AMD APU-Llano,Trinity/Richland,Kaveri/Carrizo,BristoRidge

[del42sa]

fakt je ze pokud nehodlaj do toho apu pridat napr triple channel nebo pcie je jina patice uplne zbytecna a mohli klidne zustat u fm1 patice i pro trinity. Me davala vetsi smysl nova patice u bulldozeru nez u trinity.uvidime...

[THANATOS]

del42sa ja by som povedal, ze trinity bude do oboch socketov ale otazka je ci vlastne ten novy socket bude lebo ten mal byt spolocny aj pre PD a nakoniec PD pojde do AM3+.

[del42sa]

to neni uplne presne napsane. FM2 melo byt pro trinity a pro Komodo (10core PD).Ten ale byl zrusen a misto nej je PD based Vishera,ktera je take pouze 8core cpu do AM3+ Cili to, pro co bylo FM2 puvodne naplanovano vlastne zatim nebude Visherou nahrazeno...

[flanker]

1)Trinity má mnohem větší pamětové propustnosti a tak si může dovolit silnější grafické jádro a nebude limitováno. Musíme vědět, že APU kategorie je po procesorové stránce průměrná a má lepší spíš grafický výkon.

BTW:ano,je to možná citace AMD, ale kde je původní zdroj citace? Jak je citace stará? Každopádně B3 revize bude FX 81xx, prootže PIledriver je nová subgenerace a těm vždy přísluší nové značení revizí (C, D, E...).

Jinak VIshera používá stejné jádro,jako původně plánované Komodo (Piledriver)

[del42sa]

ale o samotnem jadru neni rec Flankere,bavime se o celem cpu a tam si dovolim rict,ze uz nejake vetsi rozdily byt muzou. Napr. ze dokud bude kazde cpu ktere prijde po BD postavene na AM3+ tak urcite nebude mit triple channel nebo integrovane PCIe v cpu... Zdroj citace byl tusim z letosni hotchips konference.

[flanker]

hotchips bylo myslím v březnu a potom někde počátkem června...
On FX ale má v sobě PCIe, nicméně díky čipsetu nevyužit ...

[THANATOS]

flanker Tomshardware bol zdroj, ktory tvrdil, ze IPC je 1/3 z narastu.

[del42sa]

hotchips bylo 17-19 srpna,cili dnes to jsou dva mesice, pokud se tim snazis naznacit ze by ta informace mohla uz byt zastarala. Ano BD uz ma int. PCIe, ale duvod jeho nepouziti nevidim pouze v chipsetu jako spise v te patici s nedostatecnym poctem pinu resp.jeho omezeni plynouci ze zpetne kompatibility s AM3/AM2+ Thanatos: ano TH tuto informaci vypustili do sveta ale pochazela z hotchips.

[THANATOS]

del42sa ked to bolo z prezentacie pred 2 mesiacmi, tak v tej dobe este JF hovoril o vyssom IPC ako k10, tak teraz ten narast myslia proti ktoremu BD tomuto sucasnemu alebo tomu, ktory mal mat vyssie IPC?

[del42sa]

boze Thanatosi, ty si fakt myslis ze JF pred 2 mesici neznal vykon BD? Ale no tak...

[THANATOS]

del42sa trochu sa bavim na jeho ucet.

[del42sa]

celkem dobře je to rozebráno tady http://semiaccurate.com/forums/showthre ... 655&page=7

výcuc:

Charlie mentioned a few times a respin before Piledriver, that would increase Int performance. If it does, and then the said 3-5% on top surely would give >10% more. Or is that respin equal to Piledriver ?

In the "Bulldozer doesn’t have just a single problem" article he mentioned a respin B3/C0 coming in Q1'12 but also mentions Piledriver is a major core revision. AMD has stated trinity will use Piledriver cores and is also scheduled for Q1'12. Piledriver was also supposed to be just minor changes.All this stuff is conflicting. Why would they bother with a respin in Q1'12 if Piledriver will be out in Q1'12? Is piledriver a "major core revision" or just minor update?

its all a bit, schedual wise, just because they've taped out a Piledriver(most likely without any L3?) with a gpu on die, which could have all the fixes in, doesn't mean they have a fixed Piledriver ready to go, validated, and ready to produce in 8 core. I assume Trinity taped out LONG before an octo core Piledriver(assuming that has also taped out already), and both after Bulldozer Mk1 taped out.Bulldozer can have bugs that Piledriver never had in the first place, there could be a cache problem that they'd already changed the design for Piledriver in.
A new revision of an existing core is far less likely to offer any massive changes that could be present in Piledriver, but could be a fix they can add into a revision of Bulldozer and add into a final respin of Piledriver so its ready for launch.It does seem quite odd timing wise that they've gotten Piledriver ready for Trinity, but not ready for desktop at the same time, especially if its significantly better.But a fix for Bulldozer well before Piledriver(none APU versions) hit is more than plausible, infact on a new process and assuming AMD can't work magic to bring Piledriver 8 cores forward half a year, I'd be surprised if there weren't a couple new steppings anyway

Which is why Trinity is unlikely to slip, as for the bigger Piledriver, we'll have to see what comes of that, we don't actually know what, if anything, is wrong with Bulldozer, and we don't know what, if anything has already been fixed and accounted for in Piledriver. Everything that sucks in Bulldozer which taped out over a year ago, could have had a year being fixed on Piledriver already.

Bulldozer taped out a year ago, but the chip design started two years ago. The core design might be three years ago. I don't really think that they can fix the issue in one year. It should be an enhancement as Westmere and Ivy Bridge more than a Nehalem and Core 2. Just take an example, JF said the engineers believed a IPC increase in BD, but after the A0 tape-out... you know what happened. Piledriver should have been designed before BD taped out.Steamroller is more likely to be the big fix, or the 2015 one in worst-case scenario

[ArgCZ]

http://www.diit.cz/clanek/kdyz-jde-o-vy ... tel/38972/

[del42sa]

ta spotřeba je pěkná

[DOC_ZENITH]

Ten test mi přijde úplně mimo. Mobilní I5 s GT540M nedá ani hodinu v office? No.... z praxe jsem zvyklej na 4+ hodin. Tam je fakt něco naprosto ale naprosto mimo. To mi skoro připomíná testy Nvidie vs 3DFX v 3Dmarkách.... svatápravda! Určitě to maximálně odráží ralitu a neni to vůbec zaplacené či zmanipulované.... brr....

[ArgCZ]

že by vypnutý Optimus?

[THANATOS]

Tento test je o nicom, ak chcete pozriet, ktory chip(notebook) ma aku priemernu spotrebu(idle,wifi,video,load) tak doporucujem notebookcheck.net.

[Hladis]

Ten test me rano docela zarazil a nejak mi tezce nestymuje s beznou vydrzi. Vyhazuje to tam naprosto nesmyslny vysledky vycucany z prstu. No to se zase no-X predved a bez overni to hodil do eteru.

[flanker]

ono často záležří i na nastavení...a pokud nefungují korektně ovladače, je to problém produktu, ne technologie. Ale bohužel no....co kus, to někdy originál.

[semmtex]

ten test je ale totálně blbě