|
Which one is the proper memtest86? Plus or not-plus? Why is this even a thing?!
|
#
?
Mar 13, 2023 19:13
|
|
|
|
| # ? May 30, 2024 04:30 |
|
|
Combat Pretzel posted:Which one is the proper memtest86? Plus or not-plus? Why is this even a thing?!
|
|
#
?
Mar 13, 2023 19:33
|
|
|
Combat Pretzel posted:Which one is the proper memtest86? Plus or not-plus? Why is this even a thing?! Memtest86 was open source software. The original dev sold it to PassMark. When PassMark moved to version 5, they rewrote from scratch as closed-source. Memtest86+ was the fork that stayed open-source. It languished for a long time, and for years wasn't even usable (it didn't have UEFI booting years after UEIF-only was pretty normal). But recently has gotten updates making it better. I would go with non-plus closed-source memtest86.com. They have better and more thorough tests, run the tests better & much faster because they customize code to each architecture's hardware, and the limitations (only 3 repeats, can't save result reports, etc) on the free version are not relevant to the basic "test my ram" job. If you want fast testing, run a custom test and turn off the RowHammer test. RowHammer is IMO not terribly relevant for a home PC, and it takes a while. Klyith fucked around with this message at 20:26 on Mar 13, 2023 |
|
#
?
Mar 13, 2023 20:22
|
|
|
Those Hynix M timings from Buildzoid seem to work fine on these modules, even at slightly lower voltages than he's advertising in his video. DDR5-6000 ECC, oh yeah! --edit: Forgot guy's name. Combat Pretzel fucked around with this message at 23:19 on Mar 13, 2023 |
|
#
?
Mar 13, 2023 21:04
|
|
|
By optimize for architectures, do they mean microarchitectural optimization? Just for x86/amd64 there's over 50 of them. There's no way they can fit over 50 binaries in a zip that's only ~10MB.
|
|
#
?
Mar 13, 2023 22:30
|
|
|
BlankSystemDaemon posted:There's no way they can fit over 50 binaries in a zip that's only ~10MB. 
|
|
#
?
Mar 14, 2023 00:24
|
|
|
I'm glad someone else got something out of my joke. 
|
|
#
?
Mar 14, 2023 00:44
|
|
|
BlankSystemDaemon posted:I'm glad someone else got something out of my joke. Pain counts as something, yes
|
|
#
?
Mar 14, 2023 00:46
|
|
|
hobbesmaster posted:Pain counts as something, yes
|
|
#
?
Mar 14, 2023 00:48
|
|
|
Combat Pretzel posted:Which one is the proper memtest86? Plus or not-plus? Why is this even a thing?! Also, don't memory overclock; you will go insane. It's extremely hard to be sure it's actually stable, and you will never trust your system again. I say this as someone who spent approximately five weeks and many dozens of hours of stability system on overclocking DDR4-3600 to DDR4-4000 with much tighter timings for approximately ~10% memory benchmark gains. It was too much work to not use the settings but I have never trusted the system again since. It hasn't failed a stability test yet and I haven't seen any symptoms but after being disappointed by getting an error 9 hours into a stability test too many times (how do I know it won't fail 10 hours into a test, huh?) I am now 100% sure it'll start corrupting my files once summer comes and my apartment starts heating up. Everything affects memory stability. Oh and don't trust Buildzoid on memory stability, he is an extremely good resource and I've learned tons about memory OC from him and his discord, but he doesn't test for daily stability, all he cares about is benchmark stability which an entirely different beast. Also he'll say things like "all B-die will do X" but there are a lot of very bad chips out there and my particular B-die did not in fact do X (I've forgotten what exact timing it was; he was like "ok welp" but some dudes on his discord got into an argument with me if it was actually B-die or not but in the end had to agree that yes, it is, it's just lovely B-die). If there is one thing you take away from this it's that it is not enough with one memory stability test to declare a memory overclock stable. You want multiple different test programs and you want very long runs, ideally 10 hours or more (for 32GB; benchmark time scales almost linearly with amount of memory), and you should also try to add some thermal stress to the system if you want to be really sure. I'm not sure if these Hynix chips are temperature sensitive but many memory chips are. TheFluff fucked around with this message at 01:06 on Mar 14, 2023 |
|
#
?
Mar 14, 2023 00:57
|
|
|
The buildzoid timings being referenced are considered fairly lax timings that almost any Hynix M-die or A-die kit can run and were specifically tested for daily stability.
|
|
#
?
Mar 14, 2023 01:06
|
|
|
Dr. Video Games 0031 posted:The buildzoid timings being referenced are considered fairly lax timings that almost any Hynix M-die or A-die kit can run and were specifically tested for daily stability. I don't trust that whatsoever.
|
|
#
?
Mar 14, 2023 01:06
|
|
|
Like, just don't. Memory OC is not like CPU OC. You can set timings and it'll appear to work just fine and you run two different benchmarks for two hours each and it seems fine and the next day it corrupts your doctoral thesis (or 2000 hour Factorio save, or whatever). It is incredibly finicky, memory chips vary wildly in performance, CPU memory controllers vary wildly in performance, motherboards vary wildly in performance, heck even environmental factors vary wildly. If you're lucky Buildzoid has tested a couple dozen combinations. That's very helpful of him! It still isn't a guarantee that it'll work on your system even if it appears to work just fine at first! By all means do it if you like messing around but you need to stability test it yourself! yes I probably have mild PTSD from memory OC but it's not too late to learn from my mistakes
|
|
#
?
Mar 14, 2023 01:12
|
|
|
Do any of the memory OC guides encourage you to call it stable after just 4 hours? Then again at 4 hours stable I would save the bios preset as “stablish” and use that as a baseline for further tightening. BTW if you’re doing memory overclocking over the course of days you should run sfc /scannow somewhat regularly to see what you’re loving up during unstable boots. Do not open anything other than memory OC tools until you prove stability. Oh and don’t touch tREFI or explore aggressive tRFC unless you do not care about any storage device plugged into your PC.
|
|
#
?
Mar 14, 2023 01:24
|
|
|
I also have memory overclocking ptsd from frying my ram from overvolting. I did overclock my ram on my current ddr4 system but I was conservative and didn’t over think it. First I kept it at stock voltage and timings then went up in 100mhz increments until it crashed then backed off, 3333mhz @ 3600mhz for the last 5 years no problem.
|
|
#
?
Mar 14, 2023 01:26
|
|
|
TheFluff posted:yes I probably have mild PTSD from memory OC but it's not too late to learn from my mistakes See, the trick there is to buy ECC memory, OC that from JDEC to whatever similar retail kits are at speed wise, then forget about it forever, or at least until your machine segfaults or BSODs with an unrecoverable error message. Then you reboot, back the OC down 100 Mhz and go back to playing Factorio!
|
|
#
?
Mar 14, 2023 01:27
|
|
|
hobbesmaster posted:Do any of the memory OC guides encourage you to call it stable after just 4 hours? Not exactly, but Buildzoid likes to show a single full run of TM5 with anta's extreme config (takes around 3 hours for 32GB) in combination with a couple hours of something else as proof of stability. I was also somewhat concerned about Combat Pretzel who posted about memtest86 and then two hours later came back to say "these timings seem fine". I'm sure they do but please be more paranoid! e: actually he mentioned he has ECC on the previous page so I'm probably just an idiot and he can be a lot more nonchalant about this than I could! TheFluff fucked around with this message at 01:40 on Mar 14, 2023 |
|
#
?
Mar 14, 2023 01:37
|
|
|
K8.0 posted:Either your TV is a piece of poo poo or you're doing something very wrong. It's not any harder to get your PC to work with a TV than with a monitor, aside from TV-end stuff that you most certainly need to do for a console anyway. There's no difference between the output from a PC and the output from a console. The only thing that might be harder on PC is controlling when and how the TV wakes and sleeps. my (taichi ultimate+9900K, not AMD) and EVGA 3090 refuses to show any display output on my LG C1 until it gets into windows. It also hangs when trying to restart. So there's no way to get into BIOS with my C1  not that big an issue so I haven't taken the time to debug it
|
|
#
?
Mar 14, 2023 03:46
|
|
|
I never had a problem with memtest86 not being a good test of OC memory stability -- if it passed it argeed with OCCT, prime95 set to blend, and general observed system stability. But I never got close to "extreme" OCing in any way shape or form, just tiny improvements over XMP settings. Mainly running C15-3000 at C16-3200 with some tighter subtimings, to avoid the Ryzen odd-cas hate. I don't want to spend the time, and definitely wasn't spending the money on extreme OC RAM.
|
|
#
?
Mar 14, 2023 04:09
|
|
|
Dr. Video Games 0031 posted:The buildzoid timings being referenced are considered fairly lax timings that almost any Hynix M-die or A-die kit can run and were specifically tested for daily stability. hobbesmaster posted:Oh and don’t touch tREFI or explore aggressive tRFC unless you do not care about any storage device plugged into your PC. Combat Pretzel fucked around with this message at 07:46 on Mar 14, 2023 |
|
#
?
Mar 14, 2023 07:22
|
|
|
lol, was just looking up benchmarks for the new ryzen mobile parts and quote:Within minutes of the first, pre-release, 7000 series userbenchmark results, AMD’s marketers broadcast a 20% win over the 12900K via thousands of anonymous twitter, reddit, forum and youtube accounts. Buying new AMD products is like buying used cars: it takes time, experience and a taste for sales hype. It’s difficult for consumers to make rational choices while AMD completely dominates “sponsored news” and social media channels. Ten years ago, when AMD was the underdog, this type of marketing was understandable. Today, with a capitalization of $150 Billion USD, it’s disrespectful to AMD's own users. Even with Intel's marketing department asleep at the wheel, Ryzen will quickly end up in the same state as Radeon. Following a series of overhyped releases, consumers have little interest in the Radeon brand. The combined market share for all of AMD’s (discrete) Radeon RX 5000 and 6000 GPUs (Steam stats) is just 2%. Although the new 7000 series Zen4 CPUs are actually around 15% faster than their predecessors, they have hefty cooling requirements (TDP +60% vs 5000 series), 30 second BIOS post times, expensive DDR5 RAM requirements and only work with expensive motherboards. Despite the 7000 series struggling to match Intel’s outgoing 12th gen, AMD market it as a “future proof” platform! They want users to pay a premium for last gen performance in exchange for the shallow promise of upgrades in the future. Over the next few days, Intel’s 13th gen (Raptor Lake) will launch. Shoppers will do well to wait until then. Despite AMD’s Neanderthal marketing techniques, it’s hard not to admire their technical progress. AMD-Raptor4 and Intel-Zen13 would be better fitting product names. [Sep '22 CPUPro]
|
|
#
?
Mar 14, 2023 13:38
|
|
|
shrike82 posted:lol, was just looking up benchmarks for the new ryzen mobile parts and is that the userbenchmark guy?
|
|
#
?
Mar 14, 2023 13:41
|
|
|
yeah, notebookcheck did an article on them a while back and goddamn https://www.notebookcheck.net/UserB...n.638003.0.html quote:UserBenchmark has reduced its credibility levels further by simply attacking the latter, finishing an SKU description with: “if Zen 4 actually delivers anywhere near a 57% single core uplift, we will bow down, call AMD king, and commit seppuku”. This is in regard to the Ryzen 5 7600X sample showing an apparent +56% gain over the Ryzen 5 5600X in single-core speed. It’s likely there will be many DIY builders and desktop PC enthusiasts hoping for that to come true.
|
|
#
?
Mar 14, 2023 13:46
|
|
|
Subjunctive posted:is that the userbenchmark guy? He needs help. It was funny, but it's now concerning
|
|
#
?
Mar 14, 2023 14:12
|
|
|
HalloKitty posted:He needs help. It was funny, but it's now concerning It’s kind of funny in that it’s the most stereotypical kind of nerd rage possible but you’re right that the dude needs therapy bad.
|
|
#
?
Mar 14, 2023 14:13
|
|
|
Paul MaudDib posted:my (taichi ultimate+9900K, not AMD) and EVGA 3090 refuses to show any display output on my LG C1 until it gets into windows. It also hangs when trying to restart. So there's no way to get into BIOS with my C1 That also happens on monitors, though. I have an EVGA 3090 and it only displays on a single monitor (based on port) until the driver kicks it into multi monitor mode. If it's happening with only one monitor attached, that does seem a bit weird.
|
|
#
?
Mar 14, 2023 14:19
|
|
|
Twerk from Home posted:So I'm really hoping this isn't your problem, but my piece of poo poo Vizio overheats if you feed it 4:4:4 4k input like a PC would do. It does it faster and more often when it's warm, and less during winter. People got motherboards swapped during warranty but the problem just comes back. Oh hey, I had a similar problem with a D43-F1 - once it had been on for an hour or two, the image would get this periodic double-vision sort of flicker. Tried a couple replacement motherboards from eBay but one was for the wrong model of TV, and one was stuck in store display mode. I ended up going back to the original motherboard and using a hole saw to cut a port for a 120mm exhaust fan. With a 4-12V input range fan running off the USB port, it's basically silent but moves enough air that I haven't seen any problems since.
|
|
#
?
Mar 14, 2023 16:51
|
|
|
Interesting to see these in a server use case, I'd be curious to drop one into a 4u white box to replace some Haswell and Broadwell EP Xeons for general virtualization and FreeNAS use. https://www.servethehome.com/amd-ryzen-server-the-asrock-rack-1u4lw-b650-2l2t-review/4/
|
|
#
?
Mar 14, 2023 17:34
|
|
|
Hockenheim posted:Interesting to see these in a server use case, I'd be curious to drop one into a 4u white box to replace some Haswell and Broadwell EP Xeons for general virtualization and FreeNAS use. It's got AVX2 for offloading RAIDz matrix calculations and SHA2 for SHA256 checksums, so the only thing that's going to be using a much cputime is the compression algorithm. It's a pity that zstd doesn't really benefit from SIMD like AVX2, as Huffman and FSE compressions do, but given how fast zstd is, it probably doesn't mean much. BlankSystemDaemon fucked around with this message at 18:51 on Mar 14, 2023 |
|
#
?
Mar 14, 2023 18:46
|
|
|
If you can use the extra L3 cache, the X3D version should use similar power, especially if in Eco mode. The 3Dcache doesn't clock as high, so it doesn't consume as much power. Although, the heterogeneous nature of it might make using it more complicated than its worth.
|
|
#
?
Mar 14, 2023 19:13
|
|
|
Kibner posted:If you can use the extra L3 cache, the X3D version should use similar power, especially if in Eco mode. The 3Dcache doesn't clock as high, so it doesn't consume as much power.
|
|
#
?
Mar 14, 2023 20:35
|
|
|
KYOON GRIFFEY JR posted:It’s kind of funny in that it’s the most stereotypical kind of nerd rage possible but you’re right that the dude needs therapy bad. The guy is just a typical perpetually-online, curtains-drawn neckbeard. He doesn't need help, he needs to get a life.
|
|
#
?
Mar 15, 2023 11:44
|
|
|
Are we expecting any new AM5 boards before the 7800X3D rolls out, or is it still the introductory crop? All the chipset names blur together in my brain so I can never keep track. Have a case on the way, need motherboard and cooler and RAM and I'm all ready for that sexy, cachey baby.
|
|
#
?
Mar 15, 2023 14:14
|
|
|
Went to a local dealer today who told me that since I'm looking at On-DIMM ECC, I shouldn't be expecting availability until some time in August. 
|
|
#
?
Mar 15, 2023 14:16
|
|
|
Subjunctive posted:Are we expecting any new AM5 boards before the 7800X3D rolls out, or is it still the introductory crop? All the chipset names blur together in my brain so I can never keep track. No new major chipsets or revisions, so if any mobos come out they're just a new model. B650 = pretty much a direct continuation of the X470/B550 chipset with upgrades X670 = a second B650 chip connected to the first one to provide more PCIe, USB, and so on Add an E to the end = PCIe 5.0* *all AM5 boards have a gen 5 m.2 slot, so the ad copy for a non-E board may still say "PCIe 5.0!!!!!"
|
|
#
?
Mar 15, 2023 14:37
|
|
|
Klyith posted:No new major chipsets or revisions, so if any mobos come out they're just a new model.
|
|
#
?
Mar 15, 2023 14:47
|
|
|
There are AM5 boards with zero PCIe 5, but I don't think there are any non-E boards with an PCIe5 x16 slot. The presence of that is what allows manufacturers to use the E, and they all seem to do so.
|
|
#
?
Mar 15, 2023 14:52
|
|
|
BlankSystemDaemon posted:Plenty of AM5 boards don't have PCIe5, and some AM5 B650 boards have PCIe5 x16 as well as PCIe5 x4. for example...?
|
|
#
?
Mar 15, 2023 14:59
|
|
|
Klyith posted:for example...? Other manufacturers also have them. BlankSystemDaemon fucked around with this message at 15:20 on Mar 15, 2023 |
|
#
?
Mar 15, 2023 15:17
|
|
|
|
| # ? May 30, 2024 04:30 |
|
|
Storage Interface, CPU: 1 x M.2 connector (Socket 3, M key, type 25110/2280 PCIe 5.0* x4/x2 SSD support) (M2A_CPU) * Actual support may vary by CPU. 
|
|
#
?
Mar 15, 2023 15:26
|
|
