CA3LE

Hi Jose, When I get a moment I'll add São Paulo for you.

I wanted to point this out too... been thinking about this for a long time. This is a classic TestMy.net emoji that I use a lot. Keep in mind, it's not devil horns. It's "I love you". See the thumb sticking out. not-devil-horns-b-lq.mp4

That's the CPU I have in the gaming rig that started the idea to build this Ryzen server. Great CPU.

Sorry man, for some reason it got installed in my datacenter. 😜 This server is running TestMy.net now and I definitely feel the difference. Had a few networking issues over the weekend. My pfSense server had issues, something bugged out. It was my oldest VM, maybe it had some kind of legacy setting somewhere causing issues. The backups were jacked as well, so I just deleted it and put up a new one... Not really a big deal, that pfSense router is really only for my internal, non-production network and none of TestMy.net services run through it. So not a big deal, right? WRONG! Side effects are real. I didn't realize that before I had specifically setup pfSense without IPv6. With the default install it ran normally... until I got to my datacenter and reset my switches. I do this sometimes when I visit, just to get a fresh boot on them. Well, most things came back up normally but TMN was down. Everything indicated that it should be up, I'm getting pings all over the place as expected. Why isn't it working? Very stressful situation. This should have been a routine visit to quickly rack the new server. But now, nothing is working right! To make matters worse I'm on my laptop where I'm already not as comfortable working. Connected with hotspot on my phone and I realize that I only have a USB-C cable for my phone and no USB-C on my laptop. My battery's at like 30%. So now it feels like Mission Impossible. I was running in circles at this point, so I took my hands off the keys and thought for 5 minutes. ... then came up with... "Default Route." I look closer at the networking on the host and client. I then see an IPv6 address assigned to an adapter that should only have an internal IPv4 address. This device is connected to a VLAN. So traffic was trying to route out the adapter connected to VLAN. The gotcha is when an IPv6 address is assigned on the same adapter as an IPv4 address that takes an isolated route. Another fix is to tell the adapter "never use as default route". A different VM had a spare network adapter that wasn't configured, had been sitting like that for years. When this happened, it did DHCP, got an IP and then default routed to that adapter. In that case it was an IPv4 address. Same kind of issue, a little different. So I worked out a few kinks. When it came down to it, lambo joined right into the cluster, replicated my VMs and migrated them without issues. The lesson for me, when you're flustered... take a step back and think. The answers often surface.

Well said. TestMy.net started in 2001... but my first internet speed test became popular in 1996, on accident. Was pretty crazy trying to run it early on, not knowing anything. No real tools to even learn online yet either. I hope one day more people find TestMy.net again. I'll just keep building.

Just the integrated GPU. It's a virtual machine server with no need for GPU. I only need graphics to get into the bios, set up the machine and remotely manage at the base machine level. If I ever have a need for GPU I run it on one of my local machines. This puppy is built for a very specific workload. Will be racked at the datacenter in a few days. Just waiting on one more network card...

Check this out, the Ryzen 7 9700X is built on a 4 nm production process using 8,315 million transistors. I don't feel like it's getting the love it deserves. The latest AMD chips are truly amazing.

Over the decades I've always strived to have the fastest servers I can afford to host TestMy.net. My journey with hosting has taken me into a colocated datacenter where I pay for cabinet space, power & internet connection and bring in my own equipment. When I first started using colo I built everything out with brand new, current generation servers and networking gear. Over the years I added servers and built up a High Availability Proxmox cluster. Adding servers I found that eBay was definitely my friend. Now I could afford servers that only a few years earlier often went for nearly 10X what I was able to get them for "off lease". Each time I added a new server, it was more powerful than the previous. Over this time I'm adding more resources, things are getting faster. Even my first servers were full solid state, employing arrays of 4 Samsung 850 Pro SSDs. But things really started flying with the advent of Optane storage. A game changer... really, it was too good. (and now it's gone.) Fast forward to 2024. I get a generous donation of servers, the person doesn't care what I do with them, they're just happy to give them to me. They're newer than my servers, 2nd version into the next generation of CPU. So quite a bit newer. I get one of them setup with the same Optane storage I run in my current master server and start benchmarking. It's not really faster and doesn't justify switching servers. Sometimes it was slower. I assume because my servers are higher clock speed, highest end of the CPU SKU. Scratching my head, I really thought being so much newer I'd get a much better result. I take the storage out and pop it in a gaming rig... another machine recently donated to me by a friend. This PC happens to have hardware from the same year as the server I just tested. I run the same tests and had to do a double take. Not only was it faster on the gaming rig, it was WAY faster. This made me reimagine my server topology. Maybe it was time to build with consumer hardware. Here I had a setup that was a fraction of the cost that was out performing in real world scenarios. The better single thread performance and higher clock speed, faster memory... it all makes sense. There's more competition in the consumer market, things cost less. So I sold those servers to someone who can better utilize their resources. Extremely nice servers but for my use case here, they had too much of everything. As soon as they were sold I ordered parts for my next build. Promptly put it together and got to testing. Minimum, 2.5x faster at everything. Average is 3x faster and in some cases I've seen over 6x faster. It has completely blown me away. What is the magical setup? Nothing much really, minus storage I think it cost my about $900, including a piKVM setup to remotely control the machine. Much less than any of my used enterprise servers cost me. Here's my build. AMD Ryzen 7 9700X 8-Core, 16-Thread Unlocked Desktop Processor G.SKILL Ripjaws S5 Series (Intel XMP 3.0) DDR5 RAM 64GB (2x32GB) 5600MT/s CL28-34-34-89 Optane 905p (zpool) and Samsung 960 Pro (OS) ASUS Prime B650-PLUS WiFi ID-COOLING IS-40-XT Black - 47mm Height Low Profile CPU Cooler EVGA 750 BP, 80+ Bronze 750W PSU Rosewill 2U Server Chassis RSV-Z2900U BTF-LIGHTING WS2812B 2PCs 19in Individually Addressable 2GB Raspberry pi4 Geekworm KVM-A8 (kit for building a piKVM) 10G network card It's only a 65 watt TDP CPU! very easy to cool in a constrained (or SFF) build like this. A single half dollar sized CPU out performs all of my dual Xeon systems. Yes, I can't run a terabyte of RAM with this setup but I don't need to. My current systems are 128GB, I figure I only really really need 64GB with this CPU backing the system. Especially in single thread performance the new Ryzen system blows all my other systems away. This type of gain will be felt everywhere, especially in mysql queries and responsiveness. It still excels in multithread and even has a larger cache than my Xeons. It really wins for me on every level. And yes, RGB was a requirement. I never put RGB in a computer, not really my thing. But I figure this is a gaming PC at its core so for $13, why not. I can control them in linux but it adds a bunch of unnecessary packages, I like my base PVE to be as simple as possible. So default rainbow it is. lambo-2025-lq2.mp4 I haven't been necessarily doing it wrong all these years. Without the low TDP and high performance of the 9700X this build would be more challenging. These are recent developments. There's basically zero redundancy in this server. So, it also helps having a PVE cluster that has a bunch of reliable machines, to back up this consumer grade hardware. I can't wait to bring Lambo into production!

Awesome! Thank you. 😁 Welcome to TestMy.net!!

First, I've been writing in PM to @trustfor since this topic was posted. For anyone wondering why it took over 4 years for me to respond to this. trustfor just pinged the message and got me thinking about it again. I started to write this back in PM but feel it's better posted on this topic. I've continued to develop My Sensor but have never released anything. It's taken me way too long. I continue to run the program myself. Received a dozen or so credits on my Comcast bill as a result of running it and in some cases I wouldn't have known my connection went down without it. But I need to do more work to make it production ready. I'll be on my next round of development very soon. I added you to a new group, Beta 25 and will contact you before I open the beta for this. ...that is of course if you're interested in trying that program early. Now that I'm thinking about this and reading this original topic again... I think we can achieve your goal in a simpler way. The program I've been working on wouldn't just be the display, it would also run tests. But, the desktop display unit could be just a display... and then the tests could be performed by any other computer on the network. How about I tie in a simplified version of the new latency program (My Sensor) into the main site. That way we'll have nice, live latency results. I can make a simple page that shows just your last results live, graphed similar to your mockup. Then give you the ability to change the stylesheet to match your project. So your device only needs to have the ability to open a web browser.

https://stats.foldingathome.org/team/44352

You're not in the first 1000 members but you are an OG TMN member for sure. 1000th member to 10,000th is like a 6 month difference.

Wow! Your member number is right in between xs1 and cholla. You definitely deserve the Original Member badge. Welcome back. What kept you from signing in for over 20 years? I was just glancing over some of the first members. There are still a few registered in 2003 (the year the forums started here) who are still active. Pretty cool.

Hi Cholla! Happy to see you're still around.

You already said it best, "it's usable however only for loading websites" It will get you by until the next cycle. How much bandwidth do you have to use with Verizon before that limit is enforced? Are you using a dedicated hotspot device or is this a hotspot off your phone?

EDIT: scroll down first. First, here's my levels for comparison. Scrolling way over there are a lot more channels Maybe your levels are a little too high. In some cases, on the cable providers end, this is adjusted seasonally to compensate for cable length differences with temperature. In the summer, overhead cables become longer with the heat and shrink in the winter. This extra length causes a drop in signal, so it needs to be amplified to a degree. Then in the winter it needs to be turned back down. Again, this is all done on the providers end. High levels don't make things better, it can make things much worse. There's a sweet spot. Been a very long time since I was an actual cable guy... but when I was, I aimed for downstream power level to be near 0 dBmV. Mine is usually around -3 to 3 dBmV. There is something you can do about it... actually this is done on my connection right now because my levels were too high when they first installed me. Here's what I would do. Get a 3 way splitter. The type is very important... hey, an exsisting splitter could even be the issue. Here's mine for reference. Notice, 5-1002 MHz and then on the legs they say 5.5 dB ... that's -5.5 dB signal loss off those legs. A good thing, sometimes. 2 way splitters will drop 3.5 dB Most 3 way splitters drop 3.5 on one leg and 7 on the other 2. Because internally it is 2x 2 way splitters. To have a nice even 5.5 drop on all legs is kinda special, they didn't have those when I was a cable guy. For yours, maybe drop 5.5 to 7 dB. Get the splitter and cables from a Comcast store. If they only have 2 way splitters, get 2 of them. To be safe, only use cables and splitters that come from Comcast. Plus, they normally hook them up for free. . -- they really should only want you to use their connections, it saves service calls. Some cables and splitters are designed for antenna, they don't care as much about the lower and higher frequencies and they don't care as much about ingress/egress. So yeah, only use the cable providers cables, splitters and amplifiers. I don't think you need this diagram but I drew it for good measure. You may have a splitter along the line somewhere that could be to blame as well. I would trace all the way from where the cable hits the side of your house to the modem and replace any splitters, barrel connectors (they connect single cables together) or possibly bad cables along the route. Just something I would try myself before calling. BTW, in this case I would only be testing single thread. I think that's going to be the most helpful. You've seen up to 248 Mbps upload linear before, we know it's possible. https://testmy.net/db/uYQhpDzSj Please update us on any developments. And I'm sorry it took me a while to get to your topic here, I wanted to take time to properly respond... then days passed. EDIT: Looking again, I wouldn't do this. It will drop your upload signal level too low. You want 40 to 45 dBmV for upload power level. -15 to 15 dBmV for download. Trace back the lines like I said, if you don't find anything then call Comcast. Show them what you're seeing. You will have 200 Mbps again soon. This may help someone in the future.

Yes, according to my IP records in the forum... you've only been using 1 IP recently. So this is very weird. Please visit https://testmy.net/iptools click [Paste] then "Run Query" -- take a screenshot and send it to me in a PM. Maybe the IP detection is grabbing a forwarding address by mistake, under some edge case scenario. With your help we'll get it fixed.

Sorry, I gave you permission on the backend of the program but not on the frontend. It should be available for you now.

You're all set! Give it a try.

HI Ritan671, welcome! For ping, either select Latency from the main menu or enable the beta under My Settings. Let me know what you think or if you have any more questions.

Maybe contact that developer. Easier to just use TestMy.net instead. Try the beta under My Settings.

You'll get this error if you run too many tests too quickly. It should be there... but not unless you hit the test more than you did. It shouldn't happen on your first re-test like that. I was seeing it too, when combined test was used and re-test was quickly done. I made an adjustment and it now passes properly under that scenario. Thanks for the heads up.

thank you! And thank you to those who've made anonymous BTC donations! Very unexpected. You guys rule.

I'll correct that, I didn't update the old My RT program with that mirror's information yet. Thanks for catching that. That mirror is slightly different than the rest, not built into My Latency by default... unless you select that mirror, then you'll see the option. It was donated by WaveX ISP in Kenya, I own all of the other servers. Donated, but it's a full TMN server I maintain... first one. -- it also doesn't appear in the legacy multithread test right now.

You live in a very interesting location in the world. When you search for the "western most part of Europe" in Google you find Cabo da Roca. But that's totally not true. This rock, just off your island is the real western most part of Europe. That rock, from what I can see... is the end of Europe. Google Maps The Westernmost Point of Europe I feel like you're getting extremely good speeds to that remote location.