Black and White In The Age of Technocolor

Ace’s Hardware does an interesting video test using the Opteron. They run a video game benchmark that doesn’t use the video card.

The 1.8GHz Opteron falls about 10% short of a fully-gunned 3GHz PIV.

The test is interesting, but the comments are even more so:

The Opteron 1.8 GHz seems to be at the level of a 2.7-2.8 GHz P4 . . . .

Hmmmm, that’s pretty much what I’m thinking, at least for games.

. . . . but it wouldn’t surprise me if the Athlon 64 1.8 GHz performs better than the Opteron 1.8 GHz we tested today. [With the data I have right now, I think we can speculate that an Athlon 64 1.8 GHz will perform more or less like a 2.8 GHz – 3 GHz Pentium 4.] The reasoning is that the Athlon 64 will not have the ECC performance penalty and will most likely have access to DDR400. The lower latency of DDR400 compared to DDR333 ECC is much more important than the small bandwidth advantage the Opteron might have. Lower latency will help with each L2-cache miss, more bandwidth will only help if there are so many L2-cache misses that bandwidth becomes a bottleneck. In other words, lower latency (DDR400, Athlon 64) will always improve performance, more bandwidth (Opteron) will only improve performance, if you are running out of it. With a 1 MB L2-cache, that is less likely.

In other words, some minor improvements. I don’t have any problem with that, either.

So there seems to be little real difference between what I think the CPU can do and what they think the CPU can do.

The only real difference so far seems to be how the view gets expressed. Ace’s Hardware tends to express what they’re seeing a bit more positively than I am.

A difference, but really not a huge one, either.

What is the most interesting of all is what people are walking away with after reading the articles. Ace’s is viewed by AMDroids as the White (or maybe Green) Knight, while I’m the Black (or maybe Blue) Knight with Evil Inside.

What is really the difference between a darker and lighter shade of gray gets turned into black and white.

Hammer Is Going To Be A Weird Chip To Judge

The AMD philosophy behind Hammer appears to be, “There is more to a CPU than raw firepower.”

The Hammer generation introduces two new wild cards to the desktop CPU: seriously reduced memory latency and x86-64.

We’ve already seen enough to see that seriously reduced memory latency has much different effects on different programs. Sometimes it helps a lot, sometimes hardly at all.

The x86-64 wild card hasn’t even been explored yet, but it most certainly will have wildly varying effects, too.

This will make the kind of straight-up comparisons you could make back in the days of Athlons and PIIIs, or K6-2s and PIIs pretty much near impossible.

Intel is Getting Weirder, Too

Intel thinks “There is more to a CPU than raw firepower,” too, but they have a different vision of that. Hyperthreading is Intel’s major wild card to further complicate what is going to be a mess.

Hyperthreading will be good for two rather different reasons:

For now, it will help improve the quality of life in mundane computing, by letting the computer handle smaller multiple tasks more smoothly. This will rarely be reflected in a benchmark.

Later, it will probably boost workstation app scores as those applications get optimized for it.

Down the road, Intel will introduce SSE3 (aka Prescott New Instructions).

Sounds like Advantage Intel, but the AMDers will go, “BWAHAHAHA! Any worthy workstation guy will have two CPUs in his box. We may lose one-on-one, but what about two-on-two?”

And so on, and so on, and so on.

Taking Two Different Paths

What is really happening is that AMD and Intel are now on two separate roads. Intel got off the main road first with the PIV, now AMD is jumping off it with Hammer.

As they diverge more and more, general performance comparisons will become absolutely useless. Comparisons will start looking more like comparing PCs to Macs than what we’ve been accustomed to. This will greatly complicate one’s buying decisions, well, at least competent buying decisions.

I can easily see a situation a year from now in which you’ll find Hammer much better in Favorite Game #1, and the PIV much better in Favorite Game #2.

What do you do?

In short, there won’t be simple, ideal choices. There will be serious tradeoffs in any decision, there will be a lot of uncertainties and doubts.

I don’t know what the geeks who only see in black and white are going to do with orange and purple.

That’s not true; I do know.

Since their brains aren’t wired for anything else, they’ll just decide based on their own biases and prejudices, and throw out anything and everything from their universe that doesn’t fit.
They’ll ram everything into black and white, and make purple extra-black and orange extra-white. This will make for much argument, but little intelligent argument.

When the debating positions become more dubious, the debaters just get louder.

The CPUs are evolving beyond the capability of many users to compare them.

Can the users be rewired?

Email Ed

Be the first to comment

Leave a Reply