Intel may well be at its peak

ON THE PLANE from Beijing to Singapore, I thought again about the past week spent in the Chinese capital - my regular monthly haunt anyway.

Intel put up a powerful show of dominant force at this crowded IDF - and there was no competition in sight, very literally.

Despite Hector Ruiz and Bill Gate$ in town, the only talk about AMD at the show was how its current offerings are throughly outclassed by Intel's own current stuff, and how no one - except a few VVIP press & partners - me not included, had a chance to freely play with those famed Barcelonas in real life up to now.

Another bad sign that there was more side talk on what a combination of Larabee platform and all those good ex-3DLabs engineers could create for Intel's first serious discrete GPU effort, then about the hopefully soon very real ATI/AMD R600/RadeonHD GPU sets - or how would they run without air conditioning in the hot Tunisian desert, for instance. So, a remote Intel product that is at least a year or two away, was showered with more interest than a coming soon AMD same class one.

Well, I couldn't exactly blame the attendees and resident rumour-mongers. After all, AMD's past successes have forced Intel to be more open, pro-active and, yes, friendly.

It didn't take much effort even for an average attendee to find and actually play with a selection of upcoming Penryn-based machines on desktop, workstation and server demos, as well as the Santa Rosa mobile platform. Six month advance things that they were, only several years ago, reserved just for a chosen few.

Is this a 'Tour de Force' by Intel, showing off in the capital of the Asian future superpower, where it has its largest R&D facilities too, just a peak-time show-off, or a sign of stronger dominance to come?

As the recent quarterly results show, Intel beat AMD in the market share battle this round - the performance pre-eminence of Core 2, coupled with a 'fire sale' of remaining Pentiums, left AMD with no chance.

In hindsight, it wouldn't have that problem if, supposedly, they hadn't canned previous K9 and K10 efforts so quickly, assuming Intel would be stuck in the NetBust quagmire forever.

So, almost every new demo machine on this IDF, from mobile to server, could lay claim to being some kind of performance leader without direct competition: whether a 2.4 GHz FSB800 Merom Santa Rosa notebook, but without any DDR2-800 SO-DIMM memory - that needs a fix, or a Penryn game PC with X38 chipset, or a quad-Xeon Caneland single board from Sun with up to 128GB or 256GB with eight GB FB-DIMMs RAM in a single fat blade.

Also, quite a few new technologies like Robson, now TurboMemory, new versions of PCI-E, S-ATA, CE-ATA, FSB-FPGA, WiMAX mobile and so on, were there, in many cases already in real working products, or, like Teraflop processor, in working demos with a funny kind of speedometer, at best.

Will this 'state of nirvana' for Intel last? Well, a lot depends on how AMD works out the Barcelona launch later this or next quarter. I wouldn't be entirely surprised if, due to Penryn's improved FP, FSB and caches, AMD attempts to push even the initial Barcelona beyond 2.5GHz in the current leaked model part number lists - however, it is still kinda new to the 65 nanometre process, so can it do it?

Yes, whatever the case, if you have four threads talking a lot between each other, AMD's CPU speed inter-core communication will beat anything FSB that Intel has. So will direct memory controller if you match it with fast DIMMs - but the memory difference will be lesser than before, as Intel has faster FSB and larger caches in Penryn generation.

But, on generic integer and a lot of FP stuff, will a 2.5 GHz single-die four core Barcelona beat a 3.33 GHz dual-die 4-core Penryn? Your guess is as good as mine, but, based on all the input from Intel, AMD and 'independent sources', I'd dare say Penryn will win most integer and a sizeable chunk of FP benchmarks, if they use SSE. The old X87 FP will, as usual, be no contest in favour of AMD.

For 'enthusiasts', the question is whether the initial round of new AMD's will, at least, overclock better than their predecessors, if they can't match Intel's 30%+ leeways seen on many of their Core 2's. I hope that, with water cooling, AMD can go at least 20% above the retail clocks in the new generation. Any setback here will be a problem for the high-end community AMD cherishes.

Finally, the problem of timing. The halls of IDF were abuzz with talk of further AMD delays - but again, it is a delay only for those who hoped to see Barcelona in May. I guess AMD better deliver it by this summer latest, to at least have some time window advantage over expected Penryn October availability. And that is assuming that Intel doesn't decide to launch the smaller volume "Extreme Edition" Penryns earlier which, based on the state of readiness seen at IDF, they could even do few months earlier.

If AMD stumbles with the new core, it's putting itself into a dangerous knock down position - please AMD, don't screw it and go down the drain, we need healthy industry competition for more things to write about. If AMD gets it done right, Intel might not have another IDF with such a clear, decisive top-performance lead, or pole position, for some time, at least until Nehalem arrives and delivers - hopefully before the end of next year.

Of course, Intel could cut that out by simply officially announcing one of those "extreme" Penryns on the show - anyone still remembers the highly confidential "Pentium 4 Extreme Edition" IDF launch some years ago, first unearthed The INQ?

So, this past week at IDF, Intel was showing off the pink of health for current and future products, with decent roadmap and seemingly committed future deliveries. At the same time, AMD had to announce another badly bleeding quarter, and put all the hopes in the new CPU core this year, as well as the promise of Fusion a year or so later.

Unless AMD stumbles with the new core, it's putting itself into a dangerous knock down position. Please AMD, don't screw it and go down the drain, we need healthy industry competition for more things to write about. Intel might not have another IDF with such a clear, decisive top-performance lead, or pole position, for some time, at least until Nehalem arrives and delivers - hopefully before the end of next year.

So, this Beijing IDF was unique in more ways than one - the first global IDF outside US soil, and one with Intel firmly holding the CPU speed crown after a long while - and, AMD failure-barring, before another while. µ