Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


Contabo introducing VPS with NVMe drives
New on LowEndTalk? Please Register and read our Community Rules.

Contabo introducing VPS with NVMe drives

adnsadns Member

Contabo introduced NVME VPS.

Can anyone make YABS benchmark test on the new disk? Is any limit on I/O same as normal SSD VPSs?

«134567

Comments

  • finally keeping up with the times. likely EPYC based

    I bench YABS 24/7/365 unless it's a leap year.

  • IOPS was the only one downside of contabo. Now they are almost perfect. DDoS mitigation and network quality remains an open issue.

  • @LTniger said:
    IOPS was the only one downside of contabo. Now they are almost perfect. DDoS mitigation and network quality remains an open issue.

    Network quality in Germany works great for me

    Thanked by 1JasonM
  • @LTniger said: IOPS was the only one downside of contabo. Now they are almost perfect. DDoS mitigation and network quality remains an open issue.

    how about an oversold CPU with high steel time? they are far from perfect

    bench.monster - check your VPS performance

  • @SpeedTest said:

    @LTniger said: IOPS was the only one downside of contabo. Now they are almost perfect. DDoS mitigation and network quality remains an open issue.

    how about an oversold CPU with high steel time? they are far from perfect

    CPU wasn't an issue for me. They have very strange network with weird packet loss instances almost daily.

  • iKeyZiKeyZ Member

    Looks nice, might give them a try again. Remember having a few network blips before, so will be interesting to see what's changed.

    Also, just noticed while browsing their site, down the bottom of https://contabo.com/en/vps/ ( @jsg )

    Thanked by 2LTniger lentro
  • Pity, its only for new purchased VPS, old VPS systems are still on SSD.

  • @active8 said:
    Pity, its only for new purchased VPS, old VPS systems are still on SSD.

    Btw servers with nvme has x4 less space.

  • Pretty cool! For anyone wondering, you can still pick SSD or NVMe. The SSD will be 200 GB and NVMe will be 50 GB. Looks like you can upgrade to a 100 GB NVMe for $2 extra a month. This is for the basic plan.

    Free NAT VPS at MicroLXC

  • @active8 said:
    Pity, its only for new purchased VPS, old VPS systems are still on SSD.

    Looks like you can upgrade to the next plan up and switch to a NVMe drive that way. So if you want NVMe you'll either have to cancel and reorder, or upgrade, either way you'll be spending extra. In my usage the SSDs have been plenty fast enough after they unlocked my disk speed, I don't have any need to upgrade.

    Free NAT VPS at MicroLXC

  • JabJabJabJab Member

    Watch out for upgrades - if you have servers pre-2021 you (most probably) have pre-VAT prices and new orders comes with "extra" VAT, so more expensive. Not really gonna upgrade my 5€ VPS to 11€ VPS because NVMe :(

    Thanked by 1bdspice
  • AXYZEAXYZE Member

    Contabo fixing worst problem? I hope they raised IO limit to at least 8k. 1.3k from SSD offer is just bad. If they managed to do it then I will prob give them second(third tbh) chance.

    Favorite European providers: netcup and Hetzner (free €20, aff)

  • @adns said:
    Contabo introduced NVME VPS.

    Can anyone make YABS benchmark test on the new disk? Is any limit on I/O same as normal SSD VPSs?

    This is really a big step towards quality performance. But would definitely like to hear feedback about these NVMe drives in the long run...

  • I/O: Hey CPU bro, you there?

    Thanked by 1gapper

    Just another hosting control panel development discord server (https://discord.gg/9mepQ7frfM)

  • jsgjsg Member, Resident Benchmarker

    @iKeyZ said:
    Looks nice, might give them a try again. Remember having a few network blips before, so will be interesting to see what's changed.

    Also, just noticed while browsing their site, down the bottom of https://contabo.com/en/vps/ ( @jsg )

    That's based on an actual statement from me to them during internal communication. I gave my permission to quote that.

    The benchmark and review that I wanted to post shortly after their new product line going public actually is with those new Contabo / @contabo_m NVMe VPSs.

    Ultra short TL;DR: Yes, those drives are insanely fast. Everything else is the usual (Epyc processor, fast memory, connectivity as known, somewhat better than during my last benchmarks, but then it's known that Contabo's network team is always trying to improve connectivity when and where possible.

    For the full review of all their locations please be a bit patient; We (the LET team and I) are working on finding a way to get the benchmark and review posted ...

    The problem with democracy is that by definition > 85% of the voters are not particularly intelligent.

  • jsgjsg Member, Resident Benchmarker

    @msallak1 said:
    Btw servers with nvme has x4 less space.

    That's normal. Considering that NVMe is much more expensive than SSDs, at least the fast good quality ones, most providers give you much less NVMe space than SSD space (for the same price).

    Frankly, although I'm impressed by their NVMes and like them a lot, I think that most use cases needing lots of space can easily can get away with SSD.

    Thanked by 1msallak1

    The problem with democracy is that by definition > 85% of the voters are not particularly intelligent.

  • dnnrdnnr Member

    New York Nvme:

    ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ##

    Yet-Another-Bench-Script

    v2021-06-05

    https://github.com/masonr/yet-another-bench-script

    ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ##

    Wed 18 Aug 2021 10:33:59 PM -03

    Basic System Information:

    Processor : AMD EPYC 7282 16-Core Processor
    CPU cores : 6 @ 2794.748 MHz
    AES-NI : ✔ Enabled
    VM-x/AMD-V : ❌ Disabled
    RAM : 15.6 GiB
    Swap : 2.0 GiB
    Disk : 97.9 GiB

    fio Disk Speed Tests (Mixed R/W 50/50):

    Block Size 4k (IOPS) 64k (IOPS)
    Read 23.31 MB/s (5.8k) 331.48 MB/s (5.1k)
    Write 23.31 MB/s (5.8k) 333.22 MB/s (5.2k)
    Total 46.62 MB/s (11.6k) 664.70 MB/s (10.3k)
    Block Size 512k (IOPS) 1m (IOPS)
    ------ --- ---- ---- ----
    Read 3.60 GB/s (7.0k) 8.45 GB/s (8.2k)
    Write 3.80 GB/s (7.4k) 9.01 GB/s (8.8k)
    Total 7.41 GB/s (14.4k) 17.47 GB/s (17.0k)

    iperf3 Network Speed Tests (IPv4):

    Provider | Location (Link) | Send Speed | Recv Speed
    | | |
    Clouvider | London, UK (10G) | 370 Mbits/sec | 384 Mbits/sec
    Online.net | Paris, FR (10G) | 368 Mbits/sec | 384 Mbits/sec
    WorldStream | The Netherlands (10G) | 362 Mbits/sec | 383 Mbits/sec
    Biznet | Jakarta, Indonesia (1G) | 281 Mbits/sec | 96.5 Mbits/sec
    Clouvider | NYC, NY, US (10G) | 381 Mbits/sec | 396 Mbits/sec
    Velocity Online | Tallahassee, FL, US (10G) | 372 Mbits/sec | 392 Mbits/sec
    Clouvider | Los Angeles, CA, US (10G) | 367 Mbits/sec | 384 Mbits/sec
    Iveloz Telecom | Sao Paulo, BR (2G) | 356 Mbits/sec | 368 Mbits/sec

    Geekbench 5 Benchmark Test:

    Test | Value
    |
    Single Core | 945
    Multi Core | 4963
    Full Test | https://browser.geekbench.com/v5/cpu/9390697

    Thanked by 1Arkas
  • jsgjsg Member, Resident Benchmarker

    @dnnr

    Sorry but that "benchmark" is quite wrong. The disks are much faster on 4K and 64K reads and writes and the hypervisor flag is available.

    The problem with democracy is that by definition > 85% of the voters are not particularly intelligent.

  • dnnrdnnr Member

    @jsg I did it again and nothing change. The results are not different.

    Thanked by 1fLoo
  • jsgjsg Member, Resident Benchmarker

    @dnnr said:
    @jsg I did it again and nothing change. The results are not different.

    Then there is a problem with that script.

    The problem with democracy is that by definition > 85% of the voters are not particularly intelligent.

  • dnnrdnnr Member

    @jsg I can try nench :


    nench.sh v2019.07.20 -- https://git.io/nench.sh

    benchmark timestamp: 2021-08-19 02:17:50 UTC

    Processor: AMD EPYC 7282 16-Core Processor
    CPU cores: 6
    Frequency: 2794.748 MHz
    RAM: 15Gi
    Swap: 2.0Gi
    Kernel: Linux 5.4.0-81-generic x86_64

    Disks:
    sda 100G SSD

    CPU: SHA256-hashing 500 MB
    2.768 seconds
    CPU: bzip2-compressing 500 MB
    4.913 seconds
    CPU: AES-encrypting 500 MB
    1.078 seconds

    ioping: seek rate
    min/avg/max/mdev = 25.0 us / 36.9 us / 2.71 ms / 25.4 us
    ioping: sequential read speed
    generated 42.4 k requests in 5.00 s, 10.4 GiB, 8.49 k iops, 2.07 GiB/s

    dd: sequential write speed
    1st run: 801.09 MiB/s
    2nd run: 1525.88 MiB/s
    3rd run: 1525.88 MiB/s
    average: 1284.28 MiB/s

    IPv4 speedtests
    your IPv4: 144.126.148.xxxx

    Cachefly CDN:         46.77 MiB/s
    Leaseweb (NL):        25.87 MiB/s
    Softlayer DAL (US):   31.31 MiB/s
    Online.net (FR):      31.41 MiB/s
    OVH BHS (CA):         45.62 MiB/s
    

    No IPv6 connectivity detected

  • jarjar Provider

    I'm quite a fan of their St Louis location. I dismissed them for a long time as unsustainable but they aren't the first to go from unsustainable to "well damn, actually pretty good."

    "Note that Romania has laws agains all the illegal activities just like US, including copyright. Is not the Dracula's country or no man's land as you thought." - Random email from someone I don't know, about nothing I've done or said

  • AXYZEAXYZE Member

    @jsg said:

    @dnnr said:
    @jsg I did it again and nothing change. The results are not different.

    Then there is a problem with that script.

    4k read and write is 1:1 the same in that benchmark. It doesnt mean there's problem with yabs, it means that there's IOPS limit just like with SSD line. All customers with SSD have 1.3k IOPS. All. Completly zero variance, because its hard limit and it just wont go above that.

    Your opinion about NVMe line is based on VPS they made for you right? Sooo... maybe they just didnt limit your IOPS as youre reviewer? ... :) You will post positive review and attract customers then.

    Blaming whole yabs for results that are worse than you expected? Bruh. Its not your company and they always did it (IOPS limit), so why it is supposed to be different now? They said to you that they wont do it or what? If there is big difference between what you AS REVIEWER got and what customer got then investigate it instead of blaming whole benchmark wtf

    Favorite European providers: netcup and Hetzner (free €20, aff)

  • AXYZEAXYZE Member

    @jsg said:
    @dnnr

    Sorry but that "benchmark" is quite wrong. The disks are much faster on 4K and 64K reads and writes and the hypervisor flag is available.

    Thats enough to see something fishy. Even flags are different?

    Favorite European providers: netcup and Hetzner (free €20, aff)

  • jsgjsg Member, Resident Benchmarker

    @AXYZE said:

    @jsg said:
    @dnnr

    Sorry but that "benchmark" is quite wrong. The disks are much faster on 4K and 64K reads and writes and the hypervisor flag is available.

    Thats enough to see something fishy. Even flags are different?

    I don't know about all flags because that script (unlike my software) shows only AES and hypervisor - but hypervisor is wrong, yes.

    As for the rest, well possible, but I don't think that Contabo f*cked me. For one I have never found them to cheat me even the slightest, they played with open cards. Also: Do you really think that a cheating provider would respond to my request to really hammer their NVMes really hard by saying "Sure, yes, here you go, 5 more testing VMs for you on the same node"? I don't think so.

    I don't know about "they always did" (maybe they did, maybe not, me not know); I'm relying on hard data, plus, frankly, it seems that Contabo wants my reviews mainly for internal purposes and not to help selling via LET. Example: I contacted them and told them that my review will be published with a delay due to my difficulties with CF. I didn't even receive an answer yet (while normally they react quickly). Honestly I think they do not even care much about my review here, they mainly (or even only) care about what I find out and about an open honest and concrete data based feedback.

    Am I a fan of Contabo? Well, yes and no. With my personal need profile in mind my interest in having VMs with them is limited. I'm A NexusBytes / @seriesn fan for a reason: their offers perfectly match my needs. "But their NVMe are somewhat slower than Contabo's!" - me not care, NexusBytes' NVMes are damn fast enough for me.
    But otoh yes, I like Contabo a lot because they are damn serious about wanting good feedback and my experience as a tester with them is great and our conversations are almost shockingly open and frank. It sometimes feels as if they wanted me to break a test VM. It feels a bit like "If jsg approves of a new product it's a go because it's stable and performance numbers are backed up by professionally gathered and cold data". Who could complain about such an attitude?

    As for the other benchmark scripts I don't care and I'm in no way against their authors, so I usually don't comment a lot on them although I know about some weaknesses even from the authors themselves. But when I happen to come across results of a VPS I just benchmarked myself and I see obvious flaws or even errors I say so clearly.

    The problem with democracy is that by definition > 85% of the voters are not particularly intelligent.

  • Hi friends, a question beyond this topic... Anyone know Contabo got supporting tickets? Or only can reach them via [email protected] ?? I bought their VPS SSD and figuring how to raise a support ticket...

  • jsgjsg Member, Resident Benchmarker

    @nephilax said:
    Hi friends, a question beyond this topic... Anyone know Contabo got supporting tickets? Or only can reach them via [email protected] ?? I bought their VPS SSD and figuring how to raise a support ticket...

    AFAIK they don't have a ticket system (yet) in their panel. Support is via email and IIRC via phone.

    Thanked by 1nephilax

    The problem with democracy is that by definition > 85% of the voters are not particularly intelligent.

  • I'm curious about how much response time(ping) is between Contabo and Hetzner.

  • jsgjsg Member, Resident Benchmarker

    @Arirang said:
    I'm curious about how much response time(ping) is between Contabo and Hetzner.

    Give me (PM) a Hetzner VPS (or dedi) IP and I'll quickly find out for you ...

    The problem with democracy is that by definition > 85% of the voters are not particularly intelligent.

  • @jsg said:

    @Arirang said:
    I'm curious about how much response time(ping) is between Contabo and Hetzner.

    Give me (PM) a Hetzner VPS (or dedi) IP and I'll quickly find out for you ...

    fsn.icmp.hetzner.com
    nbg.icmp.hetzner.com
    hel.icmp.hetzner.com

    Thank you!

Sign In or Register to comment.