AMD - Bulldozer Chitchat

[Terbinator]

With the IPC improvements Llano has over the Athlon II X4,I would suspect an A8-3850 will be similar in speed to a Phenom II X4 840. So the IGP seems to add around £30 to the total price.

Aren't they the same thing anyway?

[CAT-THE-FIFTH]

Maybe they are leaving room for some higher-end parts with turbo?

The 65W parts have Turbo AFAIK. Perhaps,they are concentrating on the mobile parts more ATM.

Aren't they the same thing anyway?

Yes,they are the same.

Edit!!

It does seem that the A8-3850 is only slightly faster than an Athlon II X4 635:

http://www.anandtech.com/show/4448/a...ance-preview/2

Cinebench does seem to run 7.5% faster though so in some benchmarks the A8-3850 will be closer to a Phenom II X4 840. I suspect the A8-3850 will have similar multi-threaded performance to a Core i3 2120 although it will have lower performance in lightly threaded applications.

[Pob255]

I think all this talk over the Llano has the wrong focus, little point in looking at it from an enthusiast view point as that's not really what these are aimed at.
From the point of view of an OEM general use system they make far more sense, as pretty UI's with fancy effects and multimedia aspects of computers are growing ever more demanding, these are the areas where intel's pore gpu performance make them weaker.
Yes intel still is king of king of raw performance cpu's but the majority of general computer use doesn't need or use that extra performance and trends are currently shifting toward more gpu demand in general computing.

[Jasp]

Sounds like it is still months away, and little to go on for performance estimates.

The BD version of Llano is said to be 50% faster, but whether that means clocks are 50% up or it has 6 cores no-one seems to be able to say.

Am still happy with my 955BE, and they are very cheap now!
But then a 2600K gets you performance today.

Cheers, it's me bday on the 26th and I fancy a treat in the way of a new pc lol.

[DanceswithUnix]

BD Llano will have a completely different architecture so clocks and core counts are irrelevant for comparison.

Exactly. So saying one is 50% faster than the other, I have to ask "doing what exactly and at what clock speeds" else this little teaser of info it pretty useless.

Though from the Anandtech benchmarks, I kind of think to go 50% faster you just need to add a third memory channel

[watercooled]

I agree, teaser info is always useless but gets people excited. A third memory channel would probably make a difference with GPU performance but I'm not sure about CPU, it would depend on the application I suppose.

[Blackmage]

Exactly. So saying one is 50% faster than the other, I have to ask "doing what exactly and at what clock speeds" else this little teaser of info it pretty useless.

Though from the Anandtech benchmarks, I kind of think to go 50% faster you just need to add a third memory channel

The 990Fx AM3+ boards seem to be dual channel. Cat's thread has some boards, Could a better memory controller be the solution?

[DanceswithUnix]

The 990Fx AM3+ boards seem to be dual channel. Cat's thread has some boards, Could a better memory controller be the solution?

Better memory controller for what though, being a GPU or being a CPU?

They mention prefetcher and buffer size improvements improvements on the Llano CPU so they have already tried to make up for worse memory latency. If the GPU is hitting the memory hard (which it seems to be) then that will hurt latency for the CPU.

Faster DDR3 speed grades should help a lot though.

[scaryjim]

... Could a better memory controller be the solution?

One of the things they've done with both llano & (I assume) BD is tune the memory controller to support up to 1866MHz, so that's definitely something they're working on.

[badass]

The BD version of Llano is said to be 50% faster, but whether that means clocks are 50% up or it has 6 cores no-one seems to be able to say.

2 Modules. i.e. 4 cores so a 50% single threaded performance improvement over Llano.

[badass]

Exactly. So saying one is 50% faster than the other, I have to ask "doing what exactly and at what clock speeds"

The doing what is relevant IMO. At what clockspeeds is not. At what power consumption is.
Someone might step in and say clock speed is related to power consumption but it's certainly not a perfect relationship, so I don't care about it.

[watercooled]

I agree, clockspeed is so heavily relied upon as a measure of performance despite the fact it tells you nothing or very little about it. With plenty of examples, think of P4:C2D, Atom:Athlon Neo. It's only useful where all the other parameters are exactly the same.

[Golden Dragoon]

It's just like the good old days when Apple were harping on about the mhz myth

[DanceswithUnix]

The doing what is relevant IMO. At what clockspeeds is not. At what power consumption is.
Someone might step in and say clock speed is related to power consumption but it's certainly not a perfect relationship, so I don't care about it.

That metric would work, I only mentioned clock speed as sometimes people try to game the results of these things in their favour by choosing an unfair match.

I used to program 8-bit 0.9MHz 6809 CPUs when friends had a slower 4MHz Z80, nothing changes

[DanceswithUnix]

So AMD think that by 2014 only 2% of their CPUs will be conventional CPU only devices. Are we entering the era of the "Discrete CPU"?

http://www.theinquirer.net/inquirer/...s-intel-nvidia

[watercooled]

In regard the the title of that page, I can't see Intel even considering buying Intel - they are just as big rivals as AMD.