GeekBench Results for Cloud Servers
-
Time for some new benchmarks from some different services.
-
Vultr, 1 vCPU, 2GB RAM.
Single CPU: 3847
Multi-Core: 3694 -
Digital Ocean, 1 vCPU, 2GB RAM $10
Single CPU: 2800
Multi-Core: 2718 -
Linode. 1 vCPU, 2GB RAM. $10
Single CPU: 2792
Multi-Core: 2629 -
Vultr has been pulling ahead in both price and performance since the last time that we did a test like this.
-
@scottalanmiller That's a dubious test at best. The only thing you are really testing is how much resources you had available on your specific instance and at that specific time. It says nothing about how much resources you are going to have next week or on another instance. Unless you were running thousands of test over a period of time (did you?).
And it only tests the CPU and not disk or network speed or latency.
BTW, the CPU in your Vultr test is Intel Xeon W-2175 CPU @ 2.50GHz, from Intel's workstation CPUs.
If we are going to draw any conclusions, it would be that you on your Vultr test had about 75% of the CPU resources on that core allocated to you at that time.
-
@pete-s said in GeekBench Results for Cloud Servers:
BTW, the CPU in your Vultr test is Intel Xeon W-2175 CPU @ 2.50GHz, from Intel's workstation CPUs.
But still, a lot more performance.
-
@pete-s said in GeekBench Results for Cloud Servers:
@scottalanmiller That's a dubious test at best. The only thing you are really testing is how much resources you had available on your specific instance and at that specific time. It says nothing about how much resources you are going to have next week or on another instance. Unless you were running thousands of test over a period of time (did you?).
While that's true, it makes it excessively hard to collect statistics. It's true, any of these might have worse performance at some other time. But that is always the case. Even if we did test over a long period of time, then we'd worry that it applied only to this instance and not to others.
But the gap here is enourmous. Far larger than we'd expect from noise neighbours, especially considering that Vultr is much faster with different hardware while the other two are nearly identical, with nearly identical hardware.
-
Basically, this is how we can get performance numbers. Nothing from the vendors themselves is useful. And collecting on production boxes isn't useful. With any shared system, we can only hope for decently clean numbers and then extrapolate over provisioning.
-
@scottalanmiller said in GeekBench Results for Cloud Servers:
@pete-s said in GeekBench Results for Cloud Servers:
BTW, the CPU in your Vultr test is Intel Xeon W-2175 CPU @ 2.50GHz, from Intel's workstation CPUs.
But still, a lot more performance.
Yes, this time and that's good, but you need a lot more than one sample to draw any meaningful conclusions, except that Vultr uses workstation hardware, at least on one machine.
-
@scottalanmiller said in GeekBench Results for Cloud Servers:
Basically, this is how we can get performance numbers. Nothing from the vendors themselves is useful. And collecting on production boxes isn't useful. With any shared system, we can only hope for decently clean numbers and then extrapolate over provisioning.
Most VMs are actually doing something, like running a webserver.
Would it be better to actually monitor things like response time, over time to actually get some metric that has some use in real life?
Aren't there any services that can do this, similar to Alexa for uptime for instance?
-
@pete-s said in GeekBench Results for Cloud Servers:
BTW, the CPU in your Vultr test is Intel Xeon W-2175 CPU @ 2.50GHz, from Intel's workstation CPUs.
How did you find this? I can't see where it is referenced.
-
@pete-s said in GeekBench Results for Cloud Servers:
@scottalanmiller said in GeekBench Results for Cloud Servers:
Basically, this is how we can get performance numbers. Nothing from the vendors themselves is useful. And collecting on production boxes isn't useful. With any shared system, we can only hope for decently clean numbers and then extrapolate over provisioning.
Most VMs are actually doing something, like running a webserver.
Would it be better to actually monitor things like response time, over time to actually get some metric that has some use in real life?
Aren't there any services that can do this, similar to Alexa for uptime for instance?
Depends, if you are looking for something standard, like a web server, AND you lock it down to make sure nothing else is using it, then yes, there are services that you can pay for that. But they will measure many things other than the cloud performance itself, so you will be measuring the whole situation and make it subjective, rather than objective.
For example, if I measure with a service that measures from Manhattan, a server in Manhattan will have a specific advantage over one in New Jersey. But my end users might be anywhere. If I measure from my house, that's useful for me, but not necessarily for my users. It can lead to bad results.
It's a useful measurement, and good services measure from all over the world. But they measure the resulting performance of your cloud, application, configuration, and all of the networking in between. Good for certain things, but not good for what we want to know here - which is which cloud provider gives us the best starting point to build upon.
-
@pete-s said in GeekBench Results for Cloud Servers:
.... except that Vultr uses workstation hardware, at least on one machine.
Intel classifies that model as Server...
https://ark.intel.com/products/130046/Intel-Xeon-W-2175-Processor-19_25M-Cache-2_50-GHz
-
I think the key behind Xeon W is that they are single socket Xeons, which actually is often better for cloud computing, as you get better cost and performance for the vast majority of cloud workloads. People looking for more capacity than a single large Xeon can provide are almost never going to cloud for that, as it is too costly. And NUMA is a killer for cloud workloads. So going with higher performance, single socket systems is very logical.
-
@scottalanmiller said in GeekBench Results for Cloud Servers:
@pete-s said in GeekBench Results for Cloud Servers:
.... except that Vultr uses workstation hardware, at least on one machine.
Intel classifies that model as Server...
https://ark.intel.com/products/130046/Intel-Xeon-W-2175-Processor-19_25M-Cache-2_50-GHz
No, it's W-series and W stands for Workstation:
New Intel Xeon W Processors Deliver Optimized Performance for Mainstream Workstation Professionals
The new Intel Xeon W processors are based on the Intel Xeon Scalable processor microarchitecture, but designed into a cost-optimized1 1-socket form factor specifically for professional workstations.It has the LGA2066 socket that it shares with Intel's high-end desktop CPUs but the Xeon W-series have ECC support. But who cares as long as it gets the job done for the right price. As you said it might be the best cost performance ratio for cloud deployment.
From your benchmarks you have a little more than 35% higher performance on Vultr. The other providers are still on the older E5-2600 generation of CPUs. When your instance shows up on a machine with the latest generation Xeon the performance will be the same. Then it will be a matter of uncontrollable parameters like how the provider allocate resources.
Unless one opts for dedicated instances. I know Vultr has them.
-
@pete-s said in GeekBench Results for Cloud Servers:
@scottalanmiller said in GeekBench Results for Cloud Servers:
@pete-s said in GeekBench Results for Cloud Servers:
.... except that Vultr uses workstation hardware, at least on one machine.
Intel classifies that model as Server...
https://ark.intel.com/products/130046/Intel-Xeon-W-2175-Processor-19_25M-Cache-2_50-GHz
No, it's W-series and W stands for Workstation:
New Intel Xeon W Processors Deliver Optimized Performance for Mainstream Workstation Professionals
It says that on the group page, but look at the specific page for that processor. Says server right on it, and on its parent page.
-
@pete-s said in GeekBench Results for Cloud Servers:
From your benchmarks you have a little more than 35% higher performance on Vultr. The other providers are still on the older E5-2600 generation of CPUs. When your instance shows up on a machine with the latest generation Xeon the performance will be the same. Then it will be a matter of uncontrollable parameters like how the provider allocate resources.
35% is a massive number in CPU Performance. That the others run older generations CPUs is a big deal, and something that Vultr advertises specifically. It's not random what you get, you always get Skylake on Vultr for that instance. You don't have to select it or get lucky. And most providers state that you won't get those versions without paying more.
-
@scottalanmiller said in GeekBench Results for Cloud Servers:
@pete-s said in GeekBench Results for Cloud Servers:
@scottalanmiller said in GeekBench Results for Cloud Servers:
@pete-s said in GeekBench Results for Cloud Servers:
.... except that Vultr uses workstation hardware, at least on one machine.
Intel classifies that model as Server...
https://ark.intel.com/products/130046/Intel-Xeon-W-2175-Processor-19_25M-Cache-2_50-GHz
No, it's W-series and W stands for Workstation:
New Intel Xeon W Processors Deliver Optimized Performance for Mainstream Workstation Professionals
It says that on the group page, but look at the specific page for that processor. Says server right on it, and on its parent page.
Yes, but that's the vertical segment. There's no segment for workstation, hence the reason it doesn't say workstation. Intel only have a few segments: desktop, embedded, mobile and server.
But have a look at Dell or HPE. You'll find it next to impossible to find a rackserver with a workstation CPU like W-2175. But you'll find it in plenty of workstations (Dell Precision, HP Z). That should tell you something.
PS. Pricing is not a lot different between the workstation and server CPUs.
Comparable to the 14 core W-2175 @ 2.5GHz is 14 core Intel Xeon Gold 6132 @ 2.6 GHz.
Rec. price: $1947 versus $2111. -
@pete-s said in GeekBench Results for Cloud Servers:
@scottalanmiller said in GeekBench Results for Cloud Servers:
@pete-s said in GeekBench Results for Cloud Servers:
@scottalanmiller said in GeekBench Results for Cloud Servers:
@pete-s said in GeekBench Results for Cloud Servers:
.... except that Vultr uses workstation hardware, at least on one machine.
Intel classifies that model as Server...
https://ark.intel.com/products/130046/Intel-Xeon-W-2175-Processor-19_25M-Cache-2_50-GHz
No, it's W-series and W stands for Workstation:
New Intel Xeon W Processors Deliver Optimized Performance for Mainstream Workstation Professionals
It says that on the group page, but look at the specific page for that processor. Says server right on it, and on its parent page.
Yes, but that's the vertical segment. There's no segment for workstation, hence the reason it doesn't say workstation. Intel only have a few segments: desktop, embedded, mobile and server.
But have a look at Dell or HPE. You'll find it next to impossible to find a rackserver with a workstation CPU like W-2175. But you'll find it in plenty of workstations (Dell Precision, HP Z). That should tell you something.
PS. Pricing is not a lot different between the workstation and server CPUs.
Comparable to the 14 core W-2175 @ 2.5GHz is 14 core Intel Xeon Gold 6132 @ 2.6 GHz.
Rec. price: $1947 versus $2111.But Dell and HPE don't focus on single CPU servers, which has long been something that they lacked. Their models are long build around the higher cost dual CPU configurations. Which would explain that more than anything.