Intel takes swift action to correct NUC problems

An object lesson in making the unpalatable less palatable

Intel - logoIn the last part, we took a look at Intel’s NUC new mini-PC form factor, what it got you for the money, and how it stacked up to the competition. This second half chronicles how Intel is attempting to fix those problems, and what it says about how they can compete going forward.

If you are wondering why we say that Intel is lost, their reaction to the non-competitive nature of the NUC is quite telling. When a product is that ill-suited to the markets it is aimed at and early sales numbers look to be reflecting that fact, quick corrective action is required. In this case, Intel management did see that there was a problem, and is taking quick corrective action. What they are doing however directly illustrates the dire straights they are in as a company.

According to Tech Report, Intel has noticed that the NUC is priced out of line with reality, and is taking steps to fix that problem. The MSRP of the NUC is about $300, and the MSRP of the CPU is $225, so to drop the price of the system, there is only one knob to turn, CPU price. Intel is doing just that, and releasing a new NUC variant soon that replaces the i3-3217U CPU with the cheaper Celeron 847.

Since the MSRP of the 847 is $134, this CPU swap will shave about $100 off the price of the new system. That seems reasonable, in relative terms anyway. In absolute terms we should revert back to the first installment of this article and ask the questions, “What are they nuts? Who green-lighted this and why do they still have a job?”. These are serious questions, and you will understand why we ask them in a bit.

Changing the CPU may mean that $100 is taken off the top of a NUC barebones system, but what do you get for this new lower price of around $200 retail? The heart of the system, the CPU with integrated GPU was swapped out for something that costs about half as much, so what was given up? Net cost for a full system with Wi-Fi for the build-it-yourself crowd only drops to $500 from $600, and the price from a Tier 2 OEM now sits at $600, not exactly what you would call cheap either way. If the performance cost was minimal, this is a step in the right direction. Progress right?

Once again, ask yourself if a new $500 or $600 NUC is more tempting than the old $600 or $700 versions. Dropping $100 is a big deal, but not when your competition is $35 for a complete system. If it does everything the consumer is asking for just as well as your version, you still have just as serious a problem. Instead of 20x the cost, the Intel box is now about 17x the price at retail, something that would be hard to call a tempting offer even no performance was given up in the CPU swap.

Most people in the market for a $10K used car could possibly be convinced to buy a $15K new car, a case could be made on payments, operating costs, and resale price. The Rolls Royce dealer next door will probably not have much luck if they offer that shopper a whopping $100K off the $380K MSRP of a new Phantom. If all the buyer needs in a vehicle is to get to work and back, and a used econobox will do that for less, the Rolls would have to come very close on price to even make a case.

The real problem for Intel lies in what that $100 costs you in performance, essentially what that Celeron 847 they substituted actually is. The i3-3217U in the original NUC was a pretty awful chip all told, too slow to be useful in business with most of the business oriented features like virtualization and encryption removed as well. The Celeron 847 takes this to another level entirely. If you look at the speeds and features sets, the $100+ price only makes it more painful. How this part was released is a mystery, how it made it in to the NUC exemplifies the bind Intel is in.

The first problem is what CPU generation the Celeron comes from. The i3-3217U is a 22nm Ivy Bridge, the Celeron 847 is an older generation 32nm Sandy Bridge and has an older less-efficient GPU architecture. Shader for shader, the Sandy generation GPU is about half the speed of the 22nm Ivy GPU, and while the 3217 has 16 ‘new’ shaders, the 847 only has six of the older ones. Some of those may be fused off, but we will assume that they are not for the sake of the argument. This means the Celeron 847 has roughly 20% of the GPU power that the i3 does, and that is before clock speed caps. Both chips have a 350MHz base GPU frequency but the i3 can increase it to 1.05GHz on the fly, the 847 stops at 800MHz. This means that even at full speed, the 847 won’t actually come near the 20% figure that the shader count difference implies.

To make things worse, the 847 has several other basic CPU performance features like Hyperthreading and Turbo (CPU turbo, not GPU) turned off. In a very counter-intuitive move however, virtualization is not turned off, a feature choice we have no good explanation for. This might allow a NUC with an 847 to be a passable business machine, but that brings us to the next crippling problem, clock speed.

The i3 has two cores and four threads running at 1.8GHz clock, not enough for real world business use. The 847 has two cores as well but as we mentioned, threading is fused off, turbo is fused off, and it only runs at a pitiful 1.1GHz. If the i3 was too slow for business use with four 1.8GHz threads, how well do you think the 847 will do with two 1.1GHz threads? We won’t bring up the added efficiency of the Ivy architecture, nor the fact that Intel capped the memory speed on the Celeron to 1333MHz, much less halved the memory size limit to 16GB from 32GB on the i3. This part is untenable for a business PC, most modern cell phone chips will make a better showing in benchmarks.

In case you can’t see where this is going, it does indeed keep getting worse, far far worse. This part choice makes so little sense that you have to question whether Intel isn’t trying to purposefully fail for some reason. The NUC with the i3 was too slow for business, and the graphics are barely acceptable for media center use, but not outright unsuited. There were some other features like hardware video encode, decode, and transcode that allowed the original to not totally fail in it’s intended purpose, but only by the slightest of margins. Guess what the newer NUC-C, our term with the -C signifying the Castrati edition, has fused off?

Indeed, no points if you correctly surmised that Intel removed the hardware accelerated video playback, encode, and transcode features. Not stopping there, they also removed the Wireless Display, the Intel Insider DRM infection, Intel InTru 3D, Intel Flexible Display Interface, and ClearHD technology. When these technologies were launched, Intel was very adamant that these features were completely necessary to have a quality video experience, users would suffer without them. SemiAccurate laughed at this because most of those didn’t actually bring any benefits for the end user, but lets take Intel at their word for once.

All of these things that they consider mandatory for a passable video experience were removed in the Celeron 847 they put in the NUC-C. But Intel didn’t stop there, basically every video related function was fused off too, there is essentially nothing but two monitor support left. Normally, a computer without hardware video features can make up for it through raw CPU power, but as we pointed out, the Celeron 847 has about half the CPU performance of the i3 it replaces, half! For a device that was barely tolerable at doing media related tasks with the old CPU, removing all of the needed features for that market is just silly. But it saves $100, and they were looking for a lower price right? Did no one consider the cost?

Getting back to the serious side of the story again, you might now have a little better understanding of Intel’s rather untenable competitive position. The competition does everything the consumer wants that the NUC provides, and does it in a similar sized box. The NUC’s old CPU alone costs 6-7x the $35 total price of the competition, a functional system built with it more than doubles that. As a response, Intel lopped $100 off the price of the NUC to create the NUC-C, and in doing so ended up with a box that is simply unsuitable for anything, too slow for business, and every possible video feature removed kills the media center use case. It is completely worthless in every potential market that Intel considers the NUC to be a good fit for. Worse yet, the $35 G500A destroys the NUC-C for performance in every relevant category, and does it for 1/15th of the cost, not 1/20th of the cost anymore. Is this what you would categorize as a more competitive offering?

How can Intel compete with silicon priced like this? Unfortunately, their marketing strategy will not allow them to put out a competitive part at a reasonable price. The self-imposed market stratification dictates that anything below $150 ends up as garbage, and wouldn’t you know it, that is exactly what the Celeron 847 is. Putting it in a media center box is madness, but marketing is the only voice that is heard at Intel currently, sanity is drowned out. Any attempt to get upper management to listen can be readily countered with the simple fact that any deviation from this madness will result in crushing margin drops and the attendant stock hits. That is not an option, period, so the madness continues.

Intel reinvented the PC with the NUC. In the desktop arena Intel silicon is not challenged for the most part, and they can exclude any competition with ruthless effectiveness. Unfortunately, the market has spoken loudly and clearly that the desktop and even laptop devices that Wintel offers are not what they want or will buy. iPads, phones, tablets, and other similar things are flying off the shelves while traditional PC sales are down. The NUC is a traditional PC in a smaller box that offers up all of the things that consumers didn’t want, and most of the ones they do want are removed. And it costs 2-3x what a traditional small form factor PC does as well. It might have a place as a $99 device, but Microsoft wants that for the OS alone and Intel has margins to protect, so $99 becomes $699.

The NUC is Intel’s grand plan productized. This is how they will fix the problems of the PC. It fails at that. They admitted as much with the introduction of the NUC-C and directly attacked the price problem. Unfortunately, the result is so awful I feel pity for anyone buying it, it is simply a non-functional box at a laughable price, and that is the best Intel is able to offer. They are trapped by their own marketing. The competition is not, they offer $35 boxes, complete out the door with all extras including power cables, that does everything the consumer wants. Intel can not include power cables at 20x the price much less offer similar functionality at 15x that price, power cable or not.

Intel does not understand why they are failing, and that is the kind interpretation. If they do understand, their lack of real action means the company is utterly trapped by their own master marketing plan. ARM, AMD, and everyone else are not bound by the same self-imposed rules, marketing dicta, and margin concerns. The competition is seeing profitable niches and exploiting them where Intel only sees cracks that they are unable to fill. Attempts like the NUC are sad, attempts like the NUC-C make one question Intel’s basic competitive ability. For the moment, we can’t come up with a good reason that they will survive in this new post-PC era.S|A

The following two tabs change content below.

Charlie Demerjian

Roving engine of chaos and snide remarks at SemiAccurate
Charlie Demerjian is the founder of Stone Arch Networking Services and SemiAccurate.com. SemiAccurate.com is a technology news site; addressing hardware design, software selection, customization, securing and maintenance, with over one million views per month. He is a technologist and analyst specializing in semiconductors, system and network architecture. As head writer of SemiAccurate.com, he regularly advises writers, analysts, and industry executives on technical matters and long lead industry trends. Charlie is also available through Guidepoint and Mosaic. FullyAccurate