Recommended Posts

I was pretty certain I started a thread for this but I can't seem to find it. Anyway, I'm having a tough time understanding why my internal network speeds are not consistent and not at gigabit speeds. I checked that the nics on both ends are set to 1 gbs and the switch / router between the devices are all 1 gbs capable. Why would I sometimes get 25 mb/s and other times 80 mb/s?

Link to comment
https://www.neowin.net/forum/topic/1136090-slow-internal-network-speeds/
Share on other sites

Are you copy files to / from the same devices when you notice a difference in speed? I find larger files will transfer a lot quicker than loads of smaller files.

Also copying to an IDE hard drive from an IDE hard drive would likely be a lot slower than copying from a SATA3 drive to a SATA3 drive.

80 MB/s ?

I don't think so, I got Maximum 25 MB/s in Giga Netowrk & 12 MB/s in 100 Mb/s Network.

& that's how it works ...

You probably have poor hard drive's then, I can easily exceed 80 MB/s across my home network.

so you stated mb (megabits) are you talking MBytes?

25MBytes is clearly on a gig network, now is it great performance - no.. But its way over what is possible on a 100mbit network. As stated there are more variables that come into play than just your network speed.

But I would also check to see what your getting with wirespeed, take the HDD out of the equation. Use iperf or netio to do a test of what speeds your seeing on the wire. I would hope your seeing in the 800mbps range -- but might be as low as 300, etc. Depends on your nics, depends on your settings, etc.

When moving large files I normally see 50 to 90 MBytes per second.

Sounds like cache on HDDs or something not being able to keep up.

Then again on my 1 Gbps network (Linksys E4200v1 router/switch) I can push 90-100MB/s straight (then again my laptop has a SSD to my NAS which are big 2TB Drives). At work from my machine (standard Office desktop) to the Server (SAS RAID 10 Drives) I get about 50-60MB/s off a WD 250GB 7200RPM HDD.

So I guess it depends on defragmentation of the file you are copying as well as the head moves to get all the pieces.

Are you copy files to / from the same devices when you notice a difference in speed? I find larger files will transfer a lot quicker than loads of smaller files.

Also copying to an IDE hard drive from an IDE hard drive would likely be a lot slower than copying from a SATA3 drive to a SATA3 drive.

You probably have poor hard drive's then, I can easily exceed 80 MB/s across my home network.

I have SATA Drive, which is gives me 80 to 150 MB/s speed when transferring from one partition to other. but on the Giga network just 25MB/s

I have SATA Drive, which is gives me 80 to 150 MB/s speed when transferring from one partition to other. but on the Giga network just 25MB/s

Cant honestly say why you speeds are so low, I have an Asus RT-N16 gigabit router running the Tomato firmware, with 4x Netgear gigabit switches connected to each port on the router which reside in various rooms of the house.

Nothing really fancy, however i have always been happy with the file transfer speeds on my home network.

speedao.png

Cant honestly say why you speeds are so low, I have an Asus RT-N16 gigabit router running the Tomato firmware, with 4x Netgear gigabit switches connected to each port on the router which reside in various rooms of the house.

Nothing really fancy, however i have always been happy with the file transfer speeds on my home network.

speedao.png

WOW, & I am connected throw Cisco SG 100 Giga Switch.

I remember using jumbo frames back in the XP days but I sort of remember having issues after enabling it. I guess it's worth a shot regardless

EDIT - also the drives on my server are mostly green drives so all of them are 5400 rmp. The primary drive is a wd black drive, though. Not sure if the bottlenecks are due to the 5400rpm drives

Jumbo frames are NOT going to be a magic pill. And unless everything in your house is gig and supports jumbo your going to cause yourself more pain.

80MB is not all that.. Yes it is quite possible with gig gear.

And sorry 3 switches does not = hops ;) Are going through 3 different "routers" then you could consider it hops.. But if they are all local and the latency is couple of ms RTT doesn't matter if it was 10 hops.

Green drives are not going to be the speed demons - they are meant for storage, low power usage not moving your file across your network at 100MB plus.

So you think my drives are the culprits then? I don't really need gigabit speeds - just a little annoyed that I spent money making sure everything is gigabit compatible to still hit some bottlenecks. I'm not going to upgrade my hard drives to faster drives to overcome the issue since I am using those drives for storage / media serving purposes as you refer to

But I would at least expect some consistency. Just seems weird that sometimes I can hit 80MB/s but then can't get past 25MB/s with the same transfer from main PC to server

I enabled 4K jumbo frames before leaving my apartment. I'll see if it makes any difference later

So your switch supports jumbo? And all your devices support jumbo?

Jumbo can lower cpu usage, but since I would have to think that you have devices that don't support it - you could have other issues with communication. Do you talk to wireless devices? Do you have any devices that are only 100mbit?

So when do you hit 80? From different device to your server, or from your server to your main pc?

Lets get some details

When you copy file from your server to your main pc what are the speeds? I would really suggest you test this with iperf or netio. Then do a file copy with robocopy so you can get a printout of the speed.

Now when you copy file from your main pc TO your server what is the speeds - do the same iperf/netio test in that direction as well, and again use robocopy to get printout of details of copy with speed.

Now repeat this test to some other box on your network to and from your server - is same results?

Is this other box same OS as your main PC? Same nic? Same Driver for nic?

I won't be able to give you all the information necessary till I'm back at my place but for now...

These are the two DLink switches I have...

http://www.newegg.co...N82E16833127422

http://www.newegg.co...N82E16833127083

Looks like both support jumbo frames according to the specs

Asus RT-N66U is the router in use

The NICs are different but both are gigabit. One of them is Intel based on an Asus Sandy Bridge board. The other is Marvell on a Asus LGA 775 board

Transfers are from my main pc (that has the Asus board) to the server (LGA). I use either Teracopy or FlashFXP to send files ranging from 1mb to 30 GB files (bluray rips). FlashFXP seems to be slower for some reason. I was going to try Filezilla to see if it's more consistent. Either way, both Teracopy and FlashFXP vary in performance

Transfers to my main PC seem to be faster. Usually above 50MB/s.

All PCs on my network are set up with Windows 8 including server. Server used to be WHS 2011 but moved to Windows 8 to make it an HTPC as well. Storage is set up using Drive Bender which is a third party app that mimics / improves on Microsoft's Drive Extender. 4 2TB drives are pooled into one virtual drive.

It could also be performance issues with Drive Bender? I'll check into that with support after seeing how transfers do between my third PC and HTPC

On the network...

Main Desktop connected to one of the switches

Server / HTPC combo connected to router

Printer connected to one of the switches

mini HTPC for main bedroom connected through Wifi

Surface Pro on Wifi

Phone on Wifi

Xbox 360 currently on wifi

PS3 connected to one of the switches

Sony TV on Wifi

Yamaha receiver connected to router

Both switches connected to router

Currently setting up home theater and devices so most of the devices will be connected to a switch or router directly. Prefer only to use Wifi if no other option is possible

  • 11 months later...

Almost a year later and this is finally resolved.  My speeds ended up locked at anywhere between 4-8 MB/s and I couldn't figure out why.  Tried a different cable.  Tried everything before buying new hardware.  Then I got ###### off back in November and replaced all the green drives I had with WD reds, added a PCIE Intel network card, switched to Server 2012 R2 ( from Windows 8) and switched to DrivePool (from DriveBender).  I don't know what did it but I'm back ~80-100 MB/s.  May be the drives but I also had issues with speeds to a separate black drive I was using for game installs.  At the end of day, it really isn't that big of a deal I was locked at slower speeds but 4-8 MB/s I found pretty annoying.

Hello,

Almost a year later and this is finally resolved.  My speeds ended up locked at anywhere between 4-8 MB/s and I couldn't figure out why.  Tried a different cable.  Tried everything before buying new hardware.  Then I got ****ed off back in November and replaced all the green drives I had with WD reds, added a PCIE Intel network card, switched to Server 2012 R2 ( from Windows 8) and switched to DrivePool (from DriveBender).  I don't know what did it but I'm back ~80-100 MB/s.  May be the drives but I also had issues with speeds to a separate black drive I was using for game installs.  At the end of day, it really isn't that big of a deal I was locked at slower speeds but 4-8 MB/s I found pretty annoying.

Did you do a direct network cable between two PCs? No router, no switch, just 1 cable between two PCs....

And one year later?

I never connected the two.  And yeah, 1 year later.  I've made some changes over time but speeds just seemed to get worse instead of improve.  Not sure what from the list of stuff I did in November till now that helped but I'm back where I expect things to be

This topic is now closed to further replies.
  • Recently Browsing   0 members

    • No registered users viewing this page.
  • Posts

    • One of the strangest galaxies in our Universe could help answer some long overdue questions by Sayan Sen Image by Pixabay via Pexels | Not representative An international team of astronomers led by the Department of Astronomy at Tsinghua University has discovered an unusually metal-poor galaxy that may contain signs of first-generation star formation. The galaxy, named Metal-Pristine Galaxy COSMOS Redshift 3 (MPG-CR3), or CR3, was identified using observations from the James Webb Space Telescope (JWST), the Very Large Telescope (VLT), and the Subaru Telescope. The findings, published in The Astrophysical Journal Letters, describe CR3 as the most metal-poor galaxy known from the period known as "cosmic noon," around 11.5 billion years ago. Cosmic noon refers to a period when the universe was producing stars at its highest rate and galaxies were growing rapidly. In astronomy, "metals" refers to all elements heavier than helium, including oxygen, carbon, and iron. Because CR3 contains so few of these heavier elements, researchers say it closely resembles what scientists expect the earliest galaxies in the universe may have looked like. The discovery is significant because it could offer clues about Population III (Pop III) stars, the first generation of stars thought to have formed after the Big Bang. These stars are believed to have formed from gas made almost entirely of hydrogen and helium, before heavier elements were created inside stars and spread across the universe through supernova explosions. Hence this is why CR3 has been referred to as a "living fossil." Scientists have long believed that Population III stars existed only in the very early universe. As more generations of stars formed and died, they enriched surrounding gas with heavier elements, making the conditions needed for metal-free star formation increasingly rare. Because of this, researchers expected the formation of such stars to have largely ended after the epoch of reionization, a period when radiation from the first stars and galaxies transformed the neutral hydrogen filling the universe and made it largely transparent to ultraviolet light. CR3 appears to challenge that idea. The galaxy was observed at a redshift of z = 3.193 ± 0.016. Redshift measures how much light from a distant object has been stretched as the universe expands and helps astronomers determine how far back in time they are looking. In this case, the redshift corresponds to roughly 11.5 billion years ago during cosmic noon. Although the universe was already several billion years old by that point, CR3 shows characteristics more commonly associated with much earlier galaxies. Observations revealed exceptionally strong emissions from hydrogen and helium, including Lyα, Hα, and He I λ10830. Lyα, or Lyman-alpha emission, is a specific wavelength of light produced by hydrogen and is widely used to study distant galaxies. Hα emission is another hydrogen signature commonly used to trace active star formation, while He I λ10830 is produced by helium and can indicate the presence of very hot, young stars. The measured equivalent widths of EW₀(Lyα) = 822 ± 101 Å and EW₀(Hα) = 2814 ± 327 Å are among the highest ever observed in star-forming galaxies. Equivalent width is a measure of the strength of an emission line relative to the surrounding light, and such large values are typically associated with intense and very recent star formation. At the same time, researchers found no statistically significant detections of metal emission lines, including [O III] λλ4959, 5007 and C IV λλ1548, 1550. Emission lines act as chemical fingerprints that reveal which elements are present in a galaxy. Oxygen and carbon lines are commonly seen in galaxies that have already undergone significant chemical enrichment. Their absence in CR3 suggests an unusually pristine environment. Using abundance calibration methods developed with JWST observations, the team placed a 2σ upper limit on the galaxy's gas-phase metallicity of 12+log(O/H)<6.52, corresponding to less than 0.7% of the Sun's metallicity (Z < 7 × 10⁻³ Z⊙). Gas-phase metallicity measures the abundance of heavy elements in a galaxy's gas. A 2σ upper limit indicates that the true value is very unlikely to be higher than the quoted threshold. Even when accounting for uncertainties in the calibration methods, the most conservative limit remains 12+log(O/H)<6.95, making CR3 the most metal-poor galaxy identified at cosmic noon. The galaxy also appears to contain very little dust. Researchers measured a Lyα/Hα flux ratio of 13.9 ± 2.5, a result that suggests negligible dust attenuation, meaning very little of the galaxy's light is being absorbed or scattered by cosmic dust. Because dust is usually produced by earlier generations of stars, this finding further supports the idea that CR3 has experienced very little chemical enrichment. Further analysis using spectral energy distribution modelling, a technique that compares observed light with theoretical models, suggests that CR3 contains an extremely young stellar population only around 2 million years old. The modelling, which used Population III stellar templates, also indicates the galaxy has a stellar mass of approximately 6.1 × 10⁵ M⊙. The symbol M⊙ represents one solar mass, or the mass of the Sun. One of the key questions raised by the discovery is how such a chemically primitive galaxy could exist in a universe that had already spent billions of years producing heavier elements. To investigate this, the researchers examined CR3's surroundings. Their analysis suggests the galaxy may lie in a slightly underdense environment, with a density contrast of roughly δ ≈ −0.12. An underdense region contains less matter and fewer galaxies than average. The team suggests that this relative isolation may have helped preserve pockets of pristine gas. Metal-rich material expelled from nearby galaxies may never have reached CR3, while the lower rate of galaxy mergers and interactions could have slowed the mixing of enriched gas into the system. If future observations confirm these findings, CR3 could provide some of the strongest evidence yet that first-generation star formation continued well after the epoch of reionization. Such a result would challenge the conventional view that pristine star formation ended by z ≳ 6 and suggest that small pockets of metal-free gas survived much longer than previously thought. Researchers stress that more observations will be needed to determine the galaxy's true nature. Future spectroscopic studies with higher resolution and better signal quality could help confirm whether CR3 is genuinely hosting Population III star formation. The discovery is also expected to encourage searches for other similar galaxies, which could help astronomers better understand how the first stars formed and how galaxies evolved in the early universe. Source: Tsinghua University, IOPscience This article was generated with some help from AI and reviewed by an editor. Under Section 107 of the Copyright Act 1976, this material is used for the purpose of news reporting. Fair use is a use permitted by copyright statute that might otherwise be infringing.
    • "I think in the immediate absence of a partner to apply relief" In the words of Sterling Archer... "Phrasing!"
    • For me, the fundamental problems with these "smartglasses" is that they really don't work well for people with significant prescriptions and massively up the price if you use attached lenses if they have displays, and if they don't, then they're not actually "smart" anything, rather just connecting to your phone and relaying voice to an AI. In a few cases like this, they throw in small cameras to feed video to the AI. All around, these feel like both a solution looking for a problem, and the problems it tries to solve seem more easily solved by different approaches and designs. Oddly, if the rumours are true, Apple may actually have invented something for once and it kind of does this right: put cameras in ear buds and manage the interface to AI exactly as most of us do: tapping on an ear bud and saying "Hey Google" or "Hey Siri." That makes them compatible with almost everyone, can double up as a hearing assist device, an impaired vision assist device, a "smart" device... and answer your phone and play music. That just seems like a better solution all around.
    • Usually the bigger ones with many fixes/changes take a few, theyre an exception to the rule most likely
    • If you don’t get lucky with Valve’s Steam Machine reservation system, you can make your own Steam Machine instead. Valve says that “starting with the SteamOS 3.8 release, you can put together your own Steam Machine using whatever PC parts you want.” SteamOS 3.8.10 launched last week with a slew of updates, including “improved compatibility with recent Intel and AMD platforms.” Alongside that improved compatibility, Valve is giving gamers the green light to install SteamOS on their own desktops. In an interview with The Verge, Valve’s Pierre-Loup Griffais said Valve has been “rolling out improvements to [SteamOS] so it’s more compatible with desktop hardware,” including eventual support for Nvidia graphics. Griffais says Valve has “a growing team” working on Nvidia driver support for SteamOS, adding, “We’re collaborating with Nvidia very closely.” While he mentioned that Nvidia support might not come this year, Griffais emphasized that “it’s certainly something that we’re working on in the background.”     Subscription not needed: https://archive.fo/Tssfc Subscription needed: https://www.theverge.com/games/953411/valve-steamos-desktop-nvidia
  • Recent Achievements

    • Dedicated
      HidekoYamamoto94 earned a badge
      Dedicated
    • One Month Later
      timbobit earned a badge
      One Month Later
    • One Month Later
      nates earned a badge
      One Month Later
    • Week One Done
      Almohandis earned a badge
      Week One Done
    • Rookie
      dorf went up a rank
      Rookie
  • Popular Contributors

    1. 1
      +primortal
      452
    2. 2
      +Edouard
      161
    3. 3
      PsYcHoKiLLa
      107
    4. 4
      Michael Scrip
      84
    5. 5
      Steven P.
      70
  • Tell a friend

    Love Neowin? Tell a friend!