AMD's 4x4 solution crashes and burns

[KG_CrimsonTrooper]

I see that both Intel and AMD released their Quad core's Earlier than previously stated in their press releases from a month and half ago...
Initially they were supposed to be end of the 1st Quarter Releases, My how money changes everything...

I found an article I thought some people would enjoy:

When you place AMD's four-way solution next to Intel's, the results are obvious.
posted 8:47am EST Fri Dec 01 2006 - submitted by RickGeek

BLURB
I'm going to speak plainly here about AMD's latest quad-core offering (see our coverage), the much touted "4x4 gaming platform." AMD's temporary solution is an absolute failure--I cannot be more plain than that. On 90 nm manufacturing processes, when examining the power budget and the relative performance compared to Intel's high end quad-core Core 2 systems there is no comparison: Intel wins hands down.

AMD's dual-CPU, quad-core solution built on 90 nm technologies consumes a ravenous 530 watts minimum just to power the dual CPUs. That means each CPU's power budget is 265 watts. Can you imagine putting a single CPU in your system that consumed 265 watts? That's what you're doing with the 4x4 platform. That CPU power budget alone is over 100 watts higher than the entire quad-core Core 2 platform's total power consumption.

One good point of note: AMD's CPUs still show very powerful performance on CPU-specific benchmarks. AMD does still dominate the CPU-intensive benchmarks, even against Intel's fastest Core 2 processor. Nonetheless, while it does speak to AMD's strong internal design, something that has been true even back to the days of when AMD's offerings competed against Intel's Pentium III, it's still not enough to compensate for it losing every other real-world or practical benchmark. Not many of us have CPU-only tasks that we must perform day-in/day-out, but most of us use applications that are a balance of CPU processing and external memory requirements, meaning the Intel offering is the choice to make today, and without reservation.

What we end up with in the AMD solution is an expensive, power-consuming, lesser performing hack, if I may be so bold. At 90 nm AMD had no choice but to produe a dual-socket quad-core solution because no practical heat removal system in existence could remove 550 watts from a single CPU package efficiently.

Frankly, I'm amazed AMD released this product.

One other potentially brighter side note is this: what we do gain in this offering is the knowledge that AMD's coherent HyperTransport bus does work and can coordinate data between remote physical packages very well. The core scaling between a single core and all four cores runs about 77% on Cinebench 9.0, for example, meaning that it falls 23% short of the ideal of 100% scaling (in moving from processing data only on one core to all four cores). That figure shows us that even when all four cores need to be fed from the HyperTransport bus architecture ... well, the delay is not horrible. Intel's best offering scales at 74.8% on its shared bus architecture.

AMD's direction with 4x4 might not be one of raw performance alone--it could be a stepping stone of sorts. Whereas dual-core users will definitely see an increase in performance when replacing their dual-core system with the new quad-core 4x4 platform, the requirements of having at least a 750 watt power supply, coupled to the lower performance relative to an equally equipped and configured quad Core 2 system, result in a quick decision for most people: Choose Intel.

I'm looking forward to seeing what AMD can do at 65 nm. If its processors do not scale any better than these offerings we're seeing today then I fear for AMD's future. What we have today is an aging 90 nm process technology, dual-socket quad-core solution from AMD that simply cannot cut the mustard when placed next to Intel's single-socket quad-core solution on 65 nm process technologies. As of this day Intel is definitely the x86 platform of choice across the board with its Core 2 architecture.

Read some of the benchmarks used in coming to this conclusion at Tom's Hardware Guide, AnandTech, and The Inquirer. As always, post your comments below. I look forward to reading them as the path to success at 90 nm has come to an absolute end.

http://www.geek.com/news/geeknews/20...1201041102.htm