[citation][nom]bunz_of_steel[/nom]If AMD wants a boost first hire a CEO or Exec's worth their pay! They shouldn't be getting huge paychecks when company is not making money. Firing the working staff cuz CEO & Exec's can't make company profitable is just plain dumb. That's like credit card companies raising the bill on folks who can't afford the current bill... how's that gonna end?! Get rid of the golden parachute too! Stop putting monies in the pockets of the executive staff and put it in R&D, stop pulling punches and crunch out an APU with some SERIOUS muscle. The AMD APU A10-5800K only has 384 cores on dies with HD 7660D which is respectable enough. The HD 7750 has 512 Stream Processors why not mulch them together why water it down? Understanding that they don't want to undersell their own GPU's vs APU's but I'm thinking surely they can come up with some seriously stronger than the fail Dozer. That was just a something hashed together imho. PileDozer was so so improvement and prob wasted $$$ there for something folks barely noticed. AMD wanna make money - fire the entire executive staff and get some folks who are hungry and willing to do what it takes to put food on the table. Playing it safe is only going to get them to the end of the round, they need a contender that can put bacon on the table... sorry for the rant[/citation]
AMD put 384 VLIW4 cores in the A10-5800K because they probably didn't have GCN ready when Trinity was being designed, let alone ready with a 32nm reverse die-shrink and any other necessary alterations for the different process (I'm not sure if there's a different term for it, but either way, this seems to fit).
AMD's upcoming APU, Kaveri, will supposedly feature 512 GCN cores like the Radeon 7750. AMD isn't pulling punches this way, they're simply doing what they can. Also, Piledriver was a huge improvement over Bulldozer. Trinity has lower idle power consumption than even Intel's Ivy Bridge dual-core models in the same price range and that's with a great process node/technology disadvantage and a powerful on-die GPU. Even without any L3 cache, Trinity had about 15% higher performance per module at the same CPU frequency. That's higher performance at the same frequency despite not having the huge 8MiB L3 cache of the Bulldozer CPUs!
I do agree that AMD's executive staff (at least much of them) seems like the sorts who shouldn't be employed at AMD. Executives have been taking seemingly exorbitant pay checks considering the company has been going down on their watch and a lot of AMD's executive decisions over the last few years have been plain stupid IMO.
[citation][nom]mousseng[/nom]I certainly hope Piledriver's not their last major core - I think Steamroller might have what it takes to put them back into competition with Intel (whether or not that's their goal), in the enthusiast segment at least.I realize that's similar to what was said about Piledriver and Bulldozer, but expanding the modules' front end alone should show major performance boosts. I think the projections for Haswell's performance aren't terribly high, so if AMD can sustain their 10~15% increases, they'll be back in the game soon enough.[/citation]
Steamroller should improve performance by more than that, at least at the same frequency. So long as it gets into the market without any as of yet unforeseen flaws, it could be a huge improvement for AMD. Kaveri is on a 28nm node IIRC and with the improve Steamroller architecture, it could have a significant performance improvement with a significant power consumption improvement. AMD needs both if they want to catch Intel. Sure, Intel isn't making big performance leaps lately, but they have been hitting their power consumption further and further down and AMD has been getting flack for not doing the same.
Kaveri and Vishera's successor both seem to have the potential to put AMD in a good position, but looking at Vishera versus Trinity, I have doubts for Vishera's successor. Piledriver was a great improvement over Bulldozer, but Vishera seems to be greatly handicapped by the crap cache being used (more specifically, the L3 cache, but the L2 isn't really good either). This is something that AMD needs to improve greatly, perhaps even more than their front end.
Even worse, AMD already has had licensing for technologies that could greatly alleviate their cache issues for many years (Z-Ram and a little more recently, TRAM; if anyone wants a more detailed explanation, feel free to ask and I'll post one)... AMD has what they need, all that is left to do is use it! I wonder if their executives had anything to do with not using it... Why pay licensing fees for what isn't used anyway? Paying for licenses that they're not even using isn't the worst decision made in the last few years by a long shot IMO, so it wouldn't surprise me.