The scrap gods are fake news my friend dave sorry not sorry
Ran yabs on my moms dell inspiron from 2014 she uses for facebook why not right
# inspiron 3542, i3-4030U, suffering
CPU: Intel(R) Core(TM) i3-4030U @ 1.90GHz
fio: (R=87MB/s, W=72MB/s) tehhdd
iperf3: 38 Mbps / 12 Mbps
geekbench: SC 312, MC 678Runs words together sometimes you figure it out he says refuses to edit posts ever
Ampere Altra strikes again! Insane value! Just got my hands on a KnownHost preview instance and it RUNS CIRCLES AROUND those hot old Xeons!
# KnownHost preview, Ampere Altra Q80-30
CPU: Neoverse-N1, 80 cores
fio: (R=4.2GB/s, W=3.8GB/s)
iperf3: 18 Gbps / 18 Gbps
geekbench: SC 1287, MC 89456180W for THIS? Your E5 graveyard pulls more at idle! ARM is the future!
Actually the German Approach to Server Procurement emphasizes the Longterm Total Cost of Ownership Analysis including the Electricity price development and the Depreciation schedule, actually I have calculated that my old Haswell Boxes from 2014 still deliver acceptable Performance per Watt if one considers the Acquisition cost of zero Euros because they were already written off in the Accounting system of my former Employer, actually here is the Benchmark of my Basement server:
# rudi3 KellerMaschine, 2x E5-2697 v3
CPU: Intel(R) Xeon(R) E5-2697 v3 @ 2.60GHz
fio: (R=1.8GB/s, W=1.4GB/s)
iperf3: 940 Mbps / 920 Mbps
geekbench: SC 987, MC 23456Actually the PowerDraw is 120W at Idle which is concerning but the Replacement cost amortizes only after fortythree Months at current Electricity prices.
Member
---
- reaction: @bellaauc
Status: appreciated
Action: tested restore
Result: failed
NOTE: backup was from 2022
WARNING: test your restores regularly
- new data:
- provider: OVHcloud
- location: SIN-04
- cpu: EPYC 7402P
- observation: |
Actually decent for the price point
Network: not symmetric
Disk: better than expected
- code:
# OVHcloud SIN-04
fio: (R=2.8GB/s, W=2.1GB/s)
iperf3: 4.1 Gbps / 2.3 Gbps
geekbench: SC 1123, MC 18976
Member
Oh here we go again with the AmpereHype[1] every single time someone runs a YABS on ARM suddenly its the SecondComingOfCompute nevermind that your actual workload is probably somebodyselses bad PHP from 2017 running singlethreaded but sure lets all Pretend the Software ecosystem is there footnotes[1] see my previous seventeen posts on this exact topic in this very thread
Wesh @Berlin1 mon ami why so angry incha'allah your blood pressure is okay le server حلو does not care about your PHP drama
# RackNerd MRS-02, second run, incha'allah
CPU: Ryzen 5 5600X, 6c/12t
AES-NI: yes
fio: (R=2.2GB/s, W=1.9GB/s)
iperf3: 1.9 Gbps / 1.7 Gbps
geekbench: SC 1654 / MC 8172
Back in 2009 I had a Core 2 Quad Q6600 in a Supermicro chassis that drew 95 watts at idle and we thought that was perfectly reasonable. They don't make them like that anymore, and frankly thank goodness for that. My old Xeon E5-2670 v2 story is similar to @Izmir — ran hot, ran loud, but it never once failed me through six years of service. The capacitors on those boards were built to last. These days you get a "server grade" board and the VRMs cook themselves in eighteen months.
"back in my day the servers were on fire and we LIKED it" @YuriDavid backup from 2022 failing restore is the most on-brand thing I've seen all week. Indeed
Member
I must raise a concern regarding the collection of Geekbench scores in this thread. Under GDPR Article 5(1)(c), data minimisation principles suggest we should evaluate whether processor identifiers constitute personal data when combined with provider locations. I am not saying cease posting — merely that we should consider the implications. My own YABS follows with provider anonymised:
🍿 lol nice