this sucks, i bought one of these as well, is there a way to force the issue for testing?
Point a blow dryer at it and make sure it only goes up to 55C (/system health print).
To the OP, please try down-clocking the CPU to 600Mhz (/system routerboard settings print)
Low speeds from distant servers can be explained by packet loss.
See the formula for throughput of a TCP connection:
http://en.wikipedia.org/wiki/TCP_tuning#Packet_loss
Even small packet loss like 0.1%, combined with typical RTT over the Internet, is significant. TCP just works that way, remember it was invented when speeds were much lower than they are today. Multiple parallel TCP connections will give better results.
But why it is correlated with Mikrotik’s temperature? When it’s relatively cool, I have 0 problem. When it’s relatively hot I have problems only with distant servers. EDIT. ok, silly question. The unit itself may be the packet loss source. Esp. when the temperature is high.
But, how does the Speedtest.net show better results even with distant servers in that case? EDIT. another silly question: Speedtest.net uses multiple connections.
Well… This sucks!
Packet loss increasing with temperature looks very much like a hardware issue - marginal Ethernet PHY, not likely to be fixable by software. Could be bad chips, bad capacitors (more likely ceramic ones, as electrolytics are worse when cold) etc. See if you could get more stable throughput by changing from 1G to 100M link.
What rate of packet loss is considered as acceptable? From the router’s terminal I’ve done ping test to the server of my ISP and got approx 0.05% packet loss (out of 25000 packets of size 1460 Bytes; the average ping was 1ms; the board temperature was 54C).
The average ping to European servers (I have an FTP server in Netherlands; and I am from Armenia) is 68ms.
After entering these values (MSS 1460; RTT 68; Loss 0.05%) in this calculator (http://www.switch.ch/network/tools/tcp_throughput/) I get network limit: 7.68 Mbit/sec. Which more or less matches with the speed drop I get at that temperature.
So the question is: is this a warranty case?
Also, do the previous boards also suffer from this? Probably less emphasized because those were little bit cooler?
Ethernet, if done properly and not congested, should have practically zero packet loss (OK, nothing is perfect so let’s say 0.0001% or one in a million). The square root in the formula means for 10 times more speed, you need 100 times lower loss. Yes, I’d suggest to RMA it. I’ll have to test my own hAP-ac too to see if it’s affected. Temperatures in this range should have no effect on properly working electronics, something is seriously wrong with just these boards.
I’ve just found an interesting article about one possible cause (which may or may not be the case here, it’s just an example of possible caveats that can affect any modern electronic devices):
Just to clarify, does this only affect ethernet ↔ ethernet connectivity? Eg if using the hAP AC solely as an access point bridged onto ether1, there won’t be any issues?
received my unit, its a EU model, i changed nothing outside of the default config, short of wireless width settings and ssid names,
upgrade to 6.35.1 3.31 with no issue
ran iperf with timer for 600 seconds with 11 parallel streams started at 44C ambient temp. eth port3,4 in switch mode. test ran at 940mbit for the duration of the test, while reaching a total temperature of 58C with 0 issues
then i ran it again over wireless 7260 intel card 2x2 866 link, and it was stable around 250mbit/32MB for the entire test as well all the way up to 57C with no loss.
outside of my voltage indicating 16v when in reality its 23, and a changing color led for wireless. i don’t seem to be affected
if there is anything else to test lmk
- connect through any ethernet port using Winbox, see if it disconnects you in every few minutes.
i do not have winbox disconnect problems, winbox has been open well over a hour with 0 issues, its generally all i use for configuration
This is what it is sounding like to me, which probably means it is an issue just with transmitting on ethernet port, not receiving. RX from WAN → TX to WLAN == good speed, RX from WAN → TX to LAN == bad speed.
If this is the case, then it could also be that download from the internet is very fast, but upload is being constrained. But you might not notice a problem unless you have a very fast symmetric internet connection from your provider.
I just got my hAP and have not gone through testing yet, but you can be sure that I will try to be very thorough after reading through this thread…
– Nathan
If you experience this kind of problem, please return the unit to the seller for RMA processing. We are looking into this issue and it would help if we could have some of these devices back for more testing.
I think this might explain the low stock levels. I was hoping for an upcoming flash upgrade.
Any news on this? Was this a problem batch or a manufacturing defect or something else? I’m hoping to deploy more hAP ACs very soon and some of the environments can get quite warm in the summer.
I was just thinking about this problem.
Could people whom are seeing this problem try to turn on flow-control on all ports (Auto should be enough). I wondering if that makes a difference. I always turn flow-control on to auto on all ports per default, so that might change something?
Is there any news on fixing of this problem?. I am planning to WiFi cover a building with 65 APs for different uses connecting Internet users, corporate connections and machine conections. I have been successfully using RB951G for this kind of projects but for new investments now AC is a must and natural evolution should be hAP ac. These performance problems mixed with the supply shortage make me doubt if these is a stable product. I could not find any reference of successful deployments around Spain.
I would thank you to provide information on current stability, successful projects deployed and/or plans to evolve the product to a more stable release.
Hi LSV,
why not thinking to use wAP AC ( http://routerboard.com/RBwAPG-5HacT2HnD) instead of hAP AC ?
The wAP AC It’s more suitable for APs.
Best Regards,
Daniel
Hi pateutz
I just ordered one to make some tests.
Hi LSV,
so have you test the product ? It’s ok ?
Best Regards,
Daniel