Kitz Forum
Computers & Hardware => PC Hardware => Topic started by: Terry74 on February 24, 2022, 08:26:59 PM
-
Hi all,
I've recently had BT/Openreach install 500 Mb/s FTTP to my home.
The Openreach guy that fitted it got something like 530Mb/s on his iphone no trouble.
I struggle to see that speed with my computers which are mainly Intel Celeron processors of varying ages. eg 2957U @ 1.4GHz an HP mini desktop achieves approx 230Mb/s down 70 Mb/s up, using Firefox and 350 Mb/s down and 70Mb/s up using Chrome. A 2955U modified chromebox achieves much the same. The processors max out with both, running Ubuntu Linux connected via ethernet.
If I try speedtest-cli via the terminal I can get 440 - 450Mb/s down and 73Mb/s up.
If I run iperf3 between these 2 computers on my network I get 944 Mb/s either way.
A chromebook with N3160 (Celeron with 4 cores) processor manages 430Mb/s down and 70Mb/s up using Chrome browser on wifi(5G),as it has no ethernet port.
My question for you knowledgable guys is what spec computer do I need to see the full download speed that my circuit is capable of?
Would an Intel i3 be OK? Would an i5 manage it? If so, of what generation?
Also, what is needed for Gigabit (900Mb/s) FTTP like my neighbour has?
Just what is the minimum computer spec necessary to download at these speeds?
Do you all run i7 or i9 top of the range computers? Surely not?
My computers suited me just fine when I had 30Mb/s FTTC, but now I just wonder what I'm missing.
Just interested.
Terry
-
Getting full speed from a web browser should never be expected (without a download manager add-on/extension) even on a top-spec PC as its single-threaded (one download vs speedtests which do multiple concurrent downloads) and often the servers you are downloading from are speed limited or contended in some way.
So the real question is WHAT are you wanting to download at full speed? What software are you using for it?
If you have the url to something you want to download, you can try using the command line downloader axel which allows multi-threaded downloads. It basically starts the download multiple times from a different position in the file then combines them all at the end. Or try to find an add on for the browser, but I'm not fond of how those work personally.
-
Just for kicks I tried something, I ran 2 speed tests at same time in 2 different browsers on my PC at exactly same time, if I add the speed together its roughly my FTTP speed.
So if you want to prove your connection, try running speed test on 2 different devices at same time, add the numbers, and should be around your 500.
-
That has the added benefit that it will multi-thread the CPU process too, so if it IS CPU bound you're likely spreading that load across multiple CPU cores.
-
TCP-based speed tests are often problematic in my opinion. They are measuring the speed of the particular protocol and it’s implementation, not the speed of the link.
If you must use TCP, then consider downloading a large file using say ftp and time it. See https://www.thinkbroadband.com/download
However a much better result is obtained by performing a large number of downloads at the same time as this will thoroughly thrash the link. I suggest that the number of downloads should be max( 4, npipes × 2 ) where npipes is the number of internet connections that you have bonded together, if you do have any such bonded connections.
Remember that other processes may be using your own internet connection at the time of the test, either on your own machine or on other machines on your LAN. This includes background processes such as doing backups to the cloud.
Also a server at the remote end or the link coming from it may be overloaded. Do several tests and at least two in the middle of the night.
-
Thanks craigski,
I had already tried that by downloading two different things on two different computers simultaneously and both downloaded at around 250Mb/s, exactly as you suggested.
Thus showing the capability of my connection.
However, the point of my OP was not so much the connection but the capability of the machine. Please see my little story below.
Terry
-
Hmmm....
Perhaps I've not done a good enough job of explaining myself!
Think of me as the ordinary "Jo" who believes the hype that he reads about the internet and download speed. eg the articles like " Is my computer ready for gigabit internet " which tells him basically "look at the info in windows and if it tells you your ethernet interface is gigabit capable then you are 'good to go' ". There are lots of those.
Jo does this, is reassured and then buys 900Mb/s internet from, say, BT and with his 2010 model celeron computer, opens his Firefox browser and types in "speed test". He chooses the first thing he sees, which is probably the "Google/M-Labs" entry OR "Speedtest by Ookla" or even "Thinkbroadband.com". He runs the speed test and is surprised to find that it displays only 150 Mb/s!!
He tries a different speed test with results much the same no matter which he chooses.
He then considers this, and decides to try a different browser. Chrome. He is surprised to find that he now gets 300 Mb/s displayed. A browser is a browser, right?? Right???
Now, Jo's been around a long time and so he looks to the "command line". He closes his browser opens his terminal and types "speedtest-cli". This shows him 400 Mb/s.
But Jo sees all these little plaques from others plastered around the technical web sites like "Kitz.co.uk" showing 944Mb/s down and 110Mb/s up and wonders why he's not able to reproduce something similar.
Jo feels cheated!
Although he doesn't use the computer for much, other than surfing the net and a bit of research perhaps, Jo wants to see those results and be able to post his plaque proudly.
Jo decides it must be his old computer at fault despite what those articles told him. Perhaps it's processor is not powerful enough, perhaps it doesn't have enough memory.
But what should he buy? It's become obsessional.
There is an attractive i3 machine in the shop down the road, but it's 5 years old. Will that do what he wants? Remember that plaque! Maybe he needs an i5, an i7, an i9??
He notices that Apple have a new M1 Mac Mini machine which everyone says is fantastic.
He wants that plaque. What does he do?
If he buys a new computer what will he see?
It's easy to say "Don't be so stupid. Stick with what you've got". But, Jo is still thinking about those adverts and that plaque. He's determined to buy something, so, what does he buy?
That is the sort of "reasoning" behind my OP.
Answers please.
Terry
-
Hi Alex Atkin UK,
You'll probably have gathered from my posting above that I want a speed test result from Thinkbroadband.com for example, just like those you have in your signature. BUT, I would like to know what spec the computer was that gave you those results. I would say that it pretty certainly wasn't a Celeron that is 10 years old.
It really isn't about downloading anything in particular.
I note you have an Apple Mac Mini for example or/and a Ryzen 5 5600 or some such which look pretty powerful computers.
Just what do you consider to be a minimum spec to achieve those results from thinkbroadband? You've always struck me as pretty knowledgable from your posts.
There must be some minimal spec to achieve that end OR am I totally wrong?
Regards
Terry
-
See here:
https://speedtest.btwholesale.com/
Click the link where it says if you are running 500Mb/s
I would say that is the absolute minimum spec, assuming you are dedicating all the pc resources to the Speedtest.
-
Hi Weaver,
Yep, I've tried downloading those thinkbroadband files with my Celeron computer and get 58 MiB/s (or about 464Mib/s if I multiply by 8) using wget or aria2 -x5 .... but I'm willing to bet my processor was maxed out doing it. That was last night around midnight, so a pretty quiet time.
Thanks for the suggestion. Not sure it thrashed the link but rather my processor.
Not having anything better computerwise, I just wonder what I'm missing. All brought on by getting FTTP!!
It's all idle speculation really. I don't compile stuff like you do except for the odd Arduino playthings with ESP32's. I really don't NEED anything new.
Regards
Terry
-
Thank you craigski,
That's just the sort of thing I was looking for. Really good info.
Regards
Terry
-
Taking another look, I'm actually impressed you got as good as you did - those Celerons as well below the Pentium 3805U laptop I had to stop using because it was painfully slow.
I think BT are exaggerating, I'd think a half-decent i3 might be enough. Its also complicated as an i3 desktop will outperform an i5 laptop in a lot of cases (especially older models), so which do they mean?
If you can find out the exact CPUs in any PCs you are looking at we can guesstimate the performance from the linked site.
https://www.cpubenchmark.net/compare/Intel-i3-2120-vs-Intel-i5-2410M-vs-Intel-Celeron-N3160-vs-Intel-Celeron-2957U/752vs796vs2812vs2200
You should be able to find something that will run rings around what you have right now for relatively cheap.
-
Sounds to me like you really don't need 500Mbps, if all you want is a nice badge I've attached two below ;)
The real one, which is easy to tell, was done on an Old Intel NUC i5-6260U, and whilst it does achieve maximum line speed both CPU cores (4 threads) were at 100% during the download test. The best part is it barely draws any power, between 10 -30w depending on what its doing.
My 6 core 4.5Ghz i7-5820K barely woke from a slumber, but got the same result.
-
Thank you Alex,
I think if I were seriously looking, I'd choose the M1 Mac Mini that you "reviewed" - it looks like good value to me at £629 from a couple of places. I started with an Apple IIe a long time ago, after a Commodore Pet. Trouble is, I sort of like Linux, so... maybe not. But at least I know a bit more about what is required in performance terms.
Hi Ronski,
Thank you for the Badges and you are quite right, I don't need 500Mb/s but it was going cheap on Black Friday....... I'm kind of surprised that the i5-6260 was flat out but that is more useful info as is the i7-5820.
I've just done a couple of Badges myself..... if they appear. I don't do this too often. The second of which frankly amazes me, when I look at your result. The difference is simply the browser that I used. It seems like I can't type dates either!!
Regards to both
Terry
-
I think if I were seriously looking, I'd choose the M1 Mac Mini that you "reviewed" - it looks like good value to me at £629 from a couple of places. I started with an Apple IIe a long time ago, after a Commodore Pet. Trouble is, I sort of like Linux, so... maybe not. But at least I know a bit more about what is required in performance terms.
Its great hardware but I do hate MacOS, but then Linux for Apple Silicon is well in the works so that could be interesting once I replace it with a newer model.
I'm absolutely in love with the Macbook Pro hardware wise, its the best screen in the house. I'd kill for a screen that good on my desktop PC.
-
Nice result, and very good upload speed on the second one. I'd love to have FTTP available, but unfortunately stuck with FTTC or Virgin, so Virgin it is.
i5-6500T (4 core/4 threads), full speed test result, hits about 80% on all cores
i5-2405s (4 core/4 threads) full speed test result, briefly hits 100% then drops to about 80% on all cores
i5-10600 (6 cores/12 threads), full speed test result, can't say it makes any difference
3900x (12 cores/24 threads), full speed test result, can't say it makes any difference
Xeon E5-2699V3 (18 cores / 36 Threads), full speed test, can't say it makes any difference
Yes I do have far too many computers, there are two other PC's, a couple of laptops, and a couple three more that are passing through.
-
I certainly wouldn't be picking a PC based on a web browser speed test. ;) However I would be upgrading from Celerons purely for how much faster using websites will be.
I do have a huge selection of PCs though.
I recently upgraded my mum from an i5 2500k AIO to an Intel NUC i5 8259U.
My Torrent box is an i5 8250U (was going to be a spare appliance for in case my router dies or needs more cores once on FTTP).
My AI video upscaler box is an i5 4690.
My AI photo upscaler box is the Mac Mini M1.
Main desktop an i9 9900k.
Gaming PC a Ryzen 9 5950X.
Portable AI upscaler is the Macbook Pro (M1 Pro).
My server has the i5 8600k.
Windows laptop has an i7 8750H.
Spare Windows laptop (intending to send to a friend in Texas if I can ever figure out how to ship it safely) a Ryzen 2500U.
Not to mention a few CPUs knocking around that are worth practically nothing as its quite hard to find motherboards for now. For example I upgraded the AIO from an i3 for the better GPU. A couple of Rasberry Pis knocking around I never got round to using for anything.
-
Just a small addendum to complete this thread, I recently bought, off ebay, an HP EliteDesk 800 G2 mini PC with i5 6500 processor. An Ex business machine with 16GB of memory. So, 4 cores @ 3.2GHz which seems to perform nicely. Discarded the Windows 11 which came with it and loaded Ubuntu 20.04 LTS.
This managed to perform the speed test showing >500Mb/s down and >70Mb/s up whilst utilising about 40% on all 4 cores.
So, I guess that the answer to my original question of necessary spec to perform these tests lies somewhere between my 2014 Celeron 2957U and this 2016 i5 6th generation processor. More modern processors will have no difficulty at all.
Regards to all who helpfully replied
Terry
-
Yeah that's 6.7 times the speed of the Celeron in multi-threaded workloads, 2.7x for single-threaded. That's got to feel a lot nicer to use.
-
But Jo sees all these little plaques from others plastered around the technical web sites like "Kitz.co.uk" showing 944Mb/s down and 110Mb/s up and wonders why he's not able to reproduce something similar.
He wants that plaque.
Jo has his plaque now :)
-
Hi Alex
Yes, Firefox opens in 2 secs instead of 10secs previously, so much nicer to use. Thanks for the actual speed difference. I figured twice for the doubled clock speed plus a bit more for the double number of cores.
and
Hi Craigski,
That made me smile!!! Thanks for that.
Best Wishes to both
Terry
-
Thanks for the actual speed difference. I figured twice for the doubled clock speed plus a bit more for the double number of cores.
That's based on CPU Mark benchmark but its a reasonable estimate IMO for common workloads.
-
Hi Ronski,
We live in the wilds of Norfolk. We only got 30Mb/s (on a good day) FTTC in 2018 before that it was 4Mb/s ADSL which gradually improved with changes of modems, filters etc to a usable 8Mb/s. Then in Jan 2020 Openreach announced Fibre First towns and to our surprise, Watton was on the list.
We watched them cable up the lanes but it wasn't until Nov 2021 that FTTP became a reality. It then took until Jan 2022 and a couple of failed attempts before it arrived at my house.
So you never know!! One day it may come your way. Here's hoping.
Best Wishes
Terry
-
Thanks, hopefully one day Openreach will install FTTP, but I get the feeling it will be some time yet, at least we have decent speeds albeit with VM.
PS nice speed test.
-
I was recommended to use the fast.com website speed test.
Though I think its more accurate to try and download a file.
@Ronski when they announced my town it was upto 2025, then within months the OR markers started appearing on roadworks website. Now they have nearly completed the town.
As your area been announced yet?
-
TCP-based speed tests are often problematic in my opinion. They are measuring the speed of the particular protocol and it’s implementation, not the speed of the link.
If you must use TCP, then consider downloading a large file using say ftp and time it. (https://gbapps.net/gbwhatsapp-apk/) See https://www.thinkbroadband.com/download
However a much better result is obtained by performing a large number of downloads at the same time as this will thoroughly thrash the link. I suggest that the number of downloads should be max( 4, npipes × 2 ) where npipes is the number of internet connections that you have bonded together, if you do have any such bonded connections.
Remember that other processes may be using your own internet connection at the time of the test, either on your own machine or on other machines on your LAN. This includes background processes such as doing backups to the cloud.
Also a server at the remote end or the link coming from it may be overloaded. Do several tests and at least two in the middle of the night.
You are right in terms of pure timings, but real world performance is something else. I'd argue 16GB on the preferred channel would be faster in a lot of use cases that actually use more than 8GB at once.
More benchmarks or a follow-up post would be appreciated as the topic in itself is quite interesting.
-
Welcome to the forum, SammyAdam!
-
Hello everyone...
Neither a computers CPU or browser will have an effect on the speed the computer receives - this is down to the speed of the ethernet or wireless adapter/connection. That said, a slower computer will generally struggle with tasks including browsing. For example any images/videos displayed on a website might download quickly, but struggle to be played/rendered by the computer.
-
On the contrary, the speed the computer receives will be largely dependent on how fast it can process what it just received and request the next piece of information. eg I might be able to push 9.4Gbit to/from my NAS with perf3, but try doing that over NFS or Samba and it NEVER reaches those speeds, even between NVMe storage which is much faster than that, because there are protocol bottlenecks that eat CPU resources.
Plus every packet of data coming into the ethernet adapter has to be dealt with by the CPU, the weaker the CPU the more chance it wont keep up. The cheaper the NIC, the more CPU load its likely to incur (as higher end NICs have hardware offloading on the NIC itself) and more likely it is to only support a single CPU core to deal with all the packet processing, whereas modern and server NICs can load balance so that incoming packets don't stall on a busy CPU core.
Download speed is hugely impacted by CPU speed if the CPU is slow and the network is fast.
Another example is game downloads. A lot of services like Steam will download and decompress the data at the same time. If your CPU maxes out, your download will slow down. If your storage isn't fast enough (because even a 70Mbit/s download could easily extract to much larger than this), the download will slow down.
EVERYTHING your PC does is impacted by the CPU speed as it has to organise everything that is going on.