AMD: 'Bobcat' smaller, faster than Intel's Atom
AMD says that its upcoming "Bobcat" core for netbooks and notebooks is smaller than a single-core Intel Atom chip – and faster. According to the company, this low-power architecture could eventually follow Atom into the server market, but at the moment, that territory is still reserved for the "Bulldozer." The annual Hot Chips …
"The question is this," says Fruehe. "Is there a need for a more discrete, less-threaded chip for servers?"
For Future Secure Command and Guaranteed Control Servers are they absolutely vital.
it may be, but surely in that (netbook) market space it's all about the WATTS, which were not mentioned at all in the entire article.
I don't know whether that is because AMD aren't talking about them or the journalist failed to ask. If it was the former, then the journalist probably should have thought to say so really.
They didn't mention specific wattage, no, but they did mention better power consumption. They didn't mention specific performance figures for the Bobcat either.
And no, it's not all about the watts. It's all about performance per watt. A netbook with a 24-hour battery life is no good if you can't do anything with it during those 24 hours.
...whilst giving less and less mips for the same MHz. So the processor clocks are faster and faster with each new model, but getting-the-job-done performance is steady, or even drops.
So I'd like to see a Windows 7 tablet with AMD inside.
I'm allowed to want what I want, surely?
What and double the price of the hardware?
Just give me a pocked sized netbook with this inside and I'll be happy running most Linux on it.
@"So the processor clocks are faster and faster with each new model"
Huh? ... I don't see any 4Ghz chips and i've been waiting years for them! ... 3Ghz was a mad rush to get to it, but 4Ghz has been a painfully slow wait. (I'm talking standard clock speeds, not cryogenically cooled amazing PC mods).
What I would like is a faster clock speed so all my old apps can benefit (Many of them were written in a time when many people assumed processor speeds were going to keep increasing). I'm not talking big increases in speed, I would be happy to upgrade to just a 4Ghz chip. :)
Although 5Ghz would be very nice. :) ... with say 16 cores please, if you are listening AMD ;)
> "I don't see any 4Ghz chips and i've been waiting years for them!"
How does a 5.0 GHz POWER6 sound to you? Or an 8 core 4 GHz POWER7? If you don't see these chips you're really taking care not to follow tech news.
I was talking about x86 PC processors. That was the context of the discussion after all, about AMD chips. You are basically using a straw man argument to then make yourself feel good about putting me down with your condescension about "not to follow tech news". I do follow the news.
Clock speed tells you next to nothing about the performance of a system. Clock for clock, Pentium-M was faster than P4, which is why they used that design for the Core II series. Single threaded apps can benefit from wide execution (including SIMD), better branch prediction, larger caches, better instruction scheduling, etc. Then there's always the question of ISA, number of physical and architectural registers, compiler performance, etc. And, all that said, your single threaded apps may also benefit from an additional core if it lets them avoid context switches.
The public obsession with MHz as a figure of merit is a sad product of Intel's marketing dept. AMD started trying to break this notion during the race to 1GHz, and Intel followed a few years later, but it has unfortunately stuck. Performance is complicated, and simplistic metrics don't help.
Performance is complicated...
...but a 5Ghz core 2 duo would wreck an i7 at a great many workloads.I think global compute resources have diverged enough that there is space for three different types of x86 CPUs.
Small low power many thread doohickies, good for lots of little programs or web servers. Would make good netbooks, laptops, and maybe would be useful for application virtualisation.
Slightly more beefy midrange things with a moderate (4-8) threads for folks doing some heavy lifting. Rendering, A/V editing or VDI.
Screamingly fast per-thread low core count (2 cores?) designed for workloads that need the straight-line speed and don’t multi-thread well.
http://arstechnica.com/old/content/2002/10/hyperthreading.ars/1
The above is an article well worth reading on the topic. I think the complicated nature of performance means that the public’s “obsession” with Mhz isn’t necessarily wrong. Architecture and attempts to parallelise can only get you so far. There are still plenty of compute tasks for which what you really need is a bloody big bit cruncher going at as high a clock rate as they can make ‘em.
Put more simply: CISC !> RISC for all workloads.
Get it into the server market, but do it right. Give it an ECC-capable memory controller! Intel's approach seems to be that if you want low wattage and/or cheap, you can't care about data integrity. The only Intel CPUs that support ECC are their Xeons (expemsive, and even the low-wattage variants are energy guzzlers compared to an Atom).
Which is why the last two LInux terabyte NAS boxes I built had AMD CPUs.
As well, make some damned motherboards. As I have lamented before, I would love to sell AMD kit to my customers rather than Intel. But Intel has it all: CPU, chipset, motherboard, backplane, and RAID on-board* for low- to mid-market servers. If AMD has anything comparable for small- to medium-sized business, then I will be more than happy to push it.
*Matrix (ahem... Rapid Storage Technology) has some issues, but it is still a great system around which to build.
Paris, go vertical!
"AMD is open to the idea, but don't hold your breath. Unless ARM, Tilera, and Intel Atom chips start getting some server sales."
Could be already in motion.
http://www.semiaccurate.com/2010/08/23/facebook-first-jump-arm-servers/
This is a good move for AMD. From what I understand, Intel doesn't want to improve the performance of Atom, or allow it in anything over a certain size, etc., for fear of cannibalizing sales of the beefier chips. AMD can make it as fast as it can, and let you put it in whatever you want - definite advantages for the AMD chips.
Also, some mentioned there was no info on power consumption, well I quote you this, from the last page of the article:
"The Bobcat design is also meant to push the power envelope down, and can hit below one watt per core of power consumption, according to Hoepper."
Sure, it's about what a future chip may be able to do, and not really as specific as one might hope, but it's something. I'm sure we'll find out more before someone forces us to buy one...
How is this bulldozer pseudo-double-core different from a single core that's 8-way superscalar?
A typical superscalar processor is still a single core and executes a single thread (assuming SMT is not used), but may execute instructions out of order in order to fit as many as it can into the execution units if instruction dependencies allow it. The alternative would be a barrel processor like the UltraSPARC T1/2 cores.
Well, it's all great. But they've been mentioning Bobcat for possibly more then a year an a half. I also have the feeling that Intel is artificially restricting Atom capabilities - for fear of eating into their more expensive products. However, Intel has been selling Atoms now for, what, 2 years or more? It's not only about spec, it's about market share as well. Even if their processor will be better, showing up 3-4 years late is not going to help.
That's not just a feeling about the Atom's capabilities, and it's not an artificially restriction - it is most definitely designed to fit inside a pretty tiny box. Why else would they put screen size restrictions on it - or intentionally limit the I/O bus so you can't really get the full oomph out of that ION2 GPU in your nettop/netbook. Many of Intel's design decisions - like the ECC memory controllers mentioned earlier - are designed to keep their different processor markets tiered and distinct.
To their credit, low-power CPUs were an unexplored/unserved market when Intel created the Atom and I am quite happy with the two Atom-based machines that I own... but at the same time I do have a problem with the Intel's [Low Power] = [Cheap Crippled Shit] equation.
Don't get me wrong, I understand that to achieve significant power efficiency like the 8 hours I get with my netbook I will not be able to have the same performance as my watt-guzzling multi-core full-size laptop that might make it 1:15... but if AMD can deliver a significant improvement in the performance-to-efficiency equation vs. Intel, I can't imagine Netbook, Laptop, and non-ARM "Pad" Manufacturers would stick with Intel just because of market share.
Put it this way, at comparable price points if Mfr. A's AMD-based Netbook can run for 10 hours, effectively run more than one application at a time, and run HD video (flash or otherwise) - and Mfr. B's Intel Atom-based Netbook can run for 10 hours, runs sluggish once you run more than 2 or 3 apps at a time, and skips/stutters all to hell when you try to watch HD video... Mfr. A is going to sell more Netbooks. The real question here IMO is if AMD can deliver.
Sorry but I should have mentioned that VIA had been in the low power CPU market well before the Atom - Intel didn't create this market as I implied, they just took it from rare niche to mainstream with the Atom.
/Steve of course - who else gets more credit for "creating" markets that already exist?
AMD seems never learn the lesson well. The company should never release any details nor mention any new products before the products are ready to ship to catch competitor off guard and take times to play catch up. AMD doing is shooting itself in the foot when they ready to ship the announced products, rival will be too.
Sign up, sign up for Blocks and Files, The Reg's weekly storage newsletter
