c0t0d0s0.org

Phil ran a memory benchmark and summarized his results in Niagara 2 memory throughput according to libMicro:

[...] seeing 240ns latency, which equates to a throughput of 267 million memory reads per second (i.e. 64 / 0.240e-6). Just to set this in context, here are some data for a quad socket Tigerton system running at 2.93GHz...
[...] his shows a peak throughput of about 86 million memory reads per second (i.e. 16 / 0.186e-6), making the single chip UltraSPARC T2 processor's throughput 3x that of its quad chip rival.

It´s a micro benchmark, but it´s a part of the puzzle why an UltraSPARC T2 is able to outperform a processor with almost 2.5 times more clock cycles per seond.

As both Rock and Niagara 2 use FB-DIMMS this are really good news: FB-DIMMs have an addtional circuit called Advanced Memory Buffer. The higher consumption of FB-DIMMS can be credited to this AMB. IDT now annouced a 25 percent more efficient circuit. Okay, thats only between 1 and 2 Watt per DIMM, but you have 16, maybe 32 DIMMs in your server, and thus 16 to 32 AMBs. And from a system view, this really counts.

On ZDNet is an interesting article about the race between Intel and AMD.. I have a somewhat different opinion. In the x86 area it is usual to do technology for technolgies sake. There isn´t the question of a balanced system. For example, the normal customer doesn´t ask the question if it´s sensible to use higher bandwidth memory when the processor is incapable to use it. The customer thinks okay DDR1 is better than DDR2, lets use it (It the same like putting PCI-E into a server but coupling all slots into one port of the internal crossbar.I call this "datasheet bullet bullshitting". As more and more decisionmakes with pc-server background try to make good decisions on buying servers, it becomes more commonplace even in the server business).
I don´t really think that the internal memory controller will be a problem. The usage of DDR2 right now is only the consequence of Intels "using-something-different"-policy and has only marginal technical merits. It´s playing with the memory market to fight on a completly different field by changing the costs for systems of the competiion. Well, from my view you could work with DDR1 for a long time.
Instead of running after such tactics, you should do something more game-changing designs. For example a marriage between Niagara and Opteron(Caution: pure speculation). A quadcore-quadchannel Opteron with FB-DIMMS (Fully Buffered DIMMS with HT-3 (BTW: a ccHT-Switch in buyable,real silicon would be nice). This would be a hell of an processor. Okay, enough of dreaming today.
At last from my view the most interesting statement is the last paragraph:

Whereas AMD had to basically go it alone to get this far (from an innovative point of view), the fact that it is very good friends with Sun these days (another company that's doing some very innovative stuff in processor design), should not be lost in any big picture analysis.

Do anyone really think that we don´t talk with AMD, and AMD not with us?