End of the Road For AMD's Geode Chip
An anonymous reader writes "AMD has no replacement planned for the aging Geode low-power chip, creating uncertainty for its use in products like future XO laptops made by One Laptop Per Child. There won't be a Geode successor and the company has no core microarchitecture planned to replace the chip, AMD executives said. The comments end speculation about the future of Geode, an integrated chip used in netbooks like OLPC's XO laptop, ultramobile PCs and devices like set-top boxes."
If you look at embedded devices or set-top boxes, you realise you don't really want Intel or AMD made CPUs. Look at most mobile devices, they all run OMAP-based devices (ARM), because of their energy efficiency and price. It also makes a helluva lot more sense to go with a SoC (System on Chip), as soon as power and size are even remotely factors in the decision making.
It's not because AMD drops out of the low-power energy manufacturing that the world is going to end, it just means they're focusing on things they're good at. I don't really ever remember AMDs being particularly energy-efficient, not nearly as what some VIA CPUs manage. I'm not talking about the Atom either, which is a whole different area.
Maybe I'm going completely bonkers, but if I were to build a low-power system, Intel and AMD would be last on my list, by quite a margin.
FlightSim is dead. Cyrix is dead.
FTA: "Geode's origins can be traced back to the mid-90s when Cyrix developed the MediaGX integrated chip for sub-US$1,000 mainstream PCs, according to McCarron. Cyrix merged with National Semiconductor in 1999 and developed the first Geode chips for embedded devices from MediaGX design. AMD ultimately bought the Geode business from National Semiconductor in 2003" See also wikipedia, of course.
Sad how so many big companies buy these 'niche' technologies then 'manage' them into a smoking hole in the ground...
is no one interested in VIA's most recent offering here in this market. From what I've read it's a much better solution than the Intel Atom.. Does anyone make a system with the VIA processors though? I haven't seen any.
It makes me think that this is really just misdirection to make Intel complacent. While not a great strategy most of the time, with the current economic situation the stock price isn't going to take much of a hit. Then, when they announce a new low power CPU, their stock should get a nice boost and Intel will need to redouble their R&D to catch up.
well, the geode certainly didn't have the performance of the Atom (maybe half at best)? It did have models that targeted lower power segments - I'm pretty sure there was an 0.5W model.
Self proclaimed typo king, and inventor of the bear destroying coffee table (patent not pending).
Yeah Mr Bagina, that's a real problem...
"I've got more toys than Teruhisa Kitahara."
Pretty decent specs for mini-notebooks and such.
512 MB RAM, 20 GB disk, 200 GB transfer, five datacenters. $19.95/month.
Proocessor->silicon->sand->minerals->rock-> Geode...get it ?
Essentia non sunt multiplicanda praeter necessitatem.
I could never give it any respect with that name.
Yeah Mr Bagina, that's a real problem...
Just ask his sister Mulva.
That would likely backfire -- witness Sun's on-again, off-again support for Solaris-on-Intel. If AMD were to suddenly introduce a new, killer low-power CPU, it would not get as much adoption by hardware vendors as if they did not discontinue the chip.
Michel
Fedora Project Contribut
AMD is NOT halting production of the Geode. They are not leaving the market (RTFM!). They have decided that it serves it's niche AS IS and will be kept AS IS. That's a very different statement. They're saying that it is a mature product (a rare thing in IT).
Currently, the Geode is good enough for many applications and would be a step up for others. The embedded world tends away from the shiny object model of upgrades. If it worked last year, it works this year, and it'll work next year. Changes in the product are considered undesirable.
AMD's statement doesn't even mean there won't be a die shrink or even a faster Geode in the future, just that they won't be updating it's architecture.
It's not a bad decision either. There is a significant niche for the Geode between the Atom (too hot, too power hungry) and things like the Dragon Ball and mips (not enough power).
Geode isn't in trouble until Intel comes out with an x86 that doesn't need a heatsink (or at least doesn't need a fan).
XScale was likely chopped because it was growing up to threaten the desktop market. XScale was based on ARM and if you haven't noticed, ARM chips are showing up in all kinds of things these days and the netbook sector is set to explode with ARM based devices this year. Dell has even put an ARM chip/system in some of their laptops to fast boot into Linux so the user can get on the web quick, get email quick, and even to run a DVD player. All with something like a 7 day battery life.
XScale was is a threat to Intel's profits and marketshare so it had to go. It had nothing to do with low performance. IMO
LoB
"Anyone who stands out in the middle of a road looks like roadkill to me." --Linus
Don't forget they have hardware cryptography as well - C3 or C7 was the first to include it. It uses the cycles as a way to randomize too. OpenBSD (for sure) takes advantage of this and uses it well. Encrypted tunnels, file system encryption, random number generation, etc. all put a LOT less strain on the CPU in comparison to other processors (especially embedded).
On the other hand, demand for chips like the Atom in netbooks is so high at the moment, AMD must be mad to be pulling out of this market.
Think less about units sold and think more about gross margins, and you'll see why a company with limited R&D resources like AMD may not be mad to let the netbook market go. Despite their popularity, the actual amount of money to be made selling netbook processors isn't that big.
Though I wouldn't expect them to be out of it forever. As the size of the netbook market grows, it will make more sense to try to take a chunk out of it. Right now, though, getting whatever 10-20% of the market they could get would not result in enough revenue to warrant large expenditure of R&D.
The enemies of Democracy are
I didn't know about this feature set before, and appreciate the information. A little research on Google led me to a page on padlock, a "driver for the cryptographic functions and RNG in VIA C3, C7 and Eden processors."
From the description: "The C3 and Eden processor series from VIA include hardware acceleration for AES. The C7 series includes hardware acceleration for AES, SHA1, SHA256 and RSA. All of the above processor series include a hardware random number generator.
The padlock driver registers itself to accelerate AES operations and if available HMAC/SHA1 and HMAC/SHA256 for crypto(4). It also registers itself to accelerate other HMAC algorithms, although there is no hardware acceleration for those algorithms. This is only needed, so padlock can work with fast_ipsec(4)."
512 MB RAM, 20 GB disk, 200 GB transfer, five datacenters. $19.95/month.
That's a big deal since the Geode doesn't need even a heatsink. That puts the Geode and Atom into separate use segments.
Yes, but this is precisely the reason why AMD should drop the Geode. They haven't improved the microarchitecture much since it was purchased from NatSemi, and NatSemi just kept bolting-on crap to the MediaGX chip they bought from Cyrix.
In other words, the Geode today is the same-old architecture from 1997 (with a few tweaks and node shrinks). The problem is, this old microarchitecture targets the same market as ARM, but can't beat ARM's power consumption. In order to cash-in on the netbook craze, AMD would need a beter microarchitecture.
Take a look at the new chips from Via and Intel: the Atom isn't a speed demon, but it kicks the crap out of a Geode without using much more power. The Nano can't match the power consumption of the Atom, but it fills the gap between Atom and beefy desktop cores. What both chips bring to the table is REAL Windows on an ultraportable platform (Geode can't do this).
Intel and Via figured it out: the embedded processor cores from the 1990s were not going to cut-it. With the lessons learned from the last decade, both designed new processors for the netbook market, using two different methodologies. Unfortunately, if AMD wants to compete, they need an entirely new architecture, and they simply can't afford to make one. They can't afford two architectures like Intel, and they've decided that the server/desktop design path is more important than the netbook path.
I personally think it's the right choice - netbooks are a growing market, but there's very little money to be made, as the margins are tiny. The server market will always be growing, and will always be worth more money.
Man is the animal that laughs.
And occasionally whores for Karma.
Processor + host chipset + GPU, on the other hand, is a bigger chunk of cash. If AMD don't have a competitive CPU, then that market will go to Intel at the low end and Intel + Nvidia at the high end.
That's really only one extra chip since the GPU will be integrated on the chipset, still with razor thin margins. That's really not any better, since what's important is the margin they can get for a given piece of silicon. Selling an extra hunk of silicon with a tiny margin makes no more sense than selling just the CPU.
And there's no such thing as a 'high end' netbook. That would just be a mainstream laptop, and AMD is competing there.
The enemies of Democracy are
Why this is a non-issue
AMD is still producing Geode processors to fill existing orders. And presumably they will keep making them as long as people buy them. The Geode processors changed substantially over time, so there is nothing to suggest that the next Geode would have been a drop-in replacement for the next XO, and it is unlikely that the next XO will use the same motherboard anyway.
Geodes that Rock
Today you can do better than the Geode in terms of price-performance. But it has certain characteristics that make it a winner, and you can buy a lot of Geode-based products right now which are wonderful. The x86 compatibility means that you can use standard operating systems and drivers, et cetera. Probably the best buys in a Geode processor come from PC Engines from whom you can buy a whole geode-based system with dual MiniPCI, quad USB2, a little RAM and a CF slot ideal for a household server/AP complete with Atheros-based WiFi for about $200. We're talking audio and video here, too, although it's not going to win any performance awards.
I recently (yesterday morning) acquired a DT Research WebDT 360 tablet. It's got a really nice 8.4" TFT @ 800x600 res, a USB2 port (plus one or two more inside) and MiniPCI WiFi as well as bluetooth and some buttons. Mine came with WinCE 5 but I intend to roll my own build of Familiar Linux to keep the install size low. The system's internal storage is a 512MB Disk-On-Chip, but it also has a hidden CF slot. This system has an AMD Geode LX 800 (500MHz; it is often misadvertised as 800MHz on eBay and even in official product literature) which has MMX and 3dNow! instructions, acceleration for AES encryption, and a TDP of 3.6W. Typical consumption is 1.8W! It is horribly hard to beat this with x86 compatibility. You can pick these up under $300 right now - original starting price is over $1200 for some reason. They do have a lovely alloy case...
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"