This is going to turn the integrated GPU world on it's head. Kinda crazy to think I might be able to get a gaming laptop with longer than 2 hr batter life.
Printable View
Ontario != true gamers fusion
Llano performance better than Ontario or Sandy bridge so it would be the best fusion choice for gamers. I still regret that AMD did not add L3 on Llano game performance could have been better.
i think for best mix of performance for a simple and small pc would be to go with tri core and L3. not complaining about it being a quad, i just dont know how much it will be fully utilized. but if things are duel optimized u still have a core free. and i think if it was 3 cpu cores wide, they could have packed in more gpu performance, since thats easier to max out since its already built to for massive parallelism. but this could all be pointless if their low power states and turbo modes were incredible, which we will have to wait and see.
utilization of 4 cores will be irrelevant by the time llano launches. the only way for intel and amd to increase performance is to add more cores anyways. i think it's interesting that the die is 50% devoted for graphics and 50% for the cpu cores. if you look at a computer with an IGP it's a huge bottleneck compared to the cpu. this is quite a change in system architecture and it shows how important 3d gfx will be.
opencl comes to mind here .....
or ati stream ....
Try to think about how software will be in the next few years. 2010 will likely be the last year software devs DO NOT put an emphasis on multi-threaded design. AMD, Intel and Nvidia are all pushing parallelism on CPU cores and GPU cores. The Fusion APU is arriving at just the right time.
So... why not demo Llano in public? Why not demo BD?
I wonder... those do have 32nm HKMG GloFo process in common...
The demoed Ontario is 40nm TSMC process.
Just wondering...
Well, isn't it odd that it wasn't Llano that was demoed? Just trying to figure out why that choice was made. BTW, it's now "GloFo's 32nm process", not "AMD's". :) And as their first HKMG process, it would not be unreasonable to expect more difficulties there than on a mature 40nm process.
Another question... AVP2... isn't that from like 2002 or something, or was there an expansion pack later?
mentioning an igp and the term gamer without the word not does not compute... :p:
A igp will be a igp, eben when its renamed to a fance name like APU (well at least till it does anything more then accelerating graphics, and media content as its main purpose).
theres a new AVP (feb 2010), so na it isn't the same "old" avp. In real its the 3rd game title, aka AVP3 but the game name is only AVP... :p:
Llano taped out last November... And they demoed it at Computex (closed demo) and are sampling it to partners now.
AVP is the new DX11 game from Febr. 2010,badly optimized at that.
Ah yes... badly optimized clause again... i don't know any game that runs good with DX11 and tesselation enabled (metro2023, AVP) and even the Unigine engien makes any singlecard puke if you enable tesselation. In most games enabling tesselation usually costs between 20-40% performance.
Yes, I know they sampled Llano, they said so in the April CC. The question is, why not Llano for the public Computex demo? And still not a peep on BD status.
If not 32nm "in general" (which BD & Llano have in common) it could be this:
The GPU portion is already a known quantity on 40nm TSMC process. So for Ontario, they need to get a relatively simple in-order core up on 40nm, and voila, demoable product.
For Llano, the GPU needs to port from TSMC 40nm over to 32nm SOI HKMG. That could be tricky. Maybe the sampled Llanos have (so far) a pretty functional CPU side (that's just k10.5) with a not-so-functional GPU side.
That seems like the more difficult product: Llano, porting the GPU over to a new process while bringing up a known CPU design on a new process, vs Ontario, adding a simple Atom-like CPU to a running GPU on a known (GPU) process...
Thanks for the AVP info...
That's bobcat, wow. Sorry SKYMTL, but this is definitely going to cause a revolution, assuming of course that the bobcat fusion chip is as low power as we all think it is. What we're looking at here is the possibility of greater than Xbox360/PS3 performance in a netbook form factor. Consoles will even lose their advantage of standardized hardware optimization, since now PC developers will simply optimize their games to run perfectly on Fusion. On a related note, if everyone starts optimizing for the Fusion gfx chip those games are going to be more suited to ATI's discrete tech as well vs. an Nvidia architecture.
Terrace, you're reading waaay too much into it. Just because you'd have rather seen the other part doesn't mean there's some hidden conspiracy since they didn't show you what you wanted to see. They'll show us when they're ready to make details public.
If that was indeed Ontario, then that is quite impressive, barring the use of any kind of trickery. The rest of the presentation sucked though. 30 seconds out of an hour of PR BS isn't exactly the way to hold my interest.