Puff

引用:

原帖由 dom 於 2013-11-12 10:45 發表


Haswell 1C VS 2M2C .....SR so suck ..

This is piledriver. But anyway, it doesn't change so much I guess.

Puff

OT: Does anybody have a clue of what speed the DDR4 majority will be at its launch?
2133? 2400? 2667?

Puff

引用:

原帖由 cheungmanhoi 於 2013-11-12 17:40 發表

kaveri is pd not sr?

still have chance

The result was benchmarked on Trinity. KV has Steamroller cores.

Puff

引用:

原帖由 qcmadness 於 2013-11-12 17:39 發表

expect 2133

I wonder if anyone will launch 3200 at launch.
btw, max 31 fps on A10-5800K, so I guess AMD is focusing on power perf (e.g. lower TDP for top bin) and HSA for Kaveri, as the memory BW is definitely an obstacle of further performance growth.

[ 本帖最後由 Puff 於 2013-11-12 19:56 編輯 ]

Puff

引用:

原帖由 dom 於 2013-11-12 20:09 發表


I think they realize OEM don't take much of their product, Server is not selling well,
so they have to go back to fight in dimishing Desktop market first ( Market Share dropping )

Generally speaking, you would launch you mobile series before desktop series ONLY IF you can't manage to have enough inventory for both.
OEMs are always aligned to big events or specific seasonal time frames, like CES, Computex, Easter, BTS and Xmas. PIBs for desktop have no such limitation. But anyway, AMD's slippery roadmap since Llano may cause some impact on their design wins, and given that the low adoption rate of AMD APUs, your statement could be true either.

They definitely need sustainable execution on this product line if they really want to regain the land. For instance, accelerate their 20nm product as early as they can. Q3 is a SUPPPPER breakthrough in AMD's history of the 10s, Q4 is excellent, Q1 is nice, Q2 is dangerous, and Q3 sucks.


[ 本帖最後由 Puff 於 2013-11-12 20:34 編輯 ]

Puff

引用:

原帖由 cheungmanhoi 於 2013-11-13 10:50 發表
amd.....

睇來佢真係無plan b
真係見步行步咁
好彩d engine人爆到個勁idea就有得番身者
唔好彩就

There is no plan B in the semiconductor world, esp for such small scaled company. If they opt on Bulldozer at the beginning of the multi-year architecture planning (3-4 yrs to design a first gen product, and make evolution on that for the next N yrs of product to save cost), they will stick with it till the end of the roadmap. Now for anything wrong with Bulldozer, given that you have simulations & real chips in 2010 and also several delays and internal issues during 10-13, add 5-6 years and you will have a clue of when you will see something brand new (if they still finance it, and I guess yes as they want to reclaim land in the server world).



[ 本帖最後由 Puff 於 2013-11-13 21:56 編輯 ]

Puff

引用:

原帖由 dom 於 2013-11-13 15:30 發表


AMD 2015 都無新野既話 真係印證 AMD 2015 RIP .....

They always have new stuff. the problem is only the pace of 20nm/FF transition, esp for the later which can close the gap with Intel.

Puff

引用:

原帖由 dom 於 2013-12-6 12:23 發表
from the article , Kaveri can offer "Up to Shanghai/Deneb Level of single thread performance"

...................takes 3 years to catch up 5 years ago performance

Bye Kaveri
I am going for Haswell ...

Go ahead. Enjoy.


[ 本帖最後由 Puff 於 2013-12-6 15:52 編輯 ]

Puff

引用:

原帖由 dom 於 2013-12-6 16:21 發表


I don't bet , FX-8300 is just extend the stay with AMD platform longer
Est. end of 2014 will switch to Intel based platform if NOTHING "on-par" with Ivy Bridge class CPU performance and 6Thread (3M ...

無架啦，轉吧啦。就算要出，我估都唔會出 6C 架啦。
前景不明朗呀。買定 16 年花生期貨。



[ 本帖最後由 Puff 於 2013-12-6 16:46 編輯 ]

Puff

引用:

原帖由 dom 於 2013-12-6 16:49 發表
   
HSA WILL NOT WORK and FAIL  just like Faildozer

- AMD isn't good at Linux platform ( neither do Intel nor nVIDIA , lol )
- Mantle is just another BS thing like they use "FX" with Faildozer

They are working extremely hard on Linux drivers in terms of both HSA and graphics (not kidding, not speculation, this is real), because their new growth products NEED good Linux support. Mantle will not fail either, but eventually it will become a cross-vendor standard or die as a bridging solution (due to another emerging cross-vendor standard). Nonetheless, it is a catalyst to the PC gaming market. For HSA, I can't comment on it further. From a toolchain standpoint, it sounds not bad at all (far better situation than the OCL/CUDA ages), but it still depends on how HSA Foundation and its members will push it further.

[ 本帖最後由 Puff 於 2013-12-6 19:13 編輯 ]

Puff

引用:

原帖由 dom 於 2013-12-6 16:48 發表 What do you expect?

If they are quitting, you shouldn't have seen the release of FS1B and FM2+.

Puff

引用:

原帖由 dom 於 2013-12-6 16:48 發表
When Intel side , they catch up with APU iGPU performance in less than 4 yrs time period

It is pretty expected. Green Arrow is still struggling on 32/28nm while Intel has moved to 22nm since the second year. Now they even leverage their ability to implement leading-edge tech ahead of all competitors.



[ 本帖最後由 Puff 於 2013-12-6 21:24 編輯 ]

Puff

引用:

原帖由 dom 於 2013-12-7 00:20 發表


I didn't say AMD is quitting
the market is quitting on them

講緊 OEM/MB

Puff

引用:

原帖由 dom 於 2013-12-9 12:11 發表
Seriously , still can't even reach Sandy Bridge level .....
good luck AMD

never la. bulldozer sucks. for instance, once upon a time AMD was the king of number crunching with K10 until SNB arrived. Now after SNB, bulldozer just f**ked everything up and drive amd's HP server dept to be a sinking ship. so unless bulldozer dies completely or someone f**k bulldozer inside out, good night AMD. crystal ball masters place bet on Jim Keller to get xv done right, but I would say he is too late for xv. you may sincerely hope amd once planned a huge rocket science update on xv in their epic failed bulldozer roadmap.



[ 本帖最後由 Puff 於 2013-12-9 21:43 編輯 ]

Puff

引用:

原帖由 dom 於 2013-12-13 12:26 發表


XV ?
What I see is AMD RIP in Server market

eXcaVator.

When BD releases: PD will fix everything
When PD releases: SR will fix everything
When SR releases: XV will fix everything



[ 本帖最後由 Puff 於 2013-12-13 21:35 編輯 ]