Howdy, Stranger!

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


Benchmark & review of excellent Nexus Bytes VPS [promo til Sat.]
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.

Benchmark & review of excellent Nexus Bytes VPS [promo til Sat.]

jsgjsg Member, Resident Benchmarker
edited October 2019 in Reviews

@seriesn offered me to benchmark and review his VPS and I accepted it because I always find it interesting when a new (at least to me) provider comes up that isn't obviously just a me-too offering the usual stuff.

When I saw his promo (-> https://www.lowendtalk.com/discussion/161100/nexusbytes-megadealthursday-nvme-kvm-starts-3-25-mo-5gbram-70gbnvme-space-1-windows-server-license) I felt confirmed in my first impression, accepted his offer and started benchmarking his $3.25/mo VPS.

Here's the system info:

Machine: amd64. Arch.: amd64. Model: AMD Ryzen 7 3700X 8-Core Processor             
OS. version: FreeBSD 12.0. Mem.: 1.985 GB
CPU - Cores: 2. Family/Model/Stepping: 23/113/0
Cache: 32K/32K L1d/L1i. 512K L2. 32M L3
Std. Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat
          pse36 cflsh mmx fxsr sse sse2 sse3 pclmulqdq ssse3 fma cx16 sse4_1
          sse4_2 popcnt aes xsave osxsave avx f16c rdrnd hypervisor
Ext. Flags: fsgsbase bmi1 avx2 smep bmi2 syscall nx mmxext fxsr_opt pdpe1gb
          rdtscp lm lahf_lm cmp_legacy cr8_legacy lzcnt sse4a misalignsse
          3dnowprefetch osvw

Yay, a 3000 Ryzen! As meanwhile usual I tested under FreeBSD because it's more honest than Linux with disk performance. Linux gives bigger numbers and seems more impressive but that's actually due to very aggressive caching and some trickery which is OK for a desktop but for a server I want real hard facts.

Note the Ryzens cache which is nice (and btw. more important than clock speed) and on the high end but there are some modern Xeons with more L3 cache/core. But still with half a meg L2 cache (which is much more critical than L3) I'm expecting nice results.
Oh and the flags, have a look at them! That's a very full and nice set of flags. popcnt, aes,avx2, sse4 and more goodies; they are all available. Nice.

CPU and memory performance (Average in bold):

                   **Avg**  Min      Max          Dmin    Dmax D-Span
single core: **492.36**  485.07 498.57  1.48   1.26   2.74
multi  core:  **1181**  1090    1230    7.68   4.17   11.86

WOW! I think that's the best processor & memory result I've ever seen in a benchmark I did. Usually a single core result of about 250 is really nice and an E-26xx v3 which is about the best one can hope for with most (better) VPS is in the 300 - 350 range. This Ryzen however is very close to 500!
This means that if you get the larger 4-core promo VPS you can watch most 4 core dedis in the rear mirror!

Disk:

Seq.Wr: **1090**   1050   1118   3.69   2.55   6.24
Rnd.Wr: **599**    574     617     4.12   3.08   7.20
Seq.Rd: **1848**   1637   1988   11.43   7.56   18.99
RndRd:  **1903**   1845   1936   3.03   1.75   4.78

And again a winner. Those are very respectable results, even for an NVMe. Although I personally would suggest to try Raid 1 only rather than Raid 10. Looking at the context (Ryzen, DDR4 mem., and PCIe4 (with the NVMe almost certainly linked via v3 only anyway) I think that additionally striping doesn't add much except in cost. I guess putting just 2 NVMes but even faster ones into the node will cost much less and offer a similar (very good) performance.
Anyway, this is a very nice and fast disk subsystem and a Span below 20% (on the day of the promo at that!) for sequential reading is quite low and perfectly OK.

Network:

GR_UNK:     **93.33**   88.50   96.80   5.17    3.72    8.89
US_SJC:     **30.69**   28.60   32.00   6.80    4.28    11.08
OK_LON:     **266.14**  247.00  288.70  7.19    8.48    15.67
DE_FRA:     **529.81**  315.20  672.20  40.51   26.87   67.38
BR_SAO:     **24.32**   23.00   24.90   5.43    2.38    7.81
FR_PAR:     **267.36**  230.30  296.70  13.86   10.97   24.84
US_WDC:     **38.36**   7.30    53.20   80.97   38.67   119.64
RO_BUC:     **100.93**  95.70   104.20  5.18    3.24    8.42
AU_MEL:     **19.91**   18.00   20.70   9.61    3.95    13.56
IT_MIL:     **247.36**  159.20  298.00  35.64   20.47   56.11
SG_SGP:     **29.92**   28.90   31.20   3.41    4.27    7.69
US_DAL:     **37.61**   36.60   38.80   2.70    3.15    5.85
RU_MOS:     **117.41**  97.30   124.00  17.13   5.61    22.74
JP_TOK:     **19.98**   18.30   21.50   8.40    7.62    16.02
NO_OSL:     **115.77**  111.10  120.80  4.04    4.34    8.38
IN_CHN:     **29.70**   28.40   31.50   4.38    6.06    10.44

The network is a winner, too. Everything in Europe, incl. Moscow, above or at least around 100 Mb/s is really nice. Cross-Atlantic is OK but not great. On the other hand the typical troublesome locations (Ozziland, Chennai, Brazil) show decent - and quite stable! - results.

I'll have to make a confession: After seeing the first result sets I started to feel serious personal interest. But as I'm stubbornly set for objectivity and fairness I pushed that thought aside and decided to first finish the benchmarking and to process the result set. Now, that I'm done I've contacted seriesn to get a better feeling about the people at Nexus-Bytes. The reason: experience tells me that the machines and network are just one part - and a part of little worth when the people are not up to par. Well, I'm pleased to say that they passed that (totally unscientific) test too and that I'm ordering a VPS from them.

That VPS is of course very different from HostSolutions current promo storage VPS. One is just insanely cheap and damn good enough (I really like my new storage VPS and had no problems so far) while the other one, this here, is an insanely fast and nice "power VPS for a very decent price".

One thing however they have in common -> Go, get one! Highly recommended!
You can't go wrong with this, especially for the ridiculously low promo price. With the bigger one (which I didn't test) you can easily run a significant dynamic web site with the database and PHP flying. Highly recommended. For a mid-size (and quite busy) community site even the small one is more than enough. And that's what I buy mine for.

As usual, Sorry for the sh_tty formatting. I don't know why Vanilla doesn't understand the bold marking in a "code" section. Take my good will for the deed please.

Comments

  • It's nice to know the vps I got has jsg approval. Equally happy with it as well

    Thanked by 2uptime seriesn
  • vyas11vyas11 Member
    edited November 2019

    Gad to see an objective review. @seriesn had approached me last month with a request to review/ BM, and while I did do so (links updated in my sig), two names that I suggested right away were @jsg and @falzo.

    I always learn something new from reading reviews by @jsg . Today I learnt that they have as they put it "totally unscientific" BM test for evaluating the providers as well :+1:

    Thanked by 1ITLabs
  • Looks like I have to add Nexus Bytes to my provider list to try. Thanks @jsg

    Thanked by 1seriesn
  • jsgjsg Member, Resident Benchmarker
    edited November 2019

    @vyas11 said:
    Gad to see an objective review. @seriesn had approached me last month with a request to review/ BM, and while I (did do so, two names that I suggested right away were @jsg and @falzo. Now that I am a paying customer too, great to see this review. I always learn something new from reading reviews by @jsg . Today I learnt that they have a as they put it "totally unscientific" BM test for evaluating the providers as well :+1:

    Thanks, but that part of my testing was only because I was personally interested and wanted to get a feeling for the people there before making the final buy decision. And seriesn can be proud (if he wants to) because after doing quite many tests, some published, some private (upon providers request; usually to get an outside view) I have gained a bit of insight into the business. I can say that I'm not personally touched or interested at all with most benchmarks I do; it's just a chore, service I do for our community but I have no lack of systems and rarely get interested. I'm friendly (almost always) but I don't get excited. In this case however the VPS is so great (in all relevant aspects) and at the same time so cheap that I got personally interested. Quite an exception. Kudos to Nexus Bytes.

  • Indeed. I’ve had series of conversations with seriesn as well, with similar result- signed up for a plan.

    Thanked by 1seriesn
  • Thank so much for taking up on my offer and sharing your kind reviews @jsg and @vyas11 . You guys took time out of your busy schedule to do a review, in exchange of literally nothing and can not appreciate that enough!

    I really appreciate all those who took a chance on a relatively smaller guy/company and signed up with us recently.

    Getting you in I guess was the easy part? Can I keep you for a long time? That will be the challenge and one that I will gladly accept :)

    vyas11 said: Gad to see an objective review. @seriesn had approached me last month with a request to review/ BM, and while I did do so (links updated in my sig), two names that I suggested right away were @jsg and @falzo.

    Thank you for those recommendations. Funny thing was, I totally forgot about pming @falzo (I should reach out to him once I have space to spare again :) ) and @jsg wouldn't have been part of the Family if he/she actually wouldn't have followed up on a conversation, that I totally forgot about :neutral: . Jsg is a straight shooter and pointed out obvious things that I kinda ignored/overlooked and those were some quality feedbacks :) The same goes to you bud! Thank you for those feedbacks and tweaks suggestions. I have implemented some and will implement the rest after further evaluation :)

    A quote that I live by "Customers might not always be right, but that doesn't mean they are wrong" :smile:

    poisson said: Looks like I have to add Nexus Bytes to my provider list to try. Thanks @jsg

    One of these Thursdays, you shall be mine <3

    I will keep on trying my best to provide quality service for a price point, that is actually logical and earn your loyalty and love :)

    Thanked by 1vimalware
  • @seriesn said:
    One of these Thursdays, you shall be mine <3

    I will keep on trying my best to provide quality service for a price point, that is actually logical and earn your loyalty and love :)

    I am thinking about consolidating my collection of servers for better management. I haven't fully decided, but the customer service and product looks really good so far. The benchmarks so far show pretty good speeds to Asia where I am based! Will strongly consider :)

  • I've got nothing against @seriesn but when providers approach members with an offer to benchmark,

    That being said, bad performance on a good server is still better then good performance on bad servers.

    As for @jsp, tl:dr.

    I did see the benchmark results, and they look very good. Hope you can add memory speed test as well.

    Thanked by 1seriesn
  • @somik said:
    I've got nothing against @seriesn but when providers approach members with an offer to benchmark,

    Haha, from an outside perspective, totally agree with you!

    Now in my defense, I did ask them for an honest and unbiased review and offered 0 incentives in return and I was totally ready for a public shaming in the event things wouldn’t go smoothly. So there’s that :)

  • @poisson said:

    @seriesn said:
    One of these Thursdays, you shall be mine <3

    I will keep on trying my best to provide quality service for a price point, that is actually logical and earn your loyalty and love :)

    I am thinking about consolidating my collection of servers for better management. I haven't fully decided, but the customer service and product looks really good so far. The benchmarks so far show pretty good speeds to Asia where I am based! Will strongly consider :)


  • A fun and actively posting host is a compelling proposition all things equal. Let me work out a consolidation plan @seriesn.

    I consider the price (mega Thursday annual pricing) and specs of this machine @jsg reviewed as the ideal reference lowend box that satisfies the use cases of 90 percent of LETers with sufficient power to take on the occasional demanding task without breaking the bank. Hoping your future offers will be similarly good, if not better!

  • vyas11vyas11 Member
    edited November 2019

    @poisson said:
    A fun and actively posting host is a compelling proposition all things equal. Let me work out a consolidation plan

    Would you consider posting your pre and post consolidation setup once ready? I am sure many (self included) would be in similar situation at some point or another.

    Thanked by 1ITLabs
  • poisson said: Hoping your future offers will be similarly good, if not better!

    I try to keep every MegaDealsThursday unique and exciting. Now, are those Ok deals? Good deals? Or Better deals? That is a question that I can not answer.

  • looks good ..

  • jsgjsg Member, Resident Benchmarker
    edited November 2019

    @somik said:
    I've got nothing against @seriesn but when providers approach members with an offer to benchmark,

    Sorry, no, and I mean a clear and loud No.

    I had quite a few requests by providers (incl. well known ones, btw) for benchmarking and reviewing one or more of their products. Some of those requests are private that is, they do not want me to publish the results. One not atypical case is when ownership or management changes it seems or when they create a new product line. Getting some experienced people from the outside to test their systems is not at all a bad way to get a clearer picture.

    Similarly for a new/young provider to approached some experienced outsider to do a benchmark and/or a review is a smart thing to do.
    Reasons: For a start I do not benchmark anyone who doesn't run away fast enough. When I'm asked and don't know the provider I have a look at his website and his products. So, my accepting or declining is already kind of a first test. In this case here the provider had a product line with (IMO) some serious flaws and I told him so and also explained the flaws. At that point 8 out of 10 young providers would have argued and be p_ssed off - but not seriesn; he listened carefully, said something like "let me think over your points" ... and "vanished" for a while. When he came back it was with a much better product line (for LET. That's what I care about). In fact, he was so concentrated on fine tuning and polishing his Product line that he simply forgot me. So I sent him a friendly reminder about his review request ... and here we are. In other words: seriesn earned having his products reviewed and after I saw his promo I smilingly obliged.

    I did see the benchmark results, and they look very good. Hope you can add memory speed test as well.

    Maybe in a new version. or maybe not. Reason: people have to learn that memory speed per se is largely worthless. For a start, looking at the bus speed is simply mindless because memory is always faster than the PCIe devices. What's important is the Ram-CPU speed - and to dumbly measure that is mindless too because nobody just pulls and pushes bytes fro/to the memory. Nope. Look at crypto, look at a DB, look at ... it's always about doing something -and- pulling and/or pulling fro/to memory.
    I have damn good reasons to preach that CPU and RAM should be considered as basically 1 unit - and a complex one with many screws for fine-tuning. Example: The speed of your memory is utterly insignificant if one can't feed byte-slices, say 64 or 256 Bytes at a time fast enough to and from the L1 and the ALU. I've never seen a system that was poor in AES due to slow memory. The reason is basically always a poor cache design.

    That's why I throw those "two" tests into one and why I use a real-world and actually used algorithm. I want to see real world data, I want to see how a given system performs doing real work rather than some synthetic BS routine to squeeze good results out.
    Btw, that's also the reason why I use actual and real random in my random reading and writing across the disk.

  • jsgjsg Member, Resident Benchmarker

    @poisson said:
    of this machine @jsg reviewed as the ideal reference lowend box that satisfies the use cases of 90 percent of LETers with sufficient power to take on the occasional demanding task without breaking the bank.

    Yes, correct. In fact I'd say that the small 2 vCore VPS from the current promo would have most 4 vCore VPS from other providers in the rear view mirror.

    The point about the Ryzen systems is not performance. I'm quite sure that intel has something in the making that will beat even the current Zen systems in multiple areas.
    The point about Ryzen and other Zen systems is AFFORDABILITY. Zen makes it possible for providers to offer high-end performance at an attractive price.

  • jsgjsg Member, Resident Benchmarker
    edited November 2019

    ** --- UPDATE --- US VPS --- **

    NexusBytes kept their promise and provided me with access to their US VPS (2 vCores, 2 GB mem., 20 GB disk).

    Note: Due to Vanillas stupidity and my own lousy HTML formatting skills I changed my presentation slightly. The (probably most important for most readers) is now separated by a '-' from the detail data.

    System Info:

    Machine: amd64, Arch.: amd64, Model: AMD Ryzen 7 3700X 8-Core Processor
    OS, version: FreeBSD 12.0, Mem.: 1.985 GB
    CPU - Cores: 2, Family/Model/Stepping: 23/113/0
    Cache: 32K/32K L1d/L1i, 512K L2, 32M L3
    Std. Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat
              pse36 cflsh mmx fxsr sse sse2 sse3 pclmulqdq ssse3 fma cx16 sse4_1
              sse4_2 popcnt aes xsave osxsave avx f16c rdrnd hypervisor
    Ext. Flags: fsgsbase bmi1 avx2 smep bmi2 syscall nx mmxext fxsr_opt pdpe1gb
              rdtscp lm lahf_lm cmp_legacy cr8_legacy lzcnt sse4a misalignsse
              3dnowprefetch osvw
    

    Not much to say here. we already know their Ryzen system from above.

    Processor/memory:

                            Avg - Low High Dlow Dhigh Span
    single core:    475.84  -  396.13   482.55  16.75   1.41    18.16
    multi core: 1136.97 -  1080.00  1190.00 5.01    4.66    9.67
    

    Not really much to say either. It's just the juicy goodness of a well equipped Ryzen 3. Very tasty, love it although frankly, most users won't need that kind of speed.

    Disk:

    Seq.Wr: 1097.91 -  731.92   1200.00 33.33   9.30    42.63
    Rnd.Wr: 550.74  -  299.91   595.51  45.54   8.13    53.67
    Seq.Rd: 1894.70 -  1132.00  2054.00 40.25   8.41    48.66
    Rnd.Rd: 1887.30 -  1063.00  1981.00 43.68   4.96    48.64
    

    Very nice indeed. That NVMe based disk is a really speedy beast.

    Network:

    GR_UNK: 36.00   -  33.60    38.70   6.66    7.51    14.17
    US_SJC: 62.91   -  57.40    75.80   8.75    20.50   29.25
    OK_LON: 56.05   -  9.40 62.20   83.23   10.96   94.19
    DE_FRA: 51.86   -  51.10    52.80   1.47    1.81    3.28
    BR_SAO: 38.41   -  26.30    40.30   31.53   4.91    36.45
    FR_PAR: 56.41   -  54.00    61.70   4.27    9.38    13.65
    US_WDC: 198.13  -  7.90 455.60  96.01   129.95  225.96
    RO_BUC: 41.20   -  39.60    44.40   3.89    7.76    11.65
    AU_MEL: 23.15   -  21.50    24.70   7.11    6.72    13.83
    IT_MIL: 45.10   -  37.90    47.00   15.96   4.21    20.18
    SG_SGP: 22.06   -  7.90 23.80   64.18   7.90    72.08
    US_DAL: 97.40   -  91.70    102.90  5.85    5.65    11.50
    RU_MOS: 43.75   -  1.70 47.90   96.11   9.50    105.61
    JP_TOK: 30.35   -  28.90    32.20   4.79    6.08    10.87
    NO_OSL: 41.99   -  39.50    45.60   5.93    8.59    14.53
    IN_CHN: 21.75   -  18.60    24.00   14.50   10.32   24.82
    

    Quite nice. Even the speed across the whole country is OK at 63 Mb/s. What I find more impressive and interesting though is that the speed is quite OK across all locations incl. Ozzieland, Chennai, and Tokyo. About 50 to 55 Mb/s to the big 3 in Europe isn't exactly a dream come true from a european perspective but hey, that's quite close to the across the USA speed. All together quite nice.About the only thing not to like is a the high span at some locations.

    Pardon me if I sound a bit bored but that's what I am, or more precisely spoiled. The only reason I don't put a fat "recommended!" under my review is that I'm not very familiar with the US market so I can't fairly judge the networking results. But if you like them, go ahead! The processor, memory, and disk speed are at the high end of impressive.

  • If not for latency issues because I am in Asia, I am tempted to RDP and use my box as a cloud desktop because that box performs better than my own physical laptop. Running my data science tools will be a blast.

    Thanked by 2seriesn vimalware
  • @poisson said:
    If not for latency issues because I am in Asia, I am tempted to RDP and use my box as a cloud desktop because that box performs better than my own physical laptop. Running my data science tools will be a blast.

    Sir,
    Vision 2020 should address certain portion of that :)

  • @seriesn said:

    @poisson said:
    If not for latency issues because I am in Asia, I am tempted to RDP and use my box as a cloud desktop because that box performs better than my own physical laptop. Running my data science tools will be a blast.

    Sir,
    Vision 2020 should address certain portion of that :)

    Will start saving up for it. China optimized network (no need expensive ones) will be superb. I can run powerful analysis on dirt cheap netbook with this set up without buying my own Ryzens. :)

    Thanked by 1seriesn
  • sed s/Ryzen 3/Ryzen 7/

    Zen 3. Not criticising, I think everyone understood you meant Zen 3 (or Ryzen 7).

  • great specs, great provider

    Thanked by 2poisson seriesn
  • @tgl said:
    great specs, great provider

    I see you have an addiction to his potassium

    Thanked by 2dahartigan seriesn
  • @poisson said:

    @tgl said:
    great specs, great provider

    I see you have an addiction to his potassium

    Potassium addition can't be cured, only treated with potassium cigarettes and marijuana needles.

Sign In or Register to comment.