Go to law with the devil and hold court in hell
|
|
|
Leaked AMD roadmap shows a 10-core 'Piledriver' chip will tip up in 2012
CHIP DESIGNER AMD has supposedly seen its 2012 desktop roadmap leaked on a Chinese website.
AMD has spent the first seven months of 2011 releasing its Fusion accelerated processor unit (APU) combined CPU and GPU chips that it hopes will eventually capture the low-end and mainstream markets. It hasn't yet come out with its Bulldozer 'FX-series' enthusiast and server parts, however it's rumoured that those will hit the shelves within the next six to eight weeks.
AMD's reported 2012 roadmap has its low-end, embedded Brazos chips supplanted by Deccan, a quad-core version of Brazos, while its mainstream Lynx platform will be taken over by Virgo. This is where the Trinity APU, which was held aloft at Computex, will appear. There is no change on the chipset side of things, but the chip reportedly will have four "Piledriver" cores.
At the high-end, AMD will introduce a 10-core Komodo chip that, if the documents are correct, will use an FM2 socket. There's an updated Hudson controller hub that has the usual accoutrements including eight SATA3 ports supporting RAID 0, 1, 5 and 10.
To top all that off, AMD is said to be planning to out its Radeon HD 7000 series graphics chips for both the mainstream and high-end markets. It might refresh its APU grahics cores, too.
Of course all this is from slides that have been leaked and it is hard to nail down the validity of the details. However Trinity has already been publicly shown off, albeit not as functional silicon, and has been slated for a 2012 launch. And as for the FM2 socket for the AMD's Komodo 10-core processors, stories about such a socket have been doing the rounds for some time too.
The real question is, what is Intel going to do, because Chipzilla has already said that it expects chips based on its 22nm Tri-gate '3D' process node to tip up in 2012. Just how smoothly Intel transitions to 22nm will likely shape AMD's 2012 as much as the sales of Intel's own products. µ
Tags: Amd
Share this:
Share
"The problem for AMD is who is going to manufacture these CPUs/APUs?
Glofo is a mess. 2 year late with 28nm. Has not delivered one single 32 nm chip.
"
Llano? And I don't know where you get this idea that they are in a mess, they are ahead of TSMC in terms of Process, considering that they have done 32nm for a while now.
"AMDs 7xxx series graphic cards was ready for manufacturing month ago but no one can manufacture them."
That's TSMC's field, not like the HD7000 is late, if anything it's a bit early.
"Intel has 2-4 year advantage over AMD in manufacturing technique. I remember when it only was month between them."
lol are you kidding me? Even before finding the actual numbers I knew this was completely made up. Here's a little quote that holds up from Wikipedia: "Intel started mass producing 45 nm chips in late 2007, and AMD started production of 45 nm chips in late 2008"
Intel's Tick Tock strategy is also quite aggressive, so yeah they are probably getting their processes done faster, considered that a $120+ Bn company may be able to pull of that faster than a $10bn company? That being said, having the smallest process isn't always a saving grace, I seem to remember Prescott, a processor that should never have been released, but yeah it was 90nm, and it sucked and had leaking problems because of it, and the architecture overall being incredibly bad. And yeah AMD is producing 32nm LLano APU's Right now, they are being sold, especially in laptops, but also for Desktops, Intel's 32nm if you haven't noticed isn't particularly old, it was launched in 2010, I guess we must be in 2014 now according to you then, huh?
"Selling of its factories to GloFo would solve this problem. Instead it is worse since AMD is just one of many customers."
I thought Glofo was a mess?
"AMD will be gone in 5 years.
Probably bought for its patents
" Right, never heard that before....
The problem for AMD is who is going to manufacture these CPUs/APUs?
Glofo is a mess. 2 year late with 28nm. Has not delivered one single 32 nm chip.
AMDs 7xxx series graphic cards was ready for manufacturing month ago but no one can manufacture them.
Intel has 2-4 year advantage over AMD in manufacturing technique. I remember when it only was month between them. Selling of its factories to GloFo would solve this problem. Instead it is worse since AMD is just one of many customers.
AMD will be gone in 5 years.
Probably bought for its patents.
Construction equipment to wrestling? Let me guess, the 20-core version in 2013 will be called "DoubleWide."
TSMC might actually beat Intel to Tri-Gate chips so AMD isnt out of the game.
As well since developers can expect GPU's as part of a CPU they can code around aspects that would be better handled by the GPU instead of the CPU giving AMD a speed boost over any GPU embedded in an Intel processor.
new applications for general use that use multi-core: Firefox is working hard to isolate processes, each on a separate thread or core. Imagine each new tab you open will use a different thread... not only speed is improved - when you crash on bad content, only that tab will crash.
IE and chrome working on this too.
It worries me a bit that info is leaking on next year's models when the architecture has not been introduced yet. Is it to prepare us for a disappointment with a sweet sause of 'it is going to become better'?
If you care about system longevity multicore is the way to go. There are already performance benefits with current games and applications (at least up to 4 cores) and should you not benefit directly you will at least not have a thread jump in when you are doing something cpu intensive due to lack of resources.
The future is multicore no matter how you slice it and while dual core it enough for many users now, that story will change in the future. Used to be mom and pop were happy with a 1.6ghz sempron too with 512meg ram.
While I am no coder I see it as feasible the industry will either produce more and more multi-thread capable software and probably inevitably an OS that presents the cores as a single resource to the software while the scheduling happens in the background. Could even be done in the hardware with a separate processor (or dedicated core) doing the scheduling. Somewhat akin to the hydra chip for gfx perhaps...
As I see it,you hit the nail on the head. Most if not all the multicore chips were for the server market,big money. The general populace are media hyped into buying these chips as an upgrade to their systems or to buy new computers, because the production line can't be stopped after you have the server market cornered, the chips would be to costly for anyone. To recover your money and continue with your cash flow you have to sell to the general public. MS,Intel,AMD, etc have to make sure your investment becomes redundant sometime in the future. Consumers are the lapdogs of the business industry other then items made strictly made for consumers. Pretty obvious their are more consumers then their are businesses. By selling to us, we keep the price down for them. Lets not get into tax write offs for the reasons they pay more,eg: I just bought two 3 year old servers that origialy priced at $8x2=16 thousand for the pair.I picked them up for $700.00.
Trinity has been shown as a running chip by Rick Bergman. INQ needs to check its story.
mutli-core is still applicable in home computing. Many cpu intensive tasks can benefit from multi-core, and with the improvements in turbo boost where it can overclock one or two cores if the workload is mostly single threaded I think you get the best of both worlds. Fast single-threaded and fast multi-threaded.
While multi-core CPUs greater than dual core aren't real efficient in poorly written code they can still process more threads than a dual core CPU can unless there is a rather large diff in clockspeed. Depending on the software that you run the quad core and greater can make a noticeable to a huge improvement in performance.
Sorry but AMD showed Trinity running in a laptop.
Most recommended builds still sit on the "2 cores enough" for the 90% (embellishment) user market. Is there a predictable software projection (no sarcasm, really wondering) that makes chasing these multi core 4,6,8,10 etc etc CPUs useful for anything outside the server market, single app productivity use (ie. photoshop, video editing, 3d modelling etc, most times a single application used all the time)or science (folding@home)?