Kitz Forum
Broadband Related => Broadband Hardware => Topic started by: snadge on June 05, 2012, 08:26:35 PM
-
can SNRM be set on 6368 chipset does anyone know?
thanks
-
I tried a while back on my VDSL2 connection - no go.
"xdslcmd configure --snr" with different values had no effect.
-
can SNRM be set on 6368 chipset does anyone know?
thanks
Yes, on the DGND3700 (which was a 6368 chipset) for example, putting adslctl configure --snr VALUE set the SNRM. It also worked using xdslcmd configure --snr VALUE on a Huawei HG612 on an ADSL1 (Max) connection.
-
ive been told elsehwere you can - it depends on your connection too - of your locked to a profile it will not go above that profile speed regardless of where SNRM is - maybe this is what happened to you BA..?
-
ive been told elsehwere you can - it depends on your connection too - of your locked to a profile it will not go above that profile speed regardless of where SNRM is - maybe this is what happened to you BA..?
That's a point. It may well have been during one of my many "stuck" DLM profile periods.
I'll try it again in a few days (I just want to prove my connection's stability over quite a few days first).
I do know the HG612 can hang on at really low SNRM values before resyncing.
I have seen it running at 1dB or less for hours on end in the past.
-
the Netgear DGND3700 is a 6368 chipset too - ive read they are 400mhz but one site (WRT Wiki) says its 300Mhz - im sure it was broadcom paper too that said they were 400mhz Viper CPU's ...but cant find it
-
Here is a Broadcom page (http://www.broadcom.com/products/Broadband-Carrier-Access/xDSL-CPE-Solutions/BCM6368) on the BCM6368. Is it any help? :-\
-
it doesnt say...
but i found this from one of asbo's posts (and ive read other places that say its Dual-Core 400Mhz chip)
CFE version 1.0.37-102.6 for BCM96368 (32bit,SP,BE)
Build Date: Mon Mar 2 15:45:35 CST 2009 (root@localhost.localdomain)
Copyright (C) 2000-2008 Broadcom Corporation.
Parallel flash device: name MX29LV640BT, id 0x22c9, size 8192KB
CPU type 0x2A031: 400MHz, Bus: 160MHz, Ref: 64MHz
CPU running TP0
Total memory: 33554432 bytes (32MB)
Boot Address 0xb8000000
-
Looking closely at the first line of the quotation, I see --
CFE version 1.0.37-102.6 for BCM96368 (32bit,SP,BE)
So BCM96368 or BCM6368, that is the question. ???
-
The same question applies to other chips in the series. Broadcom refers to them without the leading 9, but there are many references to what seems to be the same chip but with the leading 9. I suspect that the 9 is just a packaging version number or something like that.
-
Ah, thank you Eric. That is useful to know. :)
-
BCM6368 = chipset
BCM96368 = reference board design
Eg BCM96368MVNGR = VDSL2, 4 x 100Mbps ethernet, 802.11n wireless, USB and two optional FXS ports (VoIP ports)
Hope that clears it up :)
-
Thanks rizla :)
-
I think the 9 prefix always designates a reference board design in the Broadcom part numbering scheme. I'm not 100% on that though as I've never seen a definitive part numbering guide from them.
Also worth noting that the BCM6368 chipset has a GigE switch embedded but the reference board design (BCM96368MVNGR) only implements fast ethernet (100Mbps) physical connections.
The reasoning behind that seems to be an EU guideline on power consumption which suggests (haven't done the maths) that you can't have GigE and FXS ports on the board and stay within the power consumption guidelines. Use of FXS ports is virtually unheard of with any UK ISP but they do get used in other EU countries (Teliasonera springs to mind).
Probably shaves a couple of pennies off the hardware cost too. Oh and it will no doubt reduce support costs to ISPs - I wouldn't want to be the one explaining to a clueless end-user why they can't get 1Gbps on the LAN ;)
-
thanks Rizla
just too let you all know I got reply from Broadcom who refused to release details about BCM 6328 ..lol
ah well, tried!!
-
Broadcom prefer to supply boards/systems to customers rather than discrete parts in my experience. As such their documentation isn't what you'd expect from a "normal" semiconductor designer - like Intel for example who have superb documentation.
Only interesting thing about the (9)6368 is the gigabit switch which appears not to be used in any implementation, the rest is pretty meh - for example the VDSL2 only goes to 100Mbps. Capable enough no-frills board, never going to be widespread outside ISP-branded routers IMHO.
-
I guess that you have already seen it, Snadge but, for completeness, the only information that Broadcom have published on the BCM6368 is a solitary web-page. [1]
:(
[1] http://www.broadcom.com/products/Broadband-Carrier-Access/xDSL-CPE-Solutions/BCM6368
-
I guess that you have already seen it, Snadge but, for completeness, the only information that Broadcom have published on the BCM6368 is a solitary web-page. [1]
:(
[1] http://www.broadcom.com/products/Broadband-Carrier-Access/xDSL-CPE-Solutions/BCM6368
There's a Broadcom Product Brief for the 6368, but SFAICS, it just repeats what is on that web page.
http://pdf.eccn.com/pdfs/Datasheets/Broadcom/CM6368.pdf
Broadcom must produce some very detailed datasheets on its products. Though it seems these documents are not supplied to the lumpenproletariat. Even OEMs like Huawei are required to sign a Non-Disclosure Agreement (NDA) before the datasheets are furnished to them.
However, peeping over the Great Firewall of China, I spy with my little eye..some leaked documents on the 6368, on a certain 汉语/漢語 website. It is a website that is censored, unusually, not to those within the great nation, but to those outwith. So as not to ruffle the corporate feathers of big Western tech firms like Broadcom, perhaps?
cheers, a
-
However, peeping over the Great Firewall of China, I spy with my little eye..some leaked documents on the 6368, on a certain 汉语/漢語 website. It is a website that is censored, unusually, not to those within the great nation, but to those outwith.
:hmm: I think we need the services of a cat-friendly special agent to leak all those documents to the Occident. :)
-
@ rizla - the Netgear DGND3700 has a Gigabit switch on it and it uses 6368..? or are you on about gigabit xDSL driver?
@ asbokid - you dunno any info on 6328 do you? speed? multi-core? just for comparison thats all
also, are these chipsets just designs licensed out that may be changed slightly by other manufacturers? bit like graphic cards etc... e.g. Speedtouch 6348 be different too Netgear 6348..?
-
@ rizla - the Netgear DGND3700 has a Gigabit switch on it and it uses 6368..? or are you on about gigabit xDSL driver?
@ asbokid - you dunno any info on 6328 do you? speed? multi-core? just for comparison thats all
also, are these chipsets just designs licensed out that may be changed slightly by other manufacturers? bit like graphic cards etc... e.g. Speedtouch 6348 be different too Netgear 6348..?
Hi Snadge,
Sorry, I've got no more info on the Broadcom 6328 SoC (System-on-Chip). If you have a device with that particular CPU in it, maybe you can get some info from the Linux kernel, via the proc filesystem..
This from an HG612..
# cat /proc/cpuinfo
system type : CHIP96368
processor : 0
cpu model : BCM6368 V3.1
BogoMIPS : 398.95
wait instruction : no
microsecond timers : yes
tlb_entries : 32
extra interrupt vector : no
hardware watchpoint : no
ASEs implemented :
VCED exceptions : not available
VCEI exceptions : not available
unaligned exceptions : 646681218
#
The BogoMIPS metric (the 'bogus' measure of million instructions per sec) is Linus Torvalds' joke at the meaningless of core speed measurements, although it's still useful for comparisons between CPUs.
The 6328 will almost certainly be a dual MIPS32 core. It should be possible to discover more about the CPU - e.g. the size of instruction and data caches - by studying the kernel logs from booting the device.
The Broadcom System on Chips are very highly integrated. From the diagram below of the 6368 we can see that the SoC integrates a DSL Analog Front End (AFE) and line driver, ethernet switch controller, USB host controller, various data bus controllers and bridges, and even power supply regulation circuitry.
The OEM, whether that's Huawei, Netgear, ZTE or whoever, only has to bolt on a PSU, an oscillator for clocking, some flash memory (a choice of serial or parallel), some volatile memory, and they have a working system. It's almost like Lego!
Broadcom supplies the OEMs with a Reference Design - these are board level designs - detailed schematics, perhaps even PCB artwork, with full pinouts for the CPU and details of the external components and requirements, and constraints, etc.
The OEM then designs a PCB to accommodate the Broadcom SoC and its support components, and to meet those specs. Even with a small team, that stage probably takes just a few weeks. And away they go, hopefully with a working router board.
Broadcom even supplies a ready-rolled cross-compiler toolchain for building a Linux kernel and drivers for all its SoCs like the 6368. Via a simple menu system, the toolchain is configured so that it builds a Linux kernel and device drivers to match the components that are on your development board, as well as the userspace binaries that will be flashed into the firmware.
So in answer to your question. Since the 6368 is so integrated, there's little scope for improving on the Reference Design. One router powered by a 6368 is likely to perform much the same as any other. There might be occasional updates to the CPU to correct silicon errors, and over time, improvements to the DSL hardware driver. And theoretically those updates could make one 6368 board perform better than another. But in practice not by much.
cheers, a
(https://forum.kitz.co.uk/proxy.php?request=http%3A%2F%2Fwww1.picturepush.com%2Fphoto%2Fa%2F8465154%2F480%2Fsnadge-stats%2Fbcm6368-system-diagram.png&hash=79f1b4e9ed64ddc107873698bc642a8c84ef4365) (http://picturepush.com/public/8465154)
-
thanks asbo for that info :)
how do you know it will be dual-core? (6328) ... is the 6328 a 'new' chipset when compared to 6338,6348,6358 ..? just it has 'lower' number as if to imply its older? 6348 (255Mhz) and 6358 (300Mhz) are single core arent they? I know the 6362 is dual-core (400Mhz) - its just ive noticed over my 3 routers that the faster/better the chip, the better ADSL and Wi-Fi performance I get.. and there was no info on the 6328 (sky's D-Link 2640S)
6348 (255Mhz / 1 core) = 1.6ms wifi ping to router / 15.5Mbps (Netgear DG834GT)
6358 (300Mhz / 1 core) = 1.1ms wifi ping to router / 15.8Mbps (Netgear DG834N)
6362 (400Mhz / 2 core) = 0.7ms wifi ping to router / 16.6Mbps (Sagem F@ST 2504N)
-
thanks asbo for that info :)
how do you know it will be dual-core? (6328) ... is the 6328 a 'new' chipset when compared to 6338,6348,6358 ..? just it has 'lower' number as if to imply its older? 6348 (255Mhz) and 6358 (300Mhz) are single core arent they? I know the 6362 is dual-core (400Mhz) - its just ive noticed over my 3 routers that the faster/better the chip, the better ADSL and Wi-Fi performance I get.. and there was no info on the 6328 (sky's D-Link 2640S)
6348 (255Mhz / 1 core) = 1.6ms wifi ping to router / 15.5Mbps (Netgear DG834GT)
6358 (300Mhz / 1 core) = 1.1ms wifi ping to router / 15.8Mbps (Netgear DG834N)
6362 (400Mhz / 2 core) = 0.7ms wifi ping to router / 16.6Mbps (Sagem F@ST 2504N)
I don't know.. it was just a guess, based on its apparent similarity to the 6368. Looks like that was wrong though - this guy says it is single core [1] It does seem like a retro-step.. Even the 6358 was/is dual core according to Ralf Baechle [2] So unless Broadcom has done some amazing streamlining, it's difficult to see how it could perform as well. The BCM4313 802.11 transceiver used in the DSL-2640s is fairly new though, so maybe that could feasibly result in better wifi performance. [3]
There are a few mentions of the 6328 in the openwrt source, so maybe checkout the development branch to see what is in there. [4]
cheers, a
EDIT:
Jonas Gorski, one of the developers who has been working on the 6328, reports that it does have a second core but it is "disabled in hardware". [5]
[1] https://forum.openwrt.org/viewtopic.php?pid=169715
[2] http://www.linux-mips.org/wiki/Broadcom_SOCs#ADSL_series
[3] http://www.broadcom.com/products/Wireless-LAN/802.11-Wireless-LAN-Solutions/BCM4313
[4] http://dev.openwrt.org
[5] http://page.mi.fu-berlin.de/jgorski/openwrt/devices.html
-
haha Ive read all those pages too - i didnt realise I had a reply on [1] hehe, thanks for bringing it too my attention, it was the 6362 that had increased wi-fi performance for me , my friend has a D-Link on 6328 which is a sky router
I also read that the 6358 is dual-core, one for OS and one for DSP - would the OS use up much CPU time..? would seem best to share everything across 2 cores..
its a shame we can know so little about the components that power our routers, you cant even find routers with a chipset you want!! I wouldnt mind a 6362 chipped router (unlocked Sagem F@ST 2504N) cos it performs great, i just cant tweak it!! :(