New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
And trust me, my feet have been in the fire for quite a while now.
Sincerely: no — or at least, not like this.
Please don’t get me wrong: I did consider a platform refresh toward the end of 2025, but not at this scale, and definitely not in this way.
By mid-2025 it became clear that continuing on the E5-v4 platform was no longer viable — not with the performance targets we want to maintain and the direction we’re heading in. At this point, paying a premium for DDR4 on E5-v4 makes little sense when the same money invested in Xeon Scalable Gen2 delivers significantly better results.
On the storage side, 12-bay RAID6 with 16–24 TB drives has become a serious issue — rebuild times and performance are simply unacceptable. This is why we moved to RAID60 in a 3×8 layout on nodes using drives larger than 16 TB.
Large-cache RAID controllers also don’t behave well on Gen13 E5-v4 Dell servers, which are now 10+ years old. Last year alone we replaced backplanes and even motherboards far too often — the age of the platform is clearly showing.
Power consumption is also a key operational factor. A fully populated Gen13 E5-v4 system can draw around 420 W, while a Xeon Scalable Gen2 system — with double the drive capacity (24 bays) — peaks at around 680 W under heavy load and averages closer to 630 W. Given the density and performance gains, this is a far better trade-off.
There’s also a clear architectural gap:
Memory bandwidth on E5-v4 tops out at 76.8 GB/s per CPU
Xeon Scalable Gen2 delivers ~140.8 GB/s
More PCIe lanes
Far better I/O behavior under sustained load
That increase in memory bandwidth and I/O capability directly translates into a snappier VPS experience, and the difference is very noticeable in real-world use.
I could go on, but the conclusion is simple: E5-v4 no longer makes sense for us in 2026.
I won’t argue that E5-v4 can still work for storage — especially given how much we’ve tuned it — but doing so requires a disproportionate amount of effort behind the scenes. And realistically, we can not offer 40, 80, or 120 TB capacities / VPS on that platform. We did try — and it was, frankly, a fiasco.
Finally, from a cost perspective, 2023–2024 is no longer comparable to mid-2025. Expenses across the board have shifted enough that continuing with E5-v4 simply doesn’t add up anymore.
I acknowledge that this doesn’t win us popularity points. But this isn’t a beauty contest.
We’re not here because we’re cute — we’re here because we want to sell a service. Doing so at the expense of quality or long-term sustainability would defeat that purpose entirely.
If we wanted to take the easy route, we could have done a presale and disappeared the next day. Unfortunately, there are plenty of examples of that here already. That’s simply not how we operate, nor wish to in the future.
Could this have been handled better? Probably.
Was it intentional? No.
I’ll leave the final judgment to you.
I think the rest of us will be more forgiving if you $7 deals.
Out of curiosity:
@host_c
How many E5-v4 machines are we talking about?
Somehow you can place some Silvers or Platinums, no?
As of now, we still have around ~8 E5-v4 nodes in service. These are currently hosting the legacy Pioneer and Nimitz plans (5 TB and 10 TB).
I’m still working on a possible upgrade or transition option for those users. There’s nothing concrete to announce yet, but we should have more clarity sometime in February.
be gentle when unracking tonite, ok?
Well, he did get the server out eventually
Oh boy, long day ahead tomorrow.
@maverick , keep an eye on your ram
a single kilobyte missing and...
@host_c my collection isn't random enough yet!
OMG OMG OMG OMG.
PANIC!!!oneeoenON!O!Noenein!
stupid maintenance :x
Why are my servers not down yet, am I not being maintained?
All the maintenance!
Hy there
We are almost finished.
We took nodes down, 1 by 1, firstly those that needed NVME changes or other hardware maintenance. ( I saw no point in taking down the whole infra, as we can only work on 2 nodes at a time max )
This way, downtime was minimal / node.
We still have ~3 nodes to go thru, but the hardware part is over since 2 hour ago, so whatever remains is ~ 35-45 minutes / node.
It was a bit more challenging then initially planned, but we managed to pull it off.
Some nodes took more because, well, grub update fuckery and a lot more
@maverick
RAM all there?
all RAM still there, CPU twice as fast after maintenance, it's a miracle!
GJ!
OK folks, last node going thru the updates.
We will start taking in the usual " my vps does not work, I am loosing trillions " shortly, tho, they will not get answered till tomorrow frankly, we need to get some food and sleep.
Enjoy well deserved sleep.
We need some "the maintenance happened and nothing exploded" celebratory deals, yo!
Not quite.
We do have a node that is misbehaving, so we will need to reboot that one again to fix the issue.
Happy happy downtime - do I hate these things.
Good luck, brace yourself for the 15 tickets.
I can't wait for the tickets to come through for the scheduled yearly reboots.
PANIC PANIC PANIC
I hadn't even run my post-maintenance-yabs on that one yet :P
I do have the option to leave the infrastructure to age and do not apply any patches or updates, that is an option too
, it would look good on the uptime threads, not sure I wish for a 2-3 year old system to run around on the net like that, tho, that is opened for debate 
@AlteredParadox
Panic time is over, shit should be up shortly.
We shall see if this fixed the issue.
I will take planned downtime and a secure and stable infra over waking up to say... a rando-script-kiddie-encrypted-vps... any day. lol. Appreciate you guys!
Also i still giggle at the banana explosion every damn time i click thanks in this thread
Yapp, same here, unfortunately not all go as planed.
That node.... , something is not right, we are still investigating. I do not like the the Thermals VS Cooler RPM VS CPU Usage VS IO on the CPU.
I am positive this is not the last reboot of that specific node. On the bright side, the modifications we did do give a slight ( slight means 1-2% not 34334x times the yabs score ) increase in better IO handling, so I might will apply these to the rest also ( scale gen 2 platform ) - but not today. Not until I figure out the root cause of the issue.
We did had to tweak a bit for Proxmox 9, yet I love the 0.00 to 0.05 IO delay on the node, especially as that one is a fat fuker ( 67.56% (241.50 TB of 357.48 TB) )
I do have to thank @fat32 for that, so
fat32 !!!!
Well i finished my yabs, it went up a smidge and bumped serververifys random cpu scores from a C to a B, so... woo? haha
old:
new:
Last time I took something down for a scheduled update and reboot, someone demanded a year's credit on a $10/mo service for the downtime. Hopefully you don't get any of those
Yapp, we still get those even today.
You should see the sub-$5/mo services. The expectations there are........ wild pornographic at best
But I love the creativity they throw in, so +1 for that.
We don’t even hold our own upstreams to the same standards, maybe we should start???? Hmmmm.......
System Inlet Temperature 16 °C (60.8 °F)
System Exhaust Temperature 31 °C (87.8 °F)
Delta is 15 °C - for this type of node it is fine, but I still don't like the cpu temps in correlation to the IO it is doing, investigating further.
PS: Node will not be rebooted anymore, so panic time is over, now me and Stu will go back to the drawing board and lick the logs for the past 12+ hours, so we can see what the heck is wrong.
Cheers!
PS:
Dell says this: System Inlet Temperature Support Limit for this configuration 32 °C (89.6 °F)
Now, if that would really happen, Barbecue on the chassis is what the server would be good for
, idiots.... 32°C, this is not a god damn oven 
🤣🤣🤣🤣🤣🤣🤣 (laugh in TEE-less space heater)