AMD's New Direction For 2012: Heterogenous Computing, Trinity, and Hondo



+ Add a Comment


They quietly whacked Intel with FX. They might do it again someday.

Did anyone not see these curves (before)? This would be THE sandpit if everything else ran on gold. This is a good time to start paving it. Critics be damned.
They may not get it right the first few times around, but who does?



Ouch. Here is a completely opposite view on what Gordon says is "impressive" and "exciting" i.e. Intel wins, AMD is out. Gordon, were you at this meeting? Did you get the same idea that AMD is getting out of the high end CPU market?


Ghost XFX

Rory Read is focused. He's competent of the market's direction, and understands the demands that need to be met within AMD's limitations...

Question now will be: will AMD users embrace those limitations or not? AMD can't compete with Intel's "Old Money", it's simply a fact that can't be denied. It would take a grandiose CPU to even break even with Intel's lineup. The chances AMD bets the farm and house on doing such is nil. What they can do is find innovative ways to bring better performance to the average users and high end gamers alike at an affordable price. So I have to agree with their line of thought of bringing APU's to the forefront.

Naturally, Intel fans love to Nerd Bomb AMD every chance they get. They'll talk down anything AMD decides, no sooner it hits the net. Everything AMD does is ridiculous or factually insane. AMD will never be good enough in their minds. Personally, their "Mandark" attitude tends to lead to delirium. So why are AMD users hoping to compete with Intel directly?



Man, if I had a dollar for every "new" computing architecture that has been touted in the past 30 years...
Lets see
Unified architecture. check
parallel capacities of the GPU. check
Unified memory structure. check.

None of this is new, there are literally dozens of labs working on these exact ideas and probably already shipping product. AMD is trying to break into the very crowded "parallel supercomputer with massive number of CPU/GPU cores" business as they have already crowed about with their latest GPU core. Enough talk, more results please.

The rest is just hyping up their solid business on the low end laptop commodity market. They "shipped a lot!" And they have more low end designs for the future. The real question is can they make money with this business?



I think from a developer's perspective there were good improvements made when comparing to their previous generation discrete GPUs. Nvidia probably will come out with a similar competitive product line, but they are late on the draw. You would not see much a difference on a standard gaming system running up to two 1920x1080 monitors. Stress all its components under a newer software environment with three+ 2560x1600 screens, then their true nature would be seen.

Purchasing should be based on your needs. Common sense for just about anything really. Anyone with the dough can throw a ridiculous amount of money at something. Driving exotics without [legally] opening the throttle or even never driving them at all is extremely dumb in my opinion. Etc. Yea, a GPC/workstation is extremely unnecessary if all you ever do is basic tasks. A netbook does that perfectly. ;)



You bring up a point. Where is NVidia during all this AMD hoopla? There are articles every day about AMD this and AMD that, you think NVidia would want to throw some vaporware or leaked specs around to turn off some of this fan love.
I have this sneaking suspicion that both NVidia and Intel are setting up AMD to take a big fall when they get spanked by Ivy Bridge and Kepler.



I think AMD is missing the point here.

Give or take all of your programs that you have opened are I/O bound. They're just sitting around waiting for you to do something. And in fact, you probably want that, because if a program were CPU bound, it'd be crunching away and your power consumption jumps up. Especially on a portable system, you want the CPU to idle as much as possible.

Thus, since most CPUs are I/O bound, it doesn't matter how many cores you throw at it or what your computational prowess. At the end of the day, the CPU is parked, not doing anything at all.

I mean, this is great for supercomputers and mainframes that do a lot of simulations and numerical analysis, but for the average Joe checking his email and reading news websites, it's a complete waste of time.



though a good face saver for AMD, it is not looking pretty for the desktop segment at all.

Notebooks are not our thing, ball grid means no upgrade. and APU still means you need dedicated graphic for gaming.

with the Thuban die being phased out and the zambesi cores (even the quad and 6 variants) not = to the the thuban you are left with Intel usally a 60-100 markup for an unlocked processor. not to mention a higher price point for motherboards that have any real feature set.

As for Opteron- well the server side chips are not fairing well either sharing too much of the high tdp issues and poor threading issues.

will be interesting to see how this plays out over the next 3 years- will amd's vision hold true?



You don't "need" dedicated graphics for gaming. The A8 APU will play nearly every game out there. Now they will only crossfire with specific cards but there is still an upgrade path there.



"Most programs don't use more than one or two threads."

I love AMD's R&D team.


h e x e n

Does kind of suck AMD is taking a short cut out of direct competition with Intel. Oh, how I long for the days of Athlon again. I had big hopes for Bulldozer until they got bulldozed when it finally came out.

Do have to hand it to AMD for always doing something a bit different than their competitor.



Wow, who didn't see all that coming?
It kind of reminds me how floating point became part of the CPU die. The only thing I'm a little puzzled by is AMD's take on what a performance system would use, a traditional multi-core CPU instead of the heterogenous one. I can't see why. I can see by 2015 the heterogenous design becoming the standard for all x86 processors it's just so much more efficient then using a lot of full x86 cores for any thing that is highly parallel in nature. So with that in mind how many full x86 cores will be the optimal amount? I mean my i7 875K only maximizes all the cores when I stress test it with P95 or something. So I can't see more then four maybe eight being needed. After that using the GPU cores would make more sense to me.
But then again I could be wrong. Wouldn't be the first time.

P.S. I wonder if Intel is paying attention this time? AMD's on die memory controller gave them a massive jump over Intel IMHO. It took Intel using a mobile design to get them back in the game. Could be really interesting.



I had resisted getting a laptop forever because to game on them you had to spend huge. I finally caved and after researching bought a $300 e-350 lenovo machine. It is good enough to play games on and having it hooked up to my TV in the living room is great.

Unfortunately it has now made me want to spend more money on a HTPC built on the llano chip!

I think this is a good route for amd.



Wow, did the editor take today off or what?

Log in to MaximumPC directly or log in using Facebook

Forgot your username or password?
Click here for help.

Login with Facebook
Log in using Facebook to share comments and articles easily with your Facebook feed.