Talk:Geode (processor)

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

MMX[edit]

Quote: "The processors lack many of the SIMD instructions such as MMX and SSE, which are present in more popular processors."
... according to the AMD website, seemingly all Geodes nowadays support at least MMX. --Maikel 15:56, 8 Apr 2005 (UTC)

Gaming device?[edit]

I'm curious... How would one of the Geodes fair in a handheld gaming device? Has anyone attempted this yet? Or is the Arm still a better choice? - Zephyrxero 10:01, 2 January 2006 (UTC)[reply]

A few issues[edit]

Note: I am not a hardware engineer; these are from the view of a programmer and a little use of valgrind's cachegrind. Cachegrind options: (--I1=16384,4,32 --D1=16384,4,32 --L2=64,2,32) and just use the number of L2 cache references as the cache miss count (Valgrind doesn't know about L2-cacheless chips).

The Geode NX chews WAY too much power, it's a repackaged Athlon. Power-critical applications will probably need a GX or LX; but neither of these have a full hardware FPU (extended prec is microcode) and they are lower performance.

The Geode GX uses 0.9, 1.0, or 1.1 watt of power, although it has limited speed. It also lacks extended precision floating point functionality; single precision floats execute in 1 cycle, anything more gets picked up by CPU microcode and pushed through the pipeline multiple times. It has 3DNow, 3DNow-A, and MMX; it has the SSE prefetch instructions PREFETCHNTA, PREFETCHT0, PREFETCHT1, and PREFETCHT2, but no more of SSE. Performance is likely to improve by some degree with a compiler that actually uses the prefetch; gcc does not, although there is a pending patch.

In large, CPU-bound workloads the Geode GX will probably have problems with cache. Geode GX has 32KiB of 4-way set associative L1 cache and no L2. Cache misses cost around 25 cycles each, so the nominal performance hit for CPU-bound loads can be 28% or even 33% (cachegrind on Python and Rhythmbox). Comparably, a desktop CPU with 512K L2 (i.e. Barton) has a 0.2% cache miss rate (i.e. miss L1, miss L2); and L2 only takes 1-2 cycles longer to access than L1. IO bound workloads (accessing disk, NAND flash, or the graphics hardware) will not really care very much.

An off-die L2 cache option so embedded systems requiring the cache could drop in a chip between 16KiB and 512KiB may have helped certain cases; on-die L2 would just be expensive and, where actually useful, would likely need to be a lot bigger than technically feasible, quickly becoming prohibitively expensive and highly wasteful on power. Still, off-die cache costs money, power, and added circuitry, and is slower than on-die so you can't likely get the 1-2 cycle difference; but that's rolled down to the IHV implementing the Geode-based board.



I think someone should add something to the article explaining how the Geode LX800 is now in the Raon Vega UMPC; this is one of the first consumer-level devices with one of these processors. http://www.dynamism.com/vega/main.shtml 67.175.75.186 06:44, 16 November 2006 (UTC)[reply]

Amusingly the OLPC Project now uses the Geode LX. I heard rumors that the 64K I1 + 64K D1 + 128K L2 brought performance up 2-4 times. Not exactly consumer level, but pretty hot anyway.

Geode design team is gone[edit]

For those interested in Geode (real geodes - GX/SCx200/LX not NX) the design team was cut or redirected by AMD in 2006. The is no real capability left to build another Geode based on the GX/LX. The parts of the Geode team, in Longmont, CO, who AMD wanted to keep were told that they'd be moving to the Fort Collins, CO site in late '06. Several silicon designers, I think about 35 people, and 20 systems/software engineers were retained. The rest took the severance package and left. Geode never found a place to make good money, so perhaps this is for the best. Ex-geode-er 21:22, 28 February 2007 (UTC)[reply]

LX900 incorrect frequency[edit]

According to the AMD Technical spec page, the LX900 processor frequency is 600 MHz, not 667 MHz. http://www.amd.com/us-en/ConnectivitySolutions/ProductInformation/0,,50_2330_9863_13022%5E13058,00.html

Geode NX2001[edit]

User:Denniss keeps on reverting edits that remove the NX2001 from the list of Geode NX processors. I asked him why, because AMD doesn't provide any reference to the processor. The only references found on the Internet are some forums in Polish and German language and a handful of sellers of this CPU, there are no official sources available. All sellers say it is an Athlon XP 2200+ Thoroughbred really.

Though OPN (part number) ANXA2001FKC3D is suggested by Denniss, it doesn't appear in AMD's Geode datasheets and a Quantispeed rating ending in "1" is never used in any of AMD's numbering schemes. The OPN doesn't make sense (see also AMD Geode datasheet, AMD Athlon XP datasheet and AMD Athlon OPN reference):

ANX = AMD Athlon-based Geode Solution
A = Geode: unknown TDP; Athlon XP: Desktop Processor
2001 = Model Number
F = OPGA Package (Geode)
K = 1.65V Vcore
C = System Bus Speed 266 FSB (133 MHz)
3 = Size of L2 Cache: 256 kb
D = Max. Die Temperature: unknown (Geode)

A 1.65V Vcore is way too high for mobile/low power use. Typically, a Thoroughbred core @1.65V would consume 60W or more power. Unless a Vcore/multiplier switching is at play, of course, which I doubt (requires special motherboards/BIOSes; the "A" in TDP suggests a desktop processor)

Specifications of the "AMD Geode NX 2001" as advertised: Clock Speed: 1.8 GHz, 1.65 Volts core operating voltage. There are no references found for low power operation, nor any figures for power consumption.

I don't think AMD is approving the sales at all, I think this is just a marketing trick to get rid of surplus stock Athlon XP processors, meant for the embedded CPU market. Unless someone can provide solid proof of power figures (25W/40W, I have only read one reference at a German forum, the sales ads do not give power figures), I would think it is just a relabeled desktop processor, or a mobile "desktop replacement" processor.

Cheers, Brinkie 14:19, 4 October 2007 (UTC)[reply]

I actually had this processor. For all means and purposes it behaved like a 2200+. It actually looked as a relabelled CPU, I didn't check this as I was afraid of doing damage, but it looked as if there was an original label (of Athlon XP 2200+ presumably) covered by Geode NX2001 quasi-nonsense label.--Azarien (talk) 20:59, 18 March 2009 (UTC)[reply]

Missing discussion of SC1200/SC2200/SC3200, 1400, and SC1100[edit]

The SCxx00 Geodes were developed by National Tel Aviv (NSTA) based on IP from Longmont and other sources. The 1400 was first and was a disaster and wasn't productized. SCx200 followed and was a full x86 PC on a chip. Very similar to the GXm+CS5540+sio. SC1100 was a headless version. There was another variant too.

This entry is also missing information about the other chipset devices for the GX1 family. These were the CS5510, CS5520, CS5530.

CS5535 was designed by National in Santa Clara. CS5535 worked with GX2, as mentioned. It also works with LX.

CS5536 was designed by AMD in Dresden (DDC). CS5536 worked with LX, as mentioned. It also works with GX2.

During the last couple years of Longmont, the team developed two or three products with Dresden which were ultimately killed by management in the late stages of development. The lack of direction by management is what many believe led to the demise of the team.

Also, wrt OLPC, someone should reference the opensource efforts supported by the Longmont AMD staff. See http://ols.108.redhat.com/2007/Reprints/crouse-Reprint.pdf There are probably others to cite.

Ex-geode-er-2 01:22, 6 June 2008 (UTC)[reply]

GX2 RTL was given to a Chinese University by AMD in approximately 2004[edit]

The RTL for the GX2 processor and peripherals was given to a university in Beijing sometime around 2004 by AMD. The verification environment, scripts, etc. were also transferred. All was done on the up and up, so there weren't any legal or government issues doing this.

A couple AMD employees were sent to China to work with the university to help them build their own version.

Sometime in 2006 the university got the processor to function in a FPGA.

Just some interesting info on what companies do to gain market access for other products.

Ex-geode-er-3 16:53, 3 September 2008 (UTC)[reply]

CS5510, CS5520, CS5530, CS5540, CS5535, CS5536,[edit]

The CS5510, CS5520, CS5530, CS5530A all worked with the MediaGX/GXm/GXlv/GX1 architecture.

The GX2 worked with the CS5530A (in a special operational mode), the CS5535, and the CS5536.

The Castle (LX) worked with the CS5536 and the CS5535 devices.

Ex-geode-er-3 17:07, 3 September 2008 (UTC)[reply]

An image on this page may be deleted[edit]

This is an automated message regarding an image used on this page. The image File:AMD Geode Processor Logo.svg, found on Geode (processor), has been nominated for deletion because it does not meet Wikipedia image policy. Please see the image description page for more details. If this message was sent in error (that is, the image is not up for deletion, or was left on the wrong talk page), please contact this bot's operator. STBotI (talk) 14:20, 28 December 2008 (UTC)[reply]

External links modified[edit]

Hello fellow Wikipedians,

I have just modified one external link on Geode (processor). Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:

When you have finished reviewing my changes, please set the checked parameter below to true or failed to let others know (documentation at {{Sourcecheck}}).

This message was posted before February 2018. After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than regular verification using the archive tool instructions below. Editors have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the RfC before doing mass systematic removals. This message is updated dynamically through the template {{source check}} (last update: 18 January 2022).

  • If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
  • If you found an error with any archives or the URLs themselves, you can fix them with this tool.

Cheers.—cyberbot IITalk to my owner:Online 12:53, 24 April 2016 (UTC)[reply]

External links modified[edit]

Hello fellow Wikipedians,

I have just modified 4 external links on Geode (processor). Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:

When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.

This message was posted before February 2018. After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than regular verification using the archive tool instructions below. Editors have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the RfC before doing mass systematic removals. This message is updated dynamically through the template {{source check}} (last update: 18 January 2022).

  • If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
  • If you found an error with any archives or the URLs themselves, you can fix them with this tool.

Cheers.—InternetArchiveBot (Report bug) 23:50, 12 October 2017 (UTC)[reply]