CPU-Usage when transfer

[ddlink]

When I transfer a file over my wireless network I get a CPU-usage around 70-90%!

Is this normal? What do you get?

My setup:
Dell Inspiron 8100 with Pentium III 1,1Ghz WinXP
D-Link DWL-650

with and without WEP.

Btw. I get alot less CPU-usage when transfering over wire with 11MBytes/s

[ddlink]

Is there anyone on this wireless 802.11 forum who could help me?

I'd like to know how much CPU-usage your machine utilizes when doing a filecopy over a wireless lan.

Thank you

[MoleStrangler]

I think your are showing up the real differences between cheap and more expensive hardware.

The cheaper kit has most of the functionallity in software (in the drivers) because the hardware is cheaper and easier to intergrate into new products.

Also the off the shelf OEM products do a lot in the drivers so companies can easly take the product and create their own product without the need to dig down into the hardware.

The more expensive hardware usually has more complex hardware that is able to handle more of the work so the driver have to do very little.

Its the same with normal Ethernet NICs. Get a cheap one and a lot of the work is done in the software driver, the more expensive the more complex the hardware and the faster the NIC.

Remember that every time data is received over the network the NIC interrupts the main computers CPU so the driver can read the data and decide if it is for that computer or not.

The more functionallity that is contained in the drivers the more times the NIC has to interrupt the main CPU to send the data to the driver which then run the logic in software.

The more logic that you can put in hardware the less the main CPU is send an interrupt to run logic in the software drivers.

WEP is a prime example. Where is WEP implemented; in the NIC or software drivers. Now I think you can see the difference it could make.


---

This goes back to the old days of customers upgrading from 10BaseT to 100BaseT and actually experiance slower network responce, because their computers were struggeling with the extra network load, actually the main problem was their PCs & NIC combinations and not the network. People did not understand the relashionship between the NIC hardware & driver with the main CPU.

[ddlink]

Very good answer! Thank you.

Now, is there some kind of benchmark out there?

I'd like my next card to have less CPU-utilization.

[MoleStrangler]

I do not think that anyone has completed any public benchmarking based on network load.

We have some internal benchmarks based on network load (Ethernet and wireless side) on our Access Points but they are internal and not for release into the wild.

The tests are specific and not just how much data you can pump through a NIC but how well a card or AP performs under loaded network conditions.

We have recently doubled the CPU, backplain & memory capacity to deal with the extra network traffic from faster and faster network speeds.

You could try seinding you NIC a lot of data (generate some broadcast traffic) and see how your wireless NIC effects your CPU. There are lots of small tricks to workout where a manufacture has implemented logic.

But who realy cares, retail is about price and price. Industrial (where I exist) is about performance and performance.