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.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
VPS with High Speed CPU from local provider in Singapore
Looking for an SGP VDS. Key requirement is that it has to be a local provider/ASN, have high-speed core and preferably dedicated CPU core.
DETAILS BELOW:
VZ Type: Prefer VDS-type architecture with dedicated core
Number of Cores: 1 (dedicated preferred)
Core type: Has to be high-speed (Like Xeon E3-1230, preferably above 3 GHz)
RAM: 2-4
Disk Space: 25-35 GB
Disk Type: ANY, SSD preferred
Bandwidth: 500 GB/mo (1 TB preferred)
Port Speed: 10 Mbps min
DDoS Protection: Good to have, not necessary
Number of IPs: 1
Location: SG
Budget: US$20-$50/mo
Billing period: Monthly
Provider/ASN: Must be from Asia (Linode, Vultr, AWS, OVH etc won't do)
Comments
@elos42 - @seriesn from NexusBytes got you covered in Japan and Singapore.
I don't exactly get this point. Big companies might have multiple ASN, do you want the company to be registered in Singapore, or just the ASN must be registered in Singapore?
Take OVH for example, their company are registered in Singapore under OVH SINGAPORE PTE LTD
there are some issues with 'global' networks like OVH and Linode with regard to my particular application in terms of routing. I guess it has to do with how these guys advertise the IP/routes. Anyway, it's too taxing for me to have to deal with such things, and therefore an SG local player is probably better. I don't mind paying more. Otherwise I would have gone with Vultr, and it would cost only 12 dollars.
Ideally, I prefer an Asia-only provider.. like GigsGigs, for example.
There are also some global players who give IPs belonging to local networks. That will also do.
I've PMed you, let me know
apc.sg
Ping me if interested
We are a registered business from Singapore, but servers are hosted in USA
We even have local Singapore IP but they are being used in US e.g. 103.114.160.*
If it works for you, let me know
thanks
You’re looking for companies like USONYX, SparkStation etc..?
if you looking for high speed core like AMD EPYC 7302P and nvme, try maxKVM
I don't know if IndoVirtue is still providing a service or not, maybe @melita can help you with this.
They are using SG.GS on their Singapore VPS
https://indovirtue.com/premium-singapore-ssd-vps
Thanks for the mention boss!!
@elos42 while not locally registered, geoip seems to be on point with us
lgsg.nexusbytes.com
Did someone mention NexusBytes?
try lookup streamline-servers
You have any experience with their SG VPS Range?
sure, I using the Singapore location for almost 2 years and its good performance, recently they upgraded the hardware. I tried the Perth location too which good as the SG location
Thanks for the info!
You can find a host which runs VPS on AMD Ryzen 9 5900 which has capacity of 4.5Ghz per core
Unless I am reading wrong, that's not base but turbo.
At that point, not really a major upper hand over 3900x
Yes, this is with overclocking.
How many VPs providers are running their system on over clocked nodes? Sounds like a nightmare with power bill and hardware lifetime.
I don't think that any VPS provider runs his hardware on overclocking as that will reduce the life of the processor and thus lower the profit.
Wouldn’t that nullify your above statement then?
Only some providers might be doing this. Many must not be. I had been using Contabo before and they used to allow me to overclock their dedicated servers but they used to do it on their own.
For whatever reason your logo keep reminds me of Treudler...
I am not an expert on virtualization, but wouldn't most customers be consuming only like 5-15% of their available CPU cycles, even if it's a dedicated core? So, even if you have overclocked it, wouldn't it run at much lower frequencies most of the time, spiking to turbo when there's demand for it? I am just talking about my own use case, as I consume only 10-20% of the CPU on a long-term basis, but I've noticed that I cannot shift to older CPUs without a marked degradation in service. I've always been puzzled about it. Why can't I move to a slower CPU if anyway, my CPU utilization never goes above 20% on a weekly average or even a daily average basis? Perhaps newer/faster hardware has some other features (network-related or memory related?) that make processing faster even when only 10% of the CPU is engaged? In other words, perhaps 10% of a 4GHz, high-end CPU is still more efficient/faster than 20% of an E2620?
Or could it be that there are very minute spikes in CPU usage that don't show up in the provider's daily or hourly CPU utilization graphs, and these are making all the difference?
Well if you are a car person or someone who dwindles in the M.E world, think of it as horsepower.
Less horsepower will still do the job, but slower. Thus requiring more time to go from point A to point B .
Shouldn't 20% effort from 10 horses be equal to 40% effort from 5 horses is what I was wondering about. {Assume that these horses have the ability to go at 80% consistently if required}
That might be true if your application is adaptable to using multiple cores efficiently. (be 5 or 10)
I think yes.
Clock speed is not everything. The instruction set supported by the CPU and your compiler makes a huge difference.
https://gcc.gnu.org/onlinedocs/gcc/x86-Options.html
An algorithm example:
I have a program that expects to receive 64 packets. Each incoming packet has a sequence number between 0 and 63. Due to the unreliable nature of the network, it's possible to receive the same packet more than once. I want to know how many unique packets have been received within 5 seconds.
Naively, I could make a
std::vector<bool>to record whether each packet is received. When a packet arrives, I turn the corresponding element to true. At the end of 5 seconds, I count the number oftrues in the container.Since there are only 64 booleans, I can optimize the memory usage by storing in a
uint64_t. It's initially zero. When a packet arrives, I turn the corresponding bit to 1. At the end of 5 seconds, I count the number of 1-bits in the variable.In an Intel Core2 CPU, the counting requires a loop that takes at least 64 clock cycles.
In an Intel Nehalem or newer CPU, there's a new CPU instruction called POPCNT, and it can do this counting in one clock cycle. Here's my real code using this instruction.
If a program is just doing this counting work, an Intel Nehalem with 10% clock frequency would still run faster than the Core2.
Same goes for every new instruction set introduced in newer generations of CPUs. Most prominently, AES-NI makes TLS crypto runs faster.
In conclusion, your apps run faster on newer generation of CPU, but not necessarily require faster clock speed.