Broadcom Vs Qualcomm Temperatures

[microchip]

Hi,

I've been wondering for some time now, but couldn't find anything on the question which goes like this:

I own both a Broadcom-based router and a Qualcomm Atheros-based one. The Broadcom is the ASUS RT-AC66u while the Atheros one is the TP-Link Archer C7 v1. Both routers have similar specs (MIPS CPU, N/AC WiFi, 256 RAM, etc).

According the the openwrt wiki page where it lists the HW details, the Archer C7 has a faster CPU (720 MHz vs 600 MHz for the AC66U). The biggest difference between the routers is that one uses Broadcom stuff, the other Qualcomm Atheros.

The Broadcom ASUS runs much much hotter than the Archer C7. The latter even doesn't have any cooling inside of it (no fan, no heatsink, etc... yes, I've opened it up). There are no sensors on the Archer C7 GUI but just touching the router while it idles, it's much much cooler than the ASUS. The AC66U is pretty hot, even while idling, when touching it and I know from pics I've seen it has a (moderately large for a router) heatsink too...

So, why the hell is the Broadcom one running so hot while the QCA one doesn't even need any cooling? Can anyone shed some light on this?

Thanks

 

[azazel1024]

At a guess the difference in radios? There could be other hardware releated reasons, just because they are both MIPS SoCs doesn't mean they are the same line up or even necessarily the same lithography.

One thing I will say in my personal experience is that TP-Link seems to generally be very low power compared to some routers, especially Asus.

Asus stuff seems to be right near the top of power consumption and TP-Link seems to be near the bottom. Of course, capability wise, that might not stack up with performance, but power consumption, that seems to be true.

Generalized numbers, but from doing some digging, the C7 archer V1 seems to use somewhere in the 7-8w range, and the AC66u is more in the 13-15w range.

I know my, granted slightly older (not THAT old) and only N600, WDR3600 only use 3.6w when nothing is plugged in at idle and 4.5w with both WAN and one LAN port plugged in. It barely gets warm even after hours of being plugged in.

 

[azazel1024]

That would be one thing I'd like to see SNB test in their reviews, power consumption figures of the various devices. Seems like it would be relatively easy to test. Just set a baseline workload for each one and test the power consumption. Baseline could simply be WAN and one LAN plugged in, all radios turned on and nothing else, or with a HDD plugged in, or streaming max data or some combination. Probably just WAN and LAN plugged in, radios on as that is probably what most routers are going to be doing 80+% of the time they are on.

Probably most users aren't going to care, but a few might, especially if they look at two routers and the other performance figures are vaguely the same, but one uses 5w and one uses 15w. That isn't the world, but for a device left on 24/7, it adds up over time. Especially if it is a user thinking of needing 2 or 3 or 4 with some of them as APs.

 

[microchip]

Thanks,

I'm not assuming that since both use MIPS SoCs they'll have comparable heat output - I'm trying to take the whole picture.

I'm not much of electrician/hardware expert, but can a few Watt difference create such a big difference in heat output? Also the Broadcom radio's seem to generate a lot of heat compared to the QCA ones so I'm a bit at a loss as to why that might be and why they need heatsinks while QCA doesn't - has QCA found a way to make their chips much more efficient/less power consuming than one of the leaders in networking stuff like Broadcom?

It appears that Broadcom is AMD and QCA is Intel wrt to consumption and heat output (just a comparing joke)

 

[azazel1024]

QCA, AFAIK, is using a smaller process node on their radios, which no doubt helps. They also might generally have a more efficient hardware offloading for a lot of functions, so the CPU itself might not be working as hard most of the time. Dunno.

For heat output, yes. It makes a tremendous difference.

If you double the energy consumption, you double the temperature. the other wrinkle is, ICs use more power the hotter they get. It isn't a hugely pronounced difference, but an IC at 40C might use 10-15% less power than one at 90C. So the hotter the chip, the more power consumption it uses.

Also depends on where the heat production is. Some of the bits just aren't going to vary that much, for example the LAN/WAN ports on most current products for gigabit silicon use roughly .5w per port, give or take a couple of tenths. So between an 8w Archer C7 and a 15w Asus AC66u, it isn't simply that everything is using double the power, only a few components are using more power, but those components are likely using more like 2.5-3x the power, as some stuff, like USB ports, switch module, LEDs, etc are all likely using almost the same power between both routers. So the increase in power is likely coming from only a couple of bits, the SoC and the radios.

So something like the SoC and/or radios might really be 2.5-3x the power consumption and thus heat.

Generally these chips have a certain maximum thermal operation temperature and need to be design to operate at the top end of their environmental range, which is often around 100-105F. In both cases they might be designed for 90C operation, IE the SoC itself shouldn't exceed that temperature. You can see where the heat sink comes in. If you have twice the thermal load, if the TP-Link chip was at, say, 80C with 105F ambient (lets say 80C for some safety overhead before throttling), if you doubled the heat, you are going to have to find a way to extract it and still have the same environmental operating conditions.

The CPU might stay at the same temp with a heat sink on it between the two, but you are still generating something like twice the heat in the casing that has to be gotten rid of, so the casing is going to run twice as hot.

A low power router might be 8-10F above ambient, which might barely be noticable to the touch. A medium powered router might be 15F above ambient, which would be rather noticable. A high powered router might be 25-30F above ambient, which will feel HOT under most circumstances.

Its a smaller area, but try grabbing the heat sink on something like a 15W LED light bulb, it feels pretty darned hot right? Just make that plastic and expand the area a little and you have a 15W router, as the router casing is acting as the heat sink, with some minimal passive convective cooling.

I can tell you my tablet gets kind of warm after an hour of gaming or so, with a z3740 in it and total power budget of maybe around 5W in the thing. It is noticably warm, but by NO means does it feel hot. It feels about as warm as my WDR3600 gets actually (slightly noticable, a little more so right over the SoC, but just feels slightly warm). My laptop on the other hand, probably is drawing around 20W all told when gaming (17w ULV Ivy chip in it, but include board components, SSD, RAM, etc, I assume over 20w while gaming, maybe up near 25w). It does have active cooling, but parts of it get HOT after an hour or two.

Granted, we are talking about a 4 fold difference in power consumption, maybe 5 fold, but at the same time, one has active cooling to dump a moderate fraction of that heat, and one does not (one also has a partial metal case to help dump the heat faster, the laptop, and the other is all plastic/glass, the tablet).

 

[microchip]

Thanks azazel, great info. Got a lot more insight now into how it all "works"