Simon Burge writes:

>I'd guess at this stage the 25MHz TurboChannel bus is the limiting
>factor.  Are there any figures on what the defta gets on an Alpha?
>Impressive none-the-less!

I'd say it's typical TCP stack overhead.

An electrical-engineering PhD student here 5 years ago built a
Turbochannel interface for frequency-agile WDM; it sustained over
800Mbit/sec across a 5000/240 Turbochannel.  Obviously, at those
speeds, the CPU wasnt even touching the data.

Again, Matt's earlier numbers were for sender-side; and a typical 4BSD
TCP is receive-limited.  With a 3maxplus, I could drive the DECWRL T3
boaprds at (at best) 28 Mbit/sec. So we can gain around 15% more from
determined tuning.