Non-existent Epyc Tokyo pre-order, still haven't decided if it's a good or bad idea. Leaning toward bad, might just send it first but I wanted to gauge the level of interest somehow so we know if we have to build more and to get all the orders pre-screened.
Seeing so many upgrades and cheaper Epyc Tokyo plans, I feel like it was a bad decision to buy a two-year Tokyo plan last year.
Now I just want to cry out loud.
I'll have to wait until 2024 to buy a new Epyc Tokyo.
Non-existent Epyc Tokyo pre-order, still haven't decided if it's a good or bad idea. Leaning toward bad, might just send it first but I wanted to gauge the level of interest somehow so we know if we have to build more and to get all the orders pre-screened.
Seeing so many upgrades and cheaper Epyc Tokyo plans, I feel like it was a bad decision to buy a two-year Tokyo plan last year.
Now I just want to cry out loud.
I'll have to wait until 2024 to buy a new Epyc Tokyo.
Not a problem for me, whenever Virmach releases an offer, just go for it blindly.
@tulipyun said: Seeing so many upgrades and cheaper Epyc Tokyo plans, I feel like it was a bad decision to buy a two-year Tokyo plan last year.
Now I just want to cry out loud.
I'll have to wait until 2024 to buy a new Epyc Tokyo.
If we can figure out a good way to allow people to upgrade/downgrade to these plans and migrate to Epyc, we'll offer that. Keep in mind though that there's obviously both positives and negatives, for example 384MB RAM, 1TB Bandwidth for $17.76 versus 768MB, 500GB Bandwidth for $15. More RAM, less bandwidth.
Epyc will also use enterprise NVMe SSD. Higher capacity, so lower I/O per customer.
What do you think the pricing should be instead for it to feel more balanced versus your offer?
@tulipyun said: Seeing so many upgrades and cheaper Epyc Tokyo plans, I feel like it was a bad decision to buy a two-year Tokyo plan last year.
Now I just want to cry out loud.
I'll have to wait until 2024 to buy a new Epyc Tokyo.
If we can figure out a good way to allow people to upgrade/downgrade to these plans and migrate to Epyc, we'll offer that. Keep in mind though that there's obviously both positives and negatives, for example 384MB RAM, 1TB Bandwidth for $17.76 versus 768MB, 500GB Bandwidth for $15. More RAM, less bandwidth.
Epyc will also use enterprise NVMe SSD. Higher capacity, so lower I/O per customer.
What do you think the pricing should be instead for it to feel more balanced versus your offer?
@tulipyun said: Seeing so many upgrades and cheaper Epyc Tokyo plans, I feel like it was a bad decision to buy a two-year Tokyo plan last year.
Now I just want to cry out loud.
I'll have to wait until 2024 to buy a new Epyc Tokyo.
If we can figure out a good way to allow people to upgrade/downgrade to these plans and migrate to Epyc, we'll offer that. Keep in mind though that there's obviously both positives and negatives, for example 384MB RAM, 1TB Bandwidth for $17.76 versus 768MB, 500GB Bandwidth for $15. More RAM, less bandwidth.
Epyc will also use enterprise NVMe SSD. Higher capacity, so lower I/O per customer.
What do you think the pricing should be instead for it to feel more balanced versus your offer?
Since there is no configuration difference between buying one and two years of last year's Tokyo plan: if you bought one year last year, you can switch to Epyc exactly this year; if you bought two years last year, you will have to wait until 2024 or pay the extra purchase cost. This seems to work against those who bought longer last year, and aren't good business plans supposed to encourage people to buy longer services?
Of course I don't think it's necessary to specifically go back and make changes to services that were already created last year, but I would appreciate it if new business plans were planned with consideration of how to help users switch to the new plan more efficiently.
For me buying a new Epyc in 2024 would be the easiest way to go.
@tulipyun - If I understand you correctly what you are in effect saying is you want VirMach to steadily increase prices from year to year, or else reduce prices for previous contract when lower prices are offered. If we are voting on how VirMach runs things, I vote against this.
@FrankZ said: @tulipyun - If I understand you correctly what you are in effect saying is you want VirMach to steadily increase prices from year to year, or else reduce prices for previous contract when lower prices are offered. If we are voting on how VirMach runs things, I vote against this.
I'm not complaining about the price, I just hope the new plan doesn't make users feel like they're stuck with the old one
Non-existent Epyc Tokyo pre-order, still haven't decided if it's a good or bad idea. Leaning toward bad, might just send it first but I wanted to gauge the level of interest somehow so we know if we have to build more and to get all the orders pre-screened.
This looks good, but I'm a new user, do I have to pay the extra $10? It doesn't look like a good deal.
Non-existent Epyc Tokyo pre-order, still haven't decided if it's a good or bad idea. Leaning toward bad, might just send it first but I wanted to gauge the level of interest somehow so we know if we have to build more and to get all the orders pre-screened.
@VirMach said:
If we can figure out a good way to allow people to upgrade/downgrade to these plans and migrate to Epyc, we'll offer that. Keep in mind though that there's obviously both positives and negatives, for example 384MB RAM, 1TB Bandwidth for $17.76 versus 768MB, 500GB Bandwidth for $15. More RAM, less bandwidth.
Epyc will also use enterprise NVMe SSD. Higher capacity, so lower I/O per customer.
What do you think the pricing should be instead for it to feel more balanced versus your offer?
Epyc and Enterprise NVMe SSD, do they mean better stability?
If they are not in Tokyo but in LAX, I will migrate a VPS with a lot of RAM and I don't know if you guys will welcome it.
Although China to Tokyo, Japan has better internet speed, it is not stable.
If I can only migrate to Tokyo, I don't think it should be more than $15. Maybe some bandwidth can be deducted.
@Caxen said: Epyc and Enterprise NVMe SSD, do they mean better stability?
Nan, Ryzen has less oversell potential due to its physical limitation of 4 RAM slots. Hope EPYC does not end up as the shitshow of ColoCrossing E5 old days with high CPU steal, poor IO & extremely oversell
I would get multiple Tokyo EPYC which serve as proxy purpose.
@Caxen said: Epyc and Enterprise NVMe SSD, do they mean better stability?
Depends. But that's the idea.
Enterprise NVMe SSDs that match consumer speeds on certain operations are insanely expensive. We'd be going with more entry level stuff, which means slower but better capabilities when it comes to simultaneous real-world operations. I also assume they're less likely to "drop off" as a result and since they're not going to be using M.2 slots on ASRock boards which we've definitely had problems with there's added stability there.
We also are going with RAID1 and RAID10 on them, and higher capacity.
Higher capacity means higher density. Which means more users sharing the same total maximum operations but in the end it should at least meet or exceed SATA SSDs in the past and same I/O fair share policy.
If they are not in Tokyo but in LAX, I will migrate a VPS with a lot of RAM and I don't know if you guys will welcome it.
We're planning one to start in Tokyo, LAX, NYC, Amsterdam, and Frankfurt (primary locations.) Tokyo will most likely be first as it is the lowest power variant that I've already built. Tokyo CPU usage is also lower than other locations so it's good as it will allow us to gauge the usage and make sure for other locations we go with more powerful variants to meet local needs.
Although China to Tokyo, Japan has better internet speed, it is not stable.
Mind telling me what location ends up being most stable and fastest for you? I've experimented a little bit with Hawaii, Sydney, and those both seem worse than Los Angeles. Then San Jose seems better but only sometimes if it gets carried over directly, although that's rare and it usually goes down to Los Angeles first. Hong Kong seems like it's the best but only for certain regions since China is a large country. I've looked into Russia, that seems be okay for northern cities but data is limited. Singapore and South Korea look like they primarily run to Hong Kong area first.
@Caxen said: Epyc and Enterprise NVMe SSD, do they mean better stability?
Nan, Ryzen has less oversell potential due to its physical limitation of 4 RAM slots. Hope EPYC does not end up as the shitshow of ColoCrossing E5 old days with high CPU steal, poor IO & extremely oversell
I would get multiple Tokyo EPYC which serve as proxy purpose.
Depends. You can argue there's lower oversell potential for memory on Ryzen, but what that ends up meaning is the CPU ends up being less oversold in relation to RAM. For Epyc, RAM technically ends up being actually less "oversold" since more physical memory can be added. But that leads to higher CPU oversell.
E5 days, RAM was basically not oversold at all. It's complicated so I say basically, because yes we would bottleneck at CPU. Poor IO was mostly from trash RAID controllers and subpar SSDs. We'd ask for 850 Evo minimum but they often would quietly get swapped out. For example if one failed, they'd say something like "do you want to wait 2 weeks for 850 Evo or we can put in something else right now?"
@Caxen said: Epyc and Enterprise NVMe SSD, do they mean better stability?
Depends. But that's the idea.
Enterprise NVMe SSDs that match consumer speeds on certain operations are insanely expensive. We'd be going with more entry level stuff, which means slower but better capabilities when it comes to simultaneous real-world operations. I also assume they're less likely to "drop off" as a result and since they're not going to be using M.2 slots on ASRock boards which we've definitely had problems with there's added stability there.
We also are going with RAID1 and RAID10 on them, and higher capacity.
Higher capacity means higher density. Which means more users sharing the same total maximum operations but in the end it should at least meet or exceed SATA SSDs in the past and same I/O fair share policy.
Each service is entitled to 80 IOPS.
NVMe can do 1000000 IOPS, but half is lost by RAID1 and let's assume another half is lost by virtualization.
Each NVMe can host 3125 services.
Each service has at least 10 GB storage capacity.
Typically enterprise NVMe has less than 16 TB capacity, so it can fit up to 1600 services.
@VirMach said:
Mind telling me what location ends up being most stable and fastest for you? I've experimented a little bit with Hawaii, Sydney, and those both seem worse than Los Angeles. Then San Jose seems better but only sometimes if it gets carried over directly, although that's rare and it usually goes down to Los Angeles first. Hong Kong seems like it's the best but only for certain regions since China is a large country. I've looked into Russia, that seems be okay for northern cities but data is limited. Singapore and South Korea look like they primarily run to Hong Kong area first.
When used as a VPN proxy, Japan is definitely the fastest, followed by San Jose, Los Angeles, and Seattle (CT and CU).
However, there may have been recent attacks on xTom or an increase in users, and packet loss on the Japan line has increased, with all the above regions experiencing around 10% packet loss at peak times (22:00, UTC+8).
Hong Kong and Singapore usually only have CM.
Europe usually only has CU.
Korea and Taiwan are also options, but they have fewer providers and more restrictions.
These are usually chosen if the VPS costs less than $10.
When used to deploy web services, CDNs usually point directly to the US West nodes (Cloudflare, Cloudfront, Gcore), and I think people with higher spec VPS usually choose this (if not used as Windows RDP).
@H2Kimi said:
I bought a 3C8G product, but why is the CPU Ryzen 3900X (crying
I just rushed to buy the 5950X and cried
Can I pay to upgrade my server to 5950X, I need higher CPU performance. Value for money is why I always trust Virmach @VirMach
They're basically the same, both score ~1000 on gb5. You don't get dedicated resources anyways, so it doesn't really matter. I have several 5950x VMs and several 3900x VMs and there has been no difference in performance between them.
@fluttershy said:
They're basically the same, both score ~1000 on gb5. You don't get dedicated resources anyways, so it doesn't really matter. I have several 5950x VMs and several 3900x VMs and there has been no difference in performance between them.
Comments
Seeing so many upgrades and cheaper Epyc Tokyo plans, I feel like it was a bad decision to buy a two-year Tokyo plan last year.
Now I just want to cry out loud.
I'll have to wait until 2024 to buy a new Epyc Tokyo.
Not a problem for me, whenever Virmach releases an offer, just go for it blindly.
It is cheaper than last year, but Tokyo is
Unavailable + $5.00 USD Setup Fee
this year...If we can figure out a good way to allow people to upgrade/downgrade to these plans and migrate to Epyc, we'll offer that. Keep in mind though that there's obviously both positives and negatives, for example 384MB RAM, 1TB Bandwidth for $17.76 versus 768MB, 500GB Bandwidth for $15. More RAM, less bandwidth.
Epyc will also use enterprise NVMe SSD. Higher capacity, so lower I/O per customer.
What do you think the pricing should be instead for it to feel more balanced versus your offer?
39.95EUR/yr
Since there is no configuration difference between buying one and two years of last year's Tokyo plan: if you bought one year last year, you can switch to Epyc exactly this year; if you bought two years last year, you will have to wait until 2024 or pay the extra purchase cost. This seems to work against those who bought longer last year, and aren't good business plans supposed to encourage people to buy longer services?
Of course I don't think it's necessary to specifically go back and make changes to services that were already created last year, but I would appreciate it if new business plans were planned with consideration of how to help users switch to the new plan more efficiently.
For me buying a new Epyc in 2024 would be the easiest way to go.
@tulipyun - If I understand you correctly what you are in effect saying is you want VirMach to steadily increase prices from year to year, or else reduce prices for previous contract when lower prices are offered. If we are voting on how VirMach runs things, I vote against this.
LES • About • Donate • Rules • Support
I'm not complaining about the price, I just hope the new plan doesn't make users feel like they're stuck with the old one
Actually it does, on an extremely leap year.
Feb 31 is equivalent to Mar 03 in Excel and many other applications.
Webhosting24 aff best VPS; ServerFactory aff best VDS; Cloudie best ASN; Huel aff best brotein.
Buy more... Cancel old one... Repeat
The circle of low end life
"I would have gotten away with it too, if it wasn't for that meddling Frankz and Mason!!"
Whaaaaaaaaaaaaaaaagh
Yo VirmAch boy, open the pre-order!
Yo, join our premium masochist club
Epyc and Enterprise NVMe SSD, do they mean better stability?
If they are not in Tokyo but in LAX, I will migrate a VPS with a lot of RAM and I don't know if you guys will welcome it.
Although China to Tokyo, Japan has better internet speed, it is not stable.
If I can only migrate to Tokyo, I don't think it should be more than $15. Maybe some bandwidth can be deducted.
from google translate
Nan, Ryzen has less oversell potential due to its physical limitation of 4 RAM slots. Hope EPYC does not end up as the shitshow of ColoCrossing E5 old days with high CPU steal, poor IO & extremely oversell
I would get multiple Tokyo EPYC which serve as proxy purpose.
Yo, join our premium masochist club
Depends. But that's the idea.
Enterprise NVMe SSDs that match consumer speeds on certain operations are insanely expensive. We'd be going with more entry level stuff, which means slower but better capabilities when it comes to simultaneous real-world operations. I also assume they're less likely to "drop off" as a result and since they're not going to be using M.2 slots on ASRock boards which we've definitely had problems with there's added stability there.
We also are going with RAID1 and RAID10 on them, and higher capacity.
Higher capacity means higher density. Which means more users sharing the same total maximum operations but in the end it should at least meet or exceed SATA SSDs in the past and same I/O fair share policy.
We're planning one to start in Tokyo, LAX, NYC, Amsterdam, and Frankfurt (primary locations.) Tokyo will most likely be first as it is the lowest power variant that I've already built. Tokyo CPU usage is also lower than other locations so it's good as it will allow us to gauge the usage and make sure for other locations we go with more powerful variants to meet local needs.
Mind telling me what location ends up being most stable and fastest for you? I've experimented a little bit with Hawaii, Sydney, and those both seem worse than Los Angeles. Then San Jose seems better but only sometimes if it gets carried over directly, although that's rare and it usually goes down to Los Angeles first. Hong Kong seems like it's the best but only for certain regions since China is a large country. I've looked into Russia, that seems be okay for northern cities but data is limited. Singapore and South Korea look like they primarily run to Hong Kong area first.
What about specific carriers?
Depends. You can argue there's lower oversell potential for memory on Ryzen, but what that ends up meaning is the CPU ends up being less oversold in relation to RAM. For Epyc, RAM technically ends up being actually less "oversold" since more physical memory can be added. But that leads to higher CPU oversell.
E5 days, RAM was basically not oversold at all. It's complicated so I say basically, because yes we would bottleneck at CPU. Poor IO was mostly from trash RAID controllers and subpar SSDs. We'd ask for 850 Evo minimum but they often would quietly get swapped out. For example if one failed, they'd say something like "do you want to wait 2 weeks for 850 Evo or we can put in something else right now?"
@VirMach A few of us have requested about when can we expect the provisioning of these
12 quantity for DALZ010 offer, 4GB RAM / 4 Core / 130GB Disk
It will be good to know a rough timeline. Not complaining but setting expectations. Original post lacked clarity.
Lack of clarity was there as I do not have any information on expected timeline.
Each service is entitled to 80 IOPS.
NVMe can do 1000000 IOPS, but half is lost by RAID1 and let's assume another half is lost by virtualization.
Each NVMe can host 3125 services.
Each service has at least 10 GB storage capacity.
Typically enterprise NVMe has less than 16 TB capacity, so it can fit up to 1600 services.
I don't see IOPS being bottleneck here.
Webhosting24 aff best VPS; ServerFactory aff best VDS; Cloudie best ASN; Huel aff best brotein.
Thank you for the update. If this is due to location/node capacity - is an alternate location possible ? If something else…then will wait
When used as a VPN proxy, Japan is definitely the fastest, followed by San Jose, Los Angeles, and Seattle (CT and CU).
However, there may have been recent attacks on xTom or an increase in users, and packet loss on the Japan line has increased, with all the above regions experiencing around 10% packet loss at peak times (22:00, UTC+8).
Hong Kong and Singapore usually only have CM.
Europe usually only has CU.
Korea and Taiwan are also options, but they have fewer providers and more restrictions.
These are usually chosen if the VPS costs less than $10.
When used to deploy web services, CDNs usually point directly to the US West nodes (Cloudflare, Cloudfront, Gcore), and I think people with higher spec VPS usually choose this (if not used as Windows RDP).
from google translate
Korea has some of the most expensive bandwidth in APAC, so generally not advised for cheap VMs.
Plus porn block in Korea, I would consider it is inferior compare to Japan DC.
Yo, join our premium masochist club
@VirMach sir, how is the ipv6 for Tokyo?
Invoice #1530719. Paid.
Waiting for server creation! Thanks
Nice offer
I bought a 3C8G product, but why is the CPU Ryzen 3900X (crying
I just rushed to buy the 5950X and cried
Can I pay to upgrade my server to 5950X, I need higher CPU performance. Value for money is why I always trust Virmach
@VirMach
They're basically the same, both score ~1000 on gb5. You don't get dedicated resources anyways, so it doesn't really matter. I have several 5950x VMs and several 3900x VMs and there has been no difference in performance between them.
Yeah, but 5950X is bigger than 3900x xD