Welcome Guest ( Log In | Register )

Bump Topic Topic Closed RSS Feed
10 Pages « < 2 3 4 5 6 > » Bottom

Outline · [ Standard ] · Linear+

Unifi Official TM UniFi High Speed Broadband Thread V36, READ 1ST PAGE FOR TURBO SPEED INFO!

views
     
JohnLai
post Nov 13 2018, 08:27 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(nyunyu @ Nov 13 2018, 08:24 PM)
Really..demmit. Is it serious?
How can you tell?
*
Oh yes......of course I can tell.

Let start with your laptop, answer these few questions:

1) Did you update the laptop bios?

2) Win10 which version? 1803 or 1809?
JohnLai
post Nov 13 2018, 08:37 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(nyunyu @ Nov 13 2018, 08:33 PM)
1) Latest BIOS as per released by Illegear.

2) 1803 (both on desktop too)
*
Very well, download appropriate 'new' microcode update for your laptop

http://www.catalog.update.microsoft.com/se....aspx?q=4346084
2018-08 Update for Windows 10 Version 1803 for x64-based Systems (KB4346084)

and install it. Reboot.
Then run RWeverything and check if the PATCH ID changed from 84 to 96. (Technically, the latest microcode from intel is version 9A.)

If it is 96, then run TM speedtest, you should be able to reach 800Mbps via wired.

JohnLai
post Nov 13 2018, 08:55 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(nyunyu @ Nov 13 2018, 08:47 PM)
It returns "The update is not applicable for your computer". Tried both x64 and x86.
*
Oh crap....
https://support.microsoft.com/en-my/help/43...crocode-updates
Probably your laptop CPU isn't listed in microsoft database.

EDIT: it had came into my attention you mentioned the laptop is i7-8750H....., 4346084 should be able to install properly....unless your win10 version isn't 1803? hmm.gif

EDIT2: Huh, found the reason why 4346084 can't be installed in your case:
https://borncity.com/win/2018/09/24/windows...7467-kb4346084/
Quoting from the blog:
QUOTE
    Summarized: The Spectre SSB Microcode Patch KB4346084 cannot be installed if the Spectre BTI Patch KB4100347 has been installed before. But you only need one, the newer 4346084, which helps for both variants.

    So, it has indeed been explained, and the solution is most astonishing: Microsoft’s own Microcode Update KB4100347 (remedy against Spectre BTI) is the cause. If this has already been installed, the newer microcode update KB4346084 cannot even be installed; and, it loses its function and protective effect in any case!

    Microsoft… never has this pointed out. I will notify Microsoft about this. Fortunately, on affected systems, uninstalling kb4100347 is enough to solve the install issue – the protection against Spectre BTI remains with update KB4346084!






Wanna go old registry method to disable it?
KB4346084 is pretty much 'beta' by microsoft.

The other alternative is registry method.
https://support.microsoft.com/en-my/help/40...rabilities-prot

QUOTE
To disable Variant 2: (CVE-2017-5715 – "Branch Target Injection") mitigation:

reg add "HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management" /v FeatureSettingsOverride /t REG_DWORD /d 1 /f

reg add "HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management" /v FeatureSettingsOverrideMask /t REG_DWORD /d 3 /f

To enable Variant 2: (CVE-2017-5715 – "Branch Target Injection") mitigation:

reg add "HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management" /v FeatureSettingsOverride /t REG_DWORD /d 0 /f

reg add "HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management" /v FeatureSettingsOverrideMask /t REG_DWORD /d 3 /f

reg add "HKLM\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Virtualization" /v MinVmVersionForCpuBasedMitigations /t REG_SZ /d "1.0" /f


Need to disable spectre protection. Reboot is required.

Normally, I would suggest deleting mcupdate_GenuineIntel.dll from system32, but since you mentioned you had updated the bios, this is not an option.

This post has been edited by JohnLai: Nov 13 2018, 09:17 PM
JohnLai
post Nov 13 2018, 09:04 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(biloxee @ Nov 13 2018, 08:51 PM)
?

Mine same like his and no issue. I don't see any of my CPU cores reaching 100%...

[attachmentid=10103473]
*
Impossible........when you run speedtest 800Mbps or LAN data transfer stuff, there will be one core/thread maxed utilization with haswell microcode 24 on win10. laugh.gif

edit: unless if spectre protection had been disabled in first place by you. tongue.gif

This post has been edited by JohnLai: Nov 13 2018, 09:05 PM
JohnLai
post Nov 13 2018, 09:47 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(cskean @ Nov 13 2018, 09:44 PM)
Already back to my own place. Anyway where can I check the remote access is enable or disable, any idea under which window/setting?
*
Disable WPS? Remember the WPS Reaver stuff?
How about changing the SSID? Remember the rainbow table for common ssid?
Use longer password? Remember GPU password brute force cracking is quite fast?
Disable TR-069? TM can't even secure its own network and you trust them to manage your router via TR 069?
Maybe change router admin login pass for obvious reason?
JohnLai
post Nov 13 2018, 09:51 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(hashtag2016 @ Nov 13 2018, 09:45 PM)
hi, KB4346084 was replaced by KB4100347
2018-09 Update for Windows 10 Version 1803 for x64-based Systems (KB4100347)
Last Modified: 10/9/2018
*
KB4100347 contains Spectre mitigation microcode.
KB4346084 contains spectre and Foreshadow mitigation microcode.

KB4100347 is the one pushed down to win10 users by microsoft.
KB4346084 is the latest, must be manually installed by user. But there is one quirk bug that prevent this installation as I quote from a blog post just now.


JohnLai
post Nov 13 2018, 10:14 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(hashtag2016 @ Nov 13 2018, 09:57 PM)
yes and no,
older version
2018-05 Update for Windows 10 Version 1803 for x86-based Systems (KB4100347)
windows10.0-kb4100347-x64_89793c02681ab9df96565ca4c5ee601e4475ec8f.msu

latest version @ 2018.10
2018-09 Update for Windows 10 Version 1803 for x64-based Systems (KB4100347)
windows10.0-kb4100347-v3-x64_8251e1f6e3d760e110b35af950f9acee5f4f6777.msu
*
Nope, that windows10.0-kb4100347-v3 only contains Spectre mitigation microcode and missing Spectre Variant 3a (CVE-2018-3640: "Rogue System Register Read (RSRE)"), Spectre Variant 4 (CVE-2018-3639: "Speculative Store Bypass (SSB)"), L1TF (CVE-2018-3620, CVE-2018-3646: "L1 Terminal Fault").

You can check the microsoft microcode version and compare it with intel microcode version. laugh.gif


EDIT:
https://support.microsoft.com/en-us/help/43...crocode-updates

https://support.microsoft.com/en-my/help/41...-windows-server

https://www.intel.com/content/dam/www/publi...te-guidance.pdf

This post has been edited by JohnLai: Nov 13 2018, 10:25 PM
JohnLai
post Nov 14 2018, 01:24 AM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(nyunyu @ Nov 13 2018, 11:21 PM)
Went ahead and uninstall update KB4100347 and install KB4346084. Checked PATCH ID and it changed to 00000096. I guess that means I'm all patched now right?

Can't say the same on my desktop, still showing 24
*
Ignore your desktop first.
Can you get full 800Mbps on TM speedtest via your patched laptop on wired now?


QUOTE(biloxee @ Nov 14 2018, 12:40 AM)
Hmmm.
[attachmentid=10103620]
*
hmm.gif confused.gif

Something is fishy here.
Cause your cpu utlization.....as if Meltdown protection also disabled. hmm.gif

What is your powershell output? I don't really trust inspectre result.
https://support.microsoft.com/en-us/help/40...ings-powershell

JohnLai
post Nov 14 2018, 10:42 AM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(hashtag2016 @ Nov 14 2018, 02:48 AM)
hi,
I was trying to explain why he failed to  install the KB4346084,
for some reason, ms-catalog replaced it with KB4100347 v3.
(have to upload the screenshoot , more crystal clear ... )

I do understand that it doesn't mean it can fully take over the role of KB4346084.

p/s: only CVE-2018-3640 is still  missing at the moment.
others CVE was covered by  latest Cumulative Update for Windows 10
(which is KB4462933 at the moment. which already replaced older Cumulative Update KB4343909 and KB4284835)
I think they will keep updating it from time to time.
*
Oh, that picture?
Let just say microsoft catalog and changelog kinda 'unreliable'. laugh.gif
Microsoft did updated KB4100347 to v3 due to overclocking+instability issue with broadwell-e.

According to https://portal.msrc.microsoft.com/en-us/sec...isory/adv180018 , "The required microcode is the same microcode that addresses CVE-2018-3639 and CVE-2018-3640"

The Summary of Intel microcode updates https://support.microsoft.com/en-my/help/40...crocode-updates
Microsoft should really fix the page formatting. lol.gif
First table = Spectre Variant 2 (CVE 2017-5715 [“Branch Target Injection”])
Second table = Spectre Variant 3a (CVE-2018-3640: "Rogue System Register Read (RSRE)"), Spectre Variant 4 (CVE-2018-3639: "Speculative Store Bypass (SSB)"), and L1TF (CVE-2018-3620, CVE-2018-3646: "L1 Terminal Fault").

QUOTE(nyunyu @ Nov 14 2018, 03:48 AM)
Speedtest via cable on the laptop gave me stable 550-ish. I think my speed might've capped at 500Mbps.

Test using both mikrotik 750Gr3 and C3150v2 also yield similar speed prior installing this microcode patch.
*
Hmmm hmm.gif , since microcode issue had been solved and you still get that same 500 Mbps to 600Mbps, you might wanna contact TM via livechat or 100 and request TM to 'refresh' your account. There were few users who got mis-capped at 450Mbps to 550Mbps (at PPPoE stage) and they got it solved after TM 'refresh' or 'reset' their unifi account status.
Hope the 'refresh' or 'reset' or whatever it is called by TM can solve your issue.

As for your haswell desktop, you can just rename mcupdate_GenuineIntel.dll at system32 and reboot. Be warned, the microcode is required for proper protection, it is not a good idea to disable it....

QUOTE(hashtag2016 @ Nov 14 2018, 04:50 AM)
Just plug the  LAN cable between your PC and notebook directly , and do some iperf  testing to test  the speed between 2 devices ...
maybe either one of your LAN canble is low quality or  broken...(u need to test the cable one by one for sure)
*
Agree, he mentioned he is using cat6 cable, but sometime shit happens. console.gif
In addition to that cat6e cable which he uses to connect his laptop to router, he also has to test the cable that connect BTU to router.
JohnLai
post Nov 14 2018, 12:43 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(blacktubi @ Nov 14 2018, 11:37 AM)
JohnLai
I always wondered how big is the impact of speed after the patch.

My 4200U laptop with SSD can speedtest at 1G just fine.
*
So far, the microcode patch issue only happen with windows 7 and 10. See for yourself on the impact of speed after latest microcode was applied.

I made some post on ivy bridge 3570k performance when the first spectre microcode was released.
https://forum.lowyat.net/topic/4547450

Then another posts on someone thread where the guy has haswell CPU.
https://forum.lowyat.net/index.php?showtopic=4649008
Had to inform the guy that latest microcode from intel somehow has less performance impact as I also have problem with the older haswell microcode for my i3-4330 and I5-4200U.

Linux and MacOS seem to have lesser performance penalty due to usage of retpoline.
However, after reading this https://bbs.archlinux.org/viewtopic.php?id=241738 , it seem as if the performance penalty also exists, although still less severe compared to windows hmm.gif .....someone might wanna reply at this archlinux forum......blacktubi, you wanna reply at that thread that the poster issue was caused by older microcode? tongue.gif Latest haswell ULT microcode is 24, version 23 is haswell ULT problematic patch.
JohnLai
post Nov 14 2018, 12:58 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(nyunyu @ Nov 14 2018, 12:52 PM)
Already called up TM about that. I think they might've capped me at 500Mbps.
https://forum.lowyat.net/index.php?act=ST&f...post&p=90909944
*
Good, remember to update us after your account 'refresh'.
Make sure you use your laptop with updated microcode for testing via wired once your unifi account is 'refreshed'! icon_rolleyes.gif



Moving on to your haswell CPU desktop.....
Disabling the microcode is not a good idea due to security concern.
Modding bios with latest microcode via UBU Tools is out of your league, maybe?

Perhaps a permanent pendrive stuck at your usb port?
https://www.techpowerup.com/forums/threads/...-loader.248858/
mega_shok.gif
JohnLai
post Nov 14 2018, 03:39 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(arrsoo @ Nov 14 2018, 03:28 PM)
Oh dang it! I tot you mentioned this microcode thing only affects win10 and not win7?  Or I could have misinterpret you wrongly.... sweat.gif
*
For one, Microsoft doesn't push microcode update for Win7, thus you shouldn't worry about it........unless if your hand is itchy and you wanna mod the bios, then you better make sure you select the correct microcode version.

Some mainboard vendors are very consumers friendly[I]......where vendors actually release bios with Spectre protection.......but they don't release bios with Spectre-NG protection (which somehow reduced the performance penalty). doh.gif
So, if you are one of 'lucky' user, the only choice is to either fully disable spectre mitigation via registry or use the Intel Microcode Boot Loader.


I modded the bios and so I know the problem also occurred for win7.


JohnLai
post Nov 14 2018, 04:00 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(arrsoo @ Nov 14 2018, 03:51 PM)
ok...cool! thanks
*
But just to be sure, you better check the microcode version via RWEverything --->Access--->CPU MSR
Then compare both CPU ID and PATCH ID (this is the microcode version) with information from Intel pdf sa00115-microcode-update-guidance. If the patch ID is same as 'Pre-Mitigation Production MCU', err......my condolences. console.gif

It is very strange that biloxee doesn't encounter this issue for his win10. hmm.gif

JohnLai
post Nov 14 2018, 04:23 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(nyunyu @ Nov 14 2018, 04:15 PM)
Just call up TM again and this time instead of straight away says "port issue", she went on and refresh the connection step by step. End result, still around 500-ish.
So, opened a report waiting for the tech to investigate.

As for my desktop, nah, I'll live with it. It's time for an upgrade anyway ^^.
*
Oh?
Now this is strange indeed.
Can you check if your laptop has temperature issue?

https://cpuid.com/softwares/hwmonitor.html
Run TM Speedtest and report back 'Max' value.

Cause there was a user who first only got 200-250Mbps, then after I dealt with the microcode issue, speedtest increased to 500-580Mbps.
Further testing revealed his CPU is overheating at 100'C = resulting in CPU throttling. Intel CPU will start throttling after reaching 100'C.
He blasted a table fan directly to the mainboard area and somehow able to keep CPU temp below 100'C, voila 800Mbps on speedtest.
JohnLai
post Nov 14 2018, 07:07 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(nyunyu @ Nov 14 2018, 05:29 PM)
CPU peaked at 93°C DL and 33°C UL. Room temperature 26°.

Laptop normal fan:
[attachmentid=10104348]

laptop with fan boost on:
[attachmentid=10104349]
*
hmm.gif
What is the peak sustained CPU frequency during DL on speedtest?

edit:
Cause laptop power management is different than desktop.
BIOS could limit your CPU frequency.
If you on battery, certain manufacturer actually set lower 'turbo boost' bins compared when power plug is connected.

edit2:
Do you ever get constant turbo frequency of 4.1Ghz during TM speedtest download?

This post has been edited by JohnLai: Nov 14 2018, 07:38 PM
JohnLai
post Nov 14 2018, 08:04 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(cskean @ Nov 14 2018, 07:41 PM)
Guys, how do I know whether my downloading speed is being throttled by unifi or not?

Unifi : 800MB
Router : TM TP Link C1200
Adapter : TP Link AC1900 T9UH
Laptop : i5-3210M, 8GB DDR3, 250GB Samsung Evo SSD

5GHz web speedtest : 280mbps down, 170mbps up.

Downloading a movie size around 800mb using IDM, speed average 9++ KB/sec ~ 1.2 MB/sec. If this is not the right speed, what is the best speed I should be getting with my current setup? I'm asking this because I hope to see great improvement when I switch my current router to Asus GT-AC5300...

I'm not quite sure is this the right speed I'm suppose to get out from my current setup. Highly appreciate sifus here and shred some lights on me~

TQVM...
*
Hmm? Have you set C1200 5ghz channel width to 80Mhz only?
JohnLai
post Nov 14 2018, 08:11 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(cskean @ Nov 14 2018, 08:07 PM)
Yes I did, just reconfirm this by logging into the router again to reassure~ tongue.gif

Any clue what is the best speed I should get from this my setup?
*
You sure? hmm.gif I am somehow skeptical because the speedtest kinda weird. What is the wifi link speed? (don't confuse this)
Next would be enabling 5ghz beamforming....which only available on EU firmware....

Other than that, nothing much you can do for wireless performance.

As for your downloading......are you seriously torrenting on IDM? doh.gif
JohnLai
post Nov 14 2018, 08:39 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(cskean @ Nov 14 2018, 08:32 PM)
Then what is the expected speedtest for such setup?

I'm sorry, but how/where can I know the "wifi link speed"

Enabling 5GHz beamforming, this is to be done on the router or my wifi adapter?

My C1200 V1 firmware is 1.0.0 Build 20180403 rel.22124 (EU)

And my T9UH V2 is using Archer T9UH(US)_V2_180319_Win, published 2018-03-27 (how do I check the driver version directly from my laptop? I got this from the website)
Is download movie considering torrenting? I've been using IDM cause it can resume downloads... Not really started to torrent yet....
*
Best condition = you can get 480Mbps via 5Ghz wifi

Wifi link speed also don't know doh.gif = https://www.tenforums.com/tutorials/75371-s...ndows-10-a.html

Your C1200 is using Malaysia firmware = give up on beamforming. Beamforming only available with EU region firmware. Your adapter supports beamforming by default. Your c1200 router.....sigh...

Nothing wrong with your T9UH, just orient the antenna correctly.

-.- You don't even know how you download the movie? DDL or BT? Even DDL server also can resume if server supports it.



JohnLai
post Nov 14 2018, 08:59 PM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(cskean @ Nov 14 2018, 08:54 PM)
Pardon me pls as I'm a light internet user all the while...

All the while I only know how to download movies from website, whereby when u watch it online, there will be a "small green colour triangle" by IDM, I click on it and download from there...

I know, there are better ways, but I didn't bother to learn previously... But now I'm picking up, hope sifus here can share & guide me.... tongue.gif
Alright, I know Unifi isn't providing quality router to their users, which is why I got mine during the 11.11 sales. While waiting for it to arrive, need to enhance my knowledge a bit so that I won't mess up when setting up my next router...

From your link, Option One : 867Mbps. Option Five : 54/867(Receive/Transmit).

Is reading from Option Five normal? Or should get higher reading on the Receive side?
The distance between my router and adapter is about 8ft+-, nothing much blocking between both of it. Will try it out once I clear up my laptop unused data. Too much files hanging around...
*
Nothing wrong with router and adapter. Everything is normal.
Guess 280mbps down, 170mbps up are your best case with current combination.

The torrent or DDL on the other hand, you gotta solve it yourself.
JohnLai
post Nov 15 2018, 10:32 AM

Skeptical Cat
*******
Senior Member
3,669 posts

Joined: Apr 2006
QUOTE(nyunyu @ Nov 15 2018, 02:43 AM)
On battery, speedtest startup with cpu peaked to 1.92Ghz and stable around 1.6Ghz during test.
On power, it peaked to 2.2Ghz and stable around 2.09Ghz during test.
Never once it goes more than 2.2Ghz during speedtest.
*
Are you serious? blink.gif
If you change the power profile to high performance, can it get at least 4.1ghz on any one of cores? (Hwmonitor)

10 Pages « < 2 3 4 5 6 > » Top
Topic ClosedOptions
 

Change to:
| Lo-Fi Version
0.1906sec    0.58    7 queries    GZIP Disabled
Time is now: 5th December 2025 - 01:48 AM