Continuing it's recently inaugurated new methodology for measuring PC power consumption, Tom's Hardware takes, this time, the most relevant Intel processors of the past few years to see just how well this so much marketed characteristic has evolved.

We all know that a Core 2 Duo (or a Core 2 Quad, for that matter) is much faster than a Pentium 4 or a Pentium D. We also know the data regarding it's power consumption, which is way better than that of the Netburst architecture, both at idle and under load. But, how do they fare under normal working conditions?

With the same motherboard, an Asus P5E3 Deluxe (X38), and two modules of 1 GB Crucial DDR3-1066 (CL 7-7-7-20 2T), they compare as much as four Intel processors: a Core 2 Duo E6850, a Core 2 Extreme QX6850, a Pentium D 830 amd a Pentium 4 630. All of them have a working frequency of 3 GHz, and while it was obvious that the single core 630 would be the worse one, it's quite noticeable the bad performance per wat the Pentium D shows here (only a 4% efficiency increase over the Pentium 4). The Core 2 Duo is more than 400% more efficient, and wins followed by the Quad (which losses a little bit of efficiency because of the two-dies-in-a-package desing).

Link: Tom's Hardware.

Tom's Hardware has dedicated an article to place the impact of the CPU, on the overall system power consumption, into a new perspective. They consider that if a system is noticeably faster than another one, it will complete tasks much quicker, so it will go back to its low-power idle state faster, requiring less power than a similarly efficient (slower) computer. This is a different approach to most reviews that deal with this issue, and one worth taking a look at.

The processors to be tested are known to be somewhat energy efficient: an Athlon 64 X2 5000+ and an Intel Core 2 Duo E6400. SYSmark 2007 Preview is used to simulate typical user behavior, and to get an idea about maximum power consumption levels, 3DMark06 and PCMark05 are looped four times.

After running the SYSmark test for more than an hour, the difference in average power consumption was 26%, but the difference in total power required to complete the entire benchmark was 50%. That can be easily explained by the fact that the Athlon 64 X2 5000+ system took 14 minutes longer to complete the benchmark, while the Intel system was running idle during this time.

The test with 3DMark06 and PCMark05 gave a consumption of 339 Wh for the Athlon 64 X2 5000+ and 269 Wh for the Core 2 Duo E6400. The Intel system took 2:15 to complete this test, while the AMD system required less than 2:20, a small difference that pointed average power consumption as guilty for the total power consumption difference.

Little more than a month and the new Intel revision will be in stores. Despite the proximity of the launch date, there's no official word on pricing or naming conventions, which isn't the same as saying that hard working people can't collect all the information scattered so far. 

There will be more models available, but these are the ones "confirmed" for sale between November 12 and January 20. No doubts about the quad cores, but it's unknown if the cheapest dual core will be an 8200 at 2.66 Ghz with a price of $163 or if the inclusion of an 8300 model at 2.83 Ghz will bring it to an even lower price.

Source: Daily Tech.

It's been said before and now that we are getting closer to the anticipated date is still holding: we may see a 3 GHz Phenom early next year, if not at the end of 2007. Barcelona is already moving to it's B2 stepping and could therefore reach that mark relatively soon.

Realistically speaking, we'll get 2.6 or 2.8 Ghz, but let's not rule out the possibility of a low volume 3 Ghz chip. If not anything else, it would help keeping prices lower for the rest of the line.

You can track this story to TG Daily among other web sites.

Link: X-bit labs.

During die testing, AMD's engineers noticed that there was an occasional frequency mis-match among cores. That meant that the core that couldn't hit higher frecuencies was actually holding the rest back. At this point, the choice to fuse the processor at the silicon level as a triple-core was made, instead of downgrading the whole pack by fusing it as a slower quad.

"Since the processors are 'fused' at the silicon level during production, our sources inside AMD say that attempts to 'unlock' the turned off core will be impossible from the outside by simply connecting pins or something like that."

All of this thanks, supposedly, to the Phenom being a native quad-core processor (unlike Intel offerings). All the dynamic L3 cache will still be available for the remaining three cores, which might help. In cases where more cache and raw clock speed is beneficial to performance, the triple-cores could actually beat the quads.

Legit Reviews also points out that "the Microsoft Xbox 360 has triple-core architecture and as a result, the games are being designed for this hardware structure". Can anybody comment on this?

Intel Developer Forum comes packed with Penryn related news, and those include overclocking. Using a Gigabyte GA-X38T-DQ6 motherboard and a Core 2 Extreme QX9650, they were able to hit 3.87GHz on air and 4.1GHz with water cooling (not showed).

From the pictures provided, we see their heatsink looks rather unimpressive, wich means 4GHz should be between reach of the best air coolers out there.

The reason for using a single channel setup, they say, is to avoid DDR3 being a limiting factor in the pursuit of the highest overclock. Whether it's that, a BIOS problem or a chipset issue remains to be seen. Hopefully, those of us with P35 and P965 chipsets running DDR2 will scape the single channel scenario.

Link: Legit Reviews.

Intel will launch eleven 45nm dual-core notebook processors in the first half of 2008, five of them as soon as January: X9000, T9500, T9300, T8300 and T8100 (ranging from 2.8GHz to 2.1GHz). Quad-core versions will arrive in the third quarter, according to sources at motherboard makers (remember that Intel "cannot comment on unannounced products and technologies").

"[The] 45nm quad-core notebook CPU codenamed Penryn QC [will be] comprised of two Penryn cores in a multi-chip package. The CPU will adopt Socket P (35mm × 35mm) and have 12MB L2 Cache, 1066MHz FSB and 45W maximum TDP."

The only mobile quad-core that we know for sure will be available next year pertains to the Core 2 Extreme series, a similar situation to the one we face now with desktop parts. Prices are usually around $1000 for these CPUs, so those wanting something less expensive will have to wait until 2008/2009 (for desktop/mobile).

Link: DigiTimes. 

After so much waiting, the new AMD architecture comes to life as a server CPU and it doesn't look like it will be enough.

DailyTech has posted a list of articles dealing with this new offering, including performance numbers, and there's were it gets nasty. It's better than the K8, no doubt about it, but not nearly as much as it should've been in orther to regain the crown it lost to Intel more than a year ago.

Core 2 was so much better than the P4 than it easily surpassed it and the Athlon64. Hoping for AMD to do just that was optimistic to say the least. Even though the charts we can see now are using server motherboards with limitted RAM speed and Phenom is still expected to do better, it won't do well enough to compete with Penryn. And that's problem number one, the delay of the processor and the slight advancement from Intel's part will have it pitted against Kentsfield's successor. That's not what AMD had in mind, they were barely in a position to fight that battle, but Penryn will be just out of reach. It has better clock for clock performance and higher overclocking capabilities.

What's AMD going to do now? Probably continue with the only war they can still wage: a price war. But it's unclear for how long the company can sustain loses, they aren't exactly swimming in Uncle Scrooge's money pool, after going out ATI-shopping.

Above you can see a Penryn die, but until those chips arrive in the November timeframe, Intel has another ace up it's sleeve: a Core 2  processor for the MP (4-way) server market.

Since the release of the Opteron, AMD has enjoyed several years of leadership in this profitable segment, and the launch of Barcelona was meant to consolidate that position. Unfortunately for them, Intel has plans to hold a launch briefing next Wednesday with Dell, Hewlett-Packard, IBM and Sun Microsystems, to preview Tigerton. Just a few days ahead of Barcelona.

Tigerton will be more powerful and power-efficient than the Netburst microarchitecture it replaces. Intel has already made the transition to Core 2 in its desktop, notebook and dual-processor server products, but is still selling chips for multiprocessor servers based on Netburst.

You can take a look at models, clock speeds, cache sizes and prices here (DigitTimes), but unless you own a server board you can forget about running one of those.

The Xeon 7300 series will have about twice the performance of Tulsa and over twice the performance per watt of these processors. The Tulsa chips are still based on the old Pentium 4 (good riddance).