Wednesday, May 17, 2006

AMD readies Socket F 1207 Opterons

Our INQ friends tell us that the long awaited Socket F (1207) Opterons are only a few weeks away, just in time to frag the Woodcrests. AMD's situation is so enviable, it's like fighting a battle with 100 divisions of reserve forces. The enemy re-enforces with one fresh regiment? No problem, bring in a few reserve divisions and mass frag the opponent.


Anonymous Anonymous said...

Woodcrest even now isn't a match for an Opteron 64 system at 2P or 4P, (btw, 4P+ is a joke for Woodcrest) and now Socket F rev. F? 1207 w/ DDR2 & HTT 3.0 is going to annihilate anything from Intel, Woodcrest has a max of 17GB/s System I/O (including memory) and simple DDR2-667 is too much for the Dual FSB's (20GB/s for 17GB/s bus's)

Not to mention FB-DIMM's add extra latency, which throw in a overworked Northbridge, add about 20w+ more to the Northbridge to manage 2 FSB's and 4-Channels of Memory, plus the added Wattage for FB-DIMM's A.B.C, you're looking at AMD owning the Server market for the next 5 years and beyond, throw on K8L and K10, and Intel needs to pull a rabbit out of it's illustrious hat to even THINK about coming to AMD.

CSI and IMC isn't going to be enough, not given that their current Quad-Core's for Itaniums can't even run all 4 at once, and Clovertown only get 40% more performance with each added core (compare that to 85%+ on current AMD64's) and Intel is in trouble.

12:06 PM, May 17, 2006  
Anonymous Anonymous said...

Can you translate this into something meaningful: "The purpose of new memory controller is solving that nasty 1GB/s read and write speed lag that company experienced when it introduced dual-cores to the product line." ?? What does it mean?

1:33 PM, May 17, 2006  
Anonymous Zetro said...

To my understanding the AMD memory controller has issues with 1GB pages to the memory which is pivotal in server setups. Basically its a tweak for a slight issue that was in the original K8 design.

3:49 PM, May 17, 2006  
Anonymous Anonymous said...

"Woodcrest has a max of 17GB/s System I/O (including memory) and simple DDR2-667 is too much for the Dual FSB's (20GB/s for 17GB/s bus's)"

Operton has only about 11GB/s max now. If you expect a lot from
AMD's future Socket F and HT3, don't expect Intel will sit still.

"Clovertown only get 40% more performance with each added core"

Clovertown is a Pentium D alike product, just a temp product. It's not native quad-core.

Anyway, only time can tell what things are, not people...

4:20 PM, May 17, 2006  
Anonymous Anonymous said...

Woodcrest will outperform all other CPUs on the market

4:16 PM, May 20, 2006  
Anonymous Edward said...

"Woodcrest will outperform all other CPUs on the market"

Please, I've seen this crap more times than I need to. Either someone is trying to spread wrong information over and over, or there are just too many moron amateurs talking about computer architecture. Here's my response:

First, Linpack is useless because it's more a cache size detector than a benchmark. How is the "80% efficiency of Woodcrest" derived? Based on what size of dataset? If he assumed everything fits into the cache, then (1)it'll not reflect the real-world, (2)it'd just reproduce freq * FP/clock. What's to shout about then?

Second, he claimed that "the number of [apps that POWER/Itanium outperforms Woodcrest] will be greatly diminished". Where does that come from? Does he know that the percentage of real-world apps whose performance correlates to that of Linpack is effectively zero? Then what's the significance of Woodcrest's "best" Linpack performance, and how does that make Woodcrest any more attractive than other server CPUs?

I mean Woodcrest might be a respectable feat of Intel, but to say it beats all existing server CPUs based on "estimated" Linpack performance is just completely ignorant and out of his mind.

10:51 AM, May 21, 2006  

Post a Comment

Links to this post:

Create a Link

<< Home