[ath9k-devel] AR Chipset Differences

[ath9k-devel] AR Chipset Differences

[Galen]

I'm trying to determine the differences in features of the various Atheros chipsets supported by ath9k. Please note, I have only chosen dual band parts with at least 2 spatial stream support, as the single band parts are generally subsets of the dual band parts and the <2 spatial stream parts are generally extremely value-oriented (or power oriented.)

Gen 1 - AR5008:
AR5416+AR5122 - 2x2 dual band, PCI
AR5416+AR5133 - 3x3 dual band, PCI
AR5418+AR5133 - 3x3 dual band, PCIe

Gen 2 - AR9001:
AR9160+AR9104 - 2x2 dual band, PCI
AR9160+AR9106 - 3x3 dual band, PCI

Gen 3 - AR9002:
AR9220 - 2x2, dual band, PCI
AR9280 - 2x2, dual band, PCIe

Did I miss any 2x2 or better 802.11n radios?

Questions:

1) What differences exist between the AR5133 and AR9106? If the AR9106 offers 3x3 with superior performance to the AR5008, why didn't they ever offer any kind of PCIe option?

2) What is the advantage of the AR9002 family over the AR9001 family? Obviously, the AR9002 is a single chip solution, likely reducing cost, power and size. But is there any improvement to radio functionality or other features?

Note, I am interested in 802.11n features supported, quality of implementation, relative effective sensitivity, pretty much everything on the radio / DSP / layer 1 angle of things. I am curious as I want to be able to make the most appropriate hardware choices, and frankly, testing these chipsets to try and figure this out isn't the best way to do this based on my experiences thus far...

-Galen

[ath9k-devel] AR Chipset Differences

[Luis R. Rodriguez]

On Fri, Feb 26, 2010 at 08:08:31AM -0800, Galen wrote:
> I'm trying to determine the differences in features of the various Atheros chipsets supported by ath9k. Please note, I have only chosen dual band parts with at least 2 spatial stream support, as the single band parts are generally subsets of the dual band parts and the <2 spatial stream parts are generally extremely value-oriented (or power oriented.)
> 
> Gen 1 - AR5008:
> AR5416+AR5122 - 2x2 dual band, PCI
> AR5416+AR5133 - 3x3 dual band, PCI
> AR5418+AR5133 - 3x3 dual band, PCIe
> 
> Gen 2 - AR9001:
> AR9160+AR9104 - 2x2 dual band, PCI
> AR9160+AR9106 - 3x3 dual band, PCI
> 
> Gen 3 - AR9002:
> AR9220 - 2x2, dual band, PCI
> AR9280 - 2x2, dual band, PCIe
> 
> Did I miss any 2x2 or better 802.11n radios?

Yeah our AR9287 is 2x2 single band IIRC.
There is also our USB 2x2 AR9170 but that uses
a Zydas MAC and Atheros Radio. But hey the firmware
is available as GPL and driver is upstream.

There is another USB 2x2 with Atheros MAC and Atheros
Radio but not yet sure of the chipset name and if it
ships. Think it might somewhere.

> Questions:
> 
> 1) What differences exist between the AR5133 and AR9106?

Beats me.

> If the AR9106 offers 3x3 with superior performance to the AR5008, why
> didn't they ever offer any kind of PCIe option?

Not sure, but the answer to these sort of questions is usually demand.

> 2) What is the advantage of the AR9002 family over the AR9001
> family? Obviously, the AR9002 is a single chip solution, likely
> reducing cost, power and size.
>
> But is there any improvement to radio functionality or other features?

Having a single chip itself yields a lot more benefits than that.
Since things are closer together it also means less complexity on
overall programming. You can tell where things got easier by looking
at AR_SREV_9280_10_OR_LATER(ah) checks on the hardware code, I had
documented some of this on phy.c.

Apart from that you also get radio and baseband enhancements,
fixes, tunings, etc, the usual life of generation changes of
chipsets for 802.11. I can't get into specifics as that would
imply giving away the actual list of differences on each of
the blocks, but to be honest I only look at that stuff when
needed and that doesn't happen that often. I'm sure marketing
would have glanced over that and put together docs about this
with what is to be shared publicly.

If you have more questions you should work with our sales or
marketing teams.

> Note, I am interested in 802.11n features supported, quality of
> implementation, relative effective sensitivity, pretty much
> everything on the radio / DSP / layer 1 angle of things. I am
> curious as I want to be able to make the most appropriate
> hardware choices, and frankly, testing these chipsets to try
> and figure this out isn't the best way to do this based on
> my experiences thus far...

I recommend the single chip families, and specificaly AR9280
is a great candidate as its dual band and uses PCI-E. From a
software perspective Atheros dedicates more of its own resources
for testing our newer chipsets, the newer gernation 802.11n
chipsets. The AR9001 didn't get formal testing but the AR9002
did. Now its AR9002, in the near future it will be AR9003 and
so on.

And that's the way the cookie crumbles.

  Luis

[ath9k-devel] AR Chipset Differences

[Galen]

On Feb 26, 2010, at 9:23 AM, Luis R. Rodriguez wrote:

> On Fri, Feb 26, 2010 at 08:08:31AM -0800, Galen wrote:
>> I'm trying to determine the differences in features of the various Atheros chipsets supported by ath9k. Please note, I have only chosen dual band parts with at least 2 spatial stream support, as the single band parts are generally subsets of the dual band parts and the <2 spatial stream parts are generally extremely value-oriented (or power oriented.)
>> 
>> Gen 1 - AR5008:
>> AR5416+AR5122 - 2x2 dual band, PCI
>> AR5416+AR5133 - 3x3 dual band, PCI
>> AR5418+AR5133 - 3x3 dual band, PCIe
>> 
>> Gen 2 - AR9001:
>> AR9160+AR9104 - 2x2 dual band, PCI
>> AR9160+AR9106 - 3x3 dual band, PCI
>> 
>> Gen 3 - AR9002:
>> AR9220 - 2x2, dual band, PCI
>> AR9280 - 2x2, dual band, PCIe
>> 
>> Did I miss any 2x2 or better 802.11n radios?
> 
> Yeah our AR9287 is 2x2 single band IIRC.
> There is also our USB 2x2 AR9170 but that uses
> a Zydas MAC and Atheros Radio. But hey the firmware
> is available as GPL and driver is upstream.
> 
> There is another USB 2x2 with Atheros MAC and Atheros
> Radio but not yet sure of the chipset name and if it
> ships. Think it might somewhere.
> 
>> Questions:
>> 
>> 1) What differences exist between the AR5133 and AR9106?
> 
> Beats me.
> 
>> If the AR9106 offers 3x3 with superior performance to the AR5008, why
>> didn't they ever offer any kind of PCIe option?
> 
> Not sure, but the answer to these sort of questions is usually demand.
> 
>> 2) What is the advantage of the AR9002 family over the AR9001
>> family? Obviously, the AR9002 is a single chip solution, likely
>> reducing cost, power and size.
>> 
>> But is there any improvement to radio functionality or other features?
> 
> Having a single chip itself yields a lot more benefits than that.
> Since things are closer together it also means less complexity on
> overall programming. You can tell where things got easier by looking
> at AR_SREV_9280_10_OR_LATER(ah) checks on the hardware code, I had
> documented some of this on phy.c.
> 
> Apart from that you also get radio and baseband enhancements,
> fixes, tunings, etc, the usual life of generation changes of
> chipsets for 802.11. I can't get into specifics as that would
> imply giving away the actual list of differences on each of
> the blocks, but to be honest I only look at that stuff when
> needed and that doesn't happen that often. I'm sure marketing
> would have glanced over that and put together docs about this
> with what is to be shared publicly.
> 
> If you have more questions you should work with our sales or
> marketing teams.
> 
>> Note, I am interested in 802.11n features supported, quality of
>> implementation, relative effective sensitivity, pretty much
>> everything on the radio / DSP / layer 1 angle of things. I am
>> curious as I want to be able to make the most appropriate
>> hardware choices, and frankly, testing these chipsets to try
>> and figure this out isn't the best way to do this based on
>> my experiences thus far...
> 
> I recommend the single chip families, and specificaly AR9280
> is a great candidate as its dual band and uses PCI-E. From a
> software perspective Atheros dedicates more of its own resources
> for testing our newer chipsets, the newer gernation 802.11n
> chipsets. The AR9001 didn't get formal testing but the AR9002
> did. Now its AR9002, in the near future it will be AR9003 and
> so on.

I think that AR9003 will really do a lot to unify things, as we will finally see 3x3 and a lot of new features and will put this whole discussion to rest! I can't wait for that day...

> 
> And that's the way the cookie crumbles.
> 
>  Luis