The Spectrum of Red, Green & Blue

April 29, 2007 | 17:51

Tags: #beijing #fab #fusion #idf #k10 #manufacturing #penryn #ps2 #ps3 #r600 #torrenza #tunisia

Companies: #amd #ati #intel #tsmc

The Intel Developer Forum has passed (including the long flights to Beijing), and so has AMD’s ATI’s event in Tunisia. On reflection, the two events were very different: AMD had a launch event with the main focus on one particular product family and Intel held a very broad event covering all of its products. However, it’s interesting to see how different companies engage the press, and I thought my experiences are worth sharing.

IDF is all about meeting the people that matter and it’s very easy to seek out people who you’d never normally get to talk to and grill them to your heart’s content. This is despite there being so much going on at the same time because of the sheer depth of Intel’s product portfolio – there’s still time to talk to the guys behind the products.

I met one of the engineers who designed SSE4 and worked on optimising DivX 6.2 with the new instruction set. I also met with process technology superstar, Mark Bohr, who has overseen Intel’s semiconductor manufacturing developments from 130nm down to 45nm. In addition, there were the many other small technical and research groups who were available for very up front discussions where they provided open answers to the questions I asked.

"The result was a somewhat awkward and interesting stand-off between the press and PR"

This is in comparison to having to suffer the usual PR spiel which requires the essential “read between the lines” decoder in order to get any actual information out. Such is the ongoing love-hate relationship between journalists and PR and depending on your position in the game, this delicate balance can become skewed.

Obviously, there are some subjects that have been strictly deemed a no-no, or are on a need to know basis: Intel’s plans to enter the discrete graphics market is a line of questioning to avoid if you want concrete answers. This is down to the fact that people either don’t know or clearly wouldn’t discuss it because it’s still in the very early stages. Answers are immediately directed to their current range of integrated GPUs, while highlighting their industry breadth.

This is in contrast to asking open questions regarding Nehalem. If you asked six different people the same question, you’d likely get six different answers. Each of these answers would give a slightly different spin on the subject, but would generally follow a similar line of thinking if you look at the bigger picture.

Because of the controlled nature of the event, the press pass is almost boundless in its ability to get you into places and in some of the better seats. But then again, we are a messenger for the consumer, even if it is with our own independent analysis and comparisons without the PR spin.

Another example is on the “Benchmarking Penryn” testing I took part in. Despite the fact that Intel went out of its way to make sure the systems were identical (while providing plenty of information on hardware and drivers), the first thing we asked was to change the clock speeds for a direct comparison. However this came up against a stubborn wall of resistance that Intel was having none of. The result was a somewhat awkward and interesting stand-off between the press and the engineers/PR.

The former wanted to do it the way they saw most scientific and applicable to their readers, whilst the latter wanted to show it exactly within the confines of what the processors are explicitly designed to run at. This included clock speeds, cache and optimisation differences together providing the maximum difference. After all, this will represent the differences between the CPUs that are going to be sold, not explicitly between the two architectures.

"The whole event is a complex equation of time management – balancing time spent getting information and speaking to the right people, with time spent writing everything up."

Obviously if we went away saying “Penryn offers a very small performance increase, just overclock your 65nm Core 2 chip to a 1333MHz FSB”, it wouldn’t do Intel much good. Those comparative tests will definitely arrive eventually, but not before the message that “Penryn is better than Conroe” has been engrained into the minds of potential purchasers.

For those unaware of the structure of events at IDF, it’s something like this:

Day zero was a press only event next door to where IDF actually takes place. With this we get some early information through keynotes that will be worked on in time for or closely thereafter the main keynote on day one. Later that day, we were then allowed to roam around the partner stands (that were still being built) to ask a few questions to those that were setting up whilst the atmosphere was a little less hectic than it is once the show gets started.

Day one sees the start of two days of absolute craziness. The volume of people has now exploded ten-fold as invited students, executive guests and all manner of other people who paid to attend now join the crowd of journalists.

In the morning, there are keynotes from the major people at Intel; these cover everything from mobile to server releases, to future products that are still in the early research and development stages. It’s all about where the company is aligning itself and maybe even repositioning from where it was just six months previous. In the afternoon, there are 'tracks sessions' which are still presentations like the keynotes, but on a much smaller scale. These are split into subdivisions of the morning’s coverage and run simultaneously all afternoon.

Alongside, there are the extremely useful one-to-one sit downs with key Intel people you can just walk into and ask away. These only run for a couple of hours each person so you have to be clever with your time management, but they offer a far more intimate chat providing detail that can offer some very unique results to your publication and your writing.

Of course, this all depends on whether you can ask the right questions, although it’s still a mix of luck and judgement. Occasionally there will be press-only invites to things like the benchmarking of Penryn or to sit down with someone even more senior at Intel, like Steve Smith – these events are organised by PR in advance. The whole event is a complex equation of time management – balancing time spent getting information and speaking to the right people, with time spent writing everything up.
However Intel doesn't do launch events. Instead we get a CPU, some slides and a phone call with someone talking through the information. This makes it just as difficult because it's hard to get the questions out you need and face-to-face time is valuable when you're dealing with a new product. Just regurgitating the slides may work for Intel, but trying to apply an independent voice is very hard.

Ideally, AMD really needs to replicate what Intel gets out of IDF for itself. The difference between the two companies is undeniably large and even though we were treated extremely well on the Tunisia trip, the former ATI still appears to very separate from its mother company.

Tunisia was very much an ATI event, with very few people from AMD’s other divisions actually present. However, we did get inklings of a generalised event rather than just a simple launch, as Giuseppe Amato offering some insights into what we can expect from K10 – we'll be telling you about these soon. As AMD prefers to hold launch events, we hope there's a similarly big shindig for the upcoming K10 launch, but we’re still yet to hear anything and the last real product-based AMD event was back in 2003 (of course, ATI also held regular launch events before it moved under AMD's wing – Ed.).

AMD may be a smaller company, but now it has a brother and they both play in the park with the big boys. Availability of its roadmaps has been cut and information regarding future products has become sparse. Does anyone know anything about AMD’s 45nm process let alone anything beyond? I did manage to get something out of someone in the know and apparently we’re looking at well into next year for 45nm, probably even approaching 2009, however this was just a whisper on a whim in my direction.

"AMD may be a smaller company, but now it has a brother and they both play in the park with the big boys."

AMD admits the last big jump it had was the move to 130nm, which required the inclusion of an additional metal layer. It also introduced a new low-k process to the first 130nm Thoroughbred-A Athlon XP processors. Intel has many patents on its 45nm process and AMD has to work around those in order to compete.

Company secrets are always necessary and keeping ones card close to your chest means that others can’t adjust their roadmaps to compete, but when you’re on the back foot against a bigger foe giving people some long term plans of what you are working on and researching is an invaluable asset.

It means your company gets written about, instilling the assumption of a long term future in readers, consumers and investors. AMD needs this more than ever after a dismal quarter only adding to the precarious level of debt, and having to deal with the aggressive pricing on Intel’s Core 2 processors, which have caused AMD to reduce its chip prices by at least 50 percent over the last year in many cases.

All we have right now are buzz words like Fusion and Torrenza, but no hard facts or progressive updates since what we originally were told what this actually entails. For all we know, AMD could also be researching flying monkeys or tools to simulate cat herding as well. It feels like there are huge gaps in what's going on at a company of that size.

Don’t get me wrong, merging two massive companies like AMD and ATI is a massive task. In six months, ATI’s former employees can’t really expect to be privy to everything their brethren are doing (and vice versa). However, AMD wasn’t even (perceived) as the most organised company to begin with, so surely having a few more people from AMD's other business units at the Tunis event wouldn’t have made matters worse?

Would it have still detracted from being a launch event? After all, you've got a ton of journalists in one place anyway, so why not let them loose a little? For those who were there but didn't need to know about the extra depth of information, giving them the opportunity to chat to people in some other areas of the company might have increased the company profile in the press.

"All we have right now are buzz words like Fusion and Torrenza, but no hard facts or progressive updates since what we originally were told what this actually entails."

While AMD still outsources manufacturing to TSMC for ATI products (and is its biggest customer), R600 allocation is going to be exceptionally tight. Because of this, either samples or retail products just don’t arrive in the quantities that are needed in order to make everyone happy. No company has ever gone bust from lack of demand, but in order to pay the growing bills you need to also sell as many as possible. But then again, what else can you do? Not everyone is lucky enough to have Fabs to spare and AMD is making the best of it that it can. At least it's not vapourware.

AMD can’t just drop a few billion dollars on building another Fab at the moment either – it doesn't work like that: an investment now will only blossom in three years time, at best. Intel may be investing in China because as an American company it can’t (easily) employ Americans in the world’s fastest growing market, and also in that respect it’s an exceptionally long wait to get some very talented Chinese into America.

As is ever the case, home grown talent gets priority for jobs and in the continual game of politics and business Intel may have held its first ever major IDF outside of America to highlight its investment in China, but the new Chinese Fab 68 will only grow chipsets from its silicon, not the latest CPUs. If you think about it this is no surprise since like the PS2, you can’t even buy a PS3 in China (I tried) because it could possibly be used for military purposes.

While I haven’t quite done the fully x86 world tour like another fellow journalist, just getting two thirds of the way around I understand it all a little better. It really does require seeing the world in red, green and blue in order to get a better picture of how the industry is moving.

Discuss this in the forums
YouTube logo
MSI MPG Velox 100R Chassis Review

October 14 2021 | 15:04