HW novinky - CES #4 (nejen mobilní GPU AMD a Nvidia) (diskuse)

[Conscript89]

> nicméně AMD bude nadále pokrčovat v rozjetých projektech

Mensi chybka v poslednim odstavci :)

[Peter Fodrek]

Derick (somacky Dirk) Meyer odviedol pre AMD a pretym pre Digial vynikajucu pracu

Alpha - Dick Sites and Rich Witek

* Dick Sites and Dirk Meyer, Alpha architecture video, April 1992

# AMD Athlon (K7), 1999 - Dirk Meyer (Dir. Engr.), Fred Weber, ...

# AMD Opteron (K8), 2003 - Jim Keller and then Fred Weber * Jim Keller left and the initial K8 design was canceled * Fred Weber led the project to revise the K7 into a 64-bit core

http://www.cs.clemson.edu/~mark/architects.html

Podla mna su to skor politicke dovody, ked Derickovo dieta K7 definitivne konci a nastupuje Bobcat a nastupi Bulldozer Dirk totiz ostal prezidentom

Mimochodom o tam ARM-e -

Dirk bol No2 u AMD

AMD Acquisition of Alchemy Semiconductor Expands Market Focus AMD Will Offer Solutions for Personal Connectivity Devices

SUNNYVALE, CA -- 2/6/2002

The founders of Alchemy Semiconductor include industry veterans Rich Witek and Greg Hoeppner. Witek was the co-architect of the Alpha processor and the architect of the StrongARM processor while at Digital Equipment Corporation. Hoeppner was the implementation lead for the Alpha processor and the site and engineering director for the Austin, Texas, design center for Digital Equipment Corporation. Both Witek and Hoeppner will join AMD, reporting directly to Edwards. http://www.amd.com/us/press-releases/Pages/Press_Release_14491.aspx

a Witek je ina kapacita PRISM (Parallel Reduced Instruction Set Machine), 1989 - Dave Cutler, Dileep Bhandarkar, Rich Witek, Dave Orbits, and Wayne Cardoza HR-32, started in 1984 - Rich Witek and Dave Dobberpuhl

* microPRISM - Rich Witek (lead microarchitect)

# Alpha - Dick Sites and Rich Witek

StrongARM

* StrongARM 110, 1996 - Rich Witek (lead), Greg Hoepnner, Ray Stephany, Jim Montanaro, + o Sribalan Santhanam, StrongArm 110: A 160MHz 32b 0.5W CMOS ARM Processor (Hot Chips 8 slides) * StrongARM 1100, 1997 - Rich Witek (lead microarchitect), Ray Stephany (implementation)

Witek bol maximalnym sefom vyvoja u AMD

SUNNYVALE, CA -- 9/23/2002 -- AMD (NYSE: AMD) recently named Rich Witek corporate fellow, the highest position in the company’s Member of Technical Staff (MTS) program. He is the first AMD employee to hold this position. http://www.amd.com/us/press-releases/Pages/Press_Release_43200.aspx

REDMOND, WASH. -- 12/21/2006 -- AMD (NYSE: AMD) today announced the formation of a new Advanced Architecture and Technology Lab led by Rich Witek, AMD corporate fellow and chief architectural officer. The new lab will focus on technology and platform development beyond the five-year time horizon, further extending AMD’s strong tradition of advanced silicon planning. http://www.amd.com/us/press-releases/Pages/Press_Release_115133.aspx

Takze realne Wittek navrhol Bobcat a asi aj Bulldozer Meyerovi mozno vycitat,ze Hoepnera presnul do dcerskej spolocnosti AMD, do spolocnosti Razza Microelectronics (teraz RMI) a Witeka pustil do Microsoftu

Wittekov nastupca Rogers totalne pohorel AMD Promotes Phil Rogers to Corporate Fellow — Graphics expert to continue driving software development and ecosystem support for new graphics and computing models —

SUNNYVALE, Calif. -- 11/26/2007 -- AMD (NYSE: AMD) announced the appointment of Phil Rogers to AMD Corporate Fellow. In this role, Rogers will continue developing advanced architectures and extend AMD’s software capabilitie http://www.amd.com/us/press-releases/Pages/Press_Release_122586.aspx

zjavne sa asi Meyer nechcel Rogersa vzdat

AMD Promotes Two Technical Leaders to Corporate Fellow Critical software and power management contributions recognized with appointments to company’s highest technical level

Sunnyvale, CA —1/24/2011

AMD (NYSE: AMD) announced the appointments of Sam Naffziger and Leendert van Doorn to AMD Corporate Fellow for their leadership in power management technology and software development, respectively. Corporate Fellow is the highest level of technical recognition at AMD, with Naffziger and van Doorn becoming only the third and fourth employees to hold the position. http://www.amd.com/us/press-releases/Pages/tech-leads-promoted-2011jan23.aspx

van Doom ma vysledky hned a zaraz Posted by Michael Larabel on February 03, 2011 If you use the open-source ATI Linux graphics driver, there's a major stable update available. At long last, xf86-video-ati 6.14.0 has been released. This open-source X.Org driver brings official support for the Radeon HD 5000 series, Radeon HD 6000 series, and AMD Fusion Ontario hardware. This release also has a plethora of bug-fixes and flips on the KMS page-flipping support.

For the Radeon HD 5000/6000 and Fusion hardware there's also open-source 3D support available that lives within the Mesa stack. Ideally you should be using the Gallium3D "R600g" driver for the best and fastest open-source ATI driver.

If you're looking to utilize a Radeon HD 5000 series GPU with the xf86-video-ati 6.14.0 driver you will need the Linux 2.6.36 kernel or later. To use the Radeon HD 6000 or Fusion hardware you will need to be using the Linux 2.6.38 kernel, which won't be released as stable for a number of weeks so until then you're left using Git or one of the weekly release candidates http://www.phoronix.com/scan.php?page=news_item&px=OTA3Nw

Takze podla mna padol Meye na chybach svojej pravej ruky Rogersa

[ET]

Zdar, rad bych se zeptal zdejsich HW guru jak je to s VRAM u APU ? Je v APU alespon trochu int. VRAMky [pochybuju] popr nejaka jina vykonostni featura ? Rek bych ze to musi bejt vykonova brzda GPU casti pouzivat "pomalou" systemovou RAM kdyz porovnam rychlosti RAM vs VRAM [tipoval bych urcite tak jednotky, ne-li desitky % ] - nebo je to jinak ?

thx

PS: VRAMem myslim dedikovanou RAMku pro GPU - pokud placam blbosti tak me opravte, videl jsem nekolik prefixu RAMky pro GPU [VRAM, GRAM...]

[Peter Fodrek]

On ju nepotrebuje.. Je tam nieco ako L2 cache pre GPU. AMD to tusim nazyva graphic buffer

Navyse AMD CPU je velmi citlive na latencie RAM. Specialne serie RAM pre AMD maju znizene latencie. je ale fakt,ze VRAM ma stale nizsiu latenciu ako RAM.

Vykonnostna feature je v principe to, ze AMD ma OpenCL exekutivu pre CPU. T.j. program napisany v OpenCL, ide na AMD CPU aj AMD GPU, podla toho, co je v systeme a, co je volne. Zacalo to tymto, kde to nie je tak jasne

http://www.amd.com/us/press-releases/Pages/amd-delivers-and-submits-2009aug04.aspx Prakticky to teda znamena, ze treba pisat len opneCL aplikacie, aby tuto vlastnost vyuzili APU na zrychlenie vypoctu

As of 2010, the fastest PC processors six-core has a theoretical peak performance of 107.55 GFLOPS (Intel Core i7 980 XE) in double precision calculations. Both NVIDIA and AMD's consumer gaming GPUs may reach higher FLOPS. For example, AMD’s HemlockXT 5970 reaches 928 GFLOPS in double precision calculations with two GPUs on board while NVIDIA GTX480 reaches 672 GFLOPS[14] with one GPU on board. https://secure.wikimedia.org/wikipedia/en/wiki/FLOPS

Tym padom vypocty mozu bezat na APU aj 10x rychlejsie ako na Core i7, ak pouzijete openCL

A o tom je APU primarne.

[ET]

Ok, dik za info - a ja myslel ze dedikovany GPU ceka kremikovy nebe :] uvidime casem jak se to vyvine

[Peter Fodrek]

ten pad dedikovanych GPU je pricip Intel Sandy Bridge

[kik]

Cože? APU je rychlejší než ATI AMD Radeon 5870?

[Peter Fodrek]

to som nenapisal, ja som napisal, ze Radeon HD je vo Fusion architektureprimarne ako dalsia FPU nie ako GPU