Black Friday EPYC Turin VPS | Top-up Bonus | from 3.90 EUR

network · December 2025

Sun 28 Dec 11:22:46 UTC 2025

Basic System Information:
---------------------------------
Uptime     : 2 days, 17 hours, 35 minutes
Processor  : AMD EPYC 9275F 24-Core Processor
CPU cores  : 12 @ 4099.996 MHz
AES-NI     : ✔ Enabled
VM-x/AMD-V : ✔ Enabled
RAM        : 51.0 GiB
Swap       : 8.0 GiB
Disk       : 540.3 GiB
Distro     : Ubuntu 24.04.3 LTS
Kernel     : 6.8.0-90-generic
VM Type    : KVM
IPv4/IPv6  : ✔ Online / ❌ Offline

IPv4 Network Information:
---------------------------------
ISP        : ROETH \u0026 BECK GbR
ASN        : AS44486 synlinq.de
Host       : Rene Roeth trading as ROETH \u0026 BECK GbR
Location   : Gelnhausen, Hesse (HE)
Country    : Germany

fio Disk Speed Tests (Mixed R/W 50/50) (Partition /dev/sda2):
---------------------------------
Block Size | 4k            (IOPS) | 64k           (IOPS)
  ------   | ---            ----  | ----           ----
Read       | 424.39 MB/s (106.0k) | 4.06 GB/s    (63.5k)
Write      | 425.51 MB/s (106.3k) | 4.09 GB/s    (63.9k)
Total      | 849.90 MB/s (212.4k) | 8.16 GB/s   (127.5k)
           |                      |
Block Size | 512k          (IOPS) | 1m            (IOPS)
  ------   | ---            ----  | ----           ----
Read       | 3.92 GB/s     (7.6k) | 6.69 GB/s     (6.5k)
Write      | 4.12 GB/s     (8.0k) | 7.14 GB/s     (6.9k)
Total      | 8.05 GB/s    (15.7k) | 13.84 GB/s   (13.5k)

iperf3 Network Speed Tests (IPv4):
---------------------------------
Provider        | Location (Link)           | Send Speed      | Recv Speed      | Ping
-----           | -----                     | ----            | ----            | ----
Clouvider       | London, UK (10G)          | 4.97 Gbits/sec  | 4.94 Gbits/sec  | 12.1 ms
Eranium         | Amsterdam, NL (100G)      | 5.20 Gbits/sec  | 4.96 Gbits/sec  | 7.98 ms
Uztelecom       | Tashkent, UZ (10G)        | 1.73 Gbits/sec  | 2.01 Gbits/sec  | 94.9 ms
Leaseweb        | Singapore, SG (10G)       | 973 Mbits/sec   | 1.44 Gbits/sec  | --
Clouvider       | Los Angeles, CA, US (10G) | 1.10 Gbits/sec  | 1.44 Gbits/sec  | 139 ms
Leaseweb        | NYC, NY, US (10G)         | 1.83 Gbits/sec  | 2.68 Gbits/sec  | 84.0 ms
Edgoo           | Sao Paulo, BR (1G)        | 846 Mbits/sec   | 1.30 Gbits/sec  | 172 ms

Geekbench 6 Benchmark Test:
---------------------------------
Test            | Value
                |
Single Core     | 2507
Multi Core      | 14579
Full Test       | https://browser.geekbench.com/v6/cpu/15808857

YABS completed in 7 min 58 sec

Depends on the price.

£18 per month.

Seems to be running better than my cheap dedicated at the same price.

Keeper?

Yes, pretty good deal if you need this amount of power.

riomartin · December 2025

@tmdr said:
3 pages of non-sense discussion.

I think we can all agree <2200 and 10%st is worth of opening a ticket with bero to check what's going on. Most other Turins here are around ~2400 with nearly 0 st.
@riomartin just open a ticket and wait, its holidays+weekend so might take a while.

We can also all agree that Value/money you get on Bero is great and mostly unmatched.

Ticket opened
last reply 5 days ago
seems like 5 years ago for me.

And probably I cannot depend so much on Turin
since it was not technically wins on performance of single core

OsirisBlack · December 2025

I've always found Bero Host to be quite reasonable and prompt with ticket replys, generelly within ~2 to 3 hours of submission, which pleasantly included the Christmas period.

That said - my tone in every ticket submitted was respectful and presented in the same manner I would expect to be spoken to myself.

itachikonoha · December 2025

@riomartin said:

@tmdr said:
3 pages of non-sense discussion.

I think we can all agree <2200 and 10%st is worth of opening a ticket with bero to check what's going on. Most other Turins here are around ~2400 with nearly 0 st.
@riomartin just open a ticket and wait, its holidays+weekend so might take a while.

We can also all agree that Value/money you get on Bero is great and mostly unmatched.

Ticket opened
last reply 5 days ago
seems like 5 years ago for me.

And probably I cannot depend so much on Turin
since it was not technically wins on performance of single core

Turin can not match ryzen 9950x in single score.

Considering you are so much concerned about GB score, you should had know it.

barbaros · December 2025

@riomartin said:

@tmdr said:
3 pages of non-sense discussion.

I think we can all agree <2200 and 10%st is worth of opening a ticket with bero to check what's going on. Most other Turins here are around ~2400 with nearly 0 st.
@riomartin just open a ticket and wait, its holidays+weekend so might take a while.

We can also all agree that Value/money you get on Bero is great and mostly unmatched.

Ticket opened
last reply 5 days ago
seems like 5 years ago for me.

And probably I cannot depend so much on Turin
since it was not technically wins on performance of single core

What is your use case for relying so much on single core performance? Ryzen has better single core performance compared to Epyc Turin.

Arirang · December 2025

@tmdr said:
My Turin has been holding steady at 2400~2600. Never got below 2400 so I'm happy with it. We will see after the holidays.

Time is UTC
=== 2025-12-27 14:21:40 ===
Single-Core: 2439

=== 2025-12-27 10:21:19 ===
Single-Core: 2621

=== 2025-12-27 04:21:19 ===
Single-Core: 2543

=== 2025-12-26 14:21:19 ===
Single-Core: 2613

=== 2025-12-26 10:21:28 ===
Single-Core: 2476

=== 2025-12-26 04:21:26 ===
Single-Core: 2546

Ryzen started taking a hit recently, holding around 2900:
=== 2025-12-27 14:21:01 ===
Single-Core: 2858

=== 2025-12-27 10:20:56 ===
Single-Core: 2866

=== 2025-12-27 04:20:56 ===
Single-Core: 2876

=== 2025-12-27 00:38:37 ===
Single-Core: 2915

=== 2025-12-26 14:20:59 ===
Single-Core: 2841

=== 2025-12-26 10:20:54 ===
Single-Core: 2936

=== 2025-12-26 04:20:52 ===
Single-Core: 2939

=== 2025-12-26 01:18:46 ===
Single-Core: 3040

You do YABS every 6 hours?
I think I finally understand why YABS score went dropping. Few or above half might YABS every some hours.

itachikonoha · December 2025

Fluctuation of 100 in GB6 shouldn't be considered as "taking a hit" in my opinion.

kenjing789 · December 2025

I wish Andreas put those Yabs Only user who create ticket to migrate better node into same node and let them fight each

itachikonoha · December 2025

@kenjing789 said:
I wish Andreas put those Yabs Only user who create ticket to migrate better node into same node and let them fight each

.... by checking YABS every hour?

network · December 2025

Is YABS at 6 hour interval really that bad? On a fast CPU and NVMe it's 5-6 minutes of 100% CPU usage and 2-3 minutes of NVMe every 6 hours.

itachikonoha · December 2025

@network said:
Is YABS at 6 hour interval really that bad? On a fast CPU and NVMe it's 5-6 minutes of 100% CPU usage and 2-3 minutes of NVMe every 6 hours.

Problem is not hours. But in the same node, if other are running the same way, sooner or later, the timing will overlap and will lessen the GB score. Then one will call it taking a hit while it is not such in reality and presents a misleading picture.

kenjing789 · December 2025

@network said:
Is YABS at 6 hour interval really that bad? On a fast CPU and NVMe it's 5-6 minutes of 100% CPU usage and 2-3 minutes of NVMe every 6 hours.

Assume they actually fit enough ram for every people, a Turin node should have around 38 people.
If you fit 38 people in 6 hours time frame, and normally people wont yabs on odd minute on cronjob, only at x hour 00 min. You have avg 6.3 people do yabs every hour, no server can survival that without score degrade.

network · December 2025

@kenjing789 said:

@network said:
Is YABS at 6 hour interval really that bad? On a fast CPU and NVMe it's 5-6 minutes of 100% CPU usage and 2-3 minutes of NVMe every 6 hours.

Assume they actually fit enough ram for every people, a Turin node should have around 38 people.
If you fit 38 people in 6 hours time frame, and normally people wont yabs on odd minute on cronjob, only at x hour 00 min. You have avg 6.3 people do yabs every hour, no server can survival that without score degrade.

Yeah if everyone runs on 00 min then it's a problem. Otherwise 6 minutes of full CPU load every 6 hours is just 1.5% average CPU usage.

DanSummer · December 2025

@network said:
Is YABS at 6 hour interval really that bad? On a fast CPU and NVMe it's 5-6 minutes of 100% CPU usage and 2-3 minutes of NVMe every 6 hours.

I think the real question is why and of what use? let's say it slows down the vps for 12 minutes each time you run it, that's about an hour a day with slow or almost no access to sites you have on that server.

network · December 2025

@DanSummer said:

@network said:
Is YABS at 6 hour interval really that bad? On a fast CPU and NVMe it's 5-6 minutes of 100% CPU usage and 2-3 minutes of NVMe every 6 hours.

I think the real question is why and of what use? let's say it slows down the vps for 12 minutes each time you run it, that's about an hour a day with slow or almost no access to sites you have on that server.

I don't think the people who run YABS on cron are actually using the VPS. If you have stuff running, YABS numbers won't be accurate anyways.

network · December 2025

Here's the GB6 (Single / Multi) of my KVM XMAS 2025:

2025-12-21  17:40:32    2455    /   7637
2025-12-24  09:07:55    2449    /   7258
2025-12-24  18:14:24    2363    /   7093
2025-12-28  20:52:53    2377    /   6983
2025-12-28  21:19:38    2320    /   6868

Slightly lower than day 1, but still good.

ascicode · December 2025

@network said:

@DanSummer said:

@network said:
Is YABS at 6 hour interval really that bad? On a fast CPU and NVMe it's 5-6 minutes of 100% CPU usage and 2-3 minutes of NVMe every 6 hours.

I think the real question is why and of what use? let's say it slows down the vps for 12 minutes each time you run it, that's about an hour a day with slow or almost no access to sites you have on that server.

I don't think the people who run YABS on cron are actually using the VPS. If you have stuff running, YABS numbers won't be accurate anyways.

In other words, the vps works better without yabsing users. But who knows that.

itachikonoha · December 2025

@network said:

@kenjing789 said:

@network said:
Is YABS at 6 hour interval really that bad? On a fast CPU and NVMe it's 5-6 minutes of 100% CPU usage and 2-3 minutes of NVMe every 6 hours.

Assume they actually fit enough ram for every people, a Turin node should have around 38 people.
If you fit 38 people in 6 hours time frame, and normally people wont yabs on odd minute on cronjob, only at x hour 00 min. You have avg 6.3 people do yabs every hour, no server can survival that without score degrade.

Yeah if everyone runs on 00 min then it's a problem. Otherwise 6 minutes of full CPU load every 6 hours is just 1.5% average CPU usage.

If one person throws a plastic bag, then it is a non issue. But if everyone starts to think like.... "it's just me, one person.... What impact it can do in the end? Surely negligible!"....

.... Now we have a problem in hand.

OsirisBlack · December 2025

Honest question : Do people just buy a vps to YABS it and that's it .....

Buy - Logi in - YABS - Moan it's not looking the same as last time - Rinse and Repeat.

Protocol903 · December 2025

@OsirisBlack said: Honest question : Do people just buy a vps to YABS it and that's it .....

i like to do yabs of my new vps for a month or two of purchase. but unless there is a significant decrease, i dont whine like @riomartin

noisycode · December 2025

@OsirisBlack said:
Honest question : Do people just buy a vps to YABS it and that's it .....

Buy - Logi in - YABS - Moan it's not looking the same as last time - Rinse and Repeat.

I thought a server's only capable of YABS.

tmdr · December 2025

@Arirang said: You do YABS every 6 hours?

GB6 only. Scheduled for 10, 14 and 4am. Wanted to check peak working hours vs night time. I'm new with Bero.

But It's a single core only test, takes 5-10 minutes. I'd think if I someone is running multiple multi-core tests per day they'll be limited fast...

If anyone here is running tests often, I'd recommend it. Just ask Claude Code

  #!/bin/bash

  # Geekbench benchmark script - runs on 1 vCPU and saves results
  GEEKBENCH="/root/geekbench/Geekbench-6.5.0-Linux/geekbench6"
  RESULTS_FILE="/root/geekbench/results.txt"
  TEMP_OUTPUT="/tmp/geekbench_output.txt"

  # Run Geekbench limited to 1 CPU (core 0)
  taskset -c 0 "$GEEKBENCH" --cpu 2>&1 | tee "$TEMP_OUTPUT"

  # Get results URL
  RESULTS_URL=$(grep -o "https://browser.geekbench.com/v6/cpu/[0-9]*" "$TEMP_OUTPUT" | head -1)

  # Fetch scores from results page
  SINGLE_CORE=""
  MULTI_CORE=""
  if [ -n "$RESULTS_URL" ]; then
      curl -s "$RESULTS_URL" > /tmp/gb_page.html
      SINGLE_CORE=$(grep -A1 "score-container-1" /tmp/gb_page.html | grep "class='score'" | grep -o "[0-9]*")
      MULTI_CORE=$(grep -A4 "score-container desktop" /tmp/gb_page.html | grep "class='score'" | tail -1 | grep -o "[0-9]*")
      rm -f /tmp/gb_page.html
  fi

  # Get current date/time
  DATETIME=$(date '+%Y-%m-%d %H:%M:%S')

  # Create new entry
  NEW_ENTRY="=== $DATETIME ===
  Single-Core: $SINGLE_CORE
  Multi-Core:  $MULTI_CORE
  URL: $RESULTS_URL
  "

  # Prepend to results file (newest at top)
  if [ -f "$RESULTS_FILE" ]; then
      echo "$NEW_ENTRY" | cat - "$RESULTS_FILE" > /tmp/results_temp.txt
      mv /tmp/results_temp.txt "$RESULTS_FILE"
  else
      echo "$NEW_ENTRY" > "$RESULTS_FILE"
  fi

  # Cleanup
  rm -f "$TEMP_OUTPUT"

  echo ""
  echo "========================================"
  echo "Benchmark complete: $DATETIME"
  echo "Single-Core: $SINGLE_CORE | Multi-Core: $MULTI_CORE"
  echo "URL: $RESULTS_URL"
  echo "========================================"

tmdr · December 2025

@itachikonoha said: Fluctuation of 100 in GB6 shouldn't be considered as "taking a hit" in my opinion.

It went from 3000~3100 to 2800~2900 which should be normal. Yesterday was bad at 2500~2600 during the day but it has recovered to 2750~2900 today. This is the Ryzen.

The Turin I've stopped tracking as It is already running production containers, and performance was stable for the days I tracked, stopped caring about it.

Again, I only run SingleCore, don't think it's wise to run MultiCore tests often.

emperor · December 2025

@OsirisBlack said: Honest question : Do people just buy a vps to YABS it and that's it

I Yabs 2 times.. 1st when i buy the server and later when is for renewal. If the score is acceptable i would renew, if its not than im looking for another deal.

Spiridon · December 2025

Guys, don't waste your time and electricity on pointless tests!

Don't worry about the falling GB6 scores. You know that the CPU has a base frequency and a TurboBoost frequency. These high GB6 scores for single-core were obtained on an unloaded CPU, so turbo frequencies were used. When working on a node where clients create load, the CPU packages are already hot, so the turbo boost is activated in pulses for a short time, or not activated at all, and the core operates at the base frequency.

So for the 9950x, the TurboBoost frequency is 5.7, and the base frequency is 4.3. The difference is exactly 25%. So if the best GB6 results are around 3100, the actual measurements will range from 3100 to 2325.

For the 9275F, the boost frequency is 4.8, and the base frequency is 4.1. The difference is 15%. So if the best GB6 results are around 2600, the actual measurements will fluctuate between 2600 and 2210.

Therefore, there is no need to worry about GB6 results - they will fluctuate depending on the CPU load. If you want to control the load, buy a dedicated server, but that's a whole different price range.

Please don't pick on the guys at Bero-host, they know their stuff, the main thing is that they help solve problems, and, by the way, they do it quickly, day and night!

alincupunct · December 2025

Good thing about Bero is that you have options.
You need 9950x? More RAM or storage on Turin nodes? They got you! Can be either VPS or VDS. Both at very competitive prices.
Support is also decently fast (for me it's not as fast as a few minutes or hours but it's definitely under 24 hours). Migrated my VM on a different node when I was hitting 7.5-10% steal while under very low load. Now I get under 1.5% steal, and honestly for 14€/mo for 8 vCores, 40GB RAM and 400GB storage I don't think I can complain.

noisycode · December 2025

@berohost As a happy customer with ONLY 4 servers here, I have to ask if there is any plan to celebrate the 10th anniversary, with double digits!

anakara · January 1

Happy new year 2026 @berohost! Thanks for providing good and stable service over the years.

hyperblast · January 1

windows 2025, 4core, 20gb ram, Epyc 9275F
gb6 singlecore: 2011 gb6 multicore: 6063

statement from the professionals here on the values

itachikonoha · January 1

@hyperblast said:
windows 2025, 4core, 20gb ram, Epyc 9275F
gb6 singlecore: 2011 gb6 multicore: 6063

statement from the professionals here on the values

Windows 2025 runs a lot of background services and GUI is heavy. So it'll be less than when ran on Ubuntu.

Howdy, Stranger!

Categories

In this Discussion

Black Friday EPYC Turin VPS | Top-up Bonus | from 3.90 EUR

Comments

Howdy, Stranger!

Quick Links

Categories

In this Discussion

Black Friday EPYC Turin VPS | Top-up Bonus | from 3.90 EUR

Comments