I thought the benefit of using cut-through is reduced latency, not higher throughputs. Besides, if all the competing models use the same chips, why does the Linksys stand out?
Maybe because it is the first router to use BCMCTF?
This is new code, not seen before in other Broadcom GPL sources.
If I'm not mistaken then there should be a setting in the 4200 web gui for disabling CTF (the source code says there is one) so it should be easy to test how much CTF does to the routing bandwidth. _________________ Kernel panic: Aiee, killing interrupt handler!
I thought the benefit of using cut-through is reduced latency, not higher throughputs. Besides, if all the competing models use the same chips, why does the Linksys stand out?
Maybe because it is the first router to use BCMCTF?
This is new code, not seen before in other Broadcom GPL sources.
If I'm not mistaken then there should be a setting in the 4200 web gui for disabling CTF (the source code says there is one) so it should be easy to test how much CTF does to the routing bandwidth.
I have been playing around with the webgui for hours, and I can assure you there is no such setting.
Maybe because it is the first router to use BCMCTF?
This is new code, not seen before in other Broadcom GPL sources.
If I'm not mistaken then there should be a setting in the 4200 web gui for disabling CTF (the source code says there is one) so it should be easy to test how much CTF does to the routing bandwidth.
Where did you get the source from? _________________ Read the forum announcements thoroughly! Be cautious if you're inexperienced.
Available for paid consulting. (Don't PM about complicated setups otherwise)
Looking for bricks and spare routers to expand my collection. (not interested in G spec models)
I have been playing around with the webgui for hours, and I can assure you there is no such setting.
Ok, the source code for the ctf driver reveals that there is a function call to turn ctf on or off.
The httpd binary blob which contains the web gui code has display strings for "ctf_disable" , and "CTF disabled" so I guessed that it was a gui setting.
I'm still quite confident that BCMCTF is the reason for the huge routing performance change.
I have followed Broadcom's driver changes during the last 2 years especially looking at the ethernet, wifi, and dma drivers.
There were last year lots of changes to the dma driver which should have given somewhat better routing performance but there is not much difference between these drivers in E4200 and E3000 or E1000v2.
Those small differences can not account for the big jump in routing performance of E4200.
I have compared the E4200 ethernet driver with the most recent one I could find from other GPL codes and that was version 5.60.120.1 from E1000v2, the only difference between them is the calls to ctf routines in E4200.
The E4200 driver is version 2010.09.20.0 so it doesn't follow Broadcom standard way of driver version numbering, a special driver made for Linksys that other manufacturers wont get through Broadcoms driver support program? _________________ Kernel panic: Aiee, killing interrupt handler!
I thought the benefit of using cut-through is reduced latency, not higher throughputs. Besides, if all the competing models use the same chips, why does the Linksys stand out?
Maybe because it is the first router to use BCMCTF?
This is new code, not seen before in other Broadcom GPL sources.
If I'm not mistaken then there should be a setting in the 4200 web gui for disabling CTF (the source code says there is one) so it should be easy to test how much CTF does to the routing bandwidth.
I have been playing around with the webgui for hours, and I can assure you there is no such setting.
Please test routing throughput with CTF enabled and disabled if you can. You'll need two gigabit nic's to do so. _________________ Read the forum announcements thoroughly! Be cautious if you're inexperienced.
Available for paid consulting. (Don't PM about complicated setups otherwise)
Looking for bricks and spare routers to expand my collection. (not interested in G spec models)
I thought the benefit of using cut-through is reduced latency, not higher throughputs. Besides, if all the competing models use the same chips, why does the Linksys stand out?
Maybe because it is the first router to use BCMCTF?
This is new code, not seen before in other Broadcom GPL sources.
If I'm not mistaken then there should be a setting in the 4200 web gui for disabling CTF (the source code says there is one) so it should be easy to test how much CTF does to the routing bandwidth.
I have been playing around with the webgui for hours, and I can assure you there is no such setting.
So my guess was more accurate than your check.
I guess you don't mind then if I take your future claims with a pinch of salt.. _________________ Kernel panic: Aiee, killing interrupt handler!
Last edited by LOM on Sat Jan 29, 2011 6:15; edited 1 time in total
I just wish I could find out how to see a list of connected wireless devices!
About the best I can say at this point is that my Roku XD|S shows the E4200 to have more signal strength than my WNDR3700 even though the 3700 is 20ft closer and has a bit less obstructions. Now to test some hd video streaming while on E4200 wireless direct vs. wired to one of my DDWRT bridged 3700s.
I thought the benefit of using cut-through is reduced latency, not higher throughputs. Besides, if all the competing models use the same chips, why does the Linksys stand out?
Maybe because it is the first router to use BCMCTF?
This is new code, not seen before in other Broadcom GPL sources.
If I'm not mistaken then there should be a setting in the 4200 web gui for disabling CTF (the source code says there is one) so it should be easy to test how much CTF does to the routing bandwidth.
I have been playing around with the webgui for hours, and I can assure you there is no such setting.
So my guess was more accurate than your check.
I guess you don't mind then if I take your future claims with a pinch of salt..
I stand corrected (I was going by memory). Take as much salt as you like
Here's what the Help says:
CTF
The CTF (Cut-Through Forwarding) option improves the efficiency of packet forwarding between the local network and the Internet. To use the CTF option, keep the default, Enabled.
Please test routing throughput with CTF enabled and disabled if you can. You'll need two gigabit nic's to do so.
Here ya go, pretty dramatic difference.
I did an SMB file transfer between my Core 2 Quad Windows 7 desktop that rocks a velociraptor drive and my new i5 6,2 Macbook Pro with a Seagate 7200RPM Hybrid HD.
That's a pretty dramatic difference. Unfortunately the CTF magic is another binary blob so we can't see what it is they're doing. I found references to the hndctf.h header file in the wireless driver blobs so it might be improving wireless throughput as well. _________________ Read the forum announcements thoroughly! Be cautious if you're inexperienced.
Available for paid consulting. (Don't PM about complicated setups otherwise)
Looking for bricks and spare routers to expand my collection. (not interested in G spec models)
That's a pretty dramatic difference. Unfortunately the CTF magic is another binary blob so we can't see what it is they're doing. I found references to the hndctf.h header file in the wireless driver blobs so it might be improving wireless throughput as well.
The CTF packet handler is used by the wl driver and the et driver.
It is also dependent on support in the dma handler for high speed shuffling of data from one place to another.
This is not different from the normal packet handler (pktpool_ or pktlist_ routines in bcmutils.c) that one is also used by wl and et, and it also uses dma.
Disabling CTF in the gui means switching back to use the normal packet handler instead of the CTF.
I don't think we will see any improvement in wireless throughput until we reach the limit for the normal packet handler and that is from what I can see somewhere around 225Mbit/s. _________________ Kernel panic: Aiee, killing interrupt handler!
I don't think we will see any improvement in wireless throughput until we reach the limit for the normal packet handler and that is from what I can see somewhere around 225Mbit/s.
Maybe so, it even got worse performance on the simultaneous 2.4GHz+5GHz throughput test in the followup review. I was just thinking about how current models exhibit high CPU load from soft interrupts that seems to cap wireless throughput, and if CTF can reduce that load then the cap should be raised. Now I'm wondering if disabling CTF in the stock firmware might increase that simultaneous wireless throughput. Perhaps it's still buggy for wireless. _________________ Read the forum announcements thoroughly! Be cautious if you're inexperienced.
Available for paid consulting. (Don't PM about complicated setups otherwise)
Looking for bricks and spare routers to expand my collection. (not interested in G spec models)