Jump to content
Search In
  • More options...
Find results that contain...
Find results in...

EtaCarinae

Member
  • Content Count

    91
  • Joined

  • Last visited

Reputation Activity

  1. Informative
    EtaCarinae reacted to pocket full a SHELL32.DLL in Holy $H!T Ep. 14   
    "stop it" couldn't have said it better myself. Please go and learn some proper enterprise architecture before you do this kind of project.   With 24 drives in a RAID, you've culled off most of your performance off the top. RAID is inherently NOT scalable with flash and with over 8 disks. With that many SSDs, you should have MILLIONS of IOPS. What you need is not a single server. For this scale of project what you need is AT LEAST three. Split up the drives (8 each), and make a cluster. With that many drives distributed on an RDMA network and using an appropriate VSAN you should have a very solid high availability storage cluster. Also, add some spinners. Yeah, I said it. With tiered storage you'll have some cold spots that can get put in the back corner on rotational media. With your scheme it's still a blown out UPS or a kernel exception away from disaster. With multiple machines, almost nothing short of a kiloton can knock your SAN offline.   What I'm saying here is that you need to read up on enterprise arch. Throwing youtube money on a bunch of expensive shit doesn't make you a sysadmin. Please call in some consultants and storage guys for this sort of project, because when you do this young techs see it and think it's the way to go. I'm not saying you need certs, but fuck man read a book.   Reading material: BTRFS documentation, Windows Storage Spaces. And if you want to go the cluster route, read up on Storage Spaces Direct, Gluster, CephFS, and the redhat vsan solution.   Seriously man, your entire business is going to be cached/stored on this box, make sure you get it right the first time.
  2. Informative
    EtaCarinae reacted to nbritton in This Server Deployment was HORRIBLE   
    I would recommend trying a low latency (1,000 Hz) kernel. The default generic kernel, in Ubuntu, is 250 Hz. From a performance standpoint 250 and even 100 Hz are better because the system has less "interruptions", but 1,000 Hz can probably offer better stability because it gives the system more opportunities to check in on the needs of the system as a whole. Also recommend enabling x2APIC mode and MSI-X, these will provide more interrupts for your system.
     
    I strongly recommend ZFS (For my work I've done extensive comparison testing with Ext4 and XFS), but I must urge you to use the latest 0.8.x branch because it's way faster; in some cases it offers 2x the performance of the 0.7.0 branch. This is available in Ubuntu 19.10 natively, and via PPA in Ubuntu 16.04 and 18.04... https://launchpad.net/~jonathonf/+archive/ubuntu/zfs
     
    In the video you said nothing about NUMA domains, you are most assuredly hitting inter-NUMA transfer bandwidth limitations. You need to ensure everything is pinned to the same NUMA domains as what the NVMe devices are attached to. You may be better off segregating all of the storage subsystem to a single NUMA domain.
     
    For a ZFS tuning perspective 128 KB record size offers the best overall net performance gain (assuming a mixed workload), however, 64 KB is a really good choice too if you work with small files with random access patterns. The qcow2 format uses 64 KB as the default cluster size so that is also a good choice if the primary use case is VM storage. For 24 NVMe I would recommend 4x 6-disk raidz (or raidz2) vdevs, this will be nearly as fast as striped mirrors but offer a lot more useable capacity. Finding the right ashift value is not straightforward, it's best to simply performance test each value (i.g. 9, 12, 13, etc.). Use atime=off
     
    zpool create -O recordsize=64k -O compression=on -O atime=off -o ashift=9 data
     
    For benchmarking I recommend the following:
     
    fio --randrepeat=1 --ioengine=libaio --direct=0 --gtod_reduce=1 --name=test --filename=test --bsrange=64k-192k --numjobs=16 --group_reporting=1 --random_distribution=zipf:0.5 --norandommap=1 --iodepth=24 --size=32G --rwmixread=50 --time_based=90 --runtime=90 --readwrite=randrw
  3. Like
    EtaCarinae got a reaction from Velcade in Best "carry everything" backpack   
    https://blackember.com/collections/the-citadel-urban-backpack-collection/products/the-minimal-kit
  4. Agree
    EtaCarinae reacted to ddennis002 in RBG Fan Madness   
    There isn't much difference really in static pressure, the ll are quieter, look better, they less then 10% difference in performance of static pressure and airflow. Overall great fans.
  5. Like
    EtaCarinae reacted to RobbinM in Can I FIX this $10,000 CPU??   
    Could be.
    Difference is that Linus is having consistent results, whereas Der8auer got different results (number of channels) every time.
    Another possibility is failing RAM, happened to my server once. That would mean he picked the dead DIMM to install in the first board, which is only a 1/12th chance. Or 1/6th if 2DIMMs are dead (he mentioned 327GB, which is 32GBx10 equivalent). CPU has 6 channels so 1 or 2 dead dimms is possible. But only if he didn't use that exact RAM for the followup video for 6 editors, 1PC. In that case, RAM is all good (because it was OK in that video, which seems to be shot later, because the Tayun board is used in that build).
     
     
     
    I searched for the mainboard he uses first in the video, I can't read the box unfortunately. It does look like the Asus Z11PA-U12/10G-2S.
    Manual indeed states memory configuration for 1 DIMM in slot A1 is a valid configuration.
    Q-code unfortunately is unreadable. Beep code I reverse engineered to: "No Memory detected" by Asus standards.
    That means that there is a problem with the very first memory channel.
    Linus then proceeds to use the 'black memory slot' but, being different than slot A1, and still using a single DIMM, every other configuration than A1 is in fact invalid (see manual). If the physical memory slot A1 was broken, D1 could have been tried because that connects to channel A although still an invalid configuration by the manual, it is the most likely to work. However I already concluded memory channel A has a problem, so using D1 would have stated the same error with this CPU.
     
    This is why the Dual socket board actually booted: for a board to boot, channel A has to at least be occupied. This was not the case with the broken CPU, but with the good one. The board must be engineered to boot with one of the CPU's at least having memory channel A occupied.
     
    TL;DR
    Board 1 (the single socket Asus) is still OK.
    RAM channel A of CPU is not connecting to motherboard.
     
    The comment I posted on YouTube:
     
    I found the actual socket LGA3647 part on Mouser, it is rated for 30 re-seatings.
     
    I would try bending the pins on the Asus board. As soon as it boots, it should be OK for all memory channels because A is giving problems.
    After succes on the Asus board, replicate the bends for the Tayun board.
     
    If there still is no succes, I still have some other ideas to make it work.
  6. Like
    EtaCarinae reacted to Mandrewoid in Can I FIX this $10,000 CPU??   
    Der8auer had a lot of problems getting all 6 memory channels on his 28 core Xeon to work, even without dropping it. I think he said he mounted it like 10 times and eventually just went with 4 channels.
     
    Perhaps the CPU is fine, and you're just having that problem?
  7. Agree
    EtaCarinae got a reaction from Computernaut in The $100,000 PC LIVES!   
    @LinusTech, please, for the love of Dawkins, put that steel lathe/cutter to good use & capitalise on @AlexTheGreatish's wizardry to create a proper case for this frankensteinian monster. You could make it the centre-piece in your LAN centre for which the displays and peripherals could surround it.
  8. Agree
    EtaCarinae got a reaction from Infiltrator in The $100,000 PC LIVES!   
    @LinusTech, please, for the love of Dawkins, put that steel lathe/cutter to good use & capitalise on @AlexTheGreatish's wizardry to create a proper case for this frankensteinian monster. You could make it the centre-piece in your LAN centre for which the displays and peripherals could surround it.
  9. Like
    EtaCarinae reacted to JoeyDM in A case WORTHY of a $100,000 PC?   
    Man, really? I loved that thing. One of my favorite videos in a while.
  10. Like
    EtaCarinae reacted to PsyDei in Building a 4K Workstation with Dennis!   
    What are the components in this computer?, I would really like to build something similar, since I edit a lot of 4k, but I'm dealing with some stuttering atm.
  11. Like
    EtaCarinae reacted to StuP in Building a 4K Workstation with Dennis!   
    What's the build spec? I'm interested in building one but without the case.
  12. Like
    EtaCarinae reacted to Haktar in Suggestions for our iMac Pro repair   
    dude the old 5K panel is not compatible, posting again won't change that, honestly half the people on here are genuinely starting to get on my nerves, the guys clearly asked for assistance and people posting the same garbage over and over will just bury any hopeful solution.
     
    If it takes you a 30 second google to find a solution its not worth posting here as I can guarantee they have seen it, Also to assume you can just rewire the connectors is ridiculous not only for the fact that it'll be almost impossible if not impossible to acquire a schematic for the pro to figure out the pinout with no guess work but I can pretty much guarantee if there's a cheaper way to repair apples device by using an older part the part will be bricked during a future apple update, case in point iPhone 8, Uses the exact same LCD as the iPhone 7, sure it'll display until you update, bye lcd or using a iPad 5 LCD in an Air 1 aye it'll work but it'll be much dimmer.
     
    I realize I'm getting irrationally annoyed but repairs is what I do and it is so frustrating to see people throwing out terrible advise that could potentially damage someones device.
  13. Informative
    EtaCarinae reacted to slav.p in Suggestions for our iMac Pro repair   
    I'm an ACMT working for CityMac in Bellingham, WA (right across the border from you). 
    We are an AASP, can get the parts you're looking for and would be glad to repair the iMac for you. 
     
    I personally have the required certification to repair the iMac Pro.
     
    Let us know!
     
    Slav P CityMac Bellingham
    Apple Certified Macintosh Technician
    Apple Premier Partner | Verizon Authorized Reseller
    360-380-3999
    CityMac.com  
  14. Agree
    EtaCarinae reacted to SolarNova in 120Hz PC Gaming - On a TV!!   
    Sorry to say Linus but your opinion on OLED gaming stinks of ignorance.
     
    Saying one should not game on OLED due to burn in, is wrong on 2 significant levels.
     
    1) OLEDs dont burn in. If a high contrats image is left for to long then those pixels degrade faster thus become dimmer than surrounding pixels, resulting in image retention. Typical 'burn in' is exactly what it sounds, physical damage to the screen caused by a static image, this is present on older Plasma's and CRT's. Old CRT's can even show said burn in with the screen completly unpowered, sicne the screen itself is dmaaged and 'burnt in'.
     
    2) OLED much like later generation Plasma's, CAN be used for gaming and dont burn in anywhere nearly as easily as you, and alot of other people make out. Check out Rtings OLED burn in torture test for an example, it takes weeks of 20/7 usuage showing the same high contrast images to start to show perma image retention.
    Typical gaming, will somtimes show temporary image retention, but thats about it. if one treats a OLED like a Plasma or CRT, using screen savers, and being aware of image retention, it can be avoided rather easily.
     
    I speak from experience, as i write this im using my Panasonic plasma TV, which iv used for over 5 years as my PC VDU, my only PC VDU. And im a heavy user. I use it every day for hours on end. i play games with static images for hours. Guess what , no perma burn in.
     
    Spreading false , or at the very least over generalized, information about OLED's , will only harm peoples perception of them and slow the 'hopefully' eventualy replacment of LCD technologiy in favor of OLED.
  15. Like
    EtaCarinae reacted to NumLock21 in Intel selling LGA1151 Heatsink   
    Intel had better stock heatsinks, that thing must be a joke. And on the box it says "Meets Intel's Stringent shock/drop test". I wonder if that massive heatsink has anything to do with it.


  16. Like
    EtaCarinae got a reaction from Joshndroid in What was the first game you ever played on PC.   
    Digger (1983)
     
    Police Quest (1987)
     
    F-15 Strike Eagle (1989)
  17. Like
  18. Like
    EtaCarinae reacted to KemoKa in When did linus become so unappreciative?   
    I get where you're coming from, but you have to understand just what it is that Linus is doing. He's not running a Tek Syndicate or a JayzTwoCents or a Barnacules Nerdgasm here, he's running a fully-fledged media company supporting himself, eleven employees and a family of four. When you're that big from this, you have to be demanding, because a slip-up for such a small company that depends on Youtube and Vessel of all things can have a big impact. I know it sounds a bit of an exaggeration, but Linus isn't saying this stuff because he's ungrateful, it's just that he has to get stuff done. What exactly would you do in this situation?

    LinusTechTips is the largest tech channel of its kind on YouTube, dwarfing NCIX Tech Tips' subscriber base along with all others, which I think is amazing, but it has to have come from somewhere and it wasn't wishy-washy, take-shit-lying-down tactics.
  19. Like
    EtaCarinae reacted to EChondo in Batman: Arkham Knight Warner Bros[Updated]   
    You mean all the people complaining that have 970's/980's and i5's/i7's are being babies? The systems that are ABOVE recommended specs and aren't using AMD hardware are still babies for complaining about shit performance?
     
    Get over yourself.
  20. Like
    EtaCarinae reacted to EChondo in Batman: Arkham Knight Warner Bros[Updated]   
    Every game is going to have issues at launch now? Just wait a week for patches? How about YOU wait a week and then release the game when it's finished you pretentious cunt.
  21. Like
    EtaCarinae reacted to Humbug in Batman: Arkham Knight Warner Bros[Updated]   
    The good thing is we know that it is really hurting them this time, and they have no choice but to learn a lesson.
     
    -The first major AAA game to have "mostly negative" rating on steam.
    -and they are getting hit hard with refunds, including people who pre-ordered
  22. Like
    EtaCarinae reacted to dalekphalm in Batman: Arkham Knight Warner Bros[Updated]   
    How so? If the game will indeed be fixed "within a week" via patches, then they should damn well just delay the game for a week, and release something that isn't a completely broken piece of shit.
     
    Well said.
     
    Such an excellent rebuttal, good sir! /s
     
    It's such a shame, because the series itself has very positive reviews in terms of the "fun" factor of the gameplay. But conversely, the series has gotten a pretty terrible reputation for their PC Ports. You think they would try and have a smooth release? But I guess they figure, people will buy the game in boat loads anyway, so who cares.
  23. Like
    EtaCarinae reacted to Dogeystyle in Batman: Arkham Knight Warner Bros[Updated]   
    Call me PC gamers babies all you want. But if I spend £50 on a meal I expect it to be cooked when it gets to my table.
  24. Like
    EtaCarinae reacted to Nexxus in Batman: Arkham Knight Warner Bros[Updated]   
    shes doing a grand job with her marketing skills there.

    Nothing like insulting your paying customers to get the name of your project out there.
  25. Like
    EtaCarinae reacted to ShadowCaptain in Batman: Arkham Knight Warner Bros[Updated]   
    People might be being babies, but equally, why should they spend £50! of their money on your broken shit
×