zweilinkehaende

It's kinda click-baitey, but i feel like it is an acceptable level. Nothing is free, so saying "This CPU is FREE!" obviously implies either freedom (like the "freedom phone") or licencing fees. It's more on the technical side, but to me the topic is WAY more interesting than seeing the 27th Wish-/AmazonBasics-/Temu-/Alibaba-/Taobao-Setup. This is a video for the tech nerd part of the audience, not necessarily the gamer consumers. They barely talk about the computer itself so they clearly don't imply that anyone should go out and buy one. This is a technology showcase about an industry trend that will in all likelihood reshape the whole tech sector in the coming years. I agree that they should have cut more of the gaming side, but they try to get their gaming audience into the video too, which i can understand, even though i disagree with the choice. Also, saying and omitting that Linus qualifies the statement about it being free as "the IDEA behind it is free", is kinda rich. This is clearly a developer kit type of product and the IDEA behind IT is the CPU's architecture and design and the CPU itself respectively. They could have spelt it out for you, but text disclaimers or voice lines addressing that would make the video less enjoyable to watch for me. You seem to have missed the point of the video, which one could argue is an indicator of the video's quality, but most of your complaints seems to stem from a misunderstanding on your part that not many people share.

Really interesting video. This is the kind of stuff that i wish we could see every day. (Both in terms of type of content and innovation) Obviously thats impossible, but i do love the video and am happy every time this kind of video gets posted. Feedback: 1.: Why did we not see the cooler removal? Did Linus (or somebody else) remove the cooler and was the photo of the CPU made by you, or did you use a photo by somebody else? If you took the photo, why didn't you get a shot of the full removal process? Giving an explanaition, even just a short text on screen, would make the video look more polished, this way it was a little jarring, i kept waiting for that shot and wondering about that editing choice in the back of my mind for the rest of the video. 2.: I like the design choice of the animated greenscreen background with the LTT logo, it works well for extended explaination segments. But even at 4K, the compression on youtube does kinda butcher the effect though, so it looks a little cheap and ugly. The circles look more like wobbling eggs decaying into LTT logos IMO. (At least at 4K on a 4K display, might look fine at 4K on lower resolutions) I don't care that much and if i did i could always subscribe to flyboat, but it could hurt the video's performance on Youtube and i want to see this kind of video do well, so more of it's kind get made. These two things make the video feel like it could be better if you put in a little more effort, even though you obviously put a lot of effort into it already. Here the presentation does the rest of the video a disservice. 3.: For the sake of "evergreen-ness" and out of personal preference i would prefer a little more technical detail. Perhaps an Alex/Riley + whiteboard/animation segment. (Something like: Complex instruction does X on x86, but needs to be done in Y simpler steps of A-, B-, and C-type operations on RISC-architectures) I would also have loved to see a very simple real world test. Writing a very simple FOR loop that adds/subtracts/multiplies numbers, times itself and takes ~30 seconds on a modern mid-range CPU (or one with comparable transistor size and same clockspeed) to complete. I assume we would see similar scaling to the one in gaming, but maybe not, since the instructions would be so simple. Would be more interesting than seeing games, with to me unknown performance and scaling characteristics, barely run, or a screenshot of a benchmark score IMO. So if you wanted to keep the runtime the same you could reduce that segment and perhaps the politics segment a bit. The politics segment is obviously very relevant to USamericans and industry insiders right now, but is comparatively detailed and loses relevance over time, while technical explanaitions will stay relevant as long as RISC-V and x86 are relevant. Overall it's a great video and you definitely renewed my interest in tech innovation in the field of chip design. This motivated me to read the wikipedia article on RISC-V again and read up further on the state of development for it, even though it doesn't effect me personally at all until it becomes a serious competitor to ARM and x86 in the consumer space. The video does a great job of stoking interest but leaves me unsatisfied on the details and made me search for more in depth content on the subject. If stoking interest is what this video was aiming to do, it's perfect; but i feel like this video itself, a followup video, or a "TechNotAsQuickie" could satisfy the curiosity resulting from that interest better than most other content on the internet.

Idea to use the mom-and-pop-computer-shop haul for a video: You could make a combined AI + buyout-haul, the extremely new + the ancient video. If that is actually possible you could use an AI-assistant to inventory, appraise and maybe even sell off your haul. Seems like an interesting old-vs-new/progress-of-technology-theme. This way you would have a callback to the feelgood video, which would be the part of the video that you could get out of figuring out how much you got hosed, and you would have a very nice real life test of the usefulness of AI-assistants for real world applications. Maybe you could milk it for a lifestream if the content-density if not high enough for a video. Cataloging and appraising seems like the kind of busy-work that AI will replace at some point, it would be interesting to see how far the tech is along so far. Maybe when the commercial Office-Copilot releases, since interest in AI-assistants is going to be high and people will want to find out what these assistants can do. Downside would be that you would have to have someone test the viability off-screens first though, first try in the video could fall through completely. Upside would be that testing the viability wouldn't take long and if it falls through you would have company experience to contribute to the discussion on assistants that will inevitable happen on WAN show when the assistants release.

Did you check your memory with memtest? 6000MT RAM is technically out of spec for your CPU so the memory controller might not hack it. Make sure your RAM is installed correctly as indicated for your MB (most often 2/4). Are you using 2 sticks or 4? With 4 sticks you often don't get the same speed and timings as with 2. Maybe check if EXPO sets everything correctly. If memtest indicates a problem and you didn't make any mistakes you can only hope that newer BIOS/chipset drivers fix your problems, otherwise you need to loosen the timings or clock down.

I bought the server as configured above and everything worked great, thanks everyone!

Is this video sponsored by GPD? It says #ad right above the title for me, but I don't see anything in the video disclosing that. Did I miss the disclosure? (In which case it should be more obvious IMO) Or is Youtube being funky with that hashtag?

You definitely saved me some research time with that. I'm really torn between 4TB and 8 TB, but i guess if i really don't want to touch it for years i should go with the 8 TB. This way i can even offer some space in that pool to relatives.

Thanks for the warning, the last server i built had an iGPU, so i wasn't thinking about that. I guess i'll have to use a PRO 4650G then and find a cooler for it. (Cheapest AM4 APU that supports ECC with this MB, if i interpret the asrock website correctly) Edited my post accordingly.

When i think about it i actually knew that, but it slipped my mind. You are right, 8 TB would probably be overkill. I adjusted my post accordingly, thank you. I wanted an M.2 ssd because there would be no wiring (no internal usb headers for a thumbdrive). I was a little loose with the terrms here, i knew that there was some something you could use an SSD for with zfs and would have looked into it in the future, but you are probably right that there would be no point. Still, the model i selected is only 24€. Going with one without a DRAM-cache and with half the capacity would save me 5€, but i limit my options if i want to expand the functionality of the server in the future. So i think i would stick with that model.

Hi. I want to build a server that is going to host samba-shares to the local network and for remote access, running 24/7. These shares will sit on a zfs raidz storage pool and will be used as backup locations for various devices. The whole storage pool will be backed up regularly to a backup-provider with a command line tool. I want to build a central server that is going to be backed up instead of backing up individual devices. Ideally i would not have to touch the server for years to come (apart from cleaning dustfilters). I am looking for feedback on my parts selection. The budget is flexible, but lower cost is obviously preferable. First the partslist: (EU-region) Case: Fractal Node 304 89,89€ CPU: Ryzen PRO 4650G, tray 144,80 € MB: Asrock a320M-ITX 91,60€ RAM: 2x 8GB (1x 8GB) Kingston Server Premier ECC DDR4-2666 DIMM CL19-19-19- Single 99,86€ Storage: 3x Seagate IronWolf NAS HDD +Rescue 4TB, SATA 6Gb/s 3x 94,94€ Boot: Patriot P300 256GB, M.2 (P300P256GM28) 24,39€ PSU: Corsair RMx Series 2021 RM550x 550W ATX 2.4 (CP-9020197-EU) 48,90€ ~= 790€ Rationale: I want to build a NAS that is going to "live" in the living room. Thats why I'm leaning towards the node 304. It has space for 6 3.5" drives (potential future expansion) and includes fans, while being somewhat inconspicous. I excluded racks, because the fan noise is typically high (right?). This means that i need an m-ITX board. Asrock supports ECC on all AM4 chipsets apparently, so i chose the cheapest one (and the MB specs on the website ECC support for this CPU-MB-combo) (For future expansion i would need a pcie-sata riser, but that shouldn't be an issue). This leads me to the cheapest AM4 chip with included cooler and ECC support in my region (EU): Ryzen 3500. (EDIT: Ryzen PRO 4650G for the iGPU, will upgrade the cooler on a different system to salvage that stock cooler bought a stock cooler for 5€) I would run it without OC and limited TDP to minimize noise. RAM is the cheapest 16GB ECC kit available. (See below) For storage I chose the cheapest 5400 rpm (noise), cmr (performance and little cost increase), 4 TB drives, rated for this kind of load from a reputable brand. x3 means 8TB usable space, which should definitely be enough for a few years, while leaving enough room for future expansion in the case. Currently the server would need ~1TB of space for existing data, but keeping multiple versions to protect against accidental deletions would probably bring it close to 2 TB. For the bootdevice i chose the cheapest M.2 pcie, dram-cache model. It's only pcie 3, but i imagine that latency will be more important than speed for caching. but that performance should be plenty and there are few savings to be had by downgrading that ssd. PSU is the cheapest gold-certified PSU 300W+, which happens to be from a good brand. Points i'm most unsure about: Are there new mITX boards for cheaper ECC supporting CPUs? The 3500 is a little overspec, right? Can i get away with buying individual RAM modules instead of a kit? (2x Kingston Server Premier DIMM 8GB, DDR4-3200, CL22-22-22, ECC = 93€ would save 40€) (lowest speeds and loosest timings save 2-3€ per module, so i don't really see a reason to not use the fastest ones) (Went with slower RAM because the price difference became greater when sourcing all components from as few suppliers as possible to save on shipping) No ECC modules are officially supported, but that should be fine, right? Sorry for any typos/spelling/gammar and thanks in advance! EDITS: Went from 3x 4TB raidz to 2x 4TB mirror back to 3x 4TB raidz CPU: Ryzen 3500 --> PRO 4650G RAM: Kit -> 2x 8GB (1x 8GB) Kingston Server Premier ECC DDR4-2666 DIMM CL19-19-19- Single Final update: I went with the configuration as above, activated ECC in the advanced settings in the BIOS and installed TrueNas. TrueNas is reporting that ECC works and everything worked flawlessly, i didn't even have to update the BIOS, it already came with a version supporting the CPU. The setup is audible but only from close range and easily drowned out by even the slightest noise. I'm very happy with the result, thanks everyone!

I have been running my RAM at 3133 MTs for over one week without any crashes, freezes or BSODs. So if anyone has the same problem, that seems to have done the trick. Don't trust memtest blindly.

I will try using the RAM at non DOCP speeds. It's a little disappointing that memtest doesn't test the memory controller at the same time.

Setup: Ryzen 2700X (stock Prism cooler) Asus X470-F Gaming on newest BIOS 2x(16GB (2x 8192MB) G.Skill Trident Z RGB DDR4-3200 DIMM CL16-18-18-38 Dual Kit) running at D.O.C.P. speeds Palit RTX 3070 version 1 750 Watt EVGA SuperNOVA G3 Modular 80+ Gold LG C10 connected via HDMI to the GPU BenQ 1080p 120Hz LG G900 mouse LogiTech Stream cam Firefox using the AMD Ryzen Balanced Power Plan Hi Everyone! I upgraded from a 1080p monitor to the C10, from a 1080 to the 3070 and from 16 GB to 32 GB a few months back. Somewhere around that time (apart from initial RAM issues, which happened immediatly) the problems started. Since then i have experienced freezes (with the iconic sound problems) that don't resolve to a bluescreen and thus don't produce crash reports (i think?), bluescreens and back to desktopcrashes specifically in Dota 2 (was always rock solid) and Star Citizen (a mess and crashes could be coincidental) without any error messages. I have been trying to analyze the bluescreen files using WinDbg, but i'm a little out of my depth there. From what i can tell it attributes the crashes to the modules: - nt (x1) - amdppm (x2) - hardware (x1) - nvlddmkm (x1) It also starts each debugging log with: *** WARNING: Unable to verify timestamp for win32k.sys These are just the results from the clean install on the new drive, so i don't know what the other errors were, but from my recollection it mostly said amdppm.sys on the bluescreen. Such diverse errors would have led me to believe that it's a memory problem, but memtest says it's fine and i don't know how else to test for errors. "Unable to verify timestamp" leads me to speculate that the MB clock is bad (battery low?) but for all i know that could be a normal warning for a bluescreen crashdump. To troubleshoot i did (roughly in chronological order): - ran bootable memtest 86 for all tests, dialled in the RAM speed at DOCP timings but 2866 MT (earliest speed without any errors on all 4 passes) -> greatly increased stability, did this directly after installing the second RAM kit, thought the memory controller on my CPU just couldn't handle the same speed on 4 modules vs 2 - updated all drivers/windows updates -> no change - changed Dota 2 renderer in an attempt to stop the btd crashes -> no change - ran sfc -scannow (it repaired LG drivers and something with onedrive) -> no change - ran chkdsk -> no errors - reset windows to earlier image using the built in function -> no change - disconnected the second display and replaced the HDMI cable (first cable sometimes showed intermittent picture loss/artifacts) -> no change but no more artifacting/picture loss - reinstalled windows on a different drive, removed (almost, windows apparently put the bootloader on my second m.2 and neither of my m.2s was previously part of the install) all other drives -> no change (didn't install the LG drivers this time, never connected the webcam) - used ryzen master to turn off precision boost overdrive -> no change - updated BIOS from factory version from 2018 to most recent and ran bootable memtest 86 for all tests, dialled in the RAM speed at DOCP timings and speed, 0 Errors 4 passes -> bios update improved RAM stability, but didn't fix the problem - ran sfc -scannow (it repaired onedrive again) The problem (bluescreen and freezes) happens most often when running a game and starting/stopping video playback in firefox, but i didn't collect rigorous data on that, so that might just be a false correlation. I can't reproduce the problem by running stresstests (furmark + ycruncher) and at this point i don't even know where the problem could be. I already sold my 1080 and i don't have a different AM4 chip, so i can't test by swapping hardware. These problems are generally hard to predict, sometimes they don't happen for multiple days, sometimes multiple times per hour. If You have a concrete idea what the problem could be that would obviously be great! Otherwise clearing up a few questions would probably go a long way towards helping me narrow down my search for a solution: - Could the RAM be unstable, even though memtest86 says it's completely fine? - Is it possible for the graphicscard to cause failures in the nt/amdppm/hardware modules? (aka can i rule the graphicscard and display out?) - Could my secondary M.2 drive that apparently hosts the bootloader be the problem or can i rule storage out? - Would it be worth it to switch out the MB battery or am i on the wrong track there? I don't know which files to make available, i obviously have some crashdumps, but i don't know if they would help/if i should censor something about them, so tell me if You think any specific file would help with diagnosing the problem. Any help would be greatly appreciated, thank You!

Topic: A double blind study if people can tell the difference between 1080p, 4k and 8k at different distances and pixel densities. Motivation: 8k seems to become the new selling feature on the highest tier of TVs, even though many people claim that you shouldn't even be able to tell the difference between 1080p and 4k at traditional TV sizes and viewing distances. The claimed reason is that the distance between pixels projected on the retina becomes smaller than the distance between photoreceptive cells. Example: But human vision is not that simple (1 pixel --> 1 photoreceptive cell), it instead works by integrating the signals of many different photoreceptors, so while the image might not get measurably sharper, people might still be able to tell the difference because 8k seems more natural or something vague like that. (light from 2 pixels hits 3 photoreceptive cells for example and the difference in activation level is processed into the image in your mind, for example) Testing this out would be a nice opportunity to show off cool tech in any case, could save people (or peoples less tech savy relatives) a lot of money (if you can't tell the difference), or be an opportunity for TV manufaturers (possible sponsors) to prove "the haters" wrong (if you can tell the difference). In the end you could lets the 8k TV "stretch it's legs" and turn on all the special features if you want to show other benefits of that class of TVs. Setup: Three calibrated TVs of equal size, brightness, refresh rate, and contrast (alternatively 1 tv set at different resolutions, but that could lead to scaling problems, since the TVs built-in scaler will upscale the image, right?) with the same or concealed stands and bezels set up by group 1, group 2 is saying what resolution they think the TV is (no approaching the screen closer than the defined distance by groups 2 and 3 obviously) and group 3 is recording. Only group 1 knows which TV is which, but isn't directly involved in the experiment, thus making this a double-blind setup, so the results should hold up to scrutiny. 8k footage prescaled to the different native resolutions is played on the TVs. Optional Bonus Points: vary the viewing distance (screen size is irrelevant since it just translates to different viewing distances. Summary: This would be interesting for anyone in the market for a TV, or monitor (you already did a video on 4k monitors, so testing monitors too might not be worth the effort, since 8k is mostly a selling point on TVs anyway).