A couple of months ago, I wondered where VIA had gone. There's an excellent comment piece on UK enthusiast site bit-tech that shows where VIA might well be heading - no place good is the likely scenario to my mind.
Worth the read.
Extremetech are reporting that the "Kentsfield" quad-core will be released 1 November with a shipping date of around 15 November.
As I've mentioned previously, the Kentsfield is more like a dual dual-core rather than a true quad-core implementation. The real quad-core goodness probably won't launch until first or second quarter next year.
AMD are expected to launch their 4x4 quadish-core solution around the same time, according to the same article.
The blogosphere has been abuzz with a prototype chip revealed by Intel at the Intel Developer's Forum that has 80 cores. 80. Apparently within 5 years as well.
Sony is having enough trouble getting people to write games that will take full advantage of the 7 extra cores that will be available on the variant of the Cell chip they're using in the Playstation 3. Even more to the point, the vast majority of the software I use doesn't even support dual core setups. 80 cores? *Pfwah* to that I say.
If compiler technology and coding techniques has advanced sufficiently to take advantage of even half that capacity, I will be very surprised (and happy - I mean, 80 cores! That would rock).
Hot on the heels of my discovery of the OpenSparc initiative of Sun, I've come late to yet another open hardware party - AMD's Torrenza. Torrenza is a "socket compatibility" technology, where AMD has opened up the design for their Opteron sockets (the CPU technology they use for servers and other high-end uses) so that other "silicon designers" can design and build in a preexisting x86 environment with supported motherboards and chipsets that will already be in production, with resulting savings in economy of scale.
(x86, for anyone that isn't entirely familiar, is the basic processor architecture for just about all CPUs currently produced for consumers today. The x86 title stems from the early chips which would be most familiar to people as the old 286, 386 and 486 processors produced by Intel. The 86 nomenclature was eventually dropped in favour of names that could be trademarked (eg Pentium, Athlon) but as the basic architecture has remained the same, the x86 designation has remained to describe the family of processor technologies. While there have been a range of additions to the architecture such as AMDs x86-64 (among others) which changed the architecture from 32bit to 64bit, the basics are still the same, even with new multi-core processors)
While Torrenza was first announced back in June 2006, it's in the news again this week as AMD announced their plans for Torrenza with a range of partners including Sun, Cray, IBM and Dell. I'm still undecided as to whether this is a forward-thinking, strategic move from AMD to make sure that their motherboards become first choice through providing greater choice of processors, or if its a move of desperation as they battle to maintain market share in the server market.
Not being tapped into the scene in any shape or form, I guess I'm just going to have to wait and see.
Quad-core has a name, and that name is Barcelona.
Barcelona
AMD spoke to [Hard]Ocp about their upcoming multi-core solutions, in particular their latest quad-core technology, the Barcelona (can tell how much fun I'm having typing Barcelona? You should hear how I'm saying it - dodgy fake Spanish accents are a lot of fun - try saying Antonio Banderas in one and you will know what I mean).
Barcelona is due for release "mid-2007" and will be a "native" quad-core solution with four cores in the one chip as compared to Intel's Kentsfield "Quadro" which looks like it will be just two Core2 Duo dual processors crammed together on the one die, soaking up Watts and pumping out heat.
Who will use four cores?
Even with all these multi-cored beasties powering around, there's precious little software that has been designed to take advantage of the latest and greatest in chip design. Most software that can take advantage of a multi-core setup is used for things such as image manipulation, video rendering or audio tracking.
AMD have said that they are looking for games designed for consoles to lead the way, as the Xbox 360 has a triple-cored PowerPC (each core is dual-threaded so it can do 6 threads at once) and the PS3 will have the Cell chip with 7 of the 8 supporting processors activated (although one will be used for the PS3 operation system) and a controlling PowerPC core. As gaming houses get used to developing titles that support multi-threaded systems, they must be hoping that this skill set and knowledge will leak across to the PC market.
There will probably be benefits in decoding high-def TV on the fly and other media applications, and there will be potential for some parallel computing. It will be good when virtualisation software also begins to take advantage of the support being provided in the hardware which could allow for different operating systems to be run on individual cores. Aw yeah.
AMD and ATI
One of the key product developments that will come out of the AMD and ATI merger will be lower power, highly optimised CPU/GPU combinations on a single die for notebooks and lower-range desktops, possibly inside the next 18 months. There is a strong market for this sort of product in the corporate area, where desktop computers have no need for a separate graphics card, as well as the ability to pack more into a notebook/laptop chassis while taking up less space and at the very least, consuming no more power or creating no more heat than present technology and configurations.
There's also the growing markets for powerful but lower-speed x86 processors in set-top boxes, mobile communication devices (phones, Blackberries etc) and ultra mobile PCs (UMPC). ATI already has a large presence in the graphics processing side of these markets. AMD will now be in a position to develop units to compete with likes of ARM or Freescale, particularly as software for these smaller and more mobile devices becomes increasingly complex. I had an idea that this is the way they AMD and ATI might be heading, it's nice to see that confirmed by people who are actually in the know.
I remember two or three years ago, Via had its C3 processors all over the place. It seemed like every third HTPC had a low wattage, fanless and "fast enough" C3 as the heart of an admittedly adequate (as compared to killer) setup. The C3 wasn't fast, but it with low power consumption and the ability to easily run without active cooling, it seemed like the ideal combination for budget HTPC goodness.
Now I know Via are still around (they have their fingers in many pies after all) and I know that the C3 is an old chip and they have better ones like the C7 and C7M, but why is it that it seems like retail HTPC setups just don't use Via anymore?
I guess Via is only going to face further competition as Intel's new Core architectures use a lot less power than they used to, while being a lot more powerful than previous Intel chips, and no doubt the AMD multi-core solutions will have similar advances.
Maybe I'm just way off base and Via is doing just fine in the embedded space, thank you very much. If anyone has an answer as to why Via just doesn't seem to be in the race anymore, I'd love to hear it.
Edit 19/9: There's now an interesting dicussion on this NYT article at Slashdot - my normal practice is to browse comments at a threshold of 4, but whatever does it for you...
According to the New York Times, a research collaboration between Intel and the University of California, Santa Barbara, has borne fruit, allowing them to transfer data between chips over laser instead of wires. This advance in technology will allow a greater density of chips in system designs, with a faster rate of transfer between the chips.
Aspects of this technology are already in general use, particularly when transmitting data over long distances using fibre optic cable. There's probably any number of mind-boggling applications (particularly in parallel processing) but I can yet find any further information from Intel or UCSB.
It has reminded me that I need to check into Intel's Technology and Research page more often. They have got some really funky stuff going on over there.
I don't expect to see this enter the consumer desktop market any time this decade, but that suits me fine. I have enough problems with core envy. I don't need to add laser envy to the mix as well.
It would seem that AMD are still on track with their CPU production, which means that their long awaited "quad-core" chips won't be arriving until sometime in the 2nd Quarter of 2007, at least according to this Inquirer article (it is the Inquirer so have your industrial-sized salt shaker handy before ingestion)
Is this too little too late for AMD? With Intel pushing out their Kentsfield quad-core options now (probably by Q4/06, based on the Tom's Hardware preview I blogged about recently) and likely to have cooler and less power hungry versions by Q2/07, AMD look like they're going to have a tough battle on their hands if they want to keep taking market share from Intel.
With any luck these duelling cores will push existing dual-core solutions even further down in price, which can only increase the chances I will have of moving away from my sad little Sempron 2500+ into some cpu goodness.
And in the end, isn't that what technological advance is all about?
Edit: Forgot to include the story link. Idiot. All fixed now.
It's becoming increasingly obvious to me that I have no idea what is going on the tech world around me.
I have only just discovered, with the launch of the Simply RISC S1 Core, that not only is there such a concept as "open hardware", but hardware designed from open sources is actually being shipped.
It is pretty cool that Sun has made the designs for their SPARC chip freely available with the OpenSPARC project. The SPARC is a complicated looking chip that appears like it has a lot in common (although very different to) the STI Cell chip (currently most well known as the powerhouse for the PS3).
What is cooler is that there is a community building up around the idea of developing an open System-On-a-Chip (SOC) infrastructure at OpenCores.org.
I'm happy to admit that understanding the potential and comprehending the possibilities are beyond my grasp, but the idea that there are freely and readily available resources out there that people can use to design and maybe even build their own CPUs or even multi-cored, integrated SOCs. This isn't really a big concept for the industry giants, but it would have to be exciting for the budding electrical engineer who has a great idea about a new system design but is not quite sure where to start. It's not just new systems that excites me. My mind boggles just trying to think about where the more creative tinkerers can get to when designing DIY and hobbyist electronics with embedded processors. That's a whole other world away from DIY printed circuit boards (which is hard core and high tech enough to scare me).
I get goose bumps just thinking about it. Now I think I understand a little of what Woz felt in those early days.
The current financial rumour mill is speculating that Freescale is on the verge of being bought out in what is potentially the largest ever in the US tech sector, with a value of around US$16 billion.
Freescale, a spin off of electronics giant Motorola, used to make the G4 PowerPC chips used by Apple before they switched to the IBM manufactured G5 chips (now more commonly refered to as the PowerPC 970 since Apple's switch to Intel's Core architecture).
Freescale are a big player in the world of embedded chips (which is a whole other world away from anything I understand) which are used in everything from mobile phones to car engines to aerospace and military applications. Freescale's newest piece of funky tech is Magnetoresistive RAM (or MRAM) which is smaller, more efficient and faster than most current RAM technologies. Because it's non-volatile it has the potential to greatly increase storage and retrieval times, which is particularly useful for things like digital cameras and for booting up devices (especially cool if you can chuck it into a laptop for "instant on").
It will be interesting (to me, anyway) to see who eventually buys out Freescale and what they plan to do with the tech at their disposal.
Tom's Hardware have a preview (in their usual "why do it in 5 pages when we can do it in 15, with ads!" style) of the incoming Intel "quad-core" Core2 CPU solution.
By the looks of it, the "Core2 Quadro" isn't so much a pure quad-core set up as a 2x2 where they have crammed two Core2 Duos into the one little package. While I have no doubt that 4 cores on one die isn't too far away, it leaves me to ponder as to why Intel are rushing out this temporary solution. Are there better reasons than just to beat AMD to the punch? Probably not.
I bet this baby runs hot, too. New chip would probably need new cooling too (not to mention that only a specific subset of currently available motherboard chipsets will support this bleeding edge tech).
Sometimes not being an early adopter has its upside.
Interesting link:
Intel Multicore Fact Sheet - August 2006
While it probably isn't news to anyone that cares, the merger between CPU maker AMD and graphics chipset maker ATI has been given final approval from competition regulators in the US (home of AMD), Canada (home of ATI) and Germany (where most of AMD's chip fabrication plants are located). As the pieces finish falling in to place, I'm still trying to work out the ramifications of the marriage between the two major market players. It's not an easy task for a faceless government administrator, so I can only build my knowledge based on what scraps of information leak into the public domain from industry insiders.
What is the merger going to do the market place? Prior to the formation of the mega-corp, the CPU and graphics markets were fairly even split: NVIDIA and ATI battled it out in the graphics and motherboard chipset market while Intel and AMD tussled for market share with CPUs. Intel was still market leader but AMD had been steadily eroding that advantage for some time, particularly in the enthusiast market where AMD was considered to have better bang for buck. Both NVIDIA and ATI made chipsets for motherboards that supported both Intel and AMD. In fact, most motherboards for the new AM2 were based on NVIDIA chipsets.
The AMD-ATI merger leaves a number of questions that I don't have the capacity to answer (and would love it if someone could point me towards someone who can) :
With all the tech goodies hidden in the latest CPUs, the one thing that interests me the most is also the one thing that I know the least about - virtualisation supported in the hardware. I understand the basic idea of virtualisation but its implementation and its possibilities creep further and further beyond the grasp of my basic comprehension.
From what I can gather (and I am a simple man so this may be way off base) but virtualisation uses something called a Virtual Machine Manager (VMM) to provide a neat encapsulation, where the idea is to convince software that it is running in a certain environment when in reality it is encapsulated by the VMM. This could be a purely software emulation (eg convincing a DOS program that it is running on DOS) or hardware emulation (convincing software that it is running on a completely different processor). A good example of virtualisation is the backwards compatibility that Apple Computers has historically used while undergoing a complete change in CPU architecture. When they moved from the 68k processor to the Power PC, there were emulators that allowed legacy programs to run. Similarly, during the recent switch from PPC to Intel chips, the Rosetta software emulator once again allows software compiled specifically for the PPC chips to run on Intel chips.
"Virtualisation in hardware" seems to be all about providing support in the hardware for techniques utilised by the current crop of software VMMs like VMWare or Virtual PC where the VMM essentially detects whenever software tries to access various parts of the hardware and "traps" it. The VMM then returns to the software whatever information it would expect the hardware to send it. The software chugs along, happily thinking it is running in whatever environment, and the VMM is able to exercise complete control of the software by trapping it inside the VMM. I think. Because the VMM is able to utilise areas of the hardware specifically designed for that purpose, supported VMMs will theoretically be able to operate with almost zero latency or overhead as they do not need to operate through a software layer like an operating system.
I can see some of the more obvious benefits of virtualisation including:
It will be announced in the coming days (according to CNET and ZDNet) that IBM will be building a supercomputer for Los Alamos National Laboratory that is a hybrid of off-the-shelf AMD Opterons and the gorgeous Sony-Toshiba-IBM Cell processor.
While I'm always excited at the announcement of yet another supercomputer, I get an even bigger thrill when off-the-shelf components are used. I haven't been this jumpy about a supercomputer design since Virginia Tech built one entirely out of PowerMac G5s a few years back.
The best part of this announcement is that they will be using the STI-Cell. While I'll ramble on later this month on why I think the Cell is such a drool-worthy piece of kit, it is enough at this stage to say that the Cell is what will be powering the upcoming Playstation 3 and that the Cell is a single processor core chip with 8 speciality cores augmenting the main processor, providing support for coolness such as physics simulation.
I wish I could afford a supercomputer. For that matter I wish I could afford a PS3.
Ars Technica has a good write up of what makes the new Intel Core tick (Into the Core: Intel's next generation microarchitecture). It's not a new article (April 06), but newness is not a high priority for a man who less than 2 months ago bought a 2500+ Sempron which was a hot chip back in July 2004.
I'm not afraid to admit that while I appreciate that the article is probably simplified the key contents, and I did understand every single word, I continue to struggle when I try and string those single words together. ZDNet has something simpler, but I actually found that even less satisfying.
However, what I think it boils down to (and AMD is doing something similar with the Athlon64x2) is shrinking the process so much that they can fit the guts of two CPUs on to the one chunk of silicon, increasing grunt without increasing the footprint of the component. I believe Core also runs cooler, which is always a good thing.
A question I hear a lot is: "Do we even need this extra power in a home computer?"
Well, your grandma mightn't need it, but I can see many legitimate uses for it, particularly if your feeding your home movies straight from your digicam to your hard drive, editing it (Cue: Star Wipe...) and then encoding it so you can burn to CD. Extra cores can only be a good thing there. I imagine it will also be a good thing as more people start building HTPCs (home theatre PCs) to watch high-def digital TV and time-shifting.
As more multi-core computers enter the market, more software is going to be written to specifically take advantage of this and support it. Probably it will largely be games, video and audio apps, but that's a lot of market right there. If it means you can start rendering or encoding your latest home movie and then switch over and shoot some bad guys while you wait without taking too much of a performance hit, I'm all for it.
This is all in theory of course. I'm guessing by time I can afford a multi-core system, we'll all be surrounded by nano-bots doing our bidding and I'll still be excited about finally being able to play that "Doom thingy all the kids have been talking about..."
I recently replaced my CPU. Actually, over the period of about 2 months I replaced my hard drives, PSU, motherboard, CPU fan and CPU (in that order) while troubleshooting random reboots, shut downs and, on bad days, a consistent refusal to boot at all.
Since late 2001 I have been using what I affectionately call "old technology", although it was the height of sophistication for at least two weeks when I originally bought. For 4 1/2 years I ran my hand picked box of increasingly old tech goodies with a Socket A Athlon Palomino chip. It might have quickly become an evolutionary dead end as the Palomino was quickly replaced by new fabrication techniques, but it worked well enough for me. I could run Quake 3, I could play Baldur's Gate (slowly, inexpertly, infrequently and never for very long, but that's my failing, not the chip's), I could surf the net and I could create oodles of 30 second pieces of music using nothing but Virtual Studio Technology, while convincing myself that the inspiration for an actual finished track would be just around the corner.
In the intervening years between when I bought my computer and when the troubles started, the tech seemed younger and younger as my bleeding edge chip became middle of the road before quickly being relegated through the stages of veteran, venerable, out-dated and finally "legacy" (obsolete but there was too much existing infrastructure to bury it completely).
When the troubles started, I did not have much money. I looked for a way to fix it all while replacing the bare minimum of parts. When I decided that the motherboard had to go, I searched extensively for another Socket A board so that I could squeeze another year or so out of my current chip. Palomino was long gone and while Socket A chips still existed if you looked hard enough, they were no longer bleeding edge as to be coated in the dust left behind by the smaller, faster and cooler (both temp and tech) wafers of silicone delights. Although I flirted with a total "budget upgrade" that would still reduce my current machine to relic status as an historical curiosity, I managed to find a Socket A board in the end at a local Computer Fair. I was triumphant! Armed with a new PSU and a new motherboard, I proudly completed my first ever serious component upgrade. Swapping out IDE peripherals like hard drives and DVD-RWs was just practice for the glory that would be earning the title of Small Time Upgrader. Once a boy, soon to be a man.
Of course, in the end it turned out that my problems stemmed from a far deeper place and I eventually had to track down a new but still very much old tech Socket A chip for my replacement motherboard. It was much better than the old chip with a slighty different architecture and a cooler running temperature, but it just wasn't awesome enough. Don't get me wrong here - it rocks, just not very hard.
Even with all my recent upgrading (including my first ever after market CPU heat sink and van - oh, the tingle of delicious thrills cascading down my spine was strong that day), I am still trapped in my buttoned down, slow lane, caravan hauling, bowler hat wearing computing lifestyle. While I putter about with my Socket A Sempron (not even Socket 754 - oh the shame), I'm surrounded by 64 bit, multi-cored, virtual machine supporting processors to the Gods. I wasn't even trying to catch the boat and yet somehow I still missed it.
Processors: they're so hot right now. Maybe one day I will get to have one of my very own.
