AMD Begins Shipping "Llano" Accelerated Processor Unit Discussion

Once again:

...and in case you didn't realize it the first time, that was sarcasm.


Several tech discussion sites have already come to a general acceptance that there's something bogus with the early stepping Bulldozer engineering sample used for those Chiphell benchmarks. There's a been at least one tip-off that the CPU tested was bugged and only running at idle clocks.

 

I don't believe in Fanboism, it is as childish as saying Superman is more powerful than Batman.
Just round out the candidates, take the performance numbers & minimium requirements divide it by cost and come to proper conclusion.

 

True. GPU power beyond what's in Sandy Bridge is absolutely useless to the general casual user who does not play games beyond Farmville, and who will never come upon a piece of software that uses CUDA or OpenCL or whatever API they come up with next. That's the majority of the computer buying public right there.

CPU performance helps everything from boot times to program startup times when you've eliminated the hdd bottleneck with a SSD, to browser rendering speeds to accelerating video that was encoded in a format that's not gpu-acceleration compatible.

I have seen first hand the absolutely dramatic general performance increase moving from a C2D at 1.7 Ghz to a Sandy Bridge at 2.6 Ghz, using the same exact SSD and same amount of memory.

 

The difference between a C2D @ 1.7Ghz and a SB processor is much greater than the difference we are actually talking about, though.

 

All indicators point to Llano being slower than a C2D at the same clock frequency. The top of the line Llano demoed by AMD had a 1.8 Ghz base clock with turbo up to 2.5Ghz. That's not much faster than my old C2D, since AMD requires more MHz to achieve the same performance.

 

What indicators are these? The engineering sample you linked to earlier points to Llano being about the same, clock-for-clock, at C2D.

A dual core at 1.7GHz is way different than a quad core at 1.8Ghz, if they perform the same clock-f0r-clock.

 

I never said there would be a huge gap, but it's still a few percent lower performance clock per clock.

Disregarding the fact that a quad core will never have the thermal envelope or battery life as a dual core, the benefit of going quad over dual core is minimal for the majority of users. If you can squeeze in a quad core llano in a laptop, you can also squeeze in a quad core SB, which will murder it in performance. AMD CPU = budget.

 

We don't know that yet. How about we wait until there are actual, provenm benchmarks before we come to that conclusion?

My N970/N950 quad-cores in my old dv6 ran cooler than the i5s in the same chassis. And AMD will offer dual-core Llano APUs, anyway. That is not the point, though.

I agree that AMD CPUs are budget options but the point is that 90% of users don't need anything faster than a C2D and would never notice the difference between one and an SB quad.

 

1. Demo chip was an quad core, thus it'll be faster than a C2D most definitely.
2. That was an ES, I wouldn't be surprised if production chips have higher clocked skus
3. I think AMD is aiming for 35 watt quad core skus. Also, battery life is more influenced by idle power if I'm correct. That's why you can get more battery life out of a 35 watt C2D compared to a 35 watt Athlon II. Of course, Llano is based off of the same designs as the Mobile Athlon II's were, though I would expect AMD to improve power consumption.

 

Only if it can Turbo. If not, we're back to the pre-Nehalem quad-core conundrum: quad-cores are almost certainly faster in heavily threaded applications, but most usage is not heavily threaded. Thus, they're great in benchmarks and not much else.

 

that, and it's now 32nm

 

35W Quad core will have to be clocked very low since even Intel can't get their quads that low, and they have better efficiency and a small GPU on die instead of a huge 400 SP unit. I don't foresee anything better than what Intel has in terms of power consumption from AMD.