Το x86 chip της Nvidia κόντρα με την Intel στα Tablets.

[grimpr]

Το x86 chip της Nvidia κόντρα με την Intel στα Tablets.

 

 

Nvidia Said to Take On Intel in Tablet Computer Chips.

 

Aug. 13 (Bloomberg)

 

-- Nvidia Corp., the maker of chips for computer graphics cards, is working on a microprocessor for tablet devices that would directly compete with Intel Corp. products, two people familiar with the matter said.

 

Nvidia has a team of engineers developing chips that could be used by computer makers instead of Intel, said the people, who declined to be identified because the project hasn’t been made public. Nvidia began working on a tablet chip after an earlier effort to create laptop processors didn’t pan out, the people said. Those chips weren’t cheap enough to compete with Intel’s Atom chips, according to the people.

 

The effort would help Nvidia tap the burgeoning market for tablets, spurred by the success of Apple Inc.’s iPad. It also may serve as a counterattack against Intel and Advanced Micro Devices Inc., the two biggest makers of computer processors, as they encroach more onto Nvidia’s home turf of graphics chips. Nvidia will need a central processing unit, which serves as the brain of a computer, to compete in the future, said Jim McGregor, an analyst at Scottsdale, Arizona-based In-Stat.

 

“They are between a rock and a hard place,” he said. “Nvidia doesn’t have that much time. Their business model is under pressure.”

 

Tom Beermann, an Intel spokesman, declined to comment.

 

The Nvidia project relies on a technology pioneered by Transmeta Corp. that uses software to replicate the way Intel chips work, the people said. Intel, whose processors run more than 80 percent of the world’s PCs, owns the so-called X86 instruction set -- technology that went into its first CPUs in the late-1970s. While Intel licenses the design to AMD, it hasn’t granted one to Santa Clara, California-based Nvidia.

 

Graphics Features

 

Intel and AMD are changing the way the processors work by adding functions previously performed by other chips, including more graphics features. That speeds up communication between components and makes the computer more efficient. It also locks out companies that can’t make those combined chips.

 

Nvidia’s latest project may not result in a product, according to the people familiar with the matter.

 

Trying to replicate the performance of Intel chips using software -- an approach Transmeta tried about a decade ago -- hasn’t worked before because it’s hard to deliver enough performance to run computer programs like Microsoft Corp.’s Windows fast enough, according to In-Stat’s McGregor. Intel’s X86 technology has taken over the PC and server industries, displacing companies such as Motorola Inc., whose chips once ran Apple computers.

 

Nvidia has concentrated on the market for graphics processing chips, becoming popular with computer gamers. Those customers add cards to their PCs to enable them to play the latest games.

 

http://www.businessweek.com/news/2010-08-13/nvidia-said-to-take-on-intel-in-tablet-computer-chips.html

[Ailuros]

Κάπου έχω χάσει τον συνειρμό σε αυτό το άρθρο.

 

1. Μιλάει ξεκάθαρα για την αγορά tablets/netbooks ειδάλλως δεν θα καθόριζε και Intel ATOM.

2. Ποιός και γιατί είπε ότι με το Tegra2 το ουσιαστικό πρόβλημα είναι ο επεξεργαστής; Το ARM CPU IP στο T2 κάθε άλλο παρά προβληματικό είναι, ειδικά αν σκεφτεί κανείς ότι πρόκειται για dual core ARM9. Μια ματιά στο floorplan του SoC δείχνει πόσο σχετικά λίγο χώρο καταλαμβάνουν οι επεξεργαστές.

3. Σε αντίθεση η GPU στο Tegra2 χρονισμένη μόνο στα 240MHz καταλαμβάνει κάτι λιγότερο από 10 τετραγωνικά χιλιοστά ενώ η GPU στο GMA600/Intel καταλαμβάνει κάτι λιγότερο από 9 τετραγωνικά χιλιοστά χρονισμένη στα 400MHz με σχεδόν διπλάσιες επιδόσεις.

 

Μήπως λέω μήπως να αναλογιζόταν κάνας χριστιανός ότι η Intel έχει τεράστιο προτέρημα όσων αφορά τα πρωτόκολλα παρασκευής; Και μιας και μιλάμε για την μελλοντική γενιά λεπτομέρειες για την Intel έχουν ήδη διαρρεύσει:

 

http://webcache.googleusercontent.com/search?q=cache:OtxcxHj54CQJ:software.intel.com/file/27519+penwell+site:intel.com&cd=3&hl=en&ct=clnk&gl=us&client=firefox-a

 

Medfield

(32nm)

Board Size: Reduced

SoC: 144 sq mm

Part count: reduced

Standby Power: Lower

Highly integrated SoC

Up to 4X graphics performance**

Further reduction in power

and improved battery life

** Compared to Menlow with Lincroft SKU at 400MHz and DDR2

Στα 32nm θα έχει λοιπόν το ίδιο μέγεθος πυρήνα όπως ένα Tegra2 σήμερα με μικρότερη κατανάλωση από το GMA600 σήμερα και μέχρι και τετραπλάσιες επιδόσεις γραφικών σε σχέση με το GMA600@400MHz.

 

Τι σκατά να γλυτώσει η προσομοίωση του x86 εδώ ακριβώς; Στο GMA600 είναι ακριβώς η ίδια GPU (SGX535) όπως iPhone3GS,iPhone4, iPad μόνο με διπλάσιο χρονισμό. To Medfield (GMA700?) στο μέλλον θα έχει τον ίδιο χρονισμό στα 400MHz μάλλον αλλά διπλοπύρινη GPU λύση με αρκετά δυνατότερο πυρήνα από το 535.

 

SGX535=

2 USC ALUs

2 TMUs

8 z/stencil units

 

SGX544 2MP =

8 USC ALUs (twice the arithmetic throughput per ALU)

4 TMUs

32 z/stencil units

 

Bon apetit....

 

΄Οσων αφορά την εμφάνιση διαφόρων λύσεων σε διάφορα προίοντα:

 

Intel GMA500 = >85 devices

http://en.wikipedia.org/wiki/PowerVR

Intel System Controller Hub US15/W/L—SGX535(Intel GMA500) + VXD370

 

 

Intel GMA600 = 4 devices

 

Intel Z6xx (Lincroft)—GMA 600 SGX535 + VXD + VXE + Atom-based CPU

 

Tegra1 & Tegra2

 

http://en.wikipedia.org/wiki/Tegra

 

Αν βγάλει κανείς τα smartphones που δεν έχουν να κάνουν με την εν λόγω σύγκριση, 9 συνολικά για 2 γενιές SoC. Μήπως να προσομοιώνανε και την GPU μήπως αλλάξει κάτι;

 

http://www.xbitlabs.com/news/mobile/display/20100816162024_Nvidia_Uses_Transmeta_Technology_to_Create_New_Chip_for_Tablets_Rumours.html

 

It is not completely clear whether Nvidia has computing architectures to emulate x86. ARM architecture-based processors generally offer lower performance compared to x86 chips from AMD or Intel, even if it is possible to emulate x86 on latest ARM designs, final performance should be dreadful. Emulating x86 on Nvidia's scalar GeForce graphics processors also does not seem to be a good idea. Still, the sources claims that the project is experimental.

It should be noted that besides code-morphing software Transmeta had very sophisticated technologies for power consumption reduction and their usage is completely logical not only for Nvidia Tegra, but also for other products from the company, including GeForce, Quadro, Tesla and so on.