Howdy, Stranger!

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


Needs some info about netcup rootservers
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.

Needs some info about netcup rootservers

gdarkogdarko Member
edited August 2019 in Providers

Hello

Recently i was researching about the netcup rootservers which seems to be pretty good deal but i noticed few things.
I have a dedicated server from hetzner that has a cpu Intel i7-4770 and the single core benchmark on geekbench v4 64 bit is around 4700~, but the single core on netcup rootserver is around 3500~ (Intel gold 6140) cpu which normally should be very similar to the 4770. So my question is why they still limit the cpu power while they advertise that the cores are dedicated. Can anyone give some more info?

Comments

  • YmpkerYmpker Member
    edited August 2019

    Usually when a vps provider limits any resource it is because of the fact that it's still a shared environment and not a bare metal machine dedicated to one customer. That being said I didn't experience any performance issues due to any of their limits.

  • Overheads?

  • angstromangstrom Moderator

    A "dedicated core" on a KVM VPS isn't necessarily directly comparable to a (dedicated) physical core on a dedicated server.

    This said, netcup's Root-Servers are good for many purposes.

  • FalzoFalzo Member
    edited August 2019

    gdarko said: (Intel gold 6140) cpu which normally should be very similar to the 4770.

    what makes you think so?

    https://www.cpubenchmark.net/compare/Intel-i7-4770-vs-Intel-Xeon-Gold-6140/1907vs3132

    it heavily depends on the type of Gold cpu and especially the 6140 has a by far lower core speed which most likely will be the cause for the single thread performance you see

    https://browser.geekbench.com/v4/cpu/search?q=gold+6140

    just check a few others, DO looks pretty much the same f.i.

    I think you might see >4.5k with the gold only on dedicated machines where it can properly boost to the full turbo. BUT that simply does not happen on virtualized environments because it rather needs to utilize all threads instead of boosting.

    so most likely no artifical limitation but the way moar threads vs. turbo capability is intended to work.
    and in the end it's kind of what you asked for by having them dedicate (pin) threads for you - no turbo.

  • williewillie Member
    edited August 2019

    gdarko said: (Intel gold 6140) cpu which normally should be very similar to the 4770.

    xeon gold 6140 = 2.30 ghz
    i7-4770 = 3.9 ghz

    (3.9/2.3)*3500 = 5932

    The Xeon actually did pretty good if the i7 only got 4700.

    Thanked by 1Falzo
  • And just to add, for many use cases, the missing CPU flags on the Netcup RS/VPS lineup will affect performance (AVX and some related flags are missing).

    Also, there is some variance in the performance of the Root servers (much less than the VPS lineup) but it is still a far cry from having the dedicated setup and who knows what kind of neighbours you end up having.

    @Falzo has already highlighted the Turbo differences (which are again not consistent depending on the node workload) as well which does make a big difference clock speed wise in case your workload is pure frequency sensitive as well.

    Thanked by 2Falzo gdarko
  • @willie said:

    gdarko said: (Intel gold 6140) cpu which normally should be very similar to the 4770.

    xeon gold 6140 = 2.30 ghz
    i7-4770 = 3.9 ghz

    (3.9/2.3)*3500 = 5932

    The Xeon actually did pretty good if the i7 only got 4700.

    comparing the base frequency of the gold with the turbo of 4770 isn't completely fair, just saying ;-)

    Thanked by 1nullnothere
  • neikneik Member

    @nullnothere said:
    And just to add, for many use cases, the missing CPU flags on the Netcup RS/VPS lineup will affect performance (AVX and some related flags are missing).

    Actually, on the RS they do passthrough AVX flags afaik.

    Thanked by 1nullnothere
  • @neik said: Actually, on the RS they do passthrough AVX flags afaik.

    Right - they pass through more flags on the RS vs the VPS but I'm not clear on what exactly they pass through (other than that they definitely dont pass through vmx as it's a ridiculously high paid add on per core).

  • jsgjsg Member, Resident Benchmarker

    I guess there are multiple reasons, like .e.g. "housekeeping" overhead and probably most importantly the fact that a dedicated core usually actually still is a vCore (read: a hw thread) and the fact that a system running a lot of diverse software (as opposed to a dedi with a largely consistent set of processes) has the caches trashed all the time.

    That said, no matter the sales slogans (e.g. "like a small dedi"), a dedi cores VPS still is a VPS and not kind of a small dedi. Also btw. the customers first priority is usually not speed but rather consistency of performance.

    I happen to have a netcup root VPS ("dedi core") and I published some numbers here and the flags, too. Here is what I found:

    Std. Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat
              pse36 cflsh mmx fxsr sse sse2 ss htt sse3 pclmulqdq ssse3 fma cx16
              pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline aes xsave osxsave
              avx f16c rdrnd hypervisor
    Ext. Flags: fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx pat
              pse36 rdseed adx smap clflushopt clwb sha umip pku syscall nx pdpe1gb
              rdtscp lm lahf_lm lzcnt
    

    And that is a very nice set of flags for a VPS.

    Thanked by 1ITLabs
Sign In or Register to comment.