One sickly sheep infects the flock - Isaac Watts
|
|
|
AMD dodges performance question for new chips
THINK TECHNOLOGY and don't worry about performance just yet, AMD told its pre-Hot Chips show teleconference about Bulldozer and Bobcat, its latest X86 core architecture iterations.
Much trialed for many months, the two upcoming AMD chip architectures are getting technology introductions at the Hot Chips conference, with Bobcat highlighted as the CPU core in the Fusion Accelerated Processing Unit (APU) package codenamed "Ontario". Ontario, and with it the Bobcat core, will be arriving in the fourth quarter.
With the Bulldozer core scheduled to appear in products next year and in particular in AMD's server package codenamed "Interlagos", AMD said that its 32nm scale design means it will deliver power efficiency and performance advantages. But don't ask what the latter will actually be, just reflect on the fact that Bulldozer chips have multiple dual core modules, with shared cache and memory and north bridge controllers. And those shared resources are key, as AMD said, "Bulldozer dynamically switches between shared and dedicated components to maximize performance per watt". Genius, just don't ask what that performance per Watt will be.
And Gregory Hoeppner, AMD corporate VP of design engineering, was equally shy about performance when pressed about Bobcat. "It's a trade off between power and performance. As we introduce product we will actually talk performance," he said.
Yet AMD has made much of its Fusion APU products and how that's the future of computing as far as it is concerned. Oddly Bobcat is only for Ontario for now, while Llano, the other APU design that has been delayed to next year, will use an older K8 core design for its CPU components. For the time being then AMD was making much of Bobcat's sub-Watt computing performance, but what that actually meant became a little vague as the teleconference wore on.
According to Hoeppner, Bobcat can operate at lower frequency and lower voltage to get down to sub-Watt range while maintaining a "fairly impressive" level of performance. However, he qualified that by saying that if you demand higher performance then frequency and voltage are increased, obviously.
So where is the advance? Bobcat achieves a performance at a "power envelope that up to now had not been available from AMD", Hoeppner said. But is that performance lower than what was previously available or higher? He didn't say.
Hoeppner's final comment, answering a question about chipset compatibility, might explain the reason for AMD's vagueness about performance. "I think there are existing chipsets that will support these products but like anything else there is a roadmap of... continuous investment." µ
Update Go here for the slides we journos got to see for the Bulldozer, Bobcat teleconference.
Share this:
Share
"Llano, the other APU design that has been delayed to next year, will use an older K8 core design for its CPU components."
At least get the basics right. Llano will use a modified K10.5 stars core not K8.
It's not at all surprising that the performance figures for a new approach to computing will be pushed off until release.
In the case of Intel's IA64 the reason was that it was a dog, and everyone knew it was a dog from the long delays in getting it even out the door (with a long list of "unexpected design features").
Here with AMD we might have a problem in the way the parts talk to each other, making the overall system a dog. Or the problem might be in finding an OS that actually takes advantage of the cool new features to show something other than incremental performance improvement.
And it's also possible that, unlike some other marketing-focused silicon peddlers, AMD is just be waiting to get real-world performance numbers from production silicon instead of extrapolating from simulations and pre-production runs.
Based on my past experience with AMD, it seems that the last alternative is far more likely than one of the other ones. Time will (soon?) tell...
I like my cpu to use as little as power as possible, because power costs money and heat kills cpu's.
Low power consumption when performance isn't needed is great, so long as there is enough performance when I want performance.
I think the new multitouch lappads, like the ipad, will be based upon this sort of cpu. Low power, long lasting cheap battery, smartphone type OS.
If we have 6 cores and each of those has 30% more performance than a single core of 4 years ago, then what we need is a way of the software using all the cores at once.
IMO it's the software that's lagging behind.
Car Touchscreen PCs. These are here in bleeding edge form, they will become more common. this type of cpu would be good in those.
AMD and Intel are going after ARM's market.
A Phenom II 555 X2 is a good processor. It has an amazing performance. Just Imagine this with ATI video integrated on the same die... Even if AMD performance remains the same, this is an absolute killer. It Would eplace my ATI 5450 and my Athlon X2 4200 for an all in one chip.
Did you really expect a performance comment?
These are AMD's crown jewels.................
but you need to consider this. We have reached saturation, performance comes at a price. Horsepower comes from burning gasoline not from wishes.
The ceiling just doesn't keep getting higher without increasing both power and gate densities.
So AMD won't disclose speeds and feeds of its Fusion chips? Well boys and girls, time to write another uniquely INQly waa-waa-waa article.
Must've taken all of 10 minutes to jot down - couldn't even be bothered to inform the 2 million (lol) Inq readers when the Hot Chips conference starts so we can go offsite and obtain all the info your article left out.
No worries, though - that other place done went and scooped you with a mighty meaty multipage article, with nary a waa-waa-waa.
Keep it up, and this longtime INQ fan will leave you just like all the INQ talent did.