Hope you don't own Intel stock

[gunderwood]

Can you say SNAFU?

http://www.anandtech.com/show/4142/inte ... ins-recall

http://www.anandtech.com/show/4143/the- ... t-sata-bug

In the end it isn't a huge deal as long as you don't have a lot of drives on those ports and need them to work 100% before replacements can be had. However, they only set aside $700M for the recall and I doubt it will cost that little. Wait and see.

[davasmith]

Why is it that our Tech fields are becoming more and more FUBAR? Doesn't anyone plan for a hiccup these days? In the end, follow the money. They'll probably need a 1 billion dollar bailout by monday or they'll have to shut down. Oh yeah, that was the gov't.

[gunderwood]

Why is it that our Tech fields are becoming more and more FUBAR? Doesn't anyone plan for a hiccup these days? In the end, follow the money. They'll probably need a 1 billion dollar bailout by monday or they'll have to shut down. Oh yeah, that was the gov't.

Intel need a bailout? Hardly. You have to admit Intel is a very well run company even if you don't like them. They revised their profit for the year down to just over $11B after estimating a $700M hit with this and $300M in lost sales while they fix it. Intel spends almost that much money on R&D every month (two if you include the lost sales). http://www.rdmag.com/Featured-Articles/ ... Computers/

This only affects their latest chip set which was only on sale officially since January 9th. It is a serious bug which has no software workaround, but in the grand scheme of things it won't hurt them too much. Wall Street will over react and their stock will drop a lot, but even if the recall costs them 2-3x what they estimate today they will still make at least $9B this year, hardly a failing company.

These things happen when you are designing a something so complex. The root of the problem appears to be that they attempted to reuse some of their older designs (nothing wrong with that) and made a design mistaking integrating the pieces. I've got one of these boards, but it works fine for now and they will replace it. A hassle? Yes. A marking nightmare? Yup. Nasty stock hit? Probably. The end of Intel and government bailouts? Nope.

[t33j]

http://en.wikipedia.org/wiki/Pentium_FDIV_bug
Hardly a new thing. They'll fix it an move on.

[zephyp]

I own Intel and will hang onto it. Doubled down on it back when it hit a low in 2009. IMO the biggest problem for these tech companies is to constantly upgrade chips/software/whatever and then find a reason to make everybody want to buy it. You can build a PC these days that will provide decent service for 3-5 years (or longer) with minimal maintenance. That dont sit well with companies like Intel/Microsoft/Dell/etc. Fortunately for them (and investors) there are enough folks that have to have the latest and fastest chip/whatever as soon as it hits the shelf.

[gunderwood]

I own Intel and will hang onto it. Doubled down on it back when it hit a low in 2009. IMO the biggest problem for these tech companies is to constantly upgrade chips/software/whatever and then find a reason to make everybody want to buy it. You can build a PC these days that will provide decent service for 3-5 years (or longer) with minimal maintenance. That dont sit well with companies like Intel/Microsoft/Dell/etc. Fortunately for them (and investors) there are enough folks that have to have the latest and fastest chip/whatever as soon as it hits the shelf.

+1

I was explaining to a college what I called the ARM strategy over the Christmas break and then CES confirmed it. The problem is that we finally reached a point where the average PC has become fast enough for most users. Honestly, you won't notice the difference between a $1k computer and a $5k one for web browsing, email checking, FB updates, etc. There still is some work to be done at the low end of the market, but not much. Going from one core to two was a great upgrade, from two to four was nice and a few tasks can take advantage of them, but not many and the average users doesn't usually do those tasks either. So what good is the average user going to get out of a 8 core or 16 core CPU? Unless they find an application where they are needed, not a thing. The average user will end up buying those CPUs because there are markets which need them and economies of scale dictate that they sell similar designs...at least for a while longer.

The biggest problem Intel has is the x86 ISA isn't the ISA of the future. It is big, complex and results in a relatively power hungry CPU. Intel has tried expanding into the lower power market with Atom, but so far that hasn't paid off because the ISA just isn't the right one for the job. Intel hasn't always have the best general architecture, nor is x86 a great ISA, but the one thing Intel has which is IMHO the best in the world is their fabs. They can make chips like no one else in the world. Sure, some of their competitors are not too far behind, but their fabs are second to none. Their manufacturing (fabs) have bailed them out many times before from bad architectures, to competitors with better ISAs, etc. It's their ace up their sleeve so to speak.

The problem for Intel is that as soon as performance reached good enough status for most users, they have a serious problem. x86 made sense because when combined with their fabs it delivered more of what people use to want, performance. Now that performance isn't the primary concern for the average user, x86 only makes sense because we already have the infrastructure in place (software and tools) to support it. I.e. if we switch to another ISA we must redo all of those software and tools. Even most company servers are not moving towards virtualization (basically consolidation) because it doesn't make any sense to only do one task per server anymore. Sure, there are hard problems and power users (like gamers) who still want more performance, but they are quickly becoming the minority.

So what is the future? Phones. Well, not phones exactly, but their usage model. Even smart phones are underpowered to completely replace the desktop/laptop market for the average user. Again, there will always be power users who need a desktop/laptop form factor, but they won't be the norm IMHO...at least not in the way we recognize them now. Smart phones are getting better and they are popular because they allow users to do the things they usually do on a PC (web, FB, email, etc.), but with decent battery life and extreme portability. Users want to do those things anywhere and anytime. Historically those devices have been ARM because lets face it, the ARM ISA is probably the best low power, just enough performance ISA ever designed. While performance was driving the market it could never compete, but as soon as batter life, portability, efficiency and just enough performance became the driver ARM is #1 IMHO.

Since smart phones were too underpowered to run regular desktop/laptop software we had to redo the software and tools anyways, it wasn't a matter of choice it was a matter of necessity. It isn't all that different in how the Chinese have nabbed two of the top 5 fastest computers in the world including #1. They didn't have the huge infrastructure investment in HPC like we do, so since they had to build most of it from scratch they just decided to use a very different approach with GPGPUs. If you have the right problem, CPUs can't and likely never will compete with GPGPUs for matrix/FLOPS crunching. ARM has been powering the low power devices for a long time now; it makes all of the consumer electronics you don't think of as computers possible...TVs, DVRs, phones, smart appliances, etc. ARM dominates that market.

So what happens when this all converges? You end up with mobile devices like smart phones growing in power and eventually being powerful enough that the average user doesn't need anything else. The portable device becomes their primary computer. CES announcements confirmed this position.

Motorola showed a phone which has a docking station. You plug it into the docking station and you instantly have a keyboard/monitor setup just like your laptop. No syncing, they two are one. http://www.anandtech.com/show/4112/anan ... -4g-webtop
This is only a first generation device, but I'll put good money on there being a lot more very soon because it just makes sense for the average user.

Nvidia has long been rumored to be designing a CPU as they were left out in the cold as the low end GPUs get integrated into CPUs (e.g. Intel's Sandy Bridge which is this chipset problem). However, that never made sense strategically. Nvidia would end up being like AMD, a small portion of the market, but Intel's fabs just couldn't be matched at that scale. This is exactly why AMD sold off their fabs, they needed to grow much larger and make ICs in order to stay competitive.

However, at CES Nvidia announced Project Denver which is a CPU design, but not x86, it's ARM. It isn't the actual ARM processor like they did for Tegra and which is what most smart phone use, they actually licensed the ARM ISA and are designing a CPU they claim will compete with x86 in the desktop/server market. ARM can't compete with x86 on max performance, but on efficiency it can. In short, Nvidia is taking a high risk move to become the next Intel. As ARM moves up into the mainstream devices more and more software will only run on it, it will have the momentum that x86 has today. Also, as that happens portable devices will challenge typical usage models for desktops/laptops. Sure, power users will still need something bigger, but the average user won't. Nvidia is banking on the economies of scale shifting towards ARM. I.e. if most users run portable ARM devices and most software is at that point in the future written for ARM, it makes sense to move the most power users to ARM as well. This is basically exactly what the various product lines of Intel leverage today. Atom at the low end, Pentium, Core i3/5/7. Nvidia isn't the only one making this move, but they do have a lot of the GPU IP which will enable portable devices to drive those graphics and larger monitors as they challenge the standard usage models.

By itself Nvidia's announcement wouldn't mean much and I suspect that is exactly why they didn't bother letting the project be known for so long. A desktop/server ARM just doesn't have a market today. Really only Linux/Unix can leverage it...which is what has Microsoft scared. All of these mobile OSs, until Windows Phone, weren't Windows. Microsoft announced at CES that "Windows 8" will come in two versions, x86 and ARM. That is huge! At the very least it means that MS is scared enough to invest a lot of resources into making an ARM OS/Office/Etc. just in case. They may or may not think ARM will succeed in knocking x86 off, but they consider it possible enough to prepare for it. http://www.microsoft.com/presspass/pres ... pport.mspx

The usage models are probably a near certainty. Who wins the next generation is anyone's guess, but ARM has a huge lead at the moment. Intel won't take that lying down, they own the IP for x86 and want everyone using it for obvious reasons. They will try and might succeed in keeping ARM at bay, but it won't be easy. If the economies of scale shift towards ARM I'm not even sure the high performance crowd will be able to resist moving to it, especially when you consider that GPGPU technology is rapidly taking over all the really hard problems (that too is uncertain, but the momentum looks promising). A GPGPU for all the hard problems and an efficient ARM for organization and overhead and easy problems sounds like a winning combo to me. Time will tell.

[gunderwood]

http://en.wikipedia.org/wiki/Pentium_FDIV_bug
Hardly a new thing. They'll fix it an move on.

Yup, there have been others and there will be more. It happens and Intel is handling it the right way so far, but it will cause them to take a hit on profit/stock for a while.