One of the funnier bits I ran across was the dueling AMD / Nvidia booths in the North Hall foyer. I’m not sure if this was intentional, or an accident, but they’re almost right next to each other.

The potentially coolest thing I saw was Nvidia’s FG100 DX11 graphics card actually running in a PC. No word on final specs, pricing or power consumption, though. Now it’s off to CES Thursday and more meetings.

The timing is certainly interesting, coming so soon on the heels of New York State’s recent filing against Intel on that same topic. Toss in the European Union judgment, and you get a picture of a beleaguered Intel, fighting a battle on multiple fronts.

What does the settlement really mean for consumers and OEMs? How will it affect AMD going forward?

There are still a lot of unknowns from this brief announcement. In many ways, it’s a classic settlement – AMD gets cash, rights to some Intel patents and an agreement from Intel to stop it’s anti-competitive behavior. Intel, on the other hand, readily agrees to ceasing and desisting said behavior because, after all, both companies want “To avoid the time and expense of litigation, and without any admission of liability or fault…”

Intel is substantially larger than its arch-rival, generating around six times the revenue of its smaller competitor. It’s also very profitable, while AMD has only recently been able to pull itself closer to profitability. Still, all the litigation is a drain on Intel’s resources, a continuing pall on Intel’s reputation and a distraction from developing new technologies and products.

According to a conversation with AMD spokesperson Drew Prairie, the way to think of the agreement is as a logical extension of past agreements with Intel – and hence only applies to the CPU side of the equation. That’s an important point, which I’ll get to shortly.

Intel had recently complained that AMD’s split into a fabless semiconductor company – AMD – and a separate, manufacturing subsidiary – Global Foundries – was a violation of past patent agreements. Intel now drops that claim, which has two key implications:

• Global Foundries can now fully split with AMD, and isn’t required to be a subsidiary even in name.
• AMD can contract with any chip manufacturing company to build any AMD product.

While the legal wheels are still turning, this logically completes the split of AMD into two separate entities. This also means that Global Foundries really needs to stay competitive – it can’t rely on a captive AMD to give it business forever.

Now let’s get to the graphics question.

As I noted earlier, you can think of this settlement as a logical extension of past agreements between AMD and Intel. According to Prairie, this gives Intel no rights to AMD’s intellectual property on the graphics side. That’s pretty major, since Intel is trying to build its own GPU in Larrabee. If Intel violates any patents AMD might have that are purely graphics related, AMD would be free to pursue legal action.

There’s one other important point: AMD and Intel will continue their separate paths. There’s no bus licensing involved, so you won’t see AMD or Intel suddenly ship pin-compatible CPUs. This is by no means a return to the socket 370 era.

Consumers probably won’t see much of a difference at first. I’ve started to see an increasing number of interesting products built around AMD technologies, including the recently Dell’s recently announced Inspiron Zino HD. Given that increasing number of AMD-based PC products shipping recently, it’s likely that Intel had pulled back from whatever anti-competitive behavior it may have engaged in (that it doesn’t admit doing, mind you.)

OEMs will win, too, having a little more freedom to develop products that they see suit their own customer’s needs. AMD is pushing forward with its plans, including a renewed strategic approach that more tightly integrates GPU and CPU sides of the company. If AMD can just stick with a single core strategic message, OEMs will feel much more comfortable doing business.

If you get right down to it, though, Intel is a big winner, too. Sure, it’s out of pocket $1.25 billion, but it avoids losing to AMD in court, along with the potential of much larger punitive damages. In some ways, Intel is actually getting off cheap. And since Intel is so profitable, it can still use big discounts to lure OEMs to use Intel-based solutions.

Let’s hope that the playing field really is leveled. AMD still has a big chore ahead of it, building its Fusion products, keeping its feet in the server side and trying to keep Nvidia at bay in the GPU business. It’s a daunting task, but at least AMD can hope that this settlement means that it has a fair shot at the market now.

Most of them will be successful. Windows 7 setup is remarkable forgiving, but some users will still have issues. Even I had a problem with one older laptop based on the original Core 2 Duo, in which Windows 7 installed fine, but the HD audio device refuses to work even though the driver reports itself as working normally.

Of course, there will be people who have issues, sometimes serious ones. That’s the nature of the PC.

The PC Ecosystem

Let’s talk about Apple’s Macs for a moment. That may seem odd, but the Mac – specifically the iMac and the Mac Pro, illustrate a key point I’m going to make shortly.

If you buy a Mac from Apple, you can get a Macbook, iMac, Mac Mini or Mac Pro. With the sole exception of the Mac Pro, all Apple PCs are completely closed. With some effort, you can upgrade hard drives and add memory, but that’s about it. Any other expansion is through USB.

The exception is the Mac Pro, which starts at $2,499. In reality, the Mac Pro is a dual Xeon workstation – not really a desktop PC.

On the other hand, companies like Dell, HP and a host of smaller companies sell a host of expandable desktop PCs, for considerably less money than a Mac Pro. In some cases, they cost less than an iMac. That’s not to denigrate the iMac. Apple pushes useful technology forward in interesting ways. For example, the latest 27-inch iMacs offer a 2560×1440 resolution display – more pixels than the average 27-inch monitor, which is typically 1920×1200. Better yet, they’re LED backlit displays built on IPS panel technology, making them excellent for photographic work and video editing.

On the other hand, what you see is what you get. Want to upgrade that Radeon HD 4850? Sorry, you can’t. Want to build  RAID array? There’s no provision for even eSATA connections.

While the major manufacturers of Windows-based PCs also sell all-in-one systems, their desktop bread and butter is the standard PC with expansion slots. And, of course, you can build your own PCs; all of the systems here in the Case house are built from scratch.

That open hardware ecosystem evolved over time, and wasn’t intentional. It all began back in the 1980s, when several companies (Compaq being the first) reverse engineered the IBM PC BIOS, and the world never looked back. IBM’s attempt to dominate using proprietary technology – Microchannel – proved to be an abject failure.

That evolution created a rich ecosystem of hardware and software manufacturers. The staggering variety of components that can be installed into any of a number of motherboards creates an infinite combination of hardware possibilities. No one, not even a company with the resources of a Microsoft, can test all the possible combinations.

So someone, somewhere, will have problems with Windows 7. It’s the nature of the PC. What will inevitably happen is that some blog or news organization will report these inevitable issues with undisguised glee, suggesting that the evil Microsoft has gotten it wrong again.

I also built one Intel Atom-based system, using a Travla C156 case. While compact, I could have used any mini-ITX motherboard with that, and the north bridge on the particular motherboard I used had a pretty noisy fan. So when Intel dropped off yet another Atom motherboard kit with me, I wasn’t initially very interested.

But It’s Really Small…

When I finally got around to opening up the box, I realized this kit was just a little different.

All You Need for a Slimline PC

According to Intel, the D945GSEJT, aka “Johnstown”, had a design goal of a 20mm maximum component height. Built around a 1.6GHz Atom N270, the board designers actually hit 20.5mm. It’s a completely fanless design that can fit inside very slim cases.

This motherboard is just 20.5mm tall.

Given the design parameters, the height of the ATX I/O panel was also necessarily limited.

The ATX I/O cluster is half height.

As you can see, the I/O stack is half-height. The real limitation, as we’ll see shortly, are the presence of only three USB ports. I do wish that Intel would give up on VGA for these sorts of systems, and go with DVI and HDMI, though. Power for auxiliary devices is on the motherboard itself, in the form of a four-pin, Molex-style connector. The motherboard also has the DC connector for the external power supply mounted on the motherboard; it’s actually part of the ATX I/O cluster.

I know a number of power users (who are not financial analysts or programmers) who use three displays. But using more than two monitors had several requirements I considered problematic. You either had to have two graphics cards, a special workstation card (that wasn’t always a good 3D card) or some type of DVI or VGA switcher, like Matrox’s TripleHead boxes.

All those introduced complexities I didn’t like. I just wanted one card with good 3D gaming performance that supported more than two monitors – and wasn’t a dual GPU card. Now I’ve got it.

There’s been a lot of copy written in the past month on ATI’s Eyefinity capability. Eyefinity is a combination of hardware and software to enable multiple displays. The key feature ATI has been touting is using multiple displays – up to six – as a single, large surface, what ATI dubs “SLS” or “Single Large Surface”. That’s all well and good, but it’s not quite the panacea ATI makes it out to be.

Consider the setup I just built.

Three Monitors, Different Orientations

Strategies for Three Display Setup

What you see is a Dell Ultrasharp 3008WFP in the center, surrounded by two Gateway FPD2485W monitors. The Dell has a DisplayPort connector, which is a critical component to this setup. More on that in a moment.

The strategy I used is what probably many users will do. Most people won’t go out and buy three brand new displays, although you’re starting to see the occasional three monitor package deal. What most users are likely to do in a similar situation is to buy a DisplayPort equipped monitor, then sub in one or two displays they have in their homes already.

In my case, I’ve had the two Gateway monitors for several years. These are actually pretty decent, S-PVA panels – a cut above what Gateway is selling these days. I’d been eyeing the Dell 3008WFP, but didn’t want to pay the full $1699. Recently, Dell had a $300 off special, so I pulled out my tired credit card one more time and sprung for it.

I did make one compromise: I swapped out my rather bulky AudionEngine 5 speakers for a set of Creative Labs Gigaworks T3’s. The T3’s, with a little tweaking, sound almost as good as the A5’s — and do offer deeper bass.

We’ve gone through three HP inkjet all-in-one printers in the past four years. One, the Officejet 7410, refused to connect to a wireless network, so that was returned and replaced with a Photosmart C7180. The Photosmart 7180 eventually died with a paper jam that could only be cleared by removing the *ahem* non-removable paper tray. (Why, exactly, is the paper tray non-removable?)

So we recently acquired an HP Photosmart Premium Fax All-in-One aka the C309a. As a printer, it works fine — even double-sided printing works as advertised. However, the printer will occasionally fire up for no good reason. The print heads will move, the printer shakes and makes a tremendous racket, then the printer will quiet down again. I suppose it’s priming the inkjet heads, but it’s way too noisy and distracting when it does this.

On top of that, the HP-supplied scanning software is simply horrendous. It’s buggy, for one thing. Whenever I scan an image, the next scan reports that the printer is offline, when it’s not. (The printer itself is connected over a wired network.) There appears to be no native 64-bit support, either. Plus, it’s incredibly, painfully slow. The actual act of scanning is fast enough. However, any time you need to navigate from one screen to the next (say, to bring up the scan settings), it takes a full 4-8 seconds to bring up a new window. This is on a Core 2 Duo 2.26GHz laptop running Windows Vista Home Premium x64 (with 4GB of RAM.)

When the scan is finished, saving the file takes almost a minute (!).

HP’s Windows 7 support isn’t available yet, though Windows 7 has a built-in printer drivers for the C309a, but no way to connect to the scanner over a network. HP’s printer software won’t even install under Windows 7. The trick of unpacking the driver and manually installing it by setting compatibility mode (which worked with the C7180 drivers), doesn’t work with the c309a drivers. That’s why I’m using the scanning software on a laptop, instead of my much faster desktop PC.

Recently, I had the opportunity to use a Canon printer. The software was slick and responsive, and the printer output quite good. I haven’t checked out Canon’s scanning software, so can’t comment on that yet. But the printer software was easier to use and set up than HP printers, and doesn’t clutter up the desktop or tray with lots of crapware.

I’m not the only one having issues. I’ve talked to a number of users who are having similar issues with HP printers. Support has been lacking, and any issues seem to be met with a “buy a new printer” mantra from HP’s support help line.

Recently, HP announced a corporate reorganization, in which the printer group is being merged with the PC division. This can only be a good thing. HP’s PC group has been kicking serious ass lately, developing innovative laptops and building interesting technology into desktop PCs (like the TouchSmart systems.)

We can hope that some of this attitude rubs off on the printer group. HP’s printer divisions seem to be coasting along, resting on past glories. It needs to do some serious navel gazing and start designing products that work well with software that’s actually useful and not painfully slow.

These systems were the last pair of systems built to upgrade my Friday Night Follies LAN party systems. Will I be good to go for a few years, or will I be writing a version of this article in a couple of years? That depends, I suppose, on the health of PC gaming a couple of years from now.

But for the present, let’s see what the midrange looks like.

Twin Flavors of Core i7

One interesting artifact of the current http://www.intc.com/priceList.cfm Intel price list is the wholesale pricing for two Core i7 CPUs: the socket 1366 Core i7 920, and the socket 1156 Core i7 860. The 920 weighs in at 2.66GHz, 4.8 gigatransfers per second over its dual QPI links and triple channel DDR3 memory support. The Core i7 860 clocks in at 2.80GHz, but only has a single QPI link, for 2.5 gigatransfers per second, and a dual channel DDR3 memory controller.

Which is faster? It’s hard to say. Both support Hyper-Threading, but the D stepping of the Core i7 920 has become somewhat of a darling among overclockers. However, you can also argue that the Core i7 860 has more Turbo Boost headroom. Turbo Boost is Intel’s auto-overclocking capability, and the 860 offers more Turbo Boost modes than the 920. That’s likely to be a bigger factor than the number of QPI channels. My next Hot Hardware column will touch on the performance comparisons of these two CPU cousins.

The Core i7 920 System

Configurations

While I plan on benchmarking these systems with identical graphics cards, the final builds will be somewhat different, more because of what I have available here in the Basement than because of any intent. Here are the final bill of materials for the two systems:

Let’s consider pricing just for motherboard, CPU and memory for a moment:

Core i7 860 Parts

As you can see, the Bloomfield / X58 system costs more – about $51 more – for just the core components. The key difference is the motherboard pricing: even the micro ATX eVGA board costs about $200, versus the $170 for the mid-level Asus P55 board. However, memory cost per gigabyte is cheaper for the triple channel kit, although the total cost of memory is higher.

Frustration Ensues

I’d recently started building a system based on the Intel Core i7 920 “D” stepping. The D stepping is known to be highly overclockable, and I thought it would be entertaining to build a small, but pretty powerful system, around an overclocked Core i7 920.

I’ll go over the system build details later this week. But the one unfamiliar component I was using is the eVGA X58 SLI Micro, a micro ATX board built around the X58 chipset. I’d been using an eVGA full size X58 board, so felt comfortable using the micro ATX version – but still, it wasn’t something I’d used before.

I was building the system to replace an aging Core 2 Duo system that had been one of the Friday Night Follies LAN party rigs. I was planning on installing Windows Vista Home Premium, in anticipation of the Windows 7 upgrades I’d pre-ordered for $50 a few weeks ago.

So I build the system. Everything goes well, and the system cruises through the power-on self-test without any apparent hitches. I hit the DEL key in order to bring up the BIOS setup.

I get distracted, because one of the two hard drives I’d configured into the system was apparently defective – I see a “SATA Drive Failure” error on POST. I swear a little, swap out the drive (a pretty beat up Western Digital 640GB drive), and reboot, and go back into BIOS setup.

As I usually do, I configure the BIOS so the system will boot from the optical drive. I discover a downside to the eVGA BIOS – it’s pretty sparse, and it’s difficult to figure out how to set the system up to use AHCI on the hard drives, so I leave them in IDE mode.

Finally, I pop in the Windows Vista Home Premium DVD and boot into Windows setup. Everything goes well, then suddenly I see a blue screen, with the following error:

IRQ_NOT_LESS_THAN_OR_EQUAL

This error is followed by some verbiage asking me to uninstall any new hardware or software I may have installed.
I ante up the blue screen error with a few blue words of my own choosing. The entire system is new, and there’s no software installed, so this particular error message is singularly useless for troubleshooting. It occurs to me that maybe the DVD is dirty, so I wash it with warm soap and water, dry it off and reboot.
Same thing.

IRQ_NOT_LESS_THAN_OR_EQUAL

Now I’m thinking: bad motherboard. Bad, bad motherboard.

The man in the picture to the right is Justin Rattner, the chief technology officer for Intel, and he seems like a fellow who really enjoys his job.  Intel puts Justin’s presentation of Intel research projects on the last day, since it is always popular and entertaining.  This time, Justin had a huge 3D screen rolled out on stage and had us reach under our seats to discover 3D glasses.  After a live 3D interview with a vendor, the screen was used to immerse the audience into a U2 concert.  Bono appeared to walk out into the auditorium, and the effect was incredibly cool.  Consumer 3D television should start to show up next year.

The Sponsors of Tomorrow

This was Intel’s theme for IDF 2009, and my MBA background usually makes me roll my eyes at the marketing hyperbole Intel sometimes evokes.  If you just focus on Intel’s investment in basic research, their conference theme message is somewhat deserved.  While Intel has a self-interest in promoting compute-intensive tasks, many of the research areas in visualization and energy management will lead to new technology that should benefit us all.  Loyd and I attended some interesting press briefings and technical sessions that highlight some of these Intel research projects.

When Facebook just isn’t immersive enough

Anyone who has attended a graphics conference like Siggraph has seen technology for converting photographs into 3D models.  Intel has a research group in Beijing that focuses on technology for immersive user content.  The team has created software that will automatically generate a 3D model from a sequence of pictures taken at approximately 15 degree increments in a circle around an object.  You could also just take a video clip, and the software will choose appropiate frames that stitch together the model.  The obvious goal is to create a 3D model of a person and then allow you to import this model into a 3D world as your personal avatar.  From the examples Intel showed, they are getting very close to making this work, and Facebook will be an entirely different experience in about 10 years.

Signal processing your home electrical system for fun and profit

Another Intel research area is exploring the “smart grid”, which will use computer technology to improve power generation, delivery and consumption.  For a press demo, Intel showed a model of the home electrical system — complete with an actual microwave oven, a toaster, and lighting systems.  The technical concept is really cool, since the researchers discovered that every electricity consumer has a unique effect on the voltage that is measured at the wall outlet.  Using a voltage transducer to sample the voltage droop as an appliance is turned on, a small wireless device can transmit this data to your home PC.  Using time-domain signal processing algorithms, this dynamic voltage information can be cross-correlated with stored profiles for various appliances.

I was amazed that this actually works, considering how much noise is on a home electrical source.  The sensor even works when the appliance is on a completely different circuit from where you’ve plugged in the sensor.  One of the media analysts asked Intel to explain the purpose for this technology. The system gives you a lot of data about the status of every electrical consumer in your home, but what do you do with this information?  I haven’t thought about it much, so jump on to comment on what you think.

Recently, I decided it was time to replace one of the gaming rigs in my office LAN. On most Fridays, my LAN party group gets together for what we affectionately call “Friday Night Follies.” I usually build systems out of older parts for the LAN party systems, which are generally good enough for the games we play.

This time around, though, I wanted to build something a little more current. So I decided to build a compact system based around Gigabyte’s P55-UD4 micro ATX motherboard and a Core i5 750. Part of my rationale was to test long term stability. You can run all the benchmarks in the world, but there’s nothing quite like hammering on a system in a current generation multiplayer game to really test system stability.

Case Imperfect

Gigabyte had sent me a P55M-UDR motherboard, the same micro ATX motherboard I showed off in my recent appearance on Tekzilla. I’d been wanting to build a system around this, as well as push the board a bit. I finally freed up some time, gathered up some component that had been lying around the basement lab, grabbed my electric screwdriver and set to work.

The case, a black NZXT Rogue, was something I’d reviewed at ExtremeTech some time ago. It’s an attractive case that would look comfortable in a Section 8 level. It looks cool, and is a fine case to build a system you plan on just running for awhile. Building a system into it does require a lot of screwdriver work. You need to remove screws to get to the motherboard tray (it’s removable), take out screws to remove the front panel, so you can unscrew the hard drive bay to install the hard drive, which requires more… well, you get the picture.

The NZXT Rogue looks cool, but system building takes a little effort.

The other glitch with this case is that some graphics cards are a tight fit. Actually, it’s long enough to handle long graphics cards. However, the printed circuit board on many high end graphic cards sticks up above the PCI slot plate just a tiny bit. With the Rogue, it means you can’t install the graphics card before sliding the motherboard tray in. What you do is slide the motherboard tray partway in, install the graphics card, then peer into the tiny space between the tray and the case, so you can screw down the graphics card. It’s a minor annoyance, but a little forethought in the design could have prevented this.

Like many compact cases, NZXT got a little creative with the power supply location. The PSU lives above the expansion slots so you need to have all the components installed before installing the PSU. That installation actually works better than you might expect, and NZXT even supplies you with an extension bracket if you use one of the more immense kilowatt PSUs.

Once up and running, the Rogue looks good, keeps the internals reasonably cool and isn’t too noisy. I would have preferred to have fans with 3-pin motherboard connectors (at least as an option.) If you’re not inside the guts of the case, replacing components often, it’s a fine little case.

Other parts consisted of gear I’ve had in the lab forever.

• I’ve had an excellent Antec Signature 650W PSU just sitting on a shelf, which is a tragic waste of a good PSU.
• A 4GB (2 x 2GB) 1600MHz HyperX DDR3 kit from Kingston.
• A Western Digital WD7500 AAYS SATA hard drive (750GB, 7200RPM)
• Lite-On 20x DVD +/- RW optical drive (IDE)
• AMD Radeon HD 4870 1GB graphics card.