Видео

They have a 30Tb HAMR which is based on CMR but using the techniques in the video to shrink the domains. And they have a 32Tb version that combines this with SMR, though I would imagine host managed, not device managed, which is a different proposition. There are also larger ones coming. However, these disks are being used by hyper scalers, but seems still ‘coming soon’ for general availability. As an aside, WD and Tosh are also working on this tech, but seems they are a bit behind. I am also testing SMR drives from all these manufactures at the moment, and initial results suggest WD SMR is actually the best working implementation from a performance perspective. Still not good, but the others look like they are worse.

🙏

I think the problem actually turned out to be with the recording software producing clicks in the audio. Took some time to root cause this but shouldn’t be an issue on later videos, but sorry if it’s distracting.🫣

Very welcome, glad it was useful. I will be doing an update on this soon also as there is more data on larger drives available now. 😎

👍

Is this just down to bad experiences with the products? Or is it around feature/pricing? And interested which you prefer.

The RUclips algorithm is a fickle beast. This video is also a bit different to other content I created recently, so that doesn’t help. But views, likes shares and subs all help give YT indicators so appreciate you taking the time to watch it. It was a fun one to make. 👍

Edit: Found the issue and fixed the chapters. Thanks!

The CMR disks (conventional magnetic recording) are the ones you will want for your NAS. The SMR (Shingled magnetic recording) are the ones that can be specifically problematic when used in things like NAS RAID arrays. And it actually isn't just WD that sell SMR disks, the vast majority of 'desktop' disks at the moment are SMR, and this includes the Barracuda's from Seagate. So for a NAS, I would check that I am getting CMR disks, and i would avoid the desktop drives and stick to NAS, or more likely Enterprise drives. Note that the WD Red's are also SMR, but the Red Plus and Red Pro are CMR. I am doing comparative testing at the moment on WD Blue vs Seagate Barracudas vs Toshiba P300 disks.

@@sometechguy thanks a lot for the info i will take extra attention when searching for the correct disk model. specially since i want to be able to buy the same model at a later moment potentially.

This is great to hear, glad this was a help. 👍

They are good drives, but make sure you have two copies of such things. 🫣

Would love to hear your experience with that also.

The only cable that is inter-changeable between the two is the power cable, and these are both a standard cable, you could google 'IEC C13' and you can use them either on a monitor or power supply (or many other things).

@@sometechguy so it doesn’t matter? The cable also says 250 volt idk if that is imported

@@ahmed8607 I don’t what cables you have, what they are rated for or what country you are in (or the mains power in that country). But monitors and power supplies would typically use the same cable type and as long as they are properly made and fused, it should not matter.

@@sometechguy thx for your time deserve a sub

@@ahmed8607 no problem, hope the move goes well and you get your pc up and running 👍

I now wish I had saved some of my old drives, just for sentimental reasons. There was something very industrial about some of them and would be a great trip down memory lane.

Thank you, appreciated! The two bay option could be good, its around 30% cheaper. Amazon.co.uk appear to have a £50 discount on the 4 bay at the moment also, at least at amzn.to/3yGmDEc. QNAP is cheaper, but doesn't have the NVMe option on it. So for the high speed storage option, the D5 and D8 do have an advantage. Which ever you go for, all the best with it. 😁

Thank you! and interesting to hear about using a PCIe 4x expander. I have a 2x PCIe 5 Expander here to test and interested in the performance if running this and a GPU in slot 1. As for temps, when I do the perf testing on the PCIe cards I am going to test the motherboard heatsink vs an external heatsink to see what impact this has. I do wonder about the impact of throttling due to heat there. Especially if multiple slots are populated, as you say.

You may have seen some compares I did of WD vs Seagate and recent reliability. WD look stronger, but there always models (or batches of a model) that are lemons. That said, I totally understand about some of the ‘transparency’ problems and how they can switch you off a brand, sometimes for ever. For the ironwolf vs pro, the pro have a better warranty but the main selection criteria may be capacity. I would personally go for Enterprise grade though. Often far better Tb per buck, but maybe a little noisier.

The hardware doesn’t support this, so I am guessing you set the RAID up in macOS instead and it’s done in software. So the bandwidth limitation is probably down to the RAID config, or hardware capabilities in the Mac. Maybe it’s worth writing something large to the array and checking what the CPU/Memory load is on the Mac, it’s possible it’s struggling with that.

@@sometechguy Thanks for the suggestion. Yes I'm using the MacOS RAID assistant. I've actually been writing lots to do the DAS and have observed that the write speed remains very consistent around 38-40MB/s, with very occasional dips to 35MB/s or spikes to 45MB/s. As for memory usage, it typically peaks out just above 4.8GB out of the 8GB I have installed. CPU usage though goes up close to 80-90% when I'm scanning and updating my Plex folders.

This performance is well below what I would expect, and I think its likely because the Mac isn't up to the task of running the RAID, especially a RAID 10. Also, I don't know if this would be a reliable way to run a RAID to be honest. Depending on your objective here, you may be better off doing RAID 1 in the DAS on the first 2 slots, and then backing up to the other disks which wouldn't be in a RAID. If you are using X18 Exos 12Tbs, those disks are capable of somewhere in the 260MBps range, so 40MBps is way below what they are capable of. I don't know how busy the Mac is, but if its 80-90%, you could try killing all unneeded processes and seeing if it affects the performance. Also, scanning Plex could be random reads which would be a bit slower, but shouldn't be that slow, but may well put even more strain on the Mac. Maybe try and straight copy of data on and off with and without plex running to check CPU loads and what sequential read and write can get to.

@@sometechguy thanks again! I'm using 2x x12 and 2x X16. I've put one pair in bays 1&2 and the other pair in bays 3&4 but I can't remember which ones are in which. They're all renewed by the way, directly from Amazon. In any case, I'm guessing that the 40MB/s is still a huge underperformance of what should be expected. I actually have a temporarily spare external HDD that I'll use to house my library, and then follow your suggestion and test the straight read/write speeds. Regarding your suggestions and questions about my objective is, can I first just confirm if you meant to suggest that I should run RAID 0 (as opposed to RAID 1) for bays 1&2? Putting bsys 1&2 into RAID 1 would only give me 12TB there, correct? I'm actually planning to mainly use it as a Plex server. I guess I could put bays 1&2 into RAID 0 (via the DAS), and then maybe put bays 3&4 into another RAID 0 (via MacOS) and use some other software to mirror the two RAID 0s? One thing that puzzles me though is why the Mac's System Info detects the DAS as a USB 2.0 hub but correctly detects another hub (that my ext HDD is connected to) as USB 3.1. Any ideas? Could the RAID 10 setup be the problem there also?

@sometechguy OK I've tested the read/write speeds in single, JBOD and RAID 0 (not 10), and the cap appears to remain there. I did get a slightly higher sustained write speed (of around 42MB/s) in single and JBOD, but that's negligible. I'm wondering if it's got anything to do with the fact that the Mac seems to be recognizing the DAS as only USB 2.0 hub?

This is a great question, I can’t see what the benefit is of using the same platter and head count and then using SMR. So I would expect the SMR drives to have less platters or heads. But the only real thing that should affect raw read performance off the disk is higher data density per track, tracks per inch should not help. More platters also should not help, as the disk should only be able to read one platter at a time, though I guess a possible small advantage in getting heads to the right place on other platters. Also, there is a 2g heavier wait on the CMR. I believe platters would weight more than that, but maybe they have one less read head and use one platter surface less? Only way to know may be to open them up and look, but that likely kills the drives and I have more testing planned for them.

@@sometechguy I naively expect a plate and at least one head to weigh more than 2g, but I could be wrong. Furthermore, with tracks being closer, seek times should be reduces. I naively expect that to make a bigger difference than increasing rotation speed from 5400 to 7200 RPM, but again - I could be wrong. Would be interesting if you opened the drives (100% your prerogative) to see what happens inside. I appreciate everything that went into making this video. The results caught me by surprise, learned something from watching it.

I have heard this, but also people say it requires additional force to be used on the tension arm. But it’s probably a habit to take it out. Do you find this works well?

@@sometechguy I've only installed 2 CPUs that way. One ancient Intel and one 7900x it went off easily

There isn’t anything in the manual that states that using any ports limits the availability or bandwidth of others, but it’s just the shared bandwidth on the DMI channel. I am going to be doing some testing however, so we will see.

@@sometechguy That would be super cool! Thanks a lot!

I would add that there is no over-provisioning on the chipset on this board. e.g. there are enough lanes available for all slots and ports on the board at the same time. Some boards over provision, where, for example, there are not enough lanes to add an M.2 slot and a PCIe slot, so they both end up sharing the lanes and then the user can choose which to populate, which then disables the other. And I think this is what you mean. So in short, yes it can handle this, but yes I will also be tested this as I have an NVMe expansion card also I am going to be putting in slot 2.

@@sometechguy Biggest regret of my current build is buying B550 motherboard. Didn't think that i can't use all possible ports & slots.

Appreciate it, thank you. 👌

If noise is the primary concern, maybe a Purple Pro, Red Plus, Red Pro or Ironwolf Pro. But at 12Tb, everything is going to be 7200 rpm. The black tops out at 10Tb, and I think there are over priced. The blue and purple max out at 8Tb. I don't know what the noise requirement is, but I would personally be going for an Ultrastar currently, which is an enterprise disk but is often cheaper than all those I listed above, comes with 5 year warranty is a really solid disk. But Ent disks are noisier when they are accessing disk. Also some disks perform wear leveling which can cause a regular sound even if they are not being accessed, you may want to check on the wear leveling schedule on the drive you look at if noise is a concern. Have a look at this video to get some ideas : ruclips.net/video/QDyqNry_mDo/видео.html On that link, I compare the WD range along with price per terabyte and what capacities are available. And this is why I come to the conclusions above. Hope you find what you are looking for.

@@sometechguy Thank you. I dont mind that it makes some noise when it's beeing written too or from. My main concern is if it's making noise when it's idle. Will these disks spin up randomly just beeing in the computer and make this noise?

You don't need to remove the fans from the shroud, you can remove the shroud with the fans in situ and either flip it to one side, or unplug the fans from the board so you can get the fan shroud properly out of the way so you can easily remove the heatsink from the GPU.

@@sometechguy thank you so much good sir

Are yours 16TB Seagates?

Break out the cable ties 🫣 Good luck, and you may get lower temps also. Interestingly, the best way I found to lower temps was to repad the cooler. Some cards, even of the same model ran at different temps and changing the pads could make a huge difference on the worst offenders.