?
Apr 20, 2023 18:13
|
|
|
kliras posted:some boards don't have enough storage for support of both older and the most recent cpu's, but if you're on a 3000 cpu, both should probably work fine IIRC it's only Bristol Ridge support that goes, and basically nobody is using those old Excavator arch APUs anyway
|
|
#
?
Apr 20, 2023 18:17
|
|
|
HalloKitty posted:It's not windows boot times, it's POST. On AM5 without restore memory context it is slow. Yeah I meant the overall boot experience. But like the other time it comes up is people who say some super-fast nvme drive is good because windows boots 2 seconds faster. HalloKitty posted:Maybe you use your PC in a different way, but I turn mine off or use hibernate every day. S3 sleep, instant on instant off. On windows I could log in before the monitor even came on. (linux the OS is much slower to wake up) I don't know why anyone would use anything else for a desktop, unless maybe you live somewhere with really rotten power reliability. HalloKitty posted:Having a black screen for an extended period is not exactly a slick experience. You're fine with it, great. It's definitely not ideal, and if I had a board with a 30 second post I guess I'd mildly annoyed just by the fact this is obviously broken and they should fix it.
|
|
#
?
Apr 20, 2023 19:19
|
|
|
I fully turn off every night, because windows sucks and there is no reason to leave it on. I have an HTPC/homeserver/NAS that is on 24/7 tho.
|
|
#
?
Apr 20, 2023 19:22
|
|
|
I dual-boot in addition to turning my computer off at night, so I reboot a fair amount. The few seconds it takes my machines to boot are short enough it's really not a problem.
|
|
#
?
Apr 20, 2023 19:24
|
|
|
Klyith posted:S3 sleep, instant on instant off. On windows I could log in before the monitor even came on. (linux the OS is much slower to wake up) That can be flaky on AM5 if the memory is set at anything but base JEDEC speed. Edit: although now I'm reading that can also happen if you disable the iGPU. Bizarre. And yeah, I wouldn't trust the power not to cut out, so hibernate gives me that peace of mind. Honestly, resuming from hibernation is so quick anyway, so that works for me. Klyith posted:Yeah I meant the overall boot experience. But like the other time it comes up is people who say some super-fast nvme drive is good because windows boots 2 seconds faster. Yeah, I agree there. Windows boot times have been a non-issue since we got even half decent SATA3 SSDs. vv There are posts about BIOSes that fixed the issue, I've seen it myself but from what I can tell it's probably related to disabling the iGPU, I guess I should test that more HalloKitty fucked around with this message at 20:10 on Apr 20, 2023 |
|
#
?
Apr 20, 2023 19:25
|
|
|
Dang, hopefully S3 sleep isn't too fucky on AM5 because I also use that over powering off every night. Haven't had issues with it since Win10 at least (aside from multi-monitor weirdness which was fixed in Win11).
|
|
#
?
Apr 20, 2023 19:54
|
|
|
If you overclock and follow the recommendations of disabling ACPI S3 STR/STD (because it's known to cause instability across S3 states) and disabling caching of memory training (this helps in maintaining system stability when overclocking), your system always takes 5 or more minutes to boot - all to get a few more FPS in a game.
BlankSystemDaemon fucked around with this message at 20:24 on Apr 20, 2023 |
|
#
?
Apr 20, 2023 20:21
|
|
|
This memory training bullshit, and caching the results or not. For real... What's the difference in training it once and keeping my box up 24/7 for a month and using memory context restore every daily boot???
|
|
#
?
Apr 20, 2023 21:13
|
|
|
Combat Pretzel posted:This memory training bullshit, and caching the results or not. For real... What's the difference in training it once and keeping my box up 24/7 for a month and using memory context restore every daily boot??? On boot, consumer systems have been relying on the caching of DDR training, in order to skip a large part of the long boot-up that's typically seen on servers (which, ideally, aren't rebooted and don't crash because they're built with system stability in mind through Redundant Array of Independent Memory and/or ECC - except for mandatory scheduled maintenance windows, where security patches can be applied). The details I know are specific to Intel, but AMD has to have something equivalent - because it's the case for anyone doing x86 of some description and everyone who uses DDR - anyway, here goes:
 I think AMDs Platform Security Processor maybe does some/all of the above, but since it's based on AMD TrustZone, it's not particularly fast either. EDIT: Oh, I forgot to mention, DDR memory has their own chips that also run their own firmware, as do the expansion busses like PCI(e) - and these are the bits that're communicated with in order to do link training. And all this hardware is proprietary, even on ARM and RISC-V CPUs. A modern computer is, quite demonstrably, 5-8 completely independent computers, depending on your definition (mine happens to be "capable of executing code in the form of firmware or software"). BlankSystemDaemon fucked around with this message at 22:49 on Apr 20, 2023 |
|
#
?
Apr 20, 2023 22:19
|
|
|
Nice post!
|
|
#
?
Apr 20, 2023 23:12
|
|
|
i leave my pc on 24/7 because it's 2023 and idle chips sip barely any power, the difference between a full blown desktop at idle with the monitor off and a NAS is maybe a dozen watts plus, the fans all turn off, what with all the giant heatsinks designed to dissipate a gorillion watts, so unless it's doing something heavy or gaming, it's silent, too instead, my pc is the only "real" computer that's in the apartment and i skip all the NAS/htpc/whatever else other people run in addition to their pc, and run that poo poo on the pc instead. it's got more than enough power for the desktop experience to not get bogged down by whatever else is happening on the network, what with the 16 cores BlankSystemDaemon posted:A modern computer great post
|
|
#
?
Apr 20, 2023 23:43
|
|
|
HalloKitty posted:I have personally found you can do exactly that by adjusting settings in device manager, miraculously enough Holy poo poo I never knew that was there. Thank you!
|
|
#
?
Apr 20, 2023 23:47
|
|
|
Truga posted:i leave my pc on 24/7 because it's 2023 and idle chips sip barely any power, the difference between a full blown desktop at idle with the monitor off and a NAS is maybe a dozen watts While you can turn off DC fans by stalling them through decreasing 12VDC, and it isn't going to immediately damage something as it's well within the tolerances, the current spike that comes with the drop in electric potential is the sort of thing that wears out electronics little by little. A PWM fan, on the other hand, turns off when you set the signal to low.
|
|
#
?
Apr 21, 2023 00:20
|
|
|
yeah i do have some old noctua case fans that run at 200rpm because positive pressure, but they will just shut off if i set them below 15%
|
|
#
?
Apr 21, 2023 00:36
|
|
|
BlankSystemDaemon posted:If your system is running all the time, there are (mostly internal, sometimes external) that keep things in check. a Good Post
|
|
#
?
Apr 21, 2023 03:47
|
|
|
BlankSystemDaemon posted:I plan on buying at least 1 stick of DIMM-wide ECC memory, even if I can't run it at the 6000MT/s that's the sweet-spot for Zen4 - just so I have something to validate the rest of the hardware against. Small correction, the dual data rate in "DDR" (which leads to the frequency mismatch) is independent of channel count. If it wasn't, we'd see higher channel count systems marketed with crazy numbers like 19,200MHz. It's actually due to transfers happening on both the rising and falling edge of the clock cycle, which is why the marketed number it's often referred to more technically as megatransfers per second. Consumer DDR5 platforms aren't even dual channel anymore, they're quad channel. We just still call them dual channel because A. Channels were conventionally 64b (and DDR5 channels are 32b), and B. There are two channels per DIMM so it is transparent to the consumer.
|
|
#
?
Apr 21, 2023 03:49
|
|
|
Kazinsal posted:We need to bring back POST cards as a standard with documented hex staging numbers so people can actually look up where in the boot process their new machines are hanging at. Most boards that do not include a 2 character 8 segment display have unpopulated headers that will either allow you to plug one in or solder connectors for one/or just rawdog it. I always thought that was one of the coolest things about AM4, I added a clock gen to my x370 board and didn't even have to patch the bios to do BLCK OC! I know I'm asking for a lot here, but it's hosed up that I know failcodes for several AM4 boards by heart at this point (or worse, what it actually means when the code says one thing but the cause is effectively unrelated). The new crosshair hero is like nine hundred and ninety nine earth dollars and it has all those goddamn LEDs yet still that same lovely display. Put a goddamn e ink display on there or something, give me words. Hell, put a frown emoji on there, anything but battleship (C4, B2, etc) Buddy of mine just built an AM5 rig, wanted to do a sesh to see what kind of timings we could get to... the board has a bug with retaining the memory training context so every single loving attempt takes minutes. I told him we will try again later, I don't have that kind of patience if it's not even my machine. Everyone I know of (legit builders, not scammer thieves like kirneill) who used to offer OC services + extended support for $150-200 outright refuses to gently caress with AM5 memory timings right now because of how time intensive it is.
|
|
#
?
Apr 21, 2023 05:50
|
|
|
BurritoJustice posted:Small correction, the dual data rate in "DDR" (which leads to the frequency mismatch) is independent of channel count. If it wasn't, we'd see higher channel count systems marketed with crazy numbers like 19,200MHz. It's actually due to transfers happening on both the rising and falling edge of the clock cycle, which is why the marketed number it's often referred to more technically as megatransfers per second. The 6000 in DDR-6000 refers to the number of MegaTransfers per second, and used to be an industry naming scheme (read: marketing bullshit) that was 8 times as high as the number of MegaTransfers - ie. DDR-800 was named PC3-6400 (the equivalent industry label would be PC5-48000). The Dual Data Rate refers to the ability of the memory to do two data transfer on the rising and falling edge of the clock signal, respectively, whereas Single Data Rate only has one data transfer on the rising edge of the clock signal, and Quad Data Rate uses a 90° phase change to manage 4 transfers per clock signal. BlankSystemDaemon fucked around with this message at 08:51 on Apr 21, 2023 |
|
#
?
Apr 21, 2023 08:39
|
|
|
Isn't DDR5 using QDR
|
|
#
?
Apr 21, 2023 08:41
|
|
|
VostokProgram posted:Isn't DDR5 using QDR Just to be clear, DDR5 is sort-of quad-channel whereas DDR4 ran true quad-channel (on some motherboards/CPU combos, not all). DDR5 achieves quad-channel by having each DIMM be two independent 32bit addressable sub-channels (for a combined data width of 64bit). However, the frequency of the memory is still only 3000MHz according to the SPD. EDIT: DIMM-wide ECC adds 4 bits of error correction per-32bit subchannel. EDIT2: This makes me wonder if, when you're addressing less than 2^32 bits in one transfer, you get the higher speed - but if you're addressing more than 2^32 bits, you don't? Although I don't think it matters, because native 64bit DMA isn't really something used anywhere outside of mainframes, if memory serves. EDIT3: Resizable PCIe BAR was made mandatory in PCIe 4.0 specification, and was supposed to fix this whole 32bit DMA nonsense, but I guess we're now stuck with a different problem? Computers were a mistake. BlankSystemDaemon fucked around with this message at 09:35 on Apr 21, 2023 |
|
#
?
Apr 21, 2023 08:46
|
|
|
EDIT: Double-post, please disregard.
|
|
#
?
Apr 21, 2023 08:53
|
|
|
BlankSystemDaemon posted:If your system is running all the time, there are (mostly internal, sometimes external) that keep things in check.
|
|
#
?
Apr 21, 2023 16:04
|
|
|
Combat Pretzel posted:I don't know. To me it seems entirely wrong that there'd be enough variance on every cold (and even warm) start, that it needs to retrain instead of just reloading previously measured values. Yet latter evidently doesn't appear to work reliably for whatever reason. Just because it's possible for it to not need to retrain doesn't mean the board vendors didn't implement it badly.
|
|
#
?
Apr 21, 2023 16:10
|
|
|
VostokProgram posted:Isn't DDR5 using QDR You might be thinking of GDDR5X/GDDR6?
|
|
#
?
Apr 21, 2023 17:02
|
|
|
apparently there might be issues with X3D chips burning out in ASUS motherboards? several examples like this were posted on reddit, all with ASUS boards and then ASUS started quietly deleting all the old AM5 BIOS downloads from their site, leaving only the newest version probably prudent to update your BIOS if you have an ASUS board repiv fucked around with this message at 13:31 on Apr 22, 2023 |
|
#
?
Apr 22, 2023 13:29
|
|
|
gn steve also purchased one cpu+mobo combo from a redditor, so you know asus are sweating asus are already on the lower end of mobo rep lately. still waiting on that pbo curve optimization support on x470
|
|
#
?
Apr 22, 2023 13:47
|
|
|
They got off to a bad start on AM5 with regards to memory compatibility too. It's just been one issue after another with them, it seems. In contrast, my gigabyte b650 board just works. I haven't had a single issue with them other than the fact that it sometimes requires me to manually restart the computer after changing some bios settings, and that's probably only because i'm on a beta bios.
|
|
#
?
Apr 22, 2023 13:59
|
|
|
Was it ASUS or ASRock that shipped with stupidly sticky substance that shredded itself when you sought to strip the sticker off the RAM slots? edit: Agh, alliterative associations inadequately apprehended afore allusions by aught hombre. SwissArmyDruid fucked around with this message at 14:07 on Apr 22, 2023 |
|
#
?
Apr 22, 2023 14:00
|
|
|
SwissArmyDruid posted:Was it ASUS or ASRock that put the stupidly sticky sticker that shredded itself when you tried to peel it off the memory slots? That was ASRock. I think they stopped doing that pretty quickly, thankfully.
|
|
#
?
Apr 22, 2023 14:01
|
|
|
I've been building PCs since 1995 and Asus has always been a crapshoot. I never understood the Asus=Quality rep that they finally seem to be losing. The only other brand I had go on me was a Gigabyte but it was DOA.
|
|
#
?
Apr 22, 2023 14:02
|
|
|
I liked Asus during the nForce chipset era, my Opteron 165 build from back then is still one of my faves. Had a few problem boards mid to late 2000s though and haven't gone back to them in a while. My last 3 builds were on MSI boards and have been great, this time around I went with a Gigabyte X670 AORUS Elite AX.
|
|
#
?
Apr 22, 2023 14:19
|
|
|
slidebite posted:I've been building PCs since 1995 and Asus has always been a crapshoot. I never understood the Asus=Quality rep that they finally seem to be losing. This might partially be because those customers are also the ones likely to buy reviewed units, which don't tend to be the absolute bottom-barrel stuff. Another explanation, of course, is that we have absolutely no idea about the actual failure rates of production lines vs what passes QA vs how much RMA any of the vendors have to process.
|
|
#
?
Apr 22, 2023 14:24
|
|
|
|
| # ? Jun 6, 2024 05:08 |
|
|
Dr. Video Games 0031 posted:They got off to a bad start on AM5 with regards to memory compatibility too. It's just been one issue after another with them, it seems.
|
|
#
?
Apr 22, 2023 14:32
|
|