I'm just thinking from a cost perspective... Most home labs are not built for paying thousands or hundreds on any sort of regular basis, from what I've seen
@Bob We have a space which can be used for server gear, but when the 50+ dBA from my old Dragon laptop under load is already an issue, I'm not sure this is even a possibility.
For my server, it's on an AMD e350 Mini-ITX motherboard with 4 hard drives. No noise except the PSU and HDDs. Running Linux, a DOS 7 VM (for old-school games), web sites, Samba AD, and mail
(and on my linux box, that's literally "I need it to run 24/7 quietly until it dies, not die too often, and cost under a hundred quid a year over its lifetime")
I have a fancy USB enclosure that will boot ISOs directly off a directory. I just downloaded a zip rather than an ISO, dropped it into a directory and... wondered why its not showing up
AMD's Vega 8 processor graphics outperforms Intel's UHD Graphics 620 by an almost comical margin, while Intel still has an edge on some CPU-bound workloads.
Application launches are also faster on the Intel processor, possibly due to their Speed Shift technology.
But that AMD is anywhere near competitive with Intel's very latest Kaby Lake-R processors in everyday use cases... shows how far they've come.
Well, that proves that making repeated, nonessential edits to messages that contain pings is still a Bad Idea™.
On to another topic...
One feature I miss in Corsair Utility Engine that ROCCAT Swarm has is the ability to change the RGB lighting of each key on events such as when another key is pressed.
While the essential benefit of per-key RGB lighting is still there, namely the ability to color-code each key, more complicated setups such as highlighting keyboard shorts when Ctrl is held down are not possible.
The RGB lighting is entirely software-controlled in Swarm, though, so if Swarm is not running, the keyboard just uses the default bluish color. On the other hand, the K95 RGB Platinum does have a limited ability to store RGB lighting effects in hardware, though I suspect conditional RGB effects are not likely to happen without software support.
This is something Corsair can definitely add in future versions of CUE.
@djsmiley2k iperf doesn't strictly require a server; there are featureful routers that support it in their router GUI
iperf, generally speaking, requires at least two computers capable of running a general-purpose operating system like Linux or BSD
there's no meaningful way to measure throughput with only a single computer because you're not sending the data to anywhere
it's like saying, I moved 3 meters. What was my speed? Well, without knowing the time it took me, you are lacking a dimension required to define a "speed".
there's no IS workload in the world that has to return that much data to the application tier because people in IS work with smaller data sets and page it out (like scrolling through a huge list of employees) - not viewing/processing the whole thing at once
I get slow-start borkedness all the time when doing Linux updates against Ubuntu: because you're downloading hundreds or thousands of files, each with its own separate HTTP Request, even though their server can in theory send data to me at near gigabit speeds, I end up downloading at like 300 KByte/s in actuality because the slow-start is constantly resetting
> TripAdvisor's top London restaurant is... a SHED: Prankster fools ratings site with fake web page, fake food and fake reviews - and even manages to 'hoodwink' top critic Jay Rayner
> Oobah Butler set up a website and got take photographs of the 'food' which in reality was of shaving foam, bleach and paint.
@djsmiley2k it's a congestion control algorithm that's more or less in the hands of the party sending the most data over the TCP connection
to see if it's a session layer problem, use a session algorithm that is completely without congestion control (at least unless a given application chooses to implement it), like UDP
you need to test for the common network impairments at the IP layer: jitter, high average latency, latency spikes, packet loss, reduction of throughput availability, or bufferbloat
as long as you're sitting on top of TCP, it may obscure any readings you get
