Tybio

They are a little high, in the low 40s during parity check. But well within the specs of the drives. The problem is more my 5-in-3's not moving enough air, I'm just too lazy to figure out why

I've been a fan of the Ironwolves for a bit now, just had my first failure in years about a month ago and got the RMA no problem. That said, the Ironwolf 10TB drives tend to have high (IMHO) seek head clicking, so if quiet is important they aren't my first reccomendation. As to speeds, I get ~250MB/sec pre-clear speeds (dropping to 190MB/sec over time)...so they are FAST, but IIRC I was about 220ish on the Reds I used to have, so it isn't an epic difference. Hope this helps a little!

I went another route, I was on a 4U 24 bay rack server and just moved to a cube. The benifit of the chassis I chose is that it seperates the cooling air for the drives and components...letting me use crazy quiet fans and really tune them to the area they cool. I was able to get 15 hot swap bays and 4 more fixed ones for Cache drives and VM OS drives etc. I'm not saying this is the right answer, but a rack mount is not /always/ the right answer...sometimes a high quality tower/cube can give a lot more and be very cool looking.

Erg, good catch. I think I'll order the MB/RAM and then figure out the cooling solution once I see the board and hold it up to the case. Then I'll wait for Newegg or Amazon to get the chip in stock (Or someone else I know and mostly trust!). Thanks for the advice on this, going to be great to have the power of an i7 with ECC and IPMI.

Nope, the Supermicro I linked in the first post is a C246...with IPMI and most/all of the features you get in an i7 board...8 SATA, 2 M.2, even a U.2 that I'll never ever use.

NextWarehouse is showing the E chips in stock, I've never heard of them before but am considering pulling the trigger http://www.nextwarehouse.com/item/?2955423_g10e&gclid=EAIaIQobChMI7amBn7Ho3QIVwlYNCh3OKwoDEAYYASABEgIXi_D_BwE

Thanks, I didn't notice how small the stepping was in this family... if going for the iGPU over power then it makes sense to step down. I'm a bit concerned that these processes aren't in stock anywhere and don't seem to be getting any buzz. Perhaps I would be better served just getting an i7 and putting it in that MB (The page says it "should work").

I'd been rather settled on a 2700x with one of the new boards for a while, but as I'm unlikely to do VM gaming and more likely to do Plex transcoding of UHD in the near future...Intel just gut-punched me with the XEON E-2186G: https://ark.intel.com/products/134855/Intel-Xeon-E-2186G-Processor-12M-Cache-up-to-4_70-GHz I found a nice server board for that chip: https://www.supermicro.com/products/motherboard/X11/X11SCA-F.cfm And now I'm wanting a sanity check to make sure I've got the trade off right. With the XEON I get less power and slightly higher cost, but end up with a server class board/proc that support ECC AND have the iGPU that can do a few 4k UHD transcodes. What does the group think? Is this as flip-of-the-coin as I'm thinking, or is there a consideration I haven't taken into account?

I think most all of them do. I know the ASRock Taichi does for example. I'd spend some time reading the Ryzen thread before making any final decisions...the problems have been worked out but you will see a lot of talk about hardware etc.

The main reason is to get ECC support, and save some money over an Intel setup. I would not expect 4k transcoding to work well in any processor of this generation...Perhaps I'm just overly cautious, but I'm keeping 4k and 1080p content separated and only letting remote users see the 1080p library. Perhaps someone can jump in here who has done 4k transcoding on the intel procs.

As you are just going to do standard work loads, I'd look at the Ryzen 2700. 65 TDP processor that can transcode like a beast. That will also let you use ECC RAM, which RAM is going to depend on your motherboard, so once you pick that just hit to QVL for it. I really like the PC-D6000 with 3 5x3.5, if you are going to have it "in view" it will look much nicer than the 1200. That said, 5 more drives in the 1200 though consider that drive size is increasing. I currently have a 64TB array on 8 disks and with 14TB drives coming out this year...I actually down-sized from a 24 bay server to my 15....I'd just rather increase drive size than deal with a rack mount or other huge build. I'm not going to comment on the HBA, my experience is the opposite of the community recommendation. I went from LSI to SAS2LP due to LSI issues.

@lixe If your have an old stick, check the bios for "Legacy Boot" or some such option. It might only be looking for EFI boot and you haven't set that on the stick. Or you could plug the USB into another system and follow the instructions on the wiki to enable EFI boot. Just a guess here BTW, never worked with that board before.

Your results are in line with mine, I still have 3x8TB drives so the slow down toward the end of those drives accounts for the extra time on my check. They are still fast drives, but any time you have different sizes you will get longer checks.

I just moved from LSI cards back to the SAS2LP, I had an issue where streaming from one disk would pause for 15-20 seconds when a second disk on the card spun up. I've never had a problem with my Supermicro cards, and only swapped because I was thinking about doing some VMs. From my experience, a big mistake to try the LSIs, now have useless cards that cost money :). For reference, on the SAS2LP I'm seeing the same speed as on the LSI, parity check starts at about 240MB/sec...obviously that's not where it ends!

That seems long, I just did a parity check on my older than dirt 1275v2 with all drives on HBAs and it took ~18 hours to do 10TB. Granted, I'm using the seagate 7200rpm drives...but even with 5400s I can't see it taking over twice as long. I went the pure HBA route as my MB has mostly SATA2 ports ;).

10gb makes sense in corner cases, such as mine where I have a Gig internet connection and can cap it out when downloading, so having head-room to be able to stream to devices in the house even when flirting with the 1G line makes sense to me. If you have 2 1G ports, bonding will also solve 99% of the corner cases, if each client maxes out at 1G including internet then with 2G bonded you would be able to download at line speed if you had a Gig plan and write at line speed to an SATA SSD while leaving head room for streaming. NVMe may change that...but really, 2x client BW should be enough in nearly all cases.

tldr: The Enterprise drives ship in 2H with the goal of 14TB Pro drives late this year: https://www.anandtech.com/show/12911/seagate-shows-off-14-tb-exos-hdd-promises-consumer-14-tb-drives-in-q4

Nice! Never thought about doing it that way

I've been considering this as well, but according to Anandtech the base clock on the 2xxx TRs is 3.0 and right now they are only able to get to 3.4 boost...that's the base of the current versions. I'm not sure if this is just because it is early in sampling or what, but for VMs I'd want at least the base clock of the current generation so I can game...the TR is already a single-core speed hit against the Intel chips, combined with 16 cores not having direct memory access and I'm firmly in the "waiting for benchmarks" camp. Now what I AM considering is getting one of the new generations of X399's that will be able to support the TR2 comfortably and putting a 1950x on it...so I have options down the road.

And it works perfectly fine on the AOC-SAS2LP-MV8, I can spin up/down disks and even spin up/down all disks and not a dropped frame. I guess I'm just a SuperMicro person from here on out. I'd consider selling on those LSI cards, but I don't want to risk passing on my problems to anyone else...guess I just eat the cost.

No luck at all with this, going to swap back to my AOC-SAS2LP-MV8 to see if that suffers the same issue. All I can think of is that something on the MB isn't happy, going backwards is the only way I can test it out...so we will see.

Just start a long read from a disk, like gzip a 1080p remux or something. That's what I did when I lost track of what was where. Since then, I've made sure that the "slot" in the system is mapped to the "disk" in the array, it can be annoying if you are in an upgrade cycle, but picking a "slot1" in the server and making sure that disk is assigned as "disk1" makes life easy, as you always know which disk is the one you are adding/replacing/moving when you are doing it :).

Interesting, I was under the impression that 3200 was an overclock on the X370 as they don't support that natively. Seems like a bit of a mess when it comes to RAM speed these days...not sure how they got so twisted around the flag pole

@killeriq, The 3200 is overclocked to that speed. If you were running at their native speed then they would likely work fine in both motherboards...which is the point I was making, compatability (QVL) is far far more important when running an OC profile than when just running stock speeds. As to the x470 vs x370, all of the critical improvements are/will be available on x370 via bios update. There isn't much penalty to running the x370, but if cost is similar then it would make sense to get the newer chipset...but it isn't worth a huge price delta IMHO.

Wow, 32 cores, 64 threads...good freaking lord.