Vegemeister
No bio...
User ID: 1906
Vegemeister
1yr ago
Probably? I don't know if that would pick JEDEC timings, re-use the XMP timings in cycles directly, or adapt the XMP timings in nanoseconds.
I think he should try it and see, but I don't think he's reported back about whether he's found any test that reliably produces a crash.
Vegemeister
1yr ago
The widely-touted Factorio benchmarks run at well over 200+ UPS. The outsized gains from the big cache don't hold up as well with larger factories that struggle to maintain 60, presumably because they overflow it. Something that does help Factorio run faster on pretty much any computer is forcing it to use a larger page size with mimalloc.
Vegemeister
1yr ago
I have "HyperX® 32 GB DDR4-3200 XMP SDRAM (2 x 16 GB)". Is that el cheapo ship? Should I upgrade those instead? I believe I can handle pulling out and inserting memory sticks... I hope.
Aha! As Herr Bernd says below, that's decently fast memory, if HP enabled the XMP preset. But also, if they enabled the XMP preset, that's technically an overclock of your processor's memory controller, which is spec'd at 2933 MT/s max. XMP almost always works, even when it's out of spec, but... sometimes it doesn't. That might be the cause of your instability.
Unfortunately, memtest86+, which I suggested before, is more aimed at finding problems with the memory itself, rather than problems with the communication between the CPU and the memory. For that you'll want one of the overclocker-preferred stress tests that I linked before.
You can poke around in the BIOS setup options and try to find out if XMP is enabled, if so disable it, and see if that makes a difference to stability tests (assuming you find a test that reliably reproduces a problem). Turning off XMP will make your computer slower, and how much slower depends on what non-XMP settings are burned into the memory sticks.
If you're up for a challenge, you can try to stabilize the XMP profile with extra voltage, or find a stable intermediate speed between XMP and default non-XMP. If you try, the rest of that DDR4 OC guide will be helpful. But I do mean challenge. Memory is kind of the last bastion of "real" overclocking, in that every other part of a modern desktop has self-test circuits to characterize its own timing margin, and is able to run near maximum performance out of the box (if only for a short turbo boost time window). And unless you have error-correcting RAM like on a server or workstation, a memory overclock is the most difficult kind of overclock to validate, and the most likely to persistently corrupt your data.
I don't know if it works for notifications (Edit: it does), but if you go to your profile settings and add the following custom CSS rule, you can ctrl+f for new comments:
.unread::before {
content: "new";
}
Just like the old slatestarcodex comment section.
Edit: those are supposed to be tildes around "new", but the markdown code block is rendering it as a strikethrough on my machine, no matter whether I use 4-space or triple-backtick format. escaped inline backtick
Vegemeister
1yr ago
In particular I think that when we think about the term grooming in the non-culture-war sense, we are generally discussing a close relationship between a particular adult and a particular minor, where that adult intends to form a sexual relationship or otherwise initiate sexual contact with that minor.
As I recall, that usage is itself somewhat recent. Kind of like how a bunch of people switched from "CP" to "CSAM" around the time of the discourse about Apple's cloud photo backup scanning.
20 years ago, the most salient meanings of "grooming" would've been, first apes picking bugs and leaves out of each other's hair, and second preparing an employee for a promotion, typically to an executive position or to replace someone soon to retire. I for one, applaud the anti-trans set for managing to make that incredibly goofy word verboten among Respectable People. (If only there were some way to sic them on "colorway" next...)
They way they use it is denotationally the succession thing, but teases at the pedo thing.
The motte is that there is a funnel that begins with a tumblr or tick-talk subculture, or a danger-haired middle school teacher fresh out of women's college (I know one), and ends with endocrine-disrupting drugs ordered from an internet pharmacy or supplied by an NGO such as planned parenthood, and the people along that funnel view transition as a desirable outcome. The bailey is that the people along it want a sexual relationship with the people sliding down it. (Probably not true in general, but if you run away to join a trans-girl group house, I refuse to believe it's a nunnery.)
When that is not present, we need to look more at what the nature of the information and motives behind it are. If you ask activists about their motives, you'll hear about reducing suicidality, encouraging people to express themselves and form healthy relationships, and so on. Often you'll hear something to the effect of "what I wish I'd heard / seen when I was growing up."
It's a stand-alone complex. The core beliefs motivating it are
-
LGBT identities are common and an inherent character trait.
-
LGBT identities can be discovered by introspection.
-
Un-discovered LGBT identities are harmful.
You can imagine a similar stand-alone complex for pedophilic groomers. "I wish I knew how to make myself cum earlier growing up."
Overly empowering young people to make decisions about gender and sexuality that could have long term repercussions is a bad idea and minors need more supervision, guidance, and control than what is the current trend.
The way I would phrase it is more like...
It is terrifyingly easy to fall off the wagon, permanently. Just as the path to greatness in industry or academia begins in adolescence and is easily derailed, so does the path to spending many healthy years surrounded by loving family. The time a person has with their family begins with the birth of their first child, and ends with death. Therefore, you really want to avoid children getting taken in by subcultures that encourage self-sterilization or inceldom.
About the lifetime of BSOD, I think I've mentally resigned to suffering from monthly strokes because basically every PC I've ever owned has suffered from it. They ranged in manufacturers: Dell, Lenovo, HP, Asus, they're laptops and desktops, they ran various versions of Windows. I am a very respectable average user, I swear. I don't subject my machines to harsh physical conditions, never spill anything on them, don't live in filth where dust covers everything, don't live with electrical surges, don't have little cousins borrowing it, don't mine crypto, don't pirate or visit sketchy sites with viruses, don't open phishing emails, don't leave it on 24/7, don't unplug USBs until I'm told it's safe to do so, I update fairly frequently, etc. The machines I buy new directly from manufacturers or Amazon/Best Buy. Anyways, you get the point. And yet I've literally never owned a single fully stable machine.
There's probably some common factor, although we can only guess at what it is. Whenever I've seen a machine behave like that it's been some combination of
-
Installed in a shed with no climate control and free access to outside air.
-
Over a decade old (chips and capacitors do degrade).
-
Manufactured during the early 2000s "capacitor plague" (rumor says one capacitor maker tried to steal a formula from another and didn't get it quite right).
-
Fixed by spraying contact cleaner in the memory slots and re-seating.
-
Showing messages in the log that match a common complaint on the bugtracker for the Linux kernel or graphics driver, and the problem goes away when that bug is reported fixed.
-
My own damn fault for overclocking/undervolting something.
Things that might be different between us:
-
We have different electrical grids.
-
We have different levels of background radiation. (EPA says gamma cross count rate in my location is ~3000/min.)
-
Almost none of my machines run Windows (only the one in the shed). But people on the internet say Windows BSoD-ing all the time is supposed to be a thing of the past.
-
All of my machines are either home-built or business grade.
-
I run one pass of memtest86 whenever I get a new machine or replace RAM. Only time this found something though, was when I was buying dodgy RAM from eBay.
If your electrical supply is spotty, you might be able to fix it with an uninterruptible power supply that has the "AVR" (automatic voltage regulation) feature. Unfortunately they're kind of expensive and the batteries usually have to be replaced every few years.
I will share one suspicion I've had about the cause of the BSODs, in case it provides any obvious clues to you as to what's the main culprit. I use a browser plugin called video speed controller to speed up all kinds of media that are too slowly paced. I think my freezes have semi-frequently coincided with when playing a video at higher speeds (say, maybe 2.5x or even 3x). Do you suspect that to be a RAM-related issue?
Playing back video at high speed is obviously a heavier load than 1x, but it could be any of CPU, RAM, power supply, or even the graphics card, assuming your browser uses hardware video decode (probably does).
The first thing you might try is to see if you can reliably reproduce the problem by cranking the video playback speed to the moon. I use a similar extension, "Enhancer for YouTube", which has no upper speed limit AFAICT. Use youtube's "stats for nerds" to detect dropped frames, which means you have reached the limits of your computer (or internet connection). This probably works best with a short video that you can re-play without having to re-download.
If you can reproduce the problem, you have a very good "my computer crashes when I do this" story to tell the warranty people.
If that didn't work, to try to differentiate between causes and maybe find a better reproducer, I would suggest...
First, install hwinfo64. This will show you a bunch of things, but the important ones are the Windows hardware error log counts and the CPU temperature and package power. Here's an example of it in use.
Then download prime95. Run the "small FFT" test for at least an hour. If your computer crashes, any of the threads crash, any of the self tests fail, or hwinfo shows any errors in the Windows log, it is probably a CPU or power supply problem. If the CPU package power is not near or above 125W while the all-thread test is running, and the CPU temperature is at or very close to 100°C, it's a cooling problem (heatsink detached in shipping?). If "small FFT" doesn't find anything, you might try blend. Keep in mind "CPU problems" are likely to be "motherboard power delivery to the CPU" problems, so replacing the CPU might not fix it.
For the graphics card, you can use any of the unigine benchmarks. Superposition is the most similar to modern AAA games, but also a large download. You have a monster graphics card with a much higher peak power draw than the CPU, but if you only play games like Rimworld and SC2 with vsync on, it's probably not being pushed close to full power. Unigine will do that. Unfortunately, I don't know any GPU tests that check their own results and are easy to run. But if it crashes, that's a fail obviously.
For the memory, memtest86+ is probably easiest. There are better tests that the overclockers use, which you can find here.
To really put the hurt on your power supply and cooling, you can run 7 threads of prime95 and unigine at the same time. This will draw more power than pretty much any real workload other than folding@home, crypto mining, or things involving custom job schedulers, but a proper computer should be able to take it.
Unfortunately, there's no guarantee that you will be able to identify the problem. But the good news is that you only need to find a reproducer, to use as ammunition against the customer service line. That's part of what you're paying for when you buy OEM computers and replace them before the warranty runs out.
I love your optimism. I can tell you that none of the machines I've owned lasted 7+ years. It's not that they always become inoperable at that point, but that they seem obsolete by the 5 year mark at the latest. I don't mean to sound like a snob. It's just that a computer is what I interact with the most both professionally and leisurely, so I think it's worthwhile to invest good money in it. Like, if I drove 8 hours a day for work and for fun, you bet I wouldn't be trying to extract every last bit of value until it qualifies for cash for clunkers. Plus, I really don't think it's that wasteful; people replace their thousand dollar smart phones every 2-3 years, so going all the way to 7 years for a $1700 computer seems comparatively overly conservative.
No doubt. But security update stoppage and battery degradation are big drivers of phone replacements, and neither is a problem for desktop computers. I am using a $200 phone, a CPU launched in 2013, and a graphics card from 2016, and they do what I need them to do.
Vegemeister
2yr ago
Aside from a few random blue screen of deaths maybe once a month, which I feel like is a feature at this point with any brand of PC, no complaints.
This isn't normal. Or rather, perhaps it's normal in the statistical sense that the average person's computer is an unreliable heap of junk, but it's not nominal, and you shouldn't put up with it. "It just does that sometimes," is a piss-poor way to relate to computers, and if a hardware problem is causing your machine to crash that hard, it might also be corrupting your data.
You can use a couple passes of memtest86+ to identify some problems with your memory. It's not great for overclocking-related instability, but if your memory chips are going bad it should be able to detect it. You can run prime95's blend test overnight to ferret out CPU/memory/motherboard problems.
In my experience, poor electrical connections are the cause of a significant fraction of weird computer problems, although this may depend on the humidity in your climate. You can try re-seating your RAM and graphics card, as long as you are careful to avoid ESD. (Touch your computer's metal chassis immediately before touching any components, and do not remove from the confines of the chassis. Pop it out of the slot and right back in again.)
If none of that fixes it or finds anything, your computer is probably still under warranty if you bought it new. BSoDs are not supposed to happen, and you should make them somebody else's problem. The ability to do that is the whole point of buying from an OEM.
But I'm tempted by a new CPU.
First off, don't. In my opinion, your current machine will be fine for at least 5 years.
The newest CPUs that might be compatible with your motherboard are Intel 11th-gen, and those were widely panned for being an insignificant improvement over 10th gen. There are some workloads where they win, but some where they lose because the 11th gen i9 has only 8 cores compared to the 10th gen's 10 cores, and the power consumption is very high. That could be a problem for upgrading, because OEM (HP/Dell/Lenovo) motherboards are typically not designed to be capable of supplying significantly more power than needed by the CPU the PC comes with.
Furthermore, even if you replace your motherboard, published benchmark results for the 13th gen CPUs are usually using the newer DDR5 memory standard unless they say otherwise, so you'd have to replace your RAM too or else have slightly (only very slightly) less performance than the internet says.
UserBenchmark suggests that a 13900K outperforms the 10700K by 33% on "effective speed", or 61% on single core speed.
Userbenchmark is notoriously terrible. The operator has a strong anti-AMD bias. That wasn't too much of a problem back when Intel had a solid lead in single-thread performance and he could just weight low-thread-count benchmarks heavily, but since they've caught up he has to put a heavier and heavier thumb on the scales. At this point it's practically an entire arm.
The 13900K has as many P-cores as the 10700K, and 16 extra E-cores on top. Therefore, it makes no sense for the "effective speed" difference to be less than the single core speed.
The tricks, in this case, are:
-
The "effective speed" does not account for workloads using >8 threads at all.
-
The "effective speed" includes memory latency in the average. Memory latency contributes to the performance of a computer, but it isn't independently observable outside of its effect on any particular benchmark. It's an implementation detail. Picking a CPU based on memory latency makes about as much sense as picking them them by clock frequency or die size (i.e., none, unless you are designing a chip).
Unfortunately, unless the application you care about (Rimworld) is directly benchmarked, reading benchmarks properly is very difficult without a decent understanding of the characteristics of your application -- how threaded it is, how big its memory working set is (this is not the memory usage task manager shows you), etc.
Also, a lot of the published benchmarks really suck. Examples include single-thread cinebench (completely fits in cache on modern CPUs, and real users don't use Cinema 4D that way), Factorio benchmarks with small factories that run way over 200 UPS (broken by large L3 cache, which won't happen for factories that struggle to maintain 60), benchmarking Civilization games for frame rate instead of turn time, benchmarking frame rate in games that aren't CPU-limited in typical play (400 FPS 720p is benchmarking the graphics driver, not the game), etc.
What I would suggest is to find the openbenchmarking.org link from a recent CPU comparison article on phoronix.com, and filter the results to show only benchmarks that have similar characteristics to your application. For example,
-
Web browser tests: lightly threaded with small-ish cache footprint (based on 5800X vs 5800X3D.
-
Compiler benchmarks: heavily threaded with moderate cache footprint.
-
Google Draco: lightly threaded with large cache footprint. Most CPU-bound games are likely to fall in this category.
Perhaps I wasn't clear. I was talking about the individual becoming a parent with a healthy family of their own. This isn't possible if the individual's life course is derailed into HRT, inceldom, drug-addled homelessness, some kind of celibate religious order, etc. Furthermore, I believe healthy individuals and healthy families are, in almost all cases, synergistic, not things that trade off against one another.
Your hypothetical contradicts mammalian biology and results in total population collapse within single-digit generations, and hopefully re-establishment of civilization by the survivors or by whatever pockets of resistance managed to escape detection.
More options
Context Copy link