RAMpage 70

Barcelona: Fuzzy Facts and Fiction

All right. AMD’s quad-core Barcelona is released and the reviews are coming in. I’ve invested more time to learning about Barcelona than I care to admit in the last month, and it’s clear to me that if you just skim this week’s headlines you stand a serious chance of misinterpreting several important things.

Taking a peek over the waters at today’s coverage on CMP’s ChannelWeb, we find “Review: AMD’s New Quad Core Barcelona.” Before we go on, I’m not here to spew Intel sunshine all over your face. I’m here to promote clarity so you can make better business decisions. Shall we?

1. CMP’s test system used a Tyan motherboard with a beta version BIOS. Which motherboard was this? Was it an old model Rev F board with an unfinished BIOS for Barcelona compatibility? If so, can we trust the results as being official enough for a business to trust? And if not, if this was an actual Barcelona-generation board, shouldn’t one have been picked with an official BIOS? Or is there a problem with finding finalized boards now that the chips are available? Hm.

2. If this was a Rev F-generation board being used, then the test system wasn’t providing the split plane power functionality that is one of Barcelona’s top draws, which in turn obscures the review’s statement that “the new metric for judging CPUs [is] performance per watt.” If it really is all about performance-per-watt, why was the only test run PassMark’s PerformanceTest, a tool that doesn’t take power consumption into account? You can look at performance scores, and you can look at draw from the wall, but unless you’re looking at both concurrently, you don’t have a meaningful conclusion.

3. “The system tested offered a CPU-Mark score of 4693, the highest score the Test Center has seen to date for a dual CPU server.” Well, that’s dandy. The last time I used DOOM III for a benchmark was on a mid-level NVIDIA 6000-series card, and I sure wouldn’t use those numbers against a GeForce 8800 GTX. Yet CMP makes this comment without having ever used that benchmark on a quad-core Xeon chip. Saying “the new chip is better than the old chip” is absurd. You have to test apples against apples. If you don’t have the time or editorial freedom to make fair comparisons, just say nothing.

4. That same performance results paragraph concludes with “odds are that an Intel Xeon E5345 (2.33Ghz) in a dual CPU configuration will out perform AMD's 2ghz quad core Opteron.” Nice backpedal. But this is also a fallacy. See, on paper, Barcelona looks like it should hammer Xeon on floating point operations, just as integer operations should heavily favor Intel. Overgeneralizing everything into a “this chip is better than that one” mentality is dumb for resellers, because you have to pick the best component for a client’s specific environment, not crown a blue ribbon champion at the CPU Benchmark Bake Off. Also, why are we talking about a 2.33 GHz Xeon when current SKUs already span up to 3.0 GHz. I mean, if we’re going to talk about performance, let’s have some performance, yes?

5. Several power saving enhancements to Barcelona are mentioned without dissing on Intel. Fair enough. AMD did a good job advancing on this front. Mind you, some of the enhancements echo features that Intel has been implementing for over a year, so it’s not like AMD has a monopoly on power savings.

6. Speaking of monopolies, funny how there seems to be a rise in buzz about AMD’s legal case(s) against Intel shortly before Barcelona’s launch. Perhaps it’s coincidence. But would someone please remind me: If AMD is selling into every major OEM—enough to cause availability shortages in the channel—and was talking for a long time about being on a path to 40% market share, how does Intel have a monopoly? Stop whining and start executing.

7. “When it comes down to it, Barcelona is a welcome addition to the CPU choices flooding the channel.” Flood implies lots and lots of product. Right, so who’s soaking wet yet? Anybody? Anybody? Let’s revisit the flood image in a month and see where the tide is at.

8. “[Barcelona’s] power sipping ways and backward compatibility with dual core Opterons will surely create opportunities for channel players in the server market.” This is marketing blather at its best. Said differently, this is what writers write when they’re regurgitating PowerPoint presentations—believe me, I know. Backward compatibility will not create channel opportunities because nobody goes around upgrading server CPUs except the HPC market, where channel presence is just shy of nil. Besides, as stated before, if you plant a Barcelona chip in a Rev F motherboard, you lose the split plane power and performance benefits, which are central to Barcelona’s messaging in the first place.

Now, after all that, don’t get me wrong. Barcelona is a damned fine chip in its own right, and it has a worthy, needed place in the channel. But get your facts straight and come to appropriate conclusions. If you want to learn more on the subject, check out RAM’s cover story on Barcelona, due to land soon in your mailbox.