Considering Xeon E-2278G, is there something newer/better/different I should be looking at?

[EvanVanVan]

I'm looking to upgrade my 7-8 year old hardware (Xeon E2-1230 v2). I've spent over 10 minutes researching this (/s lol) and I've come upon the E-2278G. My CPU budget is let's say < $1,000. Is there anything different, better, or on the horizon that I should consider? I'd rather spend a little more now or go a little overkill than the alternative. I like that the E-2278G was "only" released in Q4 2019 (although though even that is a little old for my liking).

I see there is an E-2300 series that's been released recently although I can't find it listed anywhere?

I know E-2278 is out of stock everywhere and am ok waiting for it on backorder.

Thank you

 

[jgreco]

It's worth noting that a NAS does not benefit from being the latest, greatest. The E3-1230v2 is still a pretty competent CPU, primarily limited by the 32GB RAM limit, and, of course, component age of your mainboard, which may fail at some point, probably in the next five to eight years.

Be wary of throwing cash at a CPU whose value will never be realized. If the 2019-released CPU has all the things you need or can see needing, compare it carefully to the newest, hottest CPU just to make sure you're not missing anything, and then move forward.

 

[EvanVanVan]

It's worth noting that a NAS does not benefit from being the latest, greatest.

Lol that does make sense. But the NAS is not so much my issue (that still works great for my uses), it's my plex server. I've been heavily debating whether to upgrade my TrueNAS server or build an entirely separate PC to run PMS off of. The latter of which seems more wasteful than needlessly upgrading my aging TrueNAS hardware.

Be wary of throwing cash at a CPU whose value will never be realized. If the 2019-released CPU has all the things you need or can see needing, compare it carefully to the newest, hottest CPU just to make sure you're not missing anything, and then move forward.

Yeah...10-12 times a year I watch movies at my parent's house or sister's apartment. While they have sufficiently fast internet connections and hardware capable of direct streaming UHD movies, they only 5.1 sound bars. The transcoding from 7.1 Atmos to 5.1 causes buffering and that ruins the movie watching experience for me. I'd rather just throw money at the problem...

If I don't get the best of the best and it still buffers I will beat myself up over wasting the money. Whereas if I do get the best of the best and it still buffers I can at least feel comfortable knowing there was nothing more I could have done. That being said, like you said, it is still just a NAS, so I'm looking for the best of the best CPU for <$1,000. I'll end up going with 64 GB of ram this time around, unless 128 GB is preferable? Then, yolo, 128 it is! lol

 

[jgreco]

For transcoding, please do carefully review the Plex forums to find out what's really needed for your transcoding use case. This may be a good argument for a newer CPU, but I believe that most of the latest CPU's are way overpowered for a simple single transcode.

 

[Ericloewe]

Yeah...10-12 times a year I watch movies at my parent's house or sister's apartment. While they have sufficiently fast internet connections and hardware capable of direct streaming UHD movies, they only 5.1 sound bars. The transcoding from 7.1 Atmos to 5.1 causes buffering and that ruins the movie watching experience for me. I'd rather just throw money at the problem...

The audio transcoding is what's concerning? That seems kind of weird. Might this be fixable with larger buffers?

This may be a good argument for a newer CPU, but I believe that most of the latest CPU's are way overpowered for a simple single transcode.

Around the Haswell era, a Xeon was expected to do something like three separate transcodes (Full HD or thereabouts). With better IPC and extra cores, a current Xeon E (or are they Xeon W-something now?) should run circles around that.

 

[Jailer]

Yeah...10-12 times a year I watch movies at my parent's house or sister's apartment. While they have sufficiently fast internet connections and hardware capable of direct streaming UHD movies, they only 5.1 sound bars. The transcoding from 7.1 Atmos to 5.1 causes buffering and that ruins the movie watching experience for me. I'd rather just throw money at the problem...

Transcoding the audio is not the problem. An Atmos stream contains a DD stream as well so it shouldn't be transcoding anything. Also audio takes very little CPU power to transcode. I'd suspect it's the upload speed of your internet connection on your end that's causing the problem. Of course this is just a guess but I'd bet that's the issue. The logs from Plex at the time it buffers would give you the answer.

 

[jgreco]

Around the Haswell era, a Xeon was expected to do something like three separate transcodes (Full HD or thereabouts). With better IPC and extra cores, a current Xeon E (or are they Xeon W-something now?) should run circles around that.

The older Haswell era CPU's still have a problem with higher bandwidth content, and can run out of steam. I have seen, somewhere, a nice table that shows CPU generations and what approximately they can handle, but I'm not finding it right at the moment.

One additional thing to note is that if you are running Plex on a bandwidth-limited connection, you probably want to set the congestion control algorithm to something like cubic or htcp; see

Summary of Five New TCP Congestion Control Algorithms Project | FreeBSD Foundation

Grenville Armitage has provided a summary of the completed TCP Congestion Control Algorithms project:BackgroundTCP is a crucial part of any modern operating system. FreeBSD's standard "NewReno" congestion control (CC) is not able to fully utilize the high capacity links available today. A range...

freebsdfoundation.org

And it doesn't hurt to change the client settings to expect only about half of the ISP-advertised upstream bandwidth available to your Plex server. Plex debugging often winds up with counterintuitive problems being the issue...

 

[EvanVanVan]

Thank you all for the informative input. The main reason I suspect the audio transcoding is the issue is because when there is a 5.1 compatibility track included in the file along with the 7.1 TrueHD Atmos track buffering is basically a non-issue. But I think I can test if the CPU is the issue vs the bandwidth by setting up a plex server on my desktop computer with a TR 3970x.

I have FiOS gigabit (up+down), my parents have xfinity gigabit (down only) and my sister has a 200 mbps FiOS connection and lives locally.

The logs from Plex at the time it buffers would give you the answer.

I'll check that out, thank you.

Regardless as @jgreco mention I may still want to upgrade my hardware (motherboard specifically). I see that I'm using all 32 GB of my ram and it would be nice to have more. But again I don't see the difference between a $150-$200 CPU and $500 for the E-2278 being that significant.

The older Haswell era CPU's still have a problem with higher bandwidth content, and can run out of steam. I have seen, somewhere, a nice table that shows CPU generations and what approximately they can handle, but I'm not finding it right at the moment.

One additional thing to note is that if you are running Plex on a bandwidth-limited connection, you probably want to set the congestion control algorithm to something like cubic or htcp; see

Summary of Five New TCP Congestion Control Algorithms Project | FreeBSD Foundation

Grenville Armitage has provided a summary of the completed TCP Congestion Control Algorithms project:BackgroundTCP is a crucial part of any modern operating system. FreeBSD's standard "NewReno" congestion control (CC) is not able to fully utilize the high capacity links available today. A range...

freebsdfoundation.org

And it doesn't hurt to change the client settings to expect only about half of the ISP-advertised upstream bandwidth available to your Plex server. Plex debugging often winds up with counterintuitive problems being the issue...

Ok, I'll have to investigate that when I'm a little more sober lol. Thank you..

 

[jgreco]

a little more sober lol

Don't go on a drunken NAS buying binge! ;-)

 

[EvanVanVan]

Am I going to have any compatibility problems going with bleeding edge hardware?

I.e.:

Intel CM8070804494617 Xeon E-2388G 3.20GHz 8-Core Processor - Rocket Lake

3.20GHz Xeon E-2388G 3.20GHz 8-Core (16 threads), 16MB Cache, 14nm, 8 GT/s, 95W, 1P P750 - (OEM, Tray)

www.wiredzone.com

Supermicro X12STW-TF Motherboard Proprietary WIO Single Socket LGA-1200 (Socket H5) for Intel Xeon E-2300 Processor

Motherboard Proprietary WIO Single Socket LGA-1200 (Socket H5) for 10th Gene Intel Pentium Processor, Intel Xeon E-2300 Processor Up to 128GB DDR4-3200MHz Unbuffered ECC UDIMM in 4 DIMM slots 8x SATA3 port (RAID 0,1,5,10), 2 SATA DOM power connector support 2x 10GbE LAN Ports, 1x IPMI dedicated...

www.wiredzone.com

I'm wondering because my desktop computer I built in the past 2 years couldn't run the stable version of Debian 10 when I first tried. I needed to install the Debian 11 beta.

Lol and I realizing I'm kind of ignoring all the advice about NAS's not needing to be the latest and greatest, but I look at this passmark score: 25,777 and I think wow, I could probably transcode like 20 4K movies at the same time with a score like that.

 

[Etorix]

You're going to have a compatibility issue between the form factor of this motherboard and >99% of the cases that are available at retail.
On the software compatibility side, you're a pioneer, so expect to get some arrows. Whether the iGPU is supported for transcoding in TrueNAS is the most obvious issue; the non-obvious issues, you'll be filling Jira tickets when you find them.

And, of course, you're overspending on hot new hardware, on UDIMM that is more expensive for less capacity than RDIMM, and on a CPU that will never ever be required to transcode 20 4k streams simultaneously.

 

[EvanVanVan]

You're going to have a compatibility issue between the form factor of this motherboard and >99% of the cases that are available at retail.
On the software compatibility side, you're a pioneer, so expect to get some arrows. Whether the iGPU is supported for transcoding in TrueNAS is the most obvious issue; the non-obvious issues, you'll be filling Jira tickets when you find them.

Yeah, I was trying to google what the hell a proprietary WIO form factor really meant, (although proprietary was pretty self-explanatory). I could probably scale it back a little bit lol.

And on a CPU that will never ever be required to transcode 20 4k streams simultaneously.

hahaha I know. x]

What am i missing? I'd be thinking about this more it terms of gpu hardware acceleration then getting an 18,000 watt gigajoule cpu. Why use 100 watts of cpu power when an igpu can do it in 5?

Yeah, maybe if I go with TrueNAS SCALE.. Mostly I'm just spitballing here. Weighing my options.

 

[jgreco]

what the hell a proprietary WIO form factor really meant

It's a non-ATX form factor designed for 1U and 2U server chassis that rely on riser cards. It is the heir to Supermicro's ill-fated UIO concept.

WIO, at its maximum, may allow you to add up to four full-height and three half-height cards into a 2U platform, which would normally be limited to seven half-height cards in an ATX format.

The riser cards are all available individually, and you can kinda troll around Supermicro's prebuilt WIO systems to get an idea about them. One of the interesting things is that you can build custom variations that Supermicro didn't think of or expect.

 

[jgreco]

Oh, I thought intel gpu hardware acceleration in a TrueNAS core jail was doable. I have seen posts on it anyway.

It should be eminently doable. There's nothing that would actively prevent it. That doesn't always mean something will be easy though.

 

[jgreco]

I have seen, somewhere, a nice table that shows CPU generations and what approximately they can handle, but I'm not finding it right at the moment.

Dear idiot, try harder.

Transcoding by CPU generation. I've seen another that provides estimates of simultaneous transcodes, but this seems to be much more dependent on the specific content.