• Content count

  • Joined

  • Last visited

  • Days Won


Pauven last won the day on April 2

Pauven had the most liked content!

Community Reputation

26 Good

About Pauven

  • Rank
    Advanced Member


  • Gender
  • Location
    Atlanta Metro Area
  1. Thanks guys! I think the camera running out of storage is exactly what happened! Cody (of 908 Pinball Zine) was filming and was very disappointed. He had caught me at the very end of the show, after I had pretty much packed up. I showed him under the hood, but he never actually got to see it turned on... and I couldn't be motivated to do so since his storage was full. We'll go for round 2 in a few weeks. If you want to see under the hood of Modern Firepower, here's a good post: That was before the playfield was installed, but otherwise it runs pretty much complete (just needs 3 cables to playfield: USB, 5V/12V power and 50V power). If you compare the wiring of my playfield with a typical pinball playfield, this is a HUGE improvement. Black Knight Rises will be even better! While I haven't gone to market, I have developed my own kit, and it is in Alpha Testing (Black Knight is the Alpha test). I have a new IO Controller for the LED's and Switches, both superior and cheaper to anything else I could find on the market (except for what maybe Stern uses internally). I have a new version of my Power Driver, shrunken to the size of a business card yet more capable (and Arduino/Pi compatible to boot). I've developed my own pinball LED lighting solution, though I think I need another revision to make it smaller. I'm really proud of my PCB designs. I went from through-hole 1980's style circuit boards to surface mount SMT designs, ready for mass production - but a PITA to hand assemble, especially with resistors the size of fleas!!! I can't wait to show it all off, only a handful of people have seen anything I've done. And the crown jewel is my software. I've been making it better so much better. There's a guy in the Atlanta area that makes custom pinball machines, and I have him lined up as a Beta Tester. Once that's done, then I'm ready to try and market it. Oh, and the biggest problem I had with Modern Firepower is that it looked too good! Most people walked by and didn't realize how unique and special it was in a sea of pinball machines. This time, I'm making my cabinet look very different, hopefully to draw more attention to Black Knight Rises. -Paul
  2. That has definitely been a factor, yes, but just one of many. I have started development of the next UTT release, but have not got to the testing stage yet. I also have a solid road map for coding my changes. For the moment, I've had to halt progress to work on another project. What could possibly be more important than a new unRAID Tunables Tester? Pinball, of course! Okay, okay, I get it, I have messed up priorities. I'll try to explain. For those of you in the Southeast (USA), the 4th annual "Southern Fried Gameroom Expo" (SFGE) is June 9-11 in Atlanta. Tons of pinball and arcade machines. You should go if you can. A few years ago I built my own pinball machine from parts, wrote my own software and used off the shelf, non-pinball circuitry combined with a few circuit boards of my own design. It came out great. I took it to SFGE 3 years ago, and it did really well. After that show, I designed my own control circuit boards to replace the off the shelf parts I had used. My stuff is both more capable and cheaper. I also had picked up a Williams Black Knight playfield that had been basically trashed, with the intention of building a machine around it and adding my own circuit boards and software to it. Nothing like waiting until the last minute, but I've finally started the build. With the show 3 weeks from today, I'm completely focused on building this pinball machine. My goal is to take it to SFGE. Hopefully I make it, but it is a massive project. The good news is that if I don't make the show, I've got another year to get it finished... If you interested in reading more, you can check out my blog. Currently all my posts are from my previous build, Modern Firepower, 3 years ago, but I have a lot of new posts I'm working on detailing the new build: The Black Knight Rises. And right after the show, I'm going on vacation. TL;DR: unRAID Tunables Tester development is on hold until the end of June. Sorry. -Paul
  3. Awesome, another success story makes the hard work worth it. I'm at 45 days up-time and counting. A couple noteworthy items: In an interview with Forbes, AMD confirmed that it is working to improve Ryzen DDR4 support, and has an AGESA update for May (probably the already mentioned in this thread). AMD also stressed on the need for motherboard manufacturers to improve their board designs in the future, with more PCB layers and better copper traces between the DIMM slots and the processor socket. Is that AMD throwing shade? Looks like AMD is starting to lay the blame on low memory speeds back on the manufacturers. Pure speculation, but this might mean that our first gen Ryzen motherboards have design limitations holding back memory speeds, and that AMD is not going to waste too much time trying to improve memory support for poorly designed motherboards. Also, I see unRAID 6.3.4 landed a few days ago. I looked at the release notes, but didn't see anything Ryzen related. Also, the kernel only got upgraded to 4.9.28. Very disappointing. Anybody upgrade yet? Any stability improvements? -Paul
  4. Have you disabled "Global C-state Control" yet? If you haven't read through this long forum thread, you may have missed that Ryzen has major stability issues on unRAID, and disabling "Global C-state Control" seems to resolve the issue. It's also been noted that the problems seem to happen more frequently the more "idle" your server is. So a server sitting there doing nothing (and going into lower power C-states) is more likely to crash on unRAID. On the flip side, running lots of VM's and keeping the server busy seems to make it less likely to crash (but it will still crash eventually). It's backwards from what you would expect. On my Ryzen server, I went from just a few hours of up-time before constant crashes, to now 38 days of up-time and counting with it disabled. The only drawbacks seem to be that, by disabling C-states, idle power consumption and heat increases. I also have a link in my signature below back to the original posts regarding this issue. -Paul
  5. No, nothing specific. The only thing I've seen listed so far for is the memory compatibility enhancements. My HighPoint 2760A uses the same chip as the AOC-SAS2LP-MV8 (same driver too). The difference is that the 2760A has 3 of those chips (for 24 drives instead of 8), and a PLX PCIe hub to allow all three to communicate over the PCIe bus. 3 x 8.8w = 26.4w, plus a bit more for the PLX hub, and you're at the same 27w that the 2760A consumes. Going from memory, there was a 2740A with 16 ports, and a 2720A with 8, which would be equivalent to the AOC-SAS2LP-MV8. I think all three of these are no longer in production. I chose the 2760A to go with my Celeron G1610 because the mini-ITX motherboard I used only had a single PCIe slot, and only a few SATA ports. This card was pretty much the only way I could get over 20 drives. Now with my new Ryzen motherboard, it is overkill as my motherboard has 10 SATA ports (I think), plus a couple M.2 slots, and plenty of PCIe slots. I could have gone with a 16 port card, or a couple 8 port cards. The 2760A is very expensive, so I would only recommend it if you can't find more affordable alternatives. I've done the math over and over and over again, and it always comes out that installing bigger hard drives is the cheaper solution, as you need fewer hard drives, fewer ports, fewer cages, fewer everything. If I divide the cost of my server (CPU/MB/Memory/GPU/Controller/PS/Case) by the number of drive slots (24), each bay costs about $115, and that's before adding any hard drives. Fill that up with 24 3TB drives @ $115/ea, and you get a 66TB server (dual parity) at a cost of $83/TB. Fill that up with 24 6TB drives @ $220/ea, and you get a 132TB server (dual parity) at a cost of $61/TB. Fill that up with 24 8TB drives @ $270/ea, and you get a 176TB server (dual parity) at a cost of $53/TB. Fill that up with 24 10TB HGST He10 Helium drives @ $500/ea, and you get a 220TB server at a cost of $67/TB. These are real nice drives, and should last a long time, with a similar cost to 6TB drives, but certainly not the cheapest path forward at this time. -Paul
  6. The overclock sleep bug is unrelated to the C-state issue. Basically, if you overclock your system, that bug causes the reported frequency to be really high after resuming from sleep. It was fixed in AGESA, so recent BIOS' already have this fix. I definitely spin the drives down. But even then, with 18 HD's (approx 5-10w spun down), my 2760A controller card (a massive 27w at idle), 3x 120mm case fans (about 5w at idle speeds), the six SAS backplanes in my case (another 10w) and the GT 710 GPU (guessing 5-10w at idle), combined with a Ryzen idling at 38w, and my server easily gets into the 90w-100w range at best. That was one of my surprises when I set out to build a sub 20w system around the Celeron G1610, and ended up with 70w idles, the majority of which came from my 24-port SAS adapter card and my very nice X-Case RM-424 Pro server case, two things I never even anticipated. Add those 40w+ to your 54w build, and you're in the mid 90's too. Going from the Celeron G1610 to the Ryzen 7 1800X should have added about 20-30w to my server, factoring in the higher CPU idle, 4x the memory and the external GPU. With C-states enabled I'm actually fairly close to that. -Paul
  7. New Ryzen BIOS updates will soon be in the works, as AMD is working on an AGESA microcode update with improved memory support: On a separate note, I'm happy to announce that my Ryzen server now has over 32 days of up-time, more than a full month. Disabling Global C-state Control really made a big difference for my motherboard. The one real problem I have with disabling "Global C-state Control" is that my power usage has increased. Instead of idling closer to 90w, the server is idling at almost 120w. More than anything else, this is shortening how long the UPS can keep the server running. Before upgrading to Ryzen, my Celeron G1610 idled with the same drives at about 71w, and I could get 90 minutes easy from my UPS. Now my runtime is about 60 minutes. Even getting the idle down to 90w would give me 20 more minutes of runtime on the UPS. I was prepared for a higher idle with the upgrade to Ryzen, but never anticipated it would be 45w higher, that's ridiculous. Hopefully whatever is broken in unRAID gets fixed so that we no longer have to disable the C-states. I have been able to watch a couple movies without issues, both directly from the cache drive and from the array. The one movie I couldn't finish earlier must have just been a bad ISO, but since that the movie is "Monster Trucks" and isn't worth finishing, I haven't gone back and tried it again. Perhaps Ryzen has some artificial intelligence and attempted to save me by corrupting the ISO. I've successfully pre-cleared a couple new drives (using Joe L.'s classic command line utility, not the plugin) and added them to my array without any issues, other than slowing down my monthly parity check by a couple minutes. I've also added the Docker for CrashPlan without any issues. I haven't had a chance to do any more VM testing. I found that the Lubuntu installer consistently hung the system too when installing a new VM. Haven't seen anything from Lime-Tech regarding Ryzen in a long time. Curious if they have had any progress. -Paul
  8. So I read an interesting article this morning that has me wondering if an old compiler version might be to blame for unRAID on Ryzen issues. It seems reasonable to me (without knowing for a fact) that Lime-Tech has to compile their own kernels to include the "special sauce" that makes unRAID what it is. Does anyone know what compiler version Lime-Tech uses to compile the unRAID kernel? It stands to reason that a newer compiler that is "Ryzen aware" would be important to prevent any performance or compatibility issues with the Linux kernel on AMD's newest architecture. Even the recommended compiler versions are several years old, so I would anticipate that LT is on something compatible, but who knows... Perhaps Lime-Tech would care to comment? -Paul
  9. Wait... there's a FAQ?! [searches Lime-Tech documentation... nope. Searches forum... oh, there it is!]. I guess the FAQ isn't official Lime-Tech documentation, but it would have been nice if they at least linked to it. The only FAQ they seem to be linking to is for 4.7. Anyway, thanks for the solution I am using Chrome. The next time I go through a reboot, I'll try the operation with Firefox or IE. Any reports on Edge? -Paul
  10. Thanks! It so happens to be that I'm an hour away from reaching 21 days... blackjack! With 21 days of up-time under my belt, I do have a few oddities to report. I don't know which are Ryzen specific, which are 6.3.3 specific, and which are caused by plug-ins. I figured I'd share them all here for review. 1) The Case of the Hanging Console: The main console (monitor/keyboard physically attached to the server) hangs shortly after boot. While I can still Telnet into the server, and start up multiple Screen sessions, the main console remains hung 3 weeks later. Now, this is purely a guess, but I'm starting to think that the hung console might be causing one or more of the issues described below. I'm thinking that the hung console may actually be a hung process that was executing in the console, and when the process hangs perhaps it causes other processes to be blocked from running. I know other Ryzen users have experienced the hanging main console, and I've only seen it reported on Ryzen. If anyone has a different experience, please share. I honestly didn't care too much about the hung console since I could still Telnet, but in light of the other anomalies I'm experiencing, I'm thinking the this is a bigger issue than I first surmised. 2) The Case of the Reappearing Cache Drive: In my previous build, I had a Samsung 840 1TB SSD installed as my cache drive. In my upgrade to Ryzen, I installed a faster Samsung 960 1TB M.2 drive. During this time, I left the old 840 installed and unassigned. But on every server reboot, unRAID reassigns the 840 as the cache drive (and unassigns the 960). I have to stop the array, change the assignment, and restart the array every time I reboot to get the 960 back into the cache drive role. Because I had so many crashes, I made a point of correcting this assignment and then rebooting the array before it crashed (thinking the change was somehow getting lost due to a crash), but this made no difference. Just a bit ago I physically removed the 840 from the server, but I haven't rebooted since then, so I don't know if this will allow the 960 to come up as cache drive. I'm also a bit concerned that even if I fix it for now, if I ever re-install the 840 that this bad behavior will return. 3) The Case of the Unfinishable Movie: While it might just be a bad ISO, one movie being streamed from the server kept hanging near the end at various spots. Now, it just so happens that this is the only movie I've streamed in the past 3 weeks, so I don't know if the issue is restricted to the single movie, or if it will affect more movies. So, a single bad ISO doesn't seem to be such a big deal, right? I had completely dismissed it until I saw my log file later with the following message at about the same time as the movie was hanging: Apr 13 14:09:53 Tower shfs/user: share cache full Apr 13 14:10:00 Tower shfs/user: share cache full Apr 13 14:10:00 Tower shfs/user: share cache full Apr 13 14:10:00 Tower shfs/user: share cache full Apr 13 14:22:49 Tower shfs/user: share cache full Apr 13 14:23:06 Tower shfs/user: share cache full Apr 13 14:23:06 Tower shfs/user: share cache full Apr 13 14:31:34 Tower shfs/user: share cache full Apr 13 14:31:34 Tower shfs/user: share cache full Apr 13 14:31:34 Tower shfs/user: share cache full Apr 13 14:48:16 Tower shfs/user: share cache full Apr 13 14:48:16 Tower shfs/user: share cache full Apr 13 14:48:16 Tower shfs/user: share cache full Apr 13 14:48:38 Tower shfs/user: share cache full Apr 13 14:50:45 Tower shfs/user: share cache full That message was repeated 157 times, it only occurred during the time we were watching the movie, and interestingly enough the movie being streamed was on the cache drive (it got moved onto the array later that night by the Mover). The next day, a quick forum search prompted me to Stop/Start the array (but I did not reboot) to "fix" the problem, and I haven't seen the error since (but I also haven't streamed a movie since). The cache drive was in no way full, with plenty of empty space, so I don't exactly understand the error. I also didn't see any recent examples of this error on unRAID in the forum (perhaps I missed them?), but it might be that this is the first example of the problem returning from the dead, and might be caused by Ryzen. I plan to do some more testing with this when I get some time. I'll try and finish the movie (but of course it is now in the array and not on the cache drive), and I'll try to stream other movies from both the cache and the array. Until I figure out more, it might be best just to chalk this up as a one-time anomaly. 4) The Case of the Reappearing Cache Drive, Episode 2: Just when you thought the story was over... HAH! You're so wrong, there's more to tell. When I physically removed the old 840 cache drive today, I also installed two new 3TB HD's so I could pre-clear them. But back in the GUI, unRAID still lists the 840 under "Unassigned Devices". unRAID also fails to show the two new HD's I just installed. Even more odd, I started pre-clears on both drives, and in the GUI unRAID indicates "Preclear in progress... Pre-Read. 19% @ 119MB/s (1:20:28)" for the 840 SSD. Now, to be fair, the 840 was drive ID "sdd", which is also the same drive ID for one of the two new HD's I installed (same slot), so this seems to be simply that unRAID is correctly reporting that sdd is being pre-cleared, but falsely reporting that sdd is the 840, and not one of the new drives. And it is worth mentioning that I had dlandon's "Unassigned Devices" installed until just an hour ago. I had noticed that I was seeing two listing of Unassigned Devices on the unRAID Main page, the first with a blue tic-tac-toe icon (which lists the old 840), and the second with the padlock icon (which correctly listed the two new drives). Not liking what I was seeing, I uninstalled the plug-in, and was surprised to see that the tic-tac-toe version of Unassigned Devices was still showing - I'm now thinking that this is built into unRAID stock (when did that happen?...), and by having the separate plugin installed I was seeing two lists. I don't know if having the plugin installed contributed to the problem, but it does seem that the plugin Unassigned Devices worked better (it at least showed the correct drives) but it did not show the running pre-clears. Removing it made no improvement, but I also haven't restarted the array or rebooted. 5) The Case of the Unspinnable Drives: In the GUI, if I press the small gray up-pointing triangle to the left of the drive, the drive does not spin up. If I access the drive through the shares, it will spin-up just fine. I'm also thinking that the all disks "Spin Up" button at the bottom of the Main page does work, but I might need to test again to be sure. Here's the error message that I get in the log when I try to spin-up an individual drive: 6) The Case of the Unexplained Error Messages: These error messages occurred for the very first time today, and if I'm reading the log correctly it was at the same time I was removing the 840 SSD and installing a couple new HD's. Well, I guess that's it for now. While it is nice to have achieved 3 weeks of up-time, these little issues are concerning, and taken as a whole makes me wonder if unRAID is actually running correctly on Ryzen. Perhaps one or more of these are Ryzen caused issues that are unrelated to the stability issues we addressed by disabling "Global C-states Control". -Paul
  11. Though I haven't done it myself, yes it is possible. 1 or 2 users (maybe more) have have already accomplished this. Picking the right motherboard helps, and using AMD graphics seems to make the process a bit easier and perhaps cheaper, though you can go nVidia too, just with a bit more work to do the BIOS passthrough. If you use HDMI sound passthrough that seems to alleviate much of the issues of bad IOMMU groups on sound devices and crackling on USB sound, but it seems there are solutions for both of those issues as well, all described in this thread. Just be prepared to spend a few hours fiddling with BIOS settings and unRAID settings to get it to work. In this post here ( Donach made the best first post ever, and detailed how he edited his syslinux.cfg file to include the downstream,multifunction parameters on his PCIe ACS Override. Assuming it works on other motherboards, I think this might be a huge help for those struggling with bad IOMMU groups even with ACS Override enabled. Donach's IOMMU groupings looked significantly better than any others I've seen on Ryzen. I haven't tried this myself, still testing up-time with Global C-state Control disabled. Today I reached 14 days of uptime. -Paul
  12. Yay! Stock is already up almost 3% today, thanks so much! -Paul
  13. That's really odd. I guess it's just one more example of how unRAID doesn't work quite right on Ryzen.
  14. Happy to help! Did you notice any increase in power consumption or temperatures with C-states disabled? I've never been able to get my motherboard to boot above 2400, even when running 3600 memory, so I couldn't even get far enough to get a kernel panic. I just checked and it looks like you are running QVL'd memory, so I would expect the BIOS support to be good. Have you run Memtest86? You might have bad memory. Sometimes re-installing the memory into different slots can help too. -Paul
Copyright © 2005-2017 Lime Technology, Inc. unRAIDĀ® is a registered trademark of Lime Technology, Inc.