Broadway: Is this a great improvement?

[freezamite]

I've confirmed that broadway has other improvements over the gekko apart from the speed.
In the SDK documents that are on the net (although they're from the earliest SDK) there is a table that compares Gekko and Broadway:

Broadway
*Operates at 729Mhz
*Bus from Broadway to internal main memory: 243Mhz 64bit
*L2 cache fetch modes: support for 3 modes: 32Byte, 64 Byte, 128Byte
* Bus pipeline support for depth level 2, 3, 4

Gekko
*Operates at 486Mhz
*Bus from Gekko to main memory: 162Mhz 64bit
*L2 cache fetch modes only supports 32Byte
* Bus pipeline only supports depth level 2

I think that all the improvments in the bus have the objective of not being a bottle neck for the improvments in the broadway.
Does anyone know if this is a good optimitzation or if this doesn't affect de power of the proccessor?

PD: Sorry for my bad english.

 

[ShaidarHaran]

so a 50% increase in memory bandwidth to back up the 50% increase in clock.

not terribly exciting, but a wise choice seemingly

 

[freezamite]

Yes, but what about the new fetch modes supported in the L2 cache?

 

[archie4oz]

It simply means you can fill more cachelines per fetch request. Newer processors (e.g. P4, C2D, PPE in Waternoose and Cell, etc.) generally have 128byte cachelines, Gekko and Broadway have 32byte cachelines (inherited from their 750 lineage), so this is probably just a small optimization to improve efficiency when streaming packed operands into the FPU.

 

[Rodéric]

Newer processors (e.g. P4, C2D, PPE in Waternoose and Cell, etc.) generally have 128byte cachelines,

Core 2 Duo and Nehalem cache line size are 64bytes.

 

[DeadlyNinja]

You have anything on Hollywood?

 

[Flux]

It simply means you can fill more cachelines per fetch request. Newer processors (e.g. P4, C2D, PPE in Waternoose and Cell, etc.) generally have 128byte cachelines, Gekko and Broadway have 32byte cachelines (inherited from their 750 lineage), so this is probably just a small optimization to improve efficiency when streaming packed operands into the FPU.

Well it says it supports a 128bit fetch mode above.

 

[Pete]

128byte, and archie noted that ("so this is [... an] optimization to improve efficiency [...]").

 

[freezamite]

I have nothing like this about the hollywood.

Pete:
In terms of efficiency: is there a big gain?
I mean, is the broadway a far better processor than a simple Gekko overclocked at 729Mhz (5% - 10% faster or more) or this is just a small optimitzation (less than 5% faster)?

 

[Pete]

No idea. archie & co. are the experts, I just nitpick language and alienate people.