manton
manton

Haven’t actively used EC2 in a while, but wanted to check out the pricing after this week’s announcement that Amazon is making their own CPUs. On first glance, the “a1” instances don’t look cheaper. Maybe it’s not right to compare them directly to comparable “t3” instances?

|
Embed
Progress spinner
oyam
oyam

@manton I wouldn't expect the on-demand price to be lower, but for specific types of loads, you'll save because you'll process your loads faster.

|
Embed
Progress spinner
manton
manton

@oyam Got it, thanks!

|
Embed
Progress spinner
cleverdevil
cleverdevil

@manton I was expecting lower pricing as well. Early benchmarks do show good performance for the price for specific types of work, but it’s clear that this is a 1.0. That said, I’m shorting Intel as it’s pretty clear that AWS will be able to replicate what Apple has done for mobile chips in the server space.

|
Embed
Progress spinner
In reply to
cleverdevil
cleverdevil

@manton did you see Amazon Personalize? I was thinking it might make a nice alternative recommendation engine for Micro.blog or even Microsub servers!

|
Embed
Progress spinner
manton
manton

@cleverdevil Interesting, I hadn't seen that!

|
Embed
Progress spinner
msroest
msroest

@manton the other thing to consider is a1 instances aren't burstable like a t3 is. So although they're not cheaper then a t3 you get the full CPU available 100% so if you're CPU bound the a1's might be cheaper then a m/c instance. If your workload is possible on the arm chip.

|
Embed
Progress spinner
sharding
sharding

@manton The smallest a1 is a medium, so I assume they mean pricing vs. other medium instances. However, without seeing benchmarks it's hard to know what to really compare against (and on basic specs it doesn't look comparable to other mediums).

|
Embed
Progress spinner