cM :: MORLEY|DIGITAL

Nearly a year and a half ago I wrote about the unique triple-core processor from AMD.  I felt, and still do, that it’s a great play by AMD to flank Intel’s dual-core and quad-core processors.  My focus in that article was the retail mentality of your average consumer.  Three is better than two.  That’s what a brick-and-mortar buyer will see.

And now that AMD has released its new 45nm Phenom II line-up, the press has responded much more favorably this round to AMD.  But what has received the most attention and acclaim have been the value X4 810 and the X3 parts.  Because of their low cost, overclockability, and excellent performance against Intel’s Core 2 lineup, AMD has once again found a way to position itself against its larger, cash-laden competitor.  And what may be considered AMD’s finest processor is the surprising Phenom II X3 720 Black Edition.

Recently I wrote about Intel’s Core i7.  It simply rocks.  But it simply doesn’t sell.  According to Q1 reports, the X58 chipset - the only chipset that can run the i7, will be 1.8% of its chipset sales.  Let’s give Intel some benefit of the doubt and assume that in the range of 50% of those chipset are mobile, where there is no i7, and the X58 would make up less than 4% of desktop chipset sales.  That’s well within the 5% enthusiast market I continue to talk about. (read these two posts, they’re important to understand my mindset here)

AMD knows this, and so does Intel.  The fight’s between Phenom II and Core 2 until Intel decides to drop an anvil and make Core i7 mainstream.  By all accounts that won’t be until Q4 of this year or Q1 of next year.  Intel’s best selling chipsets are those used by Core 2 processors.

Till then AMD has a very, very competitive product that combines high frequency, performance out-flanking dual-core processors, a great price, and even a bone thrown to the enthusiast with its overclockability.

Combined with AMD’s strong 780G and 790GX chipsets and fantastic Radeon HD 4000 series lineup, AMD finally has the right mix of product to make a stand in this market.  

One company I haven’t mentioned in this mix but should - NVIDIA.  Remember they make chipsets that are readily available, have design wins, and are extremly powerful for both AMD Phenom and Intel Core 2 processors.  They may be having a spat with Intel over Core i7 licensing, but for now it’s not going to hurt them in the pocket book.

[Disclosure:  I run a Core 2 Quad Q9550 in my main rig.  I currently have a GeForce GTX 280 and a Radeon HD 4870 X2 I play with.  I have two laptops with Intel chips and chipsets and one with an AMD processor with NVIDIA chipset.  My media center has an AMD processor with an AMD chipset.  I am a technologist that just loves tech and am brand-agnostic.  I think all these companies make great products that sometimes you just have to use in an manner fitting to their price and performance.  My opinions are my own and do not necessarily represent those of SolidWavePC or Fluid Digital.]

Intel’s tick-tock strategy kicks butt.  First, a new architecture.  Follow that up with a tweak and then wash, rinse, repeat.  It started with the launch of Core 2 in the summer of 2006 (tock.)  They followed up with the “tick” late last year with the move to a 45nm process.  Now here comes the first follow up “tock” and it’s a doozy.  I think they should consider renaming their process “beat you into submission and then kick you when you’re down.”  The Core i7’s got some wicked fast processing power and it brings to the table features that, while have been used by AMD for a while now, serve to propel Intel’s new chip into the stratosphere.  Without going too far into it, the key features are: integrated triple-channel memory controller, monolithic quad-core, QPI (Intel’s answer to AMD’s HyperTransport), and the return of HyperThreading.  

Of course, with all these changes, a new chipset is in order.  Enter Intel’s X58.  That’s it.  Nothing from NVIDIA, and no mainstream variant from Intel.  The integrated memory controller and increased pin count make the new Core i7 understandably incompatible with previous platforms.

The interesting thing is that the X58 chipset, without the need for an integrated memory controller, is not as expensive to make as its predecessor.  In batch pricing the cost of an X58 chip is $52.  The X48 is $70.  In comparison, the mainstream, CrossFire-capable P45 chipset for today’s Core 2 lineup is $40.  

So now we have an affordable, performance-oriented chipset that is significantly cheaper that its predecessor, a completely new CPU core that starts under $300 in batches of 1000, so this should be a game changer, right?  

It’s not that black and white.  Intel’s venerable Core 2 lineup is still good.  VERY good in fact.  I’d argue that for most of the market, Core 2 is not only enough computing power, but more than some even need.  Heck, for gamers, we’re still GPU bound.  And Intel’s able to pump Core 2 chips out like crazy and at a healthy profit.  So why kill off a good thing?  If there were P and G variant 5x chipsets, allowing for motherboard manufacturers to make products for different price brands, OEMs would en masse run towards Core i7.  But as it stands all motherboards in the channel are full ATX, high-end parts that start around $249 and run north of $300!  And keep in mind you need to buy three sticks of DDR3 memory as well!  This puts it squarely in the 5% DIY and Enthusiast crowd.

Of course, Intel has plans to introduce lower cost chipsets and Core i7 variants that include integrated graphics and support for dual instead of triple channel memory.  But those are a ways off.  Intel, in the interim, plans on keeping Core 2 around for a while.  And for good reason.  It’s a great product, serves the mainstream market well, and allows them to amortize the cost of it for a bit longer.

So keep in mind that, while totally melt-your-face-off fast the Core i7 is, you can still find awesome deals on great Intel Core 2 hardware that will keep you and your wallet happy for quite a while.  (It sure is good to be a DIY guy right now, though.)

The days of needing multiple graphics cards in your PC are officially over. While game engines have become more complex, and Crysis still confounds most people’s PCs, the fact is that we have reached a pixel density plateau when it comes to monitor sizes and the GPU battle grounds are being fought at 1680×1050 resolutions or lower¹. And the simple fact of the matter is that at that resolution nearly every game on the planet can be enjoyed by the mainstream gamer utilizing a single graphics card. The hardware side of the gaming industry has gotten that good.

But don’t tell that to AMD or NVIDIA or even Intel. Don’t tell that to the system builders, but would somebody please tell that to the tech journalists? The fact is that multi-card technology, not multiple GPUs on the same PCB, allows chipset manufacturers to stratify their offerings (read: charge more money), promising more performance and more “expandability” – but costing you more money.  The fact is that these technologies are aimed squarely at the 5% uber-enthusiast market and system builders who need to win synthetic benchmarks by 3% in order to pay their salaries that month. 

Expandability or upgradeability are the buzz words that these companies use to lure you to their higher end chipsets. They want you to think you need a second or third graphics card. But most effectively, they communicate to you that you are buying “future expandability” that will “future-proof” your PC. And they’ve effectively brainwashed the mainstream tech media into believing the same thing. It’s hard to find a review of a gaming system that doesn’t include the ability to add a second graphics card where the reviewer doesn’t “ding” the builder for it. 

But nobody has ever really studied the usage patterns and buying behaviors of customers who purchase multi-video card capable motherboards and SLI or CrossFire capable graphics cards. It is that question that intrigues me, and it is simple logic that leads me to the conclusion I have written in the opening sentence of this post. 

The main problem is, if you aren’t gaming at XHD resolutions that 24”+ monitors support, spending money on two high end graphics cards is a complete waste of money, and buying two mid-range graphics cards is pointless as a single high-end card that may feature two of the same GPU on one PCB can easily meet or beat it for around the same combined price. Factor in the total cost of ownership of forgoing a more expensive chipset designed for multi-video card support, and you really need to make sure you’re spending your money the right way.

Additionally, if you’ve spent your money on an XHD monitor, and only have a budget left over for a single high end graphics card, never mind the fact that you perhaps bought an unbalanced configuration, but the odds of a better single card solution being available by the time you can afford that second card are high. Why do you think EVGA offers a 90-day trade-up program? It’s essentially an insurance policy against the age-old truism that your system is obsolete the day you buy it. 

Don’t take my word for it. Kyle and Brent at [H]ard|OCP have been shoving the term “real world benchmarks” in our faces for so many years now that every piss-ant rag on the internet and in print uses the terminology without really understanding what it means. But if you really look at what [H] video card reviews show, you’ll find that at 1680×1050 resolutions and lower, my premise holds. From there it’s a simple matter of adding up the total cost of ownership to reach the playable settings (meaning a smooth gaming experience.) So be sure that you aren’t blindly buying a dual or triple-graphics capable motherboard for $200-$300 when a significantly cheaper, and perhaps a bit more boring, solution will suffice, allowing you to spend more money on a powerful, single video card that will better serve your needs. 

1. Source: Valve Steam Hardware survey, November 2008. 96% of players have resolutions set to 1680×1050 or lower. Sample size: ~1.7 million users.

A little birdy brought this to me last week.

For full specs, read here.  I also received a mATX version of this mini-ITX powerhouse, the DG45ID.

This fulfills all the players I mentioned in my article I wrote for DailyTech last year entitled “Video Processing: A Race To The Bottom.“

That was nearly a year ago, and our options today are fantastic.  AMD’s 780G led the charge, followed closely by NVIDIA’s 8200 chipset, and now Intel’s G45 solutions are starting to enter the fray.  It’s a good time to be an HTPC enthusiast, that’s for sure.

Just to be sure, I did test the DG45FC @ 1080p over HDMI running H.264 content ~30Mbps with 2GB of RAM and an Intel Core 2 Duo E6400 and experienced about 28% CPU utilization.  Not bad in my book!  That was with Vista 32-bit and ArcSoft’s TotalMedia Theatre.

Don’t you just love witty headlines?

AMD last week announced a previously un-leaked SKU in the family of upcoming Phenom processors.  With all eyes on what AMD is going to do to compete with Intel’s dominating Core 2 architecture, AMD may have thrown the ultimate curveball before the game even begins.

Traditionally, multiprocessor configurations come in multiples of 2; in the case of the consumer desktop, your option is 1, 2, or 4 processing cores.  Currently, the most popular from both AMD and Intel on the mainstream desktop would be the dual core processor.  Intel has the Core 2 Duo, and AMD has the Athlon 64 X2, of the venerated and aging K8 family.  AMD’s new processor family, the Phenom, which will be released in the next couple of months, will span dual, quad, and now tri-cores. 

So what is the point of a three core processor?  The easy answer is that three is one more than two and that helps AMD space out their CPU portfolio in terms of price and performance. 

It all boils down to positioning.  AMD was smart not to go head to head with NVIDIA with their Radeon HD 2900 series.  Instead, they hit on popular price bands and features not found with their competitors.  It was a winning strategy that made up for publicly perceived architectural drawbacks with their advanced GPU.  With Phenom it may turn out to be a different story.  If AMD’s next gen processor has the luxury of competing with Intel’s current lineup for a couple of months before Penryn is released, it may be enough to regain precious ground lost over the last year.  All eyes are on Intel’s Penryn with its core tweaks, SSE4, and 45nm process, and that is where the protracted battle will be.  It remains to be seen just how fast Intel can bring this processor family to market.  However, AMD has created its own blue ocean with a tri-core Stars family processor.  Thanks to AMD’s Direct Connect Architecture, adding additional cores (in this case it’s actually subtracting) only requires a prerequisite number of HyperTransport links, or the communications protocol AMD uses for core to core transfer of information.  Intel is still using the tried and true Front Side Bus design, which currently limits the type of multi-core designs they can release to market.  Intel has no monolithic core design, and will be relegated to 2 or 4 cores on the desktop for quite sometime.

It is this competitive advantage that AMD hopes to leverage, and I believe it’s a winning strategy for a very large piece of the battlefield: the retail channel.  Walk into any Best Buy at the end of 2005, and you’d see a majority of the systems featured AMD processors.  For the first time in the company’s history, AMD actually beat Intel at something.  They had the larger share of the US retail market.  This increases mindshare which increases market share.  Blue Shirts, Best Buy employees, are a key group of market influencers that any company needs to court in order to be successful there.  Driven by enthusiast support in the gaming market – many of whom work at Best Buy – AMD was winning the hearts and minds.

Now, enthusiast support is firmly behind Intel, and AMD has had to play the old game of how-low-can-we-sell-these-things-for in order to maintain any type of hold in the market.  With a tri-core processor, AMD has the luxury of flanking the Core 2 Duo or Core 2 Quad.  It can compete with the former on horsepower, and it can compete with the latter on price.  Intel will have a hard time countering with its dual core offering when benchmarks show multimedia apps and certain games favoring more cores (watch 3D Mark 06 during the CPU test and you’ll see it only uses 3 cores – and 3D Mark is an IMPORTANT benchmark for retail customers.)   And with many of the Intel quad core chips filling higher ASPs (notable exception being the Q6600), Intel will have to cut lucrative margins it now enjoys in order to fend off customers who may not be willing to spend the extra cash on an extra core.

But CPU core count is only part of an AMD full court press that focuses on total platform performance.  The next three months will see many of the fruits of the new AMD/ATI effort, and Intel will have more to worry about than just CPU strengths as the focus shifts to the total platform.  However, both sides have brilliant people working for them, and Intel’s superior resources will be brought to full steam against any perceived threats by its smaller competitor.  The question is how fast they can execute, not IF.

The bottom line is I believe the tri-core is a Hail Mary retail play on AMD’s part that is likely to pay off.  We’ll see AMD tri-core machines occupying retail shelf space that may traditionally seat an Intel dual core system.   It’ll be an interesting 2008 for both companies, I’m sure.