#off-topic-tech
1 messages Β· Page 195 of 1
And maintenance is lot more than just replacing parts.
But it does mean good upkeep
I have a nail as a wheel bearing on my four wheeler
So much this
Better than nothing
And mostly bearings are pretty standard, getting something that fits is usually possible, even when the exact part code isn't available.
I got one but i have to grind it down a smidge since it doesnβt fit
Yeah, the bearing outer and inner races must not turn in respect to their mating surface while in use.
Yep
Lots of later cars have changed to wheel bearing that you replace with bolted flange. So no press needed for the job. But more expensive bearing as new comes with the flange
Honestly im woried about the motor more than anything
its so burnt and covered in oil that its literally black like coal
Eh, that in and of itself isn't a huge concern
Small oil leak from the top end will easily get covered in dust and grime on atvs
Like this
yea well it has no manifold, runs too rich, has a foam block held on by a belt as a air filter, roals coal out the right side of the head gasket and so much more
Its 40 and has nerver been rebuilt tho so it makes sense
Weren't you just complaining about how it's broken?
When you just told lot of stuff that is broken, and "fixed".
I understand that your atv probably isn't something that matters a lot to you, but from my perspective you rarely have a good reason not to take care of something
yea since a mechanic would have a stroke at the sight of it and the patch jobs but at the end of the day everything works in a usable state other than the high,low switch
i maintain what i can but at the end of the day its hard to throw money and time into parts for something worth less that even 30% of what needs replacing in a ideal world
And that's fair
I keep up on oil changes, brakes, mechanical stuff but my car is never getting repainted even though it's down to the primer in spots
That doesn't stop me from doing my best to take care of it though
And there's a little bit of difference between paint and having a cobbled together wheel bearing on an offroad vehicle that is ridden aggressively
i changed the oil once when my grandpa put mixed gas in it thinking it was normal gas when i asked him if he would fill it up
that was βfunβ
Most atvs expect an oil change every 10-20 hours
thats it?
That's it
They really are high-performance engines after all
Work-oriented machines usually aren't quite so bad
But most atvs have small filters and an even smaller oil capacity
Usually ~.75L
ok well once in like 8 years is pretty bad i have to say
At the most, down to around .5 for some models
but it had no issues and it also sits a lot
Yeah, generally that means you need one a year no matter what
For best performance and engine longevity
Best performance is called the winter atmosphere
Put some load.
PCIe can downscale to save power.
And then load the same screen again while something 3D is running.
Oh good
we love pcie
Thank god not nvidia x8 in 2025
AMD has GPUs with 4x (laptop recycle on desktop space) and 8x max widths.
B580 is 8X but idk about nvidia
Is x16 ever useful in some sort of burst workload scenarios for gaming? I've always figured laptops should be x8 max, and reallocate the lanes to Thunderbolt.
Basically once a card has "enough" bandwidth, more doesn't matter except if you run out of VRAM.
With that AMD 4x card, not having designed 4.0x4 for it hurt, like even 3.0x4 caused performance regression.
But that didn't matter in the original laptop use, as any CPU it was to be paired with would have had PCIe 4.0.
But 5090 doesn't really care if you put it into 5.0x16 or 3.0x16 slot IIRC.
When tested in games.
CUDA use is different.
For gaming workloads, the data throughput over pcie is fairly bursty but during average gameplay is also fairly low. Most commands sent to the gpu over the interface can more or less just allow it to operate on data already in vram.
Loading new models or large datasets into vram is the biggest bottleneck, usually
Or queueing up a very large batch of commands
when that 8 gb nvidia brick runs out of vram it needs pcie lines to access system ram
I can recall gaming laptops boasting about x16 lanes for their GPUs in advertising, as if it does anything. Better to figure out how to reallocate that to USB ports and external IO so everything has more throughput, and you don't need to remember which USB port is 5 gbps or 40gbps.
Yep
Especially where modern usb/thunderbolt can do 4 lanes of gen4, you really do need it for io
Especially if you have 8 spare lanes of 5.0 which can turn into pretty much as IO as you can fit on a laptop.
Yep, it's an insane amount of bandwidth
When you really stop and think about how much just a few KB/s is, GB/s gains a lot of appreciation
PCIe 5.0x16 has about same bandwidth as dual-channel DDR5 4800-5200 @ JEDEC.
Ok, little less than 4800 JEDEC
At full-duplex, while RAM is half-duplex.
I mean that in one direction they are close, but PCIe can do both at same time, while RAM cannot.
But RAM has half-to-quarter the latency of PCIe, depending on settings.
50-100ns for RAM, around 200ns for PCIe
Which is th entire reason we use a custom interface for it(that and the fact that it's much easier to handle refreshes that way)
But off to bed ->
its not really bandwith thats the problem tho no?
isent the entire problem that its too much additional coordination to be worth it
ok now i have to go back and see if GN or HUB did 4090/5090 testing at x16 & x8
oh, literally second result is GN π€£
Well, not at 8x, but with 16x at different speeds, and 8x is same as one step lower 16x.
And correction to this:
128 bit DDR5.
As that is actually quad-channel on DDR5, not dual.
As DDR5 uses 32bit channels, two per DIMM.
DDU did not help. I think monitor be dying. Why?
- issue seems to dissapear after messing around a little bit. (Take out cable, plug back in. Mess with refresh rate untill problem dissapears).
- issue is not visible when taking screenshots (need to take external photo).
- issue is also visible when using monitor menu itself.
And it only happens on cold boot of monitor.
Yeah, if it affects the monitor menu, it is monitor internal.
And when it doesn't show up in screenshot, it means the actual frame buffer on GPU is ok, and could only be from later fixed function HW on the card anyways.
Yeah. Was thinking same.
In my head this monitor isn't that old. Too sad. Maybe I used to too much. My other one is older than this one.
Iiyama is from 2017. No issues.
AOC from 2021. Four years old.
Models matter
but yeah, older tech tends to last longer
Then again: a good monitor can noticeably decrease in quality, a bad one would need to go to absolute shits
AOC 27G2U/BK which had good user reviews back at the time. But reviews don't include longevity of 4 years 
It does, but moreso for a few particular games and for high-framerate gaming. A lot of the PCI-E bandwidth cost is per-frame, and increased latency from queueing to access the bus affects high framerates more than low ones.
e.g.
As to why it affects performance even when the PCI-E bus is idle most of the time; that's not really a good reflection of how the bus is used.
For example if you suddenly need to move 1GB of data from A to B, faster bus completes that transfer in less time. It doesn't matter if it's sitting on its hands for 90% of the time after until the next transfers happen.
And somehow GN got completely different results for Black Myth for example:
different, less intensive scene
it's a common issue in game benchmarking that different scenes can have wildly different performance characteristics, even with the same game, hardware and settings. In WoW for example some scenes see vcache CCD being 5% faster, others 70% faster.
So basically you are cherry-picking for the few cases where there is difference, when most of the time there isn't?
no, there's a significant geomean delta even if you just include everything. The outliers of +10% perf from 3.0 to 4.0 to 5.0 are also important enough to have a mention.
Not in that GN testing for games?
Sponsor: Arctic Liquid Freezer III on Amazon - https://geni.us/NrMtDT
This benchmark compares PCIe generation differences on the NVIDIA RTX 5090 GPU. We're testing PCIe 5.0 x16 vs. PCIe 4.0 x16 vs. PCIe 3.0 x16. This also helps serve as an effective PCIe 5.0 x8 vs. PCIe 5.0 x16 benchmark, as PCIe 5.0 x8 is represented by PCIe 4.0 x16. The testin...
https://www.techpowerup.com/review/nvidia-geforce-rtx-5090-pci-express-scaling/ much more comprehensive
a geomean of 3 games doesn't say much about overall perf
For playing games that are often around 60-120fps and more gpu core bottlenecked, yes. For high framerate gaming it's not.
β
So basically for most part, minimal effect, except some outliers.
but most notably, it's a lot more between 3.0 and 5.0, which is what we were talking about.
yeah.. on a 5090
On full-HD, in that TechPowerup overall, 5.0x16 is 3.4% faster than 3.0x16
Yes, Baldur specifically was talking about the difference between 3.0x16 and 5.0x16 on a 5090.
we havent a clue what OP had #off-topic-tech message 
It is, but the games affected by pci-e bottlenecks are scaling much more. There are more than a few games at >10% and some in the 20-30% range. People buying 5090's do care about randomly not having 20-30% performance on the game that they're playing, even if 9/10 of games don't lose that much. Some do, therefore i take issue with just calling it a general null result that won't make significant impact.
I deal with this sort of thing all of the time, people trying to set some benchmark or world record and running into scenarios where it's like "..this game performs 20% worse than it should, why?". In several cases it's been PCI-E perf, despite being on 3.0x16 or better. More often i suspect people lose that perf and they don't notice that the game SHOULD run better because it's not under a microscope.
And I wasn't aware of the TechPowerup testing for that and the outliers.
But those seem to be pretty rare.
And no-one is limiting that PCIe speed just because anyways.
Sure, i have talked to a few peeps who limit because of e.g. thinking it wouldn't make a difference (e.g. to run an Arc encoder card in a secondary slot), only to find sometimes moderate performance loss that they didn't expect or relate to the PCI-E perf.
Weird, usually done only to use older riser cables.
But people can do weird stuff.
Standard consumer CPU's (and thus motherboards) don't have enough pci-e lanes to run x16 on the main pci-e slot and still run greater than x4 elsewhere. It's standard for boards to have a configuration that can either run x16 in the main slot, or redirect that to x8/x8 if two things are plugged in.
That started when x8/x8 or better was a requirement for Nvidia SLI, but it's stuck around longer than SLI did.
For a very long time there were 20 lanes total, so it would be 16 for PCI-E and 4 for m.2.
There's a few more lanes on ryzen 7000/9000, but i've only seen them used for a couple of CPU m.2 slots rather than beefing up the secondary pci-e slot to an x16+x8 config.
Everything that runs through the chipset gets 4x 4.0 lanes shared, so no connection on it can be greater than 4.0 x4 and they all compete for that bandwidth (even though that can be like 5 or 10 or 20 devices).
It's basically like a highway that takes in 20 roads but only has 1 road going onwards to the destination, so they can all get there technically, but there's severe performance degradation if multiple people want to use that one road at the same time.
i found 2 games above the avarage diff at 4K out of 25 games
ones 11% other is 8% (its more at 1440 and 1080p, but you have to be insane to play at those res's with a 5090)
i dont really think anyone is ever gonna notice this
Because of AMD mandate for the "chipset" naming.
I do
without an fps counter and A t B tests ?
no you wont.
And because second 8x slot isn't really useful for anything anymore, but it is marketing point.
So instead of multiple useful 1x or 4x slots at usable positions, MBs have that split 8x slot and then M.2 slots, and nothing usable for actual PCIe cards for most part.
Because marketing got into race of who has most M.2 slots.
looking somewhere else in a game will have more impact than the diference between gen 3 and 5
And then the design side for some reason positions those stupidly, so that the actual PCIe slots that MBs do have are too close to the 16x main slot...
Instead of PCIe 16x, M2, M2, PCIe, PCIe, they do PCIe, M2, PCIe, M2, PCIe.
Where then that middle PCIe is too close to that 16x to be used...
If they even have third PCIe slot at all..
Even at 1x.
They have limited options there.
Anything faster than 4.0 x4 must be connected directly to the CPU, and there are serious length and signal integrity limitations that prevent them from putting something on the other end of an ATX board. The CPU stuff (pci-e lanes, RAM) is as close to the CPU as possible without violating keep-out zones.
I think at best it would require additional and significantly expensive hardware to relay the signals, and have more latency
Yes you get +50% FPS in some scenes compared to others, doesn't mean that you don't care about +25% FPS due to a different factor.
Like if you can have 100 vs 150fps with slow pci-e or 125 vs 187 with fast, of course you notice it.
And even stupider, when 4x PCIe slot could be converted into M.2 slot with cheap adapter, but good luck converting M.2 slot into usable PCIe card slot...
its never +25% tho
max 11% in one game
you wont notice it
Point is that they do that even with chipset lane slots...
On lower end boards that don't have switching for 8x/8x.
The chipset lane slots are all <= 4.0 x4
But that POSITIONING! where the chipset slot is too close to the main slot.
Yeah, not great π
So any card used will block normal GPU air coolers by being too close.
hey its 4 with UP
Basically makes even that 8x/8x split bad, just to save on single redriver set for costs.
My current motherboard has sata connectors pointing upwards, but they're also too close to the pci-e slot so the graphics card overhangs them. Some of them can work, but i literally have to take the card out to plug/unplug the cable.
wow wtf
no wait
heres the best one yet
8fps with FG π
I'l stop using SATA when my SATA ssd's die.
i also think FSR FG is just dead here
that frame time graph is not healthy for FSR FG
hey
at least i cant tell apart fake and real frames here
hey its 12fps at 240p
this is great guys
Hmm... seems I have forgotten to take any pictures of the insides of my current computers...
aleast kryzzp was considerate enough to export this video in 4K so we can see this 240p in all its glory
Even worse:
All three black 16x slots are 1x from chipset...
And only lowest one is really usable.
You have an accessible m.2 above the graphics card slot at least
Yeah, but point is the relative positioning of the chipset M.2 slots and the chipset PCIe slots.
Another:
That top slot once gave me 0.2mm of clearance for installing a modified cpu cooler mount for offset mount
π€π»
Another older one:
π«
Basically that has been the layout for all four MB manufacturers for all AMD AM5 boards, where they put the center PCIe too high, and M.2 in between.
When with opposite order, the PCIe would be usable.
It's because of the signal integrity and latency
The pci-e x16 from the CPU
either it's right next to the CPU or you have to buy redrivers and add latency
Yeah, but that part with M.2 from CPU, and then 16x from CPU isn't the point, that is good.
nobody wants their board to be $20 more expensive than everyone else, or have worse benchmarks, but yeah it is kind of solveable from a technical POV
just cost and ease of design is a pain
Cannot put the main slot in position 1 anymore, as CPU coolers are too large.
But the lower down stuff should be M.2, M.2, then PCIe, PCIe.
Not M.2, PCIe, M.2, PCIe.
So that those PCIe slots are in positions 6 and 7, not 4 and 7.
there are also concerns with case size, especially with larger graphics cards, although it's not an issue for me
a fair few cases are fine with 3-4 slot graphics cards in the top slot, but not bottom
Example what I would like to be used more:
But that is the ONLY Asus AM5 board with anything like that slot layout, with two slots at the bottom.
Yeah, in that case though they're unable to do x8/x8 from the CPU. Although nowadays, i would also argue that it's not a priority.
Everything else has M.2 slot(s) at the positions 5-6.
Like I said, none of the boards I posted HAVE two CPU PCIe slots.
Yeah that's bad π
Only 16x from CPU, and rest are from chipset.
And still have idiotic layout for the slots.
And even with CPU 8x/8x, you would only need set of redrivers for moving the second slot down more.
Making that second slot actually useful, and not just marketing point.
By the time I buy RX 9070 XT + Ryzen 7 9700X in 2026, it will be outdated π
Probably like: buys Ryzen 7 9700X
Next day: AMD releases AM6
am5 zen6 is in like a year
has great potential (denser and higher clocking node, move to 12c CCD instead of 8c, expected +50% L3 cache with that, expected new interconnect, expected new IOD with a much faster DDR5 controller)
am6 is a while out yet, they're doing one platform per memory generation and DDR6 isn't ready. There's even some discussion of if zen7 on AM5 might happen.
Read AM6 is planned for 2028 but hope they will delay it 
at least 1 gen am5, 2028 could go either way ye
Meanwhile, Intel is so obsessed with releasing new sockets I don't even remember which last 5 sockets have been released lol
yeah lol
Last I remember is LGA 775 π€£
the 3 sockets of DDR4 skylake
Theory is they are doing that so people keep buying new cpu + motherboard and mbo manufacturers then have to buy chipsets from intel
I will be gaming on 1080p until my monitor dies (buying a new 1440p monitor then). RX 9070 XT + Ryzen 7 9700X combo should be more than enough for this
Will have to buy a new motherboard + 32 GB of RAM though
don't do it π
Still use Asus VG248QE monitor I bought in December 2015. Don't know how it's still working considering planned obsolence lol
Probably the electronics in general which lasted for the longest in my life since 2000 (and I'm always using it)
I have a theory what might have happened to my monitor. So might by my own fault. Although I doubt it since I read the monitors are able to handle it.
I will think a bit more about PSU/Motherboard. Once I order those, I probably order new monitor I guess.
Or keep this monitor and lock it to lower refreshrate.
Current title belongs to Radio Shack multimeter which is like 40 years old. Still working and the manufacturer went bankrupt a long time ago
The future of tech. Back to old hardware. 
Reminds me of Gateway monitor. First PC brand we owned
I have a monitor that's been in constant use since 2008 and still going! My first widescreen
you can get something nowadays like 5x better than a vg248qe for the same price tho. IPS, much better resolution, refresh rate and response time
I have my PC and monitor plugged into a surge protector so maybe that also contributed to my monitor still working
APC surge protector still has a green light which means it's working but that really isn't a reliable indicator. Need it test it somehow
Btw, looks like APC went out of surge protector business and they are focusing on UPS atm
That green light basically tells that the protective element hasn't broken in non-conductive way, and if it failed conductive, breaker would blow.
Just large MOV (Metal Oxide Varistor) as the protection element, resistor that has its resistance change depending on voltage with sharp cutoff in the protective use ones, above normal mains voltage.
π
Starts conducting and heats itself up to protect anything behind it.
Yeah, APC ones were good, but expensive.
So people didn't buy them but cheaper crappy ones.
People always recommend to buy a solid PSU in order to protect hardware in case PSU goes down but they never mention spending β¬15-20 on a surge protector is also important
Good PSUs also contain that MOV at the input, and couple of other protective elements.
But the bad side is that MOVs take damage each time they are used.
Also have a working Belkin surge protector protecting modem and Unify AP. Bought in 2018 and still working. Guess I'll buy Belkin next since APC surge protector no longer exists
What does the surge protector protect me from that my breaker box and PSU can't handle?
Genuine question
Overvoltage spikes.
Like the name says.
Can deal with larger spikes than the PSU internal protections, and also keeps that internal protection from failing shorted.
As in? USB device deciding to pull 6V instead of 5V?
Ah, so electricity provider giving me 260V rather than 230V?
And my breaker box doesn't catch that cuz it's too slow to react?
Breakers don't have that kind of protection, they are current based.
I once worked in an office building and they forgot to tell me to connect a computer PSU to an UPS since it was very old and very sensitive to voltage spikes. Guess what happened when I plugged the PSU directly into a wall outlet
Caused a blackout in the whole floor and fried PSU 
It was actually a very old server PSU and we were moving to a new office
Office had elevators, HVAC and other things which probably means the voltage spike killed the PSU
Yeah, starting and stopping AC motors also cause spikes.
Normally that is dealt with filters at the motors, but when cheaped on, they cause havoc with anything else.
I was so terrified when I tripped the breaker for the whole floor. Everyone was mad at me because they forgot to save their work and I was like: "How was I supposed to know you are not allowed to plug PSU directly into a wall outlet, everyone does that!"
Turns out it was simply a very old server PSU and that's why I killled it. Modern ones wouldn't have that problem
More likely had that MOV so damaged that it was on hair trigger, basically the trip voltage to conduction lowers with each dissipation pulse, and finally shorted that MOV.
And the PSU would have been fine after cutting that MOV out.
Ah.
I'm surprised I haven't had any issues with surges then
Plot twist: the whole server was also damaged since PSU didn't protect it. Company had to pay for data recovery π
Because those don't really happen that often with good western power grid.
And most electronics contain basic protections themselves.
Actually, I did have some related electricity issue at one point:
When I had my PC on the same power strip as the monitors and everything else of my setup, turning on the strip tripped the breaker.
Moving the PC to a separate outlet fixed it
So it's a mix of "better to have it and not need it than need it and not have it" and "you most likely won't ever need it"
Yup.
But large part is protecting devices during thunderstorms etc.
Or when grid comes back up after blackout if device is still connected and so on.
I wanna build my own nas with 4 HDDs. Whats a good sfx for that?
π€
what thaaa
I don't understand
no one does except whoever created this
and maybe letterdaif but he's got me blocked for some reason so he wont respond
sometimes I think this channel would have a more fitting name of serpentarium
Exactly.
I think the developers want to say that UE4 medium settings are equal to UE5 low settings/medium settings.
Haha π€£
$1,800+ for a fridge.
And you get ads. I love it. Haha, the greed of tech companies.
Now we need adblockers for fridges too. At that point, run pihole on network I guess.
The company describes the effort as a test of "promotions and curated advertisements" that it says are designed to enhance value for owners. "As part of our ongoing efforts to strengthen that value, we are conducting a pilot program to offer promotions and curated advertisements on certain Samsung Family Hub refrigerator models in the US market," a Samsung spokesperson said in a statement to Android Authority.
And ofcourse, the PR spins it so that it's an added value for the consumer π
How about I don't need a smart fridge.
The only ad I want my fridge to give me is "you have X, Y and Z leftovers. here's a recipe for using them before they go bad"
And it's not an AI recipe but a chef curated recipe selection
Yeah, I took out my shovel and went outside to put all my food in the ground for now.
If its cooking recipes with a toggle whatever i think thats cool
If its just google ads then it can go fuck itself
It would make sense if the ads are atleast food related.
Didn't know a rocket parachute flare can be so deadly. A women recently died in a wedding because she got hit by a falling flare (or some idiot launched it and pointed it at the crowd)
Talking about this btw
5070 is best buy for a GPU under 600usd stm, right?
Any flare is very hot while still burning.
And if launched at someone, fast.
So most likely direct shot to the head.
It has to reach height quite fast while dealing with air drag
Unless you catch a 9070xt on discount
looks like gpu stopped resetting to default 'tuning'
and my profile for undecrlocking and fan stay applied for a looong time
My ssd bricked itself from the windows update
Remove power from PSU for 30 seconds, start again?
Until you trigger the bug in the SSD again with suitable transfer.
And the causes for some of the cases have been insufficient cooling for the SSD with broken throttle behavior in the firmware.
Or just old firmware.
So install the SSD makers support software and check for SSD firmware updates.
Is the skhynix pc711 hfs512gde9x73n affected and is there a fw update for it
At least one SKHynix SSD controller is affected.
Which ones
Both Dell and HP have firmware update utilities available for that SSD family.
So what machine do you have?
Victus 15
So I can link you the right support page with the updater.
That is branding, not model.
0121nr
Pretty sure all of the victus 15 with the fans on the left and cpu and gpu on the right come with a overheating problem
(With the battery disconnected it gets even worse and leads to thermal shutdown)
That isn't right code?
Or at least not the complete code.
HP doesn't offer any updates for SKHynix SSDs for that model.
Only for WD or Samsung OEM models.
And as that is OEM model, SKHynix doesn't give direct updates either.
then im fucked?
Contact HP support as first step.
pretty sure they wont help
And tell you are hitting the current SSD bug where some models after suitable load soft-brick until full power cycle.
and what will they do with that
How would I know?
See what happens and ask for firmware update for the SSD if possible?
But that is the only possible source for certainly valid SSD firmware for the drive.
There is Dell released firmware from 2021 for that drive family, but might not be right for HP version.
if thats the only source then im swap the ssd entirely
SK Hynix PC711 SSD Firmware Version 4100.1131 from early january 2021, via Dell.
and when did the broken update issues start
And you can check, if the "Drive Manager" on this page supports that SSD or not:
https://ssd.skhynix.com/download/
weapons specialists call that stuff "less-lethal" for a reason. can still definitely kill
even with how loose our weapons laws are here, in my state it's massively illegal to fire a weapon into the air
so much so that we helped develop the hardware & software that law enforcement uses to triangulate weapons fire
Anyone could buy a flare in the past here but they recently made laws strict and you can't buy more dangerous firecrackers/fireworks if you are younger than 18 but then parents buy them for their kids in order to bypass restrictions...
With the exception regarding a signal flare. You can only buy it if you own a company doing fireworks or if you have a proof of owning a boat
Also, 40% of cities here gave up doing public fireworks display on new year in order to protect animals
We might even completely ban firecrackers/fireworks for citizens
I'd rather have a ban on citizen fireworks with one big organized firework display by the city (which can probably be substituted for a light/drone show)
The thing with firecrackers and fireworks has gone too much imo and only companies doing public fireworks should be allowed to buy them
Not to mention lots of kids lost their hand or even a limb because firecrackers exploded in their hand
German regulations:
Category 1 - small stuff like little fountains or spinners. Legal all year round to people aged 12+
Category 2 - most New Year's fireworks, stuff like firecrackers, rockets, batteries, big fountains. Only legal to buy/sell three business days before New Year and only for age 18+
There are separate categories for stage pyrotechnics.
Professional pyrotechnicians can apply for a special license which allows them to buy Cat 2 fireworks all yea round as well as Cat 3 or higher.
Every piece of pyrotechnics sold in Germany has to have a test number by the federal agency for material safety (BAM-Nummer)
It's actually surprising how anti-fireworks I am given my username.
I guess it's another one of those examples for "once you learn more about it, you see how bad it actually is"
I think there is a EU-wide category: F1, F2, F3 and maybe F4 (not sure)
Probably, I definitely saw those categorization in the past
But aside from sparklers, I haven't lit any fireworks in the last years
"Category F1 refers to fireworks with a very low potential for harm (such as sparklers) and those used in enclosed areas, including outside of residential buildings;
Category F2 fireworks are low-risk devices meant for usage in a contained area outside of residential buildings;
Category F3 fireworks are meant to be used outside of residential buildings in a roomy open area and represent an average level of risk;
Category F4 fireworks are extremely dangerous and are only meant to be used by those with specialised understanding; they are restricted to professionals throughout the EU. "
I think we banned people under 18 from buying F2 and F3 (F4 is for professionals only)
idk what falls under F2 and F3
Does a simple rocket or a battery need a roomy open area?
Or are they low risk devices?
Although I hate how both firecrackers and fireworks and sometimes stuffed under the same category. I've seen a lot of dangerous F2 and F3 firecrackers and trust me: they definitely aren't "low-risk" or "average-risk"
Firecrackers are even more dangerous than fireworks imo (stronger ones)
Cuz I can tell you: if that battery is light and not securely on the ground, it can tip over and fire sideway
Rockets or batteries are supposed to be launched in an open area with some distance from the nearby buildings
Firecrackers that are legally sold in Germany range from "severe burns with minor tissue damage" to "will remove fingers"
So German cat 2 probably includes F2 and F3
Same here, some of them literally sound like a military grenade
I miss my cat 1 bombette tubes.
Shoots a white glowing orb into the sky that makes a bang after a few seconds
Also, when I used a fireworks battery once, I sprayed it with water after being done in order to prevent accidental fire but people rarely do that
Plus these we call them Knallerbsen (bang peas) which are just a few small pebbles and a bit of silver fulminate wrapped in a paper bag
Which are cat 1
I know what you mean, we call it "Devil's bang" π
I count the shots and if they all fired, I leave it be for a few minutes to cool off
I want coppernitrotetrazole to follow the naming system of other Nitrotetrazoles
Silver Nitrotetrazole: AgNTZ
Gold Nitrotetrazole: AuNTZ
Copper Nitrotetrazole: DBX-1
COMMIT TO THE FUCKING NAME!!!
Last is commercial branding for product using that compound.
Here is a video of a fireworks battery I launched a long time ago:
I know, but it became the common shorthand for that compound - unlike AuNTZ
This is probably the largest battery a civlian can buy: 484 shots in 2:40 minutes and costs β¬475
And the battery is called "Sodomah and Gomorrah" π
What shots tho?
Cuz many "hurr durr bigger number better" high shit count batteries just shoot a bunch of white lighted screechers with no effect other than annoyance
Okay, damn, there's more than just screeching
I've seen some that continued what this one did at the beginning
But I guess for 400+β¬ you'd wanna see some bang for your buck, literally
419.95β¬ in Germany btw, sold under a different brand but seems to be identical
shop.roeder-feuerwerk.de/rubro-event-sodom-and-gomorrah.html
This one has the same number of shots but I think the long box I posted is probably modified
It has the screeching rockets but there are different rockets later as far as I can tell
at β¬2400 you get a lot of bang for your buck
Wha
as an american.. I don't personally care for them either.. the kids love them... whatever.. the only ones I actually enjoyed were sparklers... for reasons
You stop loving fireworks once explosions are associated with danger
Its a nice connector, just fully plug it in and enjoy high fps?
You also stop loving fireworks once you see your beloved pet shiver in fear during New Year's fireworks
maybe there's demon inside of it?
Then the fireworks aren't working in scaring off the demons
no, they do?.. causing shiver
https://www.youtube.com/watch?v=xQTGp7rk33M
Scientists Explain the Origin of Mysterious Siberian Holes
based on paper:
sciencedirect.com/science/article/abs/pii/S0048969725016821?via%3Dihub
This study critically reviews the existing models and concludes that focused deep heat and gas from below the permafrost may be the key factor allowing the formation of GECs, while atmospheric heating indirectly triggers their formation by accelerating cryogenic process rates and the formation of new lakes and rivers.
focused deep heat
Support this channel on Patreon to help me make this a full time job: https://www.patreon.com/whatdamath (Unreleased videos, extra footage, DMs, no ads)
Alternatively, PayPal donations can be sent here: http://paypal.me/whatdamath
Get a Wonderful Person Tee: https://teespring.com/stores/whatdamath
More cool designs are on Amazon: https://amzn.to...
while it doesn't explain why there's focused deep heat, there are interesting references
They are scaring the demons, but they're not scaring them off.
They're obviously still there, otherwise the pets wouldn't be shivering constantly
Bombs away
sigh ...
Python round 4.5 down. Typescript rounds it up.
And then we sprinkle some magic over it and voila.
For some reason Python 3 changed rounding from normal rounding to bankers rounding for that round() function.
The round() function rounding strategy and return type have changed. Exact halfway cases are now rounded to the nearest even result instead of away from zero. (For example, round(2.5) now returns 2 rather than 3.) round(x[, n]) now delegates to x.round([n]) instead of always returning a float. It generally returns an integer when called with a single argument and a value of the same type as x when called with two arguments.
And for some reason they don't provide any built-in way to do it the classic way.
But seems to be the spec for IEEE 754
Basically attempt to remove the bias from rounding.
And in turn Javascript doesn't seem to have that gaussian/bankers rounding as easy option either.
Ah, no, IEEE 754 had both that and the classic away from zero rounding, and then three direct rounding modes, (truncate, floor, ceiling).
So Python using that gaussian by default without any option to change is just weird.
Saw a little comment of floats in the round() documentation.
number 1: down.
number 2: down.
number 3: up.
number 1: down.
number 2: up. (would expect to go down here)
number 3: up.
Makes sense.
Note
The behavior of round() for floats can be surprising: for example, round(2.675, 2) gives 2.67 instead of the expected 2.68. This is not a bug: itβs a result of the fact that most decimal fractions canβt be represented exactly as a float. See Floating Point Arithmetic: Issues and Limitations for more information.
print(round(2.4)) # DOWN
print(round(2.5)) # DOWN
print(round(2.6)) # UPprint(round(2.474, 2)) # DOWN
print(round(2.475, 2)) # UP
print(round(2.476, 2)) # UPprint(round(2.674, 2)) # DOWN
print(round(2.675, 2)) # DOWN
print(round(2.676, 2)) # UP

I did say the exact rule:
Towards even.
So 2.5 is towards 2, 3.5 is towards 4.
etc.
So not precision issue, but using gaussian rounding with "midpoints towards even" rule.
.
That was from Python changelog.
My coworker just witnessed the horror of AI assistant first-hand
He is not receiving paypal 2FA sms code and can't login. Asked AI about it and the AI said he needs to disable 2FA
Then he got angry and asked how is he supoosed to disable it if he can't login. AI response: You need to disable 2FA π€£
Told him he should use authenticator app next time
microsoft just outright stopped sending me confirmation emails
idk if its ai related or not but its still ironic for such a big company
Paypal AI assistant is one of the dumbest AI assistants out there lol
Also told my coworker he should have backup codes but he lost them π
print(round(2.675, 2)) # DOWN shouldn't this be up then?
2.68 ... I think python documentation is right and it has to do with Floating Point Arithmetic
No, as it is between 75 to 7 or 8.
So 8 is even, so towards that.
Ah, sorry...
So it was 2.68 after the rounding? So what I said...
wait.
Python is now using "Rounding half to even", instead of most commonly used "Rounding half away from zero":
https://en.wikipedia.org/wiki/Rounding#Rounding_half_to_even
For IEEE 754: That first listed, instead of the more common second listed:
https://en.wikipedia.org/wiki/IEEE_754#Rounding_rules
Ok, then there are in addition floating point errors in play.
But the base is that rounding half to even, as the changelog says.
But then the implementation has errors because of floating point.
So this specific spot seems to be floating point error sourced difference from the rule.
As it isn't actually stored as 2.675, but tiny bit below.
Yeah, 2.675 is stored as 2.6749999523162841796875 (0b01000000001010110011001100110011)
But 2.475 is stored as 2.474999904632568359375 ... so doesn't explain that difference.
For IEEE 754 FP32.
This isn't funny anymore π
im just waiting to see if theres a spherically challenged american to say "this is why we use cars!1!"
They are doing train track reconstruction so there are delays
The problem is that the reconstruction has been in progress for 10 years and it's not finished....
Anyone know a good CPU pair for 4070 once its time to upgrade? Lower budget range.
Alright listen Im stupid okay
at least hes smart enough to understand that 
π
either way if i recall you were asked esentially the usual
stuff like what cpu do you have already
And why you wanna upgrade
Which would be partly answered by the question of what CPU they already own
Can't be using a 5600 without being on the am4 platform
because its only a 9800x3d and not a 9950x3d 
I will probably get RX 9070 XT, Ryzen 9700X, Asus TUF B650 Plus mbo and corsair 2x16 GB DDR5 (6000 MHz) next year or maybe something else if new gpu or cpu is released by then
Had no negative experience with Asus motherboards so far
Which wouldn't be enough of a reason. It's missing the "bigger number better" logic
Why TUF board, Corsair ram and 9700x?
And why plan this far ahead anyway?
the 19950x4d is better then 
I've had no positive experience with Asus motherboards so far
Except the free ones
ryzen 4070 is so last gen
were all on the ryzen 5070
model, don't brandshame.
Maybe I'll consider 7800x3D
ASUS TUF Gaming B650-PLUS has good reviews.
Went with TUF board because I own one from a previous series and don't have issues
Might have messed up with Corsair because I don't have much experience with RAM in general
For my setup, more than 6000 MHz would be overkill imo
Goldensoldier recommended my some ram sticks.
Pretty happy with them. Never heard of the brand.
Patriot Memory Viper Venom Cl36 6000 2X16GB
would have never bought them on my own without someone telling it's leggit.
Patriot has been on a RAM market for some time I think (heard of them before)
before that, I also had corsairs.
Well, you didn't buy em either
(Since you said you went for Corsair)
Won't be buying RGB RAM next time though. Don't want to pick the LTT way 
RGB makes it go faster though
That's why I want the Asus TUF motherboard, it has no RGB.
And is more expensive than current motherboard, with RGB 
At least with Asus, you can disable that all in BIOS with single toggle.
I can too with Assrock. It's (BIOS, no software) in control of my RGB.
I had to set RGB to automatically turn off when the PC is not on since RGB messes up with my sleep
Looks pretty when you are not using a PC but terrible at night
That's a non-tech option 
we should remove all evs and just put the exhaust pipe back into the engine intake 
Known as EGR (Exhaust Gas Recirculation), that current cars etc. do as part of emission control.
By who?
Professional third party reviewers?
Or customers who could be utterly clueless?
We have a popular racer here and the first thing he checks when reviewing a car if it has a fake exhaust since he hates fake exhaust
TT Smart PSUs have 4+ star Amazon reviews too
just remove the exhaust port if thats the case 
I heard some rumors, from car mechanics, that EV is not going to last.
Either whishfull thinking. Or they know more than we do. It had to do with producing the batteries for the EV's and the mines for minerals for the batteries are running dry.
So they think, EV demand will be higher than mines can produces minerals for the batteries, for the EV.
both.
What's the LTT way?
Linus tech tips
if it works it works
xD
Linus Teth Tipth*
Probably selecting by looks?
ASRock B650M-HDV/m.2
Not just no RGB, but also no RGB headers.
So even if there was an RGB part elsewhere in your build, you couldn't use it
Yes.
Linus is the reason we have RGB chairs now
probably a way to bypass that if needed.
them existing is a tragic event anyways
I know who LTT is, but how do they pick their parts?
Which sponsor pays them the most?
I think the room RGB was a step too far for me. The lamps you could buy for RGB. I think it was triangles etc.
probably dbrand with what they do
they have to be getting top dollar for that
also known as "mybossisveryshort.com" "shortlinus.com" "betterlttstore.com"
He advertises Nord VPN like crazy
For anything am5, going above 6000 would cause your system to be unstable or perform worse than 6000 - until you get to like the 7200/7600 mark
Do note however, that Patriot include a 6000 MT/s fallback xmp profile on their 6400 MT/s RAM
i have not seen a single nord vpn ad on wanshow
Nanoleaf
btw, if you buy the ASUS TUF Gaming B650-PLUS WIFI you get RGB headers.
Nah, above 6000 can work, just has lower probability because of silicon lottery, but need to set UCLK:MCLK to 1:1, because it defaults to 1:2 for above 6000.
Maybe I have mistaken it for some other sponsor
Booooring
non wifi is β¬ 148,99 on Amazon.com.be and ASUS TUF Gaming B650-PLUS WIFI is β¬ 162,90 onAmazon.com.be
proton by chance?
Okay, not guaranteed unstable, but probable unstable
So for 14 euro's you get wifi and RGB headers on same board π€·ββοΈ
No, can't remember right now but it wasn't Proton
What about AssCock B850 Pro-A?
6200 used to be about 90% of CPUs, 6400 about 50% of CPUs, 6600MT/s used to be very low percentage.
I don't want asscock anymore. Don't want to replace one ass with another ass.
private internet access?
Before the various improvements to memory training in newer AGESA versions.
i seen them on wanshow fairly recently
and like you said, you __never __had good experiences with Asus.
I __only __had good experiences with Asus. 
You get a shitty wifi module tho
ASRock just has an m.2 E-key on their boards so you can either replace the included wifi module or add your own one to a non-wifi board
Yeah, my main issue with both models. I don't want WIFI. I could use RGB headers
Why you considering Ass-sus then?
They still a thing?
Asus has standard M.2 E-keys too, just inside extra shell in some of the boards.
Thought they went bust or smth
It's on the top of my tongue but I can't remember π
yes? they have been doing a bunch of sponsors lately
Because like I said, reviews are π
HDV is the exception in that it doesn't have RGB headeds
i think if you start listening to 67 mango phonk it will come off because it is so fucking ass 
On some boards they are bare, on some they are those "WiFi Go" modules, but those are actually just shells for standard M.2 E-key.
Makes for better WiFi reception, from the extra shielding of that shell, from the computer case internal EMI, and much shorter antenna cables.
It's fine, just rarely worth its moeny
Somehow I know ...
The other ones recommended are:
- ASRock A620M-HDV/M.2+
- ASRock B650M-HDV/M.2
... the rest are all above 200-400 euro.
So relatively, it's not a expensive board π€£
the bios's which gets banged together with no testing or fucks given
is enough not to get them alone
same for asrock
Never had that experience with lot of different Asus MBs...
dosent asus have an ongoing priority issue with every single laptop bios they have released since 2021
Same here, no issue with BIOS so far
and the oc presets are pushed too far aswell not that you should use them
Which is the same thing with any auto-OC?
Because they all do insufficient testing?
Who's making those recommendations?
A620 is quite a barebones chipset
And why is it 600 series only?
Compared to asus PC hardware I had no issues, I generally don't like their laptops
Had issues with their hinge design which disables webcam on 2 last laptops
Not a single Asus component has ever worked inside my PC.
And the only peripheral I ever had from them (ROG Chakram) is a piece of shit
They can seriously go fuck themselves for tricking me into installing their Armoury Crate malware
Oh, I removed that crap a long time ago
My experience with asus so far: GPU: good, MBO: good, Monitor: Good, Laptop: bad
Their hinge design on laptops really suck
Also have some cheap asus keyboard but I got it for free as a replacement for another broken electronics so I can't complain.
better than Hinge Problemβ’ probably
i can confirm some wooden sticks with hotglue probably make better hinges than my hp hinges
Hinge problem is basically the result of their poor design. I own 2 asus laptops and can't use webcam because the hinge is probably pressing on the cable somewhere
laptop webcams suck mostly anyways
its also probably better for privacy for it not to work
Have a webcam cover on my work laptop and almost never open it 
then why even care
just feel proud about it more than anything then
you have a blessing
Home laptop is another thing though
the hinge problem is so prevailent in laptops. even good brands have it. and none of them have support for them. warranty be like "not our problem" in every case i've seen
Don't want my employer spying on me through webcam
i only buy cheap laptops and do everything on desktops as a result. steam deck is the first portable gaming pc i've bought in a lnog while
oh you have far more problems than that
Like the Xiaomi PTZ camera so I can monitor my nephew's crib. Hope CCP isn't interested into spying on my nephew 
just the roll of the dice I guess, there are others.
Although Google and other companies aren't any different when it comes to spying
its just the same pile of spyware
Now that you mentioned it, might actually be a Chad Jepity list
No, it isn't afaik.
It's Tomas Hochstenbach.
but maybe if people use same test methodology, they have same conclusions.
Seems like Tomas did his best.
Seems like at least some of that is just reflections on the MB, and he didn't lock the heat to color scale, and none of those temperatures are dangerous yet.
By just looking at the highest temp in each subpics scale.
Who said anything about dangerous temperatures.
Just showing a part of the effort they go through for their reviews.
But it's never good enough for people in here unless it's a review they like and post themselves in here 
Main points were possible reflections in the image, causing temperature to be measured as higher than it really was.
And second main point was not locking the scale, so comparing pictures of different MBs is very hard.
Third point was just general note, as I didn't know about the point of that image.
And that in some pics the backplate is cold and in some very hot is just weird.
Ah, different sockets.
btw y'all, just tried a new cooler for the first time & I was very impressed
Scythe mugen 6, was very happy w/ it
VRM temperatures.
As long as you get it on sale for <40usd I'd pick that over competing options
I know it's been around for a while, but this was my first hands-on experience with it
Very happy
And in most of those pics, the max temp is near edge of the CPU backplate., or even in the backplate center hole.
And nowhere near the VRM, which is why I expect there are reflections from some external light or like, messing the results completely up.
Because in many of the pics with specific socket, random capacitors are the hottest objects.
So it seems they might be aiming a studio light at the MB, and using the same setup for both visual range images, and the thermal images, and that studio light is messing the thermal images up with reflections.
Just as theory.
What makes you happy about it? Quiet? Good cooling results? Quality (which is too soon) and price? Or cool looking π
Incredibly quiet for the cooling performance on a single-tower cooler at that price
Seemed well made
Uses a less dense fin stack with much more depth than a lot of other coolers
So you get more airflow at the same fan rpm, but slightly less cooling per volume of air moved
Mainly the point is that using thermal cameras correctly is suprisingly hard, and most people don't even know there is anything to take into account, and not just point and click.
Yeah, but these guys should know what they are doing.
They still humans, so they can make mistakes. But there is peer reviews in place for that too.
People can make topics about reviews and start a discussion with the journalist/reviewer themselves.
Pretty open source way of writing reviews.
Entire build is almost inaudibly quiet under load and entirely air cooled
Whereas my current pc is actually loud enough to be bothersome even at idle
Mine isn't quiet under load. But I have headset on, so personally it doesn't bother me.
It keeps my CPU cool so π
I have headset one that has better passive noice cancelling than any other pair I've owned and the fans are still audible while listening to music
Damn.
Well, I got 3 massive fans on my 6950XT making noise also lol.
Lol
I mean, not massive, but still, if you look at how big the card is you know how much the fans need to cover.
My 9007xt has three fans but I do know it's dead silent at everything <300w of cooling load
I could disable all fans besides GPU to hear, but I rather not 
What do y'all reccomend for thermal paste currently?
I've always gone with hydronaut, but I ran out of my most recent tube
You inspired me to check, yeah it's my case fans
I think the bios control on this mobo is wonky...
Seems like it was really not running hte case fans at the percentage that I'd want
How did you come to that conclussion?
Set up fan control with the same curve set and behavior was very different
I think it's just because the bios fan control curves are bad
But either way, I'll just roll with fan control
brand motherboard?
gigabyte
Ah, no experience with gigabyte. But I also had bad experiences (asus) with settings curves in BIOS. A bit of a wonky way to do it in my opinion.
Yeah
I just prefer to not need software running in the background.
Btw, other news to report. Installed bazzite and got it set up, very pleased so far.
I also prefer to handle stuff as close to hardware as possible. BIOS being in theory, closer than software going through API's etc.
Yep
Give us an update in a year or two π
Lol will do
Why do you say that though?
Had some minor system stability issues, but I think it was actually memory issues @ expo 6000MT/s
Because, the longer you use, the better of an opinion you will have on user experience, different games, different updates, performance etc.
Gotcha π
yeah, from testing there was only one game in the entire library that didn't work as expected immediately, and I just had to change some steam controller settings for that
So from that perspective, it's good at least
Do you play many non steam games?
And any games without a launcher, mmorpg's?
Games without a launcher, yes
For those, the easiest solution was just to add them to steam and then tell steam to launch them with the linux runtime and proton
It's the one and only thing holding me back. Fear of having issues with a game I like and thus cannot play.
From what I saw, protondb seems to be overly conservative with their ratings
So as long as they don't say it specifically doesn't work, you're likely good
Yeah, I know of it. Still, the fact it's there means where we come from π
For sure
What I mean, afaik, there isn't a website that keeps track for that for windows because ... monopoly more or less π To be precise to everyone.
This pc isn't going to be my daily, after I got it all set up I gave it away
But it went to a close family member, so I'll be able to keep touch on the user experience
For myself I probably wouldn't use bazzite if I were to switch my daily to linux since I use other distros a lot already
But that doesn't really affect much other than making it a little less dummy-proof
Ah, I understand
For all my headless machines I run debian
And until now I've run ubuntu on my laptop
The only software I use that isn't compatible with linux is adobe lightroom
I just did a quick little test, and grab the first 10 game shortcuts from my desktop, they all hit gold/platinum on protondb.
Nice
This was my first real time using steam big picture though, I"m very impressed with that
Actually, can't find any game I have installed that is not that.
Maybe I only play mainstream games π€£
But good for me! In theory, they should work out of the box?
That's my experience
I give it some though. Last thing I need to do right now, is changing anything OS related untill I got system stable. If ever.
The only thing that really causes issues is anticheat that specifically blocks proton/linux like valorant
indeed
Not that I think it will add instability. But it will not help my cause either π
But I am pretty much done with Windows/Microsoft. As many in here I think.
Oh good point.
Absolutely
Opinion; if AC is that intrusive, I am good without it and the game.
I generally agree
My favorite fps(the finals) just switched to denuvo anticheat(but not drm)
And thankfully it's still alright, but I really don't like that company
the company behind denuvo or the finals?
Plenty of games have it.
I'm excited for their next game as well
Yep
and I appreciate that they're not running the drm package, as that's quite invasive compared to even the anticheat
Ok, less than I though: https://store.steampowered.com/curator/26095454-Denuvo-Watch/ but I see some big names in there. Like Borderlands 4.
native linux ports are also just becoming more popular
(BG3 just got one)
I wonder what triggered this change.
Which?
Is it game companies wanting their games to be compatible with Linux even if the OS is a small % of the market.
Or is it because more people have Linux?
Honestly, I think it's steam deck
Oh right!
The bg3 port was not even compatible with normal linux desktops at first, it was meant to be entirely specific to the deck
Indeed
Without the deck, we wouldn't have proton
We wouldn't have dxvk
And the whole environment would be way worse
would be so cool if steam deck was able to make linux mainstream for gaming π€£
It's getting a lot closer
The steam deck definitely broke out of "niche product" catergory
I'm not sure if it was quite mainstream yet though
still impressed that steam even pulled it off. Getting a handheld with already a bit of popular competition with more experience
Yep
And that handheld also being the first pc handheld to run linux
And also the first pc handheld w/ an oled iirc?
GPD might have beat them to that
the default are ehhh but if you change it to manual curves it works fine mostly unless your fans use some wako power or smt
Well I did change to manual
Also are you just assuming you know what mobo & bios version I have?
no but fan control is fan control and its almost always the same or has the same core
eh
Asus was like this.
so is gigabytes for the most part
Maybe they updated it. But since then. I never looked at fan curves and just went YOLO and trust defaults.
although that curve is a instant hardware kill on my aio
and I monitor my temperatures and saw it was good.
huge no no, if my fans go over 90% they destabilize and shit hits the fan quick so having such a aggressive curve would be instant failure prettymuch
I grabbed picture of the internet π
So take it with a grain of salt?
still shows asus dosent think before doing
not every fan can handle 100% speed
let me find my bios level curve before i keep trying to remember what it was
I think 70< fans 100% trying to cool below 70 and then back to less aggressive curve.
The ramp up does look a bit scary. going from 50% at what seems 50-60 degree to 100% at 70 degrees.
My chip is running hotter than I'd like, for now I'm blaming the crappy thermal paste I had to use
But I'm seeing ~70 degrees at 70w on the cpu
Which is warmer than I'd like on an x3d chip
under load?
Yep
Thats kinda normal for intel to be fair
x3d tjmax is 85c though, so it's already downclocking by then
which sku
7800
Heres what i run for an aio curve
It's not that I'm worried about hurting the chip, hotspot doesn't break 80c
But I am leaving a bit of performance on the table
they're a lot hotter at the same wattage than zen5 x3d, and also have lower tjmax and much lower voltage limit
Yep
70c@70w, it depends how the CPU is loaded. If that's an nT workload, it's not good. If it's 1 core, you can see above 70c with like 30 watts PPT.
7800x3d was really an incremental increase over the 58x3d
Even running p95 small ffts the numbers are around there
I don't remember exactly what i had on my zen4 x3d, but better than that for sure
using arctic liquid freezer II / III
i gained 100mhz on precision boost going from nh-d15 to offset liquid freezer II
Which could be making up to a 5c difference
Compared to something like mx6 or kryonaut
yeah, i am hearing rumors of ptm7950 being significantly better than those too, but i am not really sure. Need more in depth testing
i guess your air cooler isn't one with offset mount? That's a free 5c, since the overwhelming majority of heat generation and the hot spots are in the bottom third of the cpu
meanwhile me gaming on 25W π€£
Wanted to see what my temperature is compared to you, but it's irrelevant if not same load on it.
Yeah, my gaming heat output is usually more like 50-60w
70w is the highest I've ever seen it sustained
Damn, and I am playing a MMORPG.
mine is irrelevant since its probably tripple or more lol
the hell?
i idle at 40-70W
I am not sad that my power draw is low π
zen5 x3d (2ccd enabled, only vcache ccd loaded) on liquid freezer III for reference
thats good but still a hearing a desktop chip only draw 25W is disapointing lol
Not dissapointing if it's effecient
yea but if it cant heat itself to its stable point then being unstable and underperforming in the 40c's all the time isent very usefull
it sits saround 57C
"heat itself to its stable point" ?
i know both my intel cpus loose performance at sub 50C
I've had RAM that performed better at 25c than at 15c, but never reached a point where anything else was too cold
and i remember my 1660 super would only performe at its full between 75-82C
all consumer cpu's and gpu's since i got into tech have performed best subzero, usually at like -150c, but there are quirks where they sometimes e.g. won't boot while that cold.
The other thing i've heard it on is SSD controllers or flash, i forgot which, that don't like to be down at 20c
Flash performs best when warm
yea cause its making enough heat to stay running
well thats sure how everything iv owned has behaved
i remember i literally had to let my 1660 super heat up before i could use my or else it would be unbearably sluggish at like 25C/roomtemps until the heatsink and die warmed up
then explain the science behind why temperature would have no affect on electronics mr science man
I didn't say temperature didn't have any effect
you said inderectly that its bs that temperature has an affect at those levels so explain...
However, the effect of temperature on a semiconductor is
- Reduce lifespan and possibly cause damage when high enough
- Decrease stability as resistance goes up, voltage drop increases, and the voltage at gates is also lowered.
#2 means that a higher voltage is required for operation at a higher temperature and also that gate switching times can increase, limiting clock speeds. Higher voltage leads to more heat generation.
that doesn't explain cold, only effects of overheating
How so?
The inverse of #2 is true as temperature decreases
Btw @night girder in my satis save with cpu at 58w average draw, hotspot is at 69c
maybe give CPU bit time to cooldown before checking.
Hm?
If it's comming from 70W to 60W might need some time to cooldown
not sure how quick these things actually can cool down π€
Oh no, it's been sitting in the same spot in my save without being touched for 10+ minutes
im checking mine because now you have me curios
im not starting up the train tho so your getting a biased test
but yea i do have to agree 79C even for a hotpsot isent great at that power draw
After heatsoaking, here are the average numbers
Columns are current, minimum, maximum, average for anyone that doesn't use hwinfo regularly
now I am curious to see how much a repaste does.
Me too
I'd hope quite a lot
I knew going in that this paste was bad, it was just all I had on hand
My other tube was empty and I didn't realize
I don't think paste can make a major difference. But eh what do I know π€£
if it drops more than 10C I will be a bit surprised
for different paste?
Yep
or bad paste vs good paste job? π€£
Different paste
I do think I did a good paste job, but like I said earlier this is known bad thermal paste
If it's not up to your satisfaction, after repasting it, you could look into the fans to set it up more aggressive.
Already done lol
The temps above were with the fans at ~1900rpm
Which is about as loud as I want them to be
(These ones top out at around 2250rpm)
They're really quiet until around 1800 and then they start to scream by the time you hit 2k
2250 is a lot yeah π
Indeed
The fans on my gpu report over 4k at 100% pwm
That sounds about how you'd expect
I've 6x140 running push/pull on cpu rad now, at 1450 rpm. Quite happy with that noise level (they can do sub 1000 but i'd rather drop a few C than be silent)
Oh nice
My builds are usually way too budget constrained for any sort of water cooling
The intro deals for the liquid freezer III's for 6 months were actually insane
I always think; do I need it?
Rarely
full watercooling is practically nerver really
There are some things you really can't cool quietly without it
and aio is like way over 100W kinda thing
Unless you have some insane need like 4x 4090 in a desktop form factor
yea or you put them on risers so they can breath lol
the only advantage i could see to watercooling is if you go the way of having a big enough water volume you nerver need active cooling
for total silence outside of maybe a pump
Lower temps give better boosts, on my 5900x i got 100mhz more from lfII AIO than from nh-d15, and that was there even in single core with <30w.
and then they released the LF III for under the price of the nh-d15 with even more performance
they kinda have a price floor of like Β£50 though. Peerless assassin 120 is Β£29.
Y'all see that EA was just bought out? Saudi public investment fund was involved.
yeah and someone else. Political related.
Reading MX4 reviews:
Translation: needs account to install plugins. π€
This is the new MX Master 4
Oh nice, they moved the screws from underneath the pads to be accessible. My old one died, my thumb button got stuck, the big one you see at the side.
I didn't want to remove the mousepads, because you have to either get em back on or buy new ones, so I never got to open the mouse and repair it.
In theory, this should be possible for the MX Master 4. Although it's still 130 euro's for a mouse.
Always wanted to try Keychron M6 as an MX Master alternative.
And M7 to see if it gives me that shape feel I miss from my G502
MCHPRS is essential for running CraftGPT within a reasonable amount of time. It's built using vanilla redstone mechanics, and should work in vanilla, but it could take upwards of 10 years to generate a response __without __increasing the tick rate. So if you don't have that long to wait, follow the instructions to install MCHPRS first.
I built a small language model in Minecraft using no command blocks or datapacks!
The model has 5,087,280 parameters, trained in Python on the TinyChat dataset of basic English conversations. It has an embedding dimension of 240, vocabulary of 1920 tokens, and consists of 6 layers. The context window size is 64 tokens, which is enough for (very...
@mental oriole Up for a challenge? π
Also I have to agree with this article. The lack of communication/transparancy is frustrating.
I check bios update every x days now, saw there is 3.50. Wanted to know what it does. π€·ββοΈ
Shit out of luck.
Nah am good.
Booh. Only takes 439 million redstone.
Bruh
lets do it
I believe that they don't know what they have done to cause the failures
but they have clearly done something which is common across many boards
it may be voltage regulation related
this whole thing reminds me of the asrock z77 extreme4 - a motherboard that was briefly famed for its "really good overclocking headroom" until the community realised that when you set 1.3vcore and it said 1.27v on the sensor, it was actually supplying 1.55 vcore. They never admitted to or fixed that, and it was just as serious of a problem.
But, can't they just tell in the patchnotes what they do?
Apparently not. Doesn't help that the guy working on it likely doesn't speak english
there is usually like 1 bios engineer for a huge company, and they live in taiwan or china because they get samples out of factories there. Some only speak chinese.
I just went to check Asus. For fun and giggles. It's not better when it comes to patchnotes.
Version 3280
14.8 MB
2025/09/17
SHA-256 οΌ23353CC49DA8253E63A7A80E1A6D07AAC7CFC4939EF7F432BB3ACA5F1E86A137"1. Improved system performance.
2. Improved device compatibility.
Before running the USB BIOS Flashback tool, please rename the BIOS file (TGB650P.CAP) using BIOSRenamer."
Yeah, gigabyte is even worse
Imagine if games pulled this off.
- Added features.
- Removed bugs.
- Improved performance.
Just those notes over and over again for every patch π€£
Oh they do. Blizzard does it all of the time π
you wake up one day and are like.. why the fuck is a quarter of my damage gone? Undocumented hotfix that broke a main ability
fixed 3 weeks later with no note again
It's a symptom of understaffing and "live service", where the employees are always under pressure to go work on the next thing to sell rather than polish the old adequately.
and communication breakdown / language barriers / timezone barriers
AGESA is basically just a file that AMD drops on them, they don't actually have much info or great understanding of what is inside of it
but AMD could provide update notes and don't
usually there is nothing, sometimes we're left piecing together unofficial forum comments from employees
Yeah i don't like it either
it's just the same or worse everywhere else and we can't not buy CPU's
So, we kind blindly dropping bios updates on our mobo's.
With the idea; these companies know what they doing.
With the experience that; these companies not always know what they doing.
Yeah exactly. Earlier this year Asus shipped a change that caused 30% performance loss in some workloads for no good reason, without documenting it. Took community weeks to track down the particular variable.
And that's the good side, because that variable did not affect safety at all
there are regular shitshows with safety impacting things like microcode and auto voltages when overclocking (incl xmp/expo)
every major motherboard vendor applied unsafe settings on intel 13'th/14'th gen out of the box (microcode aside), and every major motherboard vendor applied unsafe settings on AM5 cpu's when auto overclocks (xmp/expo) were enabled
I hope atleast they can get the 9series for AMD stable on asrock motherboards.
I went to reddit forums today to check if they know what bios update did. And I saw a lot, and recent, posts about dying hardware.
Yeah i would not trust any am5 cpu that was ever plugged into an asrock board honestly
it sucks but it is what it is. They're dying, asrock almost certainly doesn't understand why and so they can't explain or fix it, and this kind of damage can present weeks or months later. It also cannot be pinned 100% on the board itself as it doesn't leave clear evidence, although the board is obviously implicated
Maybe reddit is not the best source. But it didn't give me more confidence in asrock π€£
Best source for hundreds of people reporting dead cpu's
Yeah, not everyone would agree with that. But I think so too.
Lucky it's all 9-series π
Yeah, that is curious
they have a different IO die stepping but it's almost the same
the main difference i would say between the 7000 x3d and 9000 x3d is the VID limits. It's 1.2v on 7000 and 1.42v on 9000
if there's spiking like +5% over VID then that might be relevant
need to get buildzoid measuring the rails π
Thing is: is it just ASRock?
Or is it some kind of statistical shenanigans that make it look like ASRock is disproportionately affected when they are actually not?
It is just asrock.
It's not like the early 7000 series where people blamed x3d due to having 90% of the failures, but that was just because they had 90% of the sales.
Asrock isn't popular, they have like 20% of the sales and 95% of the failures. They're failing at >10x the rate of other vendors.
No clue, like I said; I was looking into bios update patch notes. Went to reddit. And all I found was posts of dead 9xxx CPU's on assrock.
The advisors in the servers I frequently visit all state that it's not exclusive to ASRock and that it's unlikely to be brand-spefific
*AssCock
I would say that the evidence that we do have almost certainly confirms otherwise
I am browsing Asus reddit. No hits on dead CPU's (yet).
Ton of other problems though π€£
My 9800x3d on a MSI board is (still) fine
These spikes. Are when I reload reddit.
Bit crazy if you ask me.
And these are when I reload YT. Also a website once could see as ... a lot. But way more efficient.
I haven't found out why loading reddit is so heavy compared to other websites.
Youtube (www.youtube.com)
33 requests
4.23 MB / 1.04 MB transferred
Finish: 1.55 s
DOMContentLoaded: 5 ms
load: 331 ms
Reddit (www.reddit.com):
285 requests
36.30 MB / 30.79 MB transferred
Finish: 6.44 s
DOMContentLoaded: 535 ms
load: 4.80 s
Reddit, wtf?!?
Oh, context. I get desync in a game. So I started to monitor my network. Then I noticed that it happened when I reload reddit, I saw massive spikes and desync ingame.
That's probably buffer bloat
youtube deliberately throttles its connections to save bw, reddit does the more "correct" thing, but if your internet connection is the limit then it can cause ping spikes.
No, YT is only 4MB.
Reddit throws 10x the data at the user. I think it's loading in way more content than needed.
And YT is better lazy loading.
I mean, bufferbloat can still impact this. Not going to deny that.
But I think the main culprit is the 36.30/30.79 being transfered on landing on reddit.
because "new" reddit loads images for every link, and that blows up the latency and data costs.
I posted the url. just regular reddit url. don't have account and not much of a redditer.
That goes to a different place depending on your settings, since it's unspecified
Ah, well it's all default (clean browser).
Disable Cache
58 requests
2.56 MB / 837.16 kB transferred
Finish: 2.39 s
DOMContentLoaded: 1.55 s
load: 1.90 s
For that old.reddit come. Crazy.
Yeah, it loads 140x98 thumbnails instead of fullsize images π
"new" reddit is awful, pretty much everybody uses "old" reddit, and if they ever break it then the site will die immediately
re: buffer bloat, this video taught me a lot about it
https://www.youtube.com/watch?v=o-g2P3R84dw
I never really fully escaped the issue until i got an internet connection which is literally faster than my gigabit ethernet plug can transfer. That dropped my loaded latency to single digit milliseconds, with download and/or upload maxed out.
In this video I show you how to setup an Edge Router X and a Unifi Access Point. The ER-X is especially interesting for gamers as it supports SQM, which helps to prevent ping spikes caused by congestion.
Giveaway ended on 2017-12-17
βΊ Support Battle(Non)Sense:
https://www.patreon.com/battlenonsense
βΊ Battle(non)sense Wallpaper: https://go...
I know about that. I know i have it bad. Tested it last year.
let me see if I still have results saved, for fun and giggles.
Found it 
this is set to check loaded latency on upload, which is higher. Got 0ms and 2ms with DL.
This site uses netflix CDN, so ISP's cannot throttle it without screwing up netflix.
Date: 2024-11-22
Oh yeah the download bloat is bad but that upload will just outright break things
It's a difficult issue to fix.. but ISP settings sometimes affect it a lot, router has a ton of control, and faster connections generally suffer less often and severely
Yeah, lot of room of improvement. But I got stuck on that department.
There is some technical specifications on how we are allowed to buy our own routers from our ISP.
Instead the ISP tellings; this model and this model is good.
They just gave technical specs, that are way above my head.
It's a new law from a few years that forces ISP to let consumers buy own hardware.
And ISP wants to put up a middle finger to goverment and make it as hard as possible for consumer.
So I need to still find my own router, without breaking the network here.
gl
It's a few pages of this shit.
I tried to tell to my goverment what was going on. Never got a reply back 
"Hey, love the new law. But euhm... this is what the ISP's are doing. The minimal effort, technical specs, so in theory, yes a consumer can swap hardware. In practice, this will never happen because it's freaking rocket science to 99% of our population. So what was the point of the law?"
Man, that's all Dutch to me
doesn't matter.
That was supposed to be a joke 
Ooh, sorry my mind was totally somewhere else. Haha, I get it now.
wait, what is mind?
The phrase "my mind was somewhere else" typically means that a person is distracted or not fully present in the moment, often thinking about other things instead of focusing on their current situation. This can happen during conversations, meetings, or other activities where attention is required.
Asked for some info on AssCock boards - not known good yet, definitely disproportionately affecting ASRock, not exclusive either
And they have anything to back that up? Or trust me bro when "it's happening on other boards too still".
But thanks for checking it out. I had a hunch it woulnd't have any fixes in them. Else they would note it, one would assume?!?
yes, I understand that. a phrase as whole
but still, between brain and consciousness, where's mind? or is it something outside of this spectrum?