Slashdot Mirror

← Back to Stories (view on slashdot.org)

AMD Considered GDDR5 For Kaveri, Might Release Eight-Core Variant

Posted by timothy on from the things-that-might-have-been dept.

MojoKid writes "Of all the rumors that swirled around Kaveri before the APU debuted last week, one of the more interesting bits was that AMD might debut GDDR5 as a desktop option. GDDR5 isn't bonded in sticks for easy motherboard socketing, and motherboard OEMs were unlikely to be interested in paying to solder 4-8GB of RAM directly. Such a move would shift the RMA responsibilities for RAM failures back to the board manufacturer. It seemed unlikely that Sunnyvale would consider such an option but a deep dive into Kaveri's technical documentation shows that AMD did indeed consider a quad-channel GDDR5 interface. Future versions of the Kaveri APU could potentially also implement 2x 64-bit DDR3 channels alongside 2x 32-bit GDDR5 channels, with the latter serving as a framebuffer for graphics operations. The other document making the rounds is AMD's software optimization guide for Family 15h processors. This guide specifically shows an eight-core Kaveri-based variant attached to a multi-socket system. In fact, the guide goes so far as to say that these chips in particular contain five links for connection to I/O and other processors, whereas the older Family 15h chips (Bulldozer and Piledriver) only offer four Hypertransport links."

87 of 120 comments (clear)

  1. Latency vs bandwidth by evilsofa · · Score: 5, Informative

    DDR3 is low latency, low bandwidth. GDDR5 is high latency, high bandwidth. Low latency is critical for CPU performance while bandwidth doesn't matter as much. On video cards, GPUs need high bandwidth but the latency doesn't matter as much. This is why gaming PCs use DDR3 for system RAM and GDDR5 on their video cards. Video cards that cut costs by using DDR3 instead of GDDR5 take a massive hit in performance. The XBox One and PS4 use GDDR5 shared between the CPU and GPU, and as a result have the rough equivalent of a very low-end CPU paired with a mid-range GPU.

    1. Re:Latency vs bandwidth by symbolset · · Score: 2

      We had this argument long ago, and decided we don't like Rambus not because they don't have some good tech, but because they are fucktards.

      --
      Help stamp out iliturcy.
    2. Re:Latency vs bandwidth by dagamer34 · · Score: 2

      The Xbox One doesn't use GDDR5 but DDR3 memory with 32MB eSRAM. They made the decision early on to have 8GB of RAM (regardless of type) before they knew it would be feasible to have 8GB GDDR5 memory at an easily attainable cost.

    3. Re:Latency vs bandwidth by Anonymous Coward · · Score: 4, Interesting

      False, XBox one uses pure DDR3.

      It is also one of the key reasons why many games on XBox one cannot do 1080p (that, and the lack of ROPs - PS4 having twice as many ROPs for rasterization)

      XBox One tries to "fix" the RAM speed by using embedded sRAM on-chip as a cache for the DDR3 for graphics. Remains to be seen how well the limitations of DDR3 can be mitigated. Early games are definitely suffering from "developer cannot be assed to do a separate implementation for Xbox One".

      Kaveri, while related to the chips inside the consoles, is decisively lower performing part. Kaveri includes 8 CUs. XBox one has 14 CUs on die, but two of those are disabled (to improve yields), so 12. PS4 has 20 CUs on-die, with again two CUs disabled to improve yields, so 18.

      On the other hand, Kaveri has far better CPU cores (console chips feature fairly gimpy Jaguar cores, tho both consoles feature 8 of those cores vs 4 on Kaveri)

      Any integrated graphics setup that uses DDR3 is bound to be unusable for real gaming. Kaveri has a good integrated graphics setup compared to the competition, but it is far behind what the new consoles feature - boosting it with GDDR5 without also at least doubling the CU count wouldn't do much. Either way, it really isn't usable for real gaming. It beats the current top offering from Intel, but that's bit like winning in Special Olympics when compared to real graphics cards (even ~$200-250 midrange ones)

    4. Re:Latency vs bandwidth by Anonymous Coward · · Score: 2, Informative

      Somewhat false. Latency is approximately the same for DDR3 vs GDDR5, at least in terms of nanoseconds from request to response. GDDR5 is impractical for use in CPUs due to the need to solder to a board and high power consumption (enough to need cooling). That CPUs don't need so much bandwidth just makes it useless in addition.

      Also, the Xbox One uses DDR3, and makes up for the lack of bandwidth with four channels and 32MB of ESRAM for graphics use.

    5. Re:Latency vs bandwidth by Sockatume · · Score: 4, Insightful

      Latency in cycles is higher for GDDR5, but the clock speed's a lot faster, isn't it? As the real-time latency is the product of the number of cycles and the length of a cycle, I think it's pretty much a wash.

      --
      No kidding!!! What do you say at this point?
    6. Re:Latency vs bandwidth by arbiter1 · · Score: 1

      One point it seems being missed is " implement 2x 64-bit DDR3 channels alongside 2x 32-bit GDDR5 channels,". So 2 half width channel for gddr5, it wouldn't help a bit if its only 32bit channel. Keep in mind width of gpu its 256bit and higher.

    7. Re:Latency vs bandwidth by Firethorn · · Score: 1

      Indeed. All the variations of memory today seem to have about the same latency. RAM that cycles faster simply takes more cycles to start returning requests, though once it starts it IS faster.

      --
      I don't read AC A human right
    8. Re:Latency vs bandwidth by wagnerrp · · Score: 1

      Remember, these are CPUs with integrated GPUs, and there are several benchmarks that show nearly linear improvement in graphics performance on AMD APUs as you increase the memory bus speed. They are severely bottle-necked by a mere dual-channel DDR3 controller.

    9. Re:Latency vs bandwidth by spire3661 · · Score: 1

      Xbox One is using DDR3 with eDRAM, only Sony went with GDDR5.

      --
      Good-bye
    10. Re:Latency vs bandwidth by Bengie · · Score: 3, Informative

      DDR3 and GDDR5 have nearly the same latency when measured in nanoseconds. When measured in clock cycles GDDR5 is higher latency, but it has more cycles. This has to do with the external interface, which is more serial in nature than the internal. The data path for the internal is quite wide, but the external datapath is not because traces on a motherboard are expensive. To crank up the bandwidth, you increase the frequency. If the internal frequency remains fixed and the external frequency goes up, the external "latency" seeming goes up.

  2. AMD could do a 24 core desktop chip right now by symbolset · · Score: 1, Interesting

    They don't care to because it would cut into their server revenue where margins are higher. Personally I think that really sucks. Intel is the same way. Maybe the migration to mobile where we don't have these margin protection issues is a good thing.

    --
    Help stamp out iliturcy.
    1. Re:AMD could do a 24 core desktop chip right now by guacamole · · Score: 1

      Why do you need so many cores on the desktop though?

    2. Re:AMD could do a 24 core desktop chip right now by Anonymous Coward · · Score: 2, Interesting

      No, they don't do it because it considerably raises the cost of the chip and it doesn't help improve the "average" user's workload. Many core processors have a bunch of inherent complexity dealing with sharing information between the cores or sharing access to a common bus where the information can be transferred between processes. There are tradeoffs to improving either scenario.

      Making the interconnection between cores stronger means that you have a high transistor count and many layers in silicon. Even if you do this, the scenarios in which the cores need to intercommunicate means that either the processes are being dispatched to the processor and the processor has a scheduling algorithm built in (which has its own issues), or you have a new set of instructions allowing software to dispatch code from cpu to cpu. Then race conditions and all sorts of other nonsense are inherently in the CPU itself and you have a whole bunch of problems trying to write code for the mess.

      Even if you go the above route, you still have to get the data you process out of the CPU and into RAM. Then you have a bus contention problem: how can multiple cores access different sections of RAM simultaneously? Is the CPU responsible for settling a conflict when two different cores want to write to the same section of RAM at the same time? These issues are largely easily settled with only 2, 3, or 4 cores now (and currently the responsibility for preventing these scenarios is on the OS not the CPU), but they would explode with a 24 core (or more).

      It's currently possible to build a CPU with several hundred cores inside (and both Intel and AMD have done it). No one has settled the issues above to make it practical, or invented the new software paradigm to make it easy to fix them.

    3. Re:AMD could do a 24 core desktop chip right now by symbolset · · Score: 2

      I like to use Blender to compose my videos.

      --
      Help stamp out iliturcy.
    4. Re:AMD could do a 24 core desktop chip right now by symbolset · · Score: 1

      No, they do it because it would compete with their $1000 server chips. ARM is about to give them some correction.

      I've been an AMD booster from way back, but they think they're still competing with Intel and that is a serious mistake.

      --
      Help stamp out iliturcy.
    5. Re:AMD could do a 24 core desktop chip right now by symbolset · · Score: 1

      We're about 10 years past that argument, and we keep buying more. Stop giving us more and we stop buying your stuff.

      --
      Help stamp out iliturcy.
    6. Re:AMD could do a 24 core desktop chip right now by TheLink · · Score: 3, Interesting

      They don't care because a desktop with a 24 core AMD CPU is likely to be slower than a 4 core Intel CPU for most popular _desktop_ tasks which are mostly single threaded. They're already having problems competing with 8 core CPUs, adding more cores would either make their chips too expensive (too big) or too slow (dumber small cores).

      Sad truth is for those who don't need the speed a cheap AMD is enough - they don't need the expensive ones. Those who want the speed pay more for Intel's faster stuff. The 8350 is about AMD's fastest desktop CPU for people who'd rather not use 220W TDP CPUs, and it already struggles to be ahead of Intel's mid range for desktop tasks: http://www.anandtech.com/bench/product/697?vs=702

      A few of us might regularly compress/encode video or use 7zip to compress lots of stuff. But video compression can often be accelerated by GPUs (and Intel has QuickSync but quality might be an issue depending on implementation). The rest of the desktop stuff that people care about spending $$$ to make faster would be faster on an Intel CPU.

      A server with 24 cores will be a better investment than a desktop with 24 cores.

      --
    7. Re:AMD could do a 24 core desktop chip right now by symbolset · · Score: 1

      No, AMD's issue is that they underestimate what is required to take the desktop, and they underestimate tech Intel has in reserve. They have management control issues, partner communication issues, and their recent layoffs have left their message garbled. I'd love to be able to unleash AMD's potential on an unsuspecting world, but I'm unlikely to be given the chance. I would call a press conference and say "Hey, Krzanich: bite me!" and unleash a 16 core desktop CPU with 192 core GPU. I would open a division in Valve and open the tap: what do you need?.

      And If I was Intel and trying to compete I would quit with the herpaderp Windows crap, since obviously that isn't moving units.

      --
      Help stamp out iliturcy.
    8. Re:AMD could do a 24 core desktop chip right now by Sockatume · · Score: 1

      Where else are Intel going to go? AMD's got the console space wrapped up this generation, and the Steambox isn't far enough along to make for a solid revenue stream. That leaves low-power applications where they're making progress but not yet ready to dive in. Like it or not, Intel are going to be a Windows and MacOS house for another five years or so.

      --
      No kidding!!! What do you say at this point?
    9. Re:AMD could do a 24 core desktop chip right now by symbolset · · Score: 3, Informative

      Fine. You do it your way. I want this technological achievement so I can do it mine.

      --
      Help stamp out iliturcy.
    10. Re:AMD could do a 24 core desktop chip right now by Lonewolf666 · · Score: 1

      An 8-core Steamroller would be an improvement too, now that computer games finally start scaling well with multiple cores. I might even be willing to re-purpose a server part for my next desktop, even if it is a tad more expensive.

      If AMD does not bother though, the Xeon E3-1230 v3 from Intel looks nice as well, only the mainboards that actually support ECC RAM are a bit expensive.

      --
      C - the footgun of programming languages
    11. Re:AMD could do a 24 core desktop chip right now by symbolset · · Score: 2

      Your problem - and Intel and AMD have this problem also - is thinking the specifics of the device have something to do with its utility. It doesn't. What matters is how the thing enables people to do what they want to do, and we passed that level of utility in devices a decade ago. I can be telepresent with my children without resorting to either Intel or AMD technologies - all day. Compared to exclusively special uses of spreadsheets and slideshow programs, compatibility differentials of Office suites, that is a win.

      It's about the people, not the thing. Wrap your head around that. Make it your religion.

      --
      Help stamp out iliturcy.
    12. Re:AMD could do a 24 core desktop chip right now by Lonewolf666 · · Score: 1

      For some applications, in particular games, performance still matters. My current PC will run older games just fine, but some newer releases tend to demand a fairly powerful machine.

      For example, I might be interested in Star Citizen but Chris Roberts has already announced that he is aiming for maximum eye candy on high end machines. It is unlikely that my current machine will handle that game, even on low settings.

      If the applications you run do well on a machine from a decade ago, fine. But that is not the case for everybody.

      --
      C - the footgun of programming languages
    13. Re:AMD could do a 24 core desktop chip right now by dreamchaser · · Score: 1

      It depends on what you're doing. In my case I'm often running 4-6 VMs all busy doing various tasks. Having multiple cores helps greatly, as does having a lot of RAM. Any use case that benefits from multiple threads can, if the software is written properly, take advantage of multiple cores.

    14. Re:AMD could do a 24 core desktop chip right now by symbolset · · Score: 1

      No, it is not different. How people feel about your game is the freaking point. Making people care about your virtual world is the goal. You are an author not unlike all the other authors that ever were.

      --
      Help stamp out iliturcy.
    15. Re:AMD could do a 24 core desktop chip right now by citizenr · · Score: 1

      I think they know, they do have ARM license for 64bit stuff and will make ARM64 based Opterons.

      --
      Who logs in to gdm? Not I, said the duck.
    16. Re:AMD could do a 24 core desktop chip right now by symbolset · · Score: 1

      That is going wrong for all the right reasons. It still ends in death.

      --
      Help stamp out iliturcy.
    17. Re:AMD could do a 24 core desktop chip right now by Sockatume · · Score: 1

      I'm not one who buys into the whole "we should be designing games for $2000 systems" madness, but it's still obvious that improved technical performance is something that game creators want and can exploit to improve their art and design, much as film-makers like having new varieties of lens, film, and camera, new techniques for practical effects, new rigging systems and dollies, and so on.

      You couldn't do the gameplay of something like MGS3, for instance, on an original PlayStation. For that gameplay to work you need a lot of variation in the shape and nature of the environment and you need the environment to be large. It's not that the game's full of whizz-bang visual effects, but drawing even crude grass and trees and characters eats up enormous amounts of polygons. You could barely do it on a PS2, and it turns out that a game with those creative goals is hugely improved by the locked framerate and booster resolution that it got when it was re-released on the systems that came afterwards.

      --
      No kidding!!! What do you say at this point?
    18. Re:AMD could do a 24 core desktop chip right now by Sockatume · · Score: 1

      You know what a fast, multi-core parallelised CPU optimised for video encoding looks like?

      A GPU.

      --
      No kidding!!! What do you say at this point?
    19. Re:AMD could do a 24 core desktop chip right now by Lonewolf666 · · Score: 1

      A good point from the perspective of a game designer, and I support the sentiment.

      But most of us are consumers most of the time. Even those of us who work on one or two community software projects will typically use a bunch of software where they are not involved in the making. Which means taking the software as it is, and if its creators went for a design that requires a beefy PC you have one or you don't use that particular software.

      --
      C - the footgun of programming languages
    20. Re:AMD could do a 24 core desktop chip right now by MachineShedFred · · Score: 1

      Oh, Windows is moving units for Intel, as Intel pretty much owns the enterprise. Walk into pretty much any Fortune-100, and you'll see the blue logo stickers on practically every single laptop and desktop in the place, and the data center will be floor to ceiling with Xeon except specialty stuff (SPARC boxes, IBM pSeries, mainframes). Thin Clients running Linux? Likely on Atom, but AMD is making a bit of an inroad there.

      Don't fool yourself - big business is still suffering from 20 years of Microsoft lock-in, and Intel rules that particular roost for performance and manageability through their AMT and vPro chipset tech.

      vPro alone will save the company I work for an estimated $1M a year in tech dispatches and eliminated remote control agent licensing, and that doesn't even include the increased uptime of not having to wait for the tech to travel to the site to do the job.

      --
      Slashdot still doesnâ(TM)t support Unicode after it was added to the HTML standard in 1997.
    21. Re:AMD could do a 24 core desktop chip right now by edxwelch · · Score: 1

      Maybe they should start with the server version though, since AMD currently have no server with 24 steamroller cores.

    22. Re:AMD could do a 24 core desktop chip right now by wagnerrp · · Score: 1

      I'll give you the RAM, considering the isolation of a VM necessitates all those shared libraries be wastefully loaded and cached multiple times, but why would running 4-6 VMs result in high CPU load compared to not running VMs and just having a monolithic server?

    23. Re:AMD could do a 24 core desktop chip right now by hairyfeet · · Score: 2

      But the "dirty little secret" that doesn't get brought up enough frankly is a lot of those "single threaded loads" is as rigged as quack.exe was back in the day thanks to every Intel compiler made since 2002 being made to put out crippled code for any chip that Intel doesn't want to push. Oh and for those that use the "Intel just knows their own chips and optimizes for them" excuse that lie has been disproved and the proof was the last gen Pentium 3. You see the last gen P3 was curbstomping the Netburst P4s in early benchmarks, yet when the cripple compiler comes out? Suddenly the very same Netburst chips are winning by 30%!

      And the bitch is that any of these so called review sites could test for rigging trivially but they won't for fear of losing Intel advertising revenue. To see if a program is rigged all one has to do is run the code on a Via CPU, Via CPUs allow one to softmod the CPUID so if you change the CPUID from "Centaur Hauls" to "Genuine Intel" and suddenly the chip scores 20%-30%+ higher on the test? Then the program has been rigged by ICC, simple as that.

      All of these sites like Anandtech and Tom's have more than enough money to pick up a Via chip and keep it around for testing but they won't bite the hand that feeds so everyone should consider their tests to be tainted and as worthless as Quake tests were with the rigged drivers. It would be nice if someone would run some real tests so we could see real numbers, wouldn't be hard to do as GCC is free and there are plenty of FOSS programs like Firefox one could compile with GCC to give accurate tests but so far no review site will do this for fear of pissing off Intel. How they didn't get busted for antitrust is beyond me, this is every bit as bad as "Windows isn't done until lotus won't run" but they were allowed to bribe AMD to the tune of 2 billion to drop the lawsuit and with it the investigation.

      --
      ACs don't waste your time replying, your posts are never seen by me.
    24. Re:AMD could do a 24 core desktop chip right now by K.+S.+Kyosuke · · Score: 2

      The rest of the desktop stuff that people care about spending $$$ to make faster would be faster on an Intel CPU.

      You mean something like this? (Or similar solutions for other languages?)

      --
      Ezekiel 23:20
    25. Re:AMD could do a 24 core desktop chip right now by dreamchaser · · Score: 1

      It depends on what the VMs are doing. Having more physical processors to run threads on helps a lot in my case, where they usually involve running multiple hosts simulating client networks, with a few of said VMs doing lots of packet processing.

    26. Re:AMD could do a 24 core desktop chip right now by Bengie · · Score: 1

      Increasing the core count reduces per core performance and increases total power consumption and production cost. With most cores idle, it is a bad thing to have too many cores.

    27. Re:AMD could do a 24 core desktop chip right now by K.+S.+Kyosuke · · Score: 1

      You seriously think typical desktop java stuff will run faster on a GPU?

      Oracle certainly thinks that. Ditto for servers. I bet that they have more heads than the two of us to think about the problem, and they seem already hell-bent on implementing just that. Now I kindly suggest that you sit back and leave that to the experts.

      --
      Ezekiel 23:20
    28. Re:AMD could do a 24 core desktop chip right now by Bengie · · Score: 1

      The issue with Java for GPUs is that GPUs have objects and love structures. GPUs puke when random memory access is involved. Right now, Java treats everything as an object, even structs. Java will need to be revamped a bit to work on GPUs. Get ready to start doing buffers and structs in Java, it'll be like C.

    29. Re:AMD could do a 24 core desktop chip right now by K.+S.+Kyosuke · · Score: 1

      I think you missed the part where AMD has been continuously tweaking their GPU hardware in the past few years to make it increasingly happy with random access and C/C++/Java-like data layouts. Namely the switch to superscalar RISC cores, the conception of the HSA in the first place, and now the advent of shared access to paged CPU memory, C++11-compatible atomics included. I didn't notice anything in the HSAIL spec that would prevent the implementation of "traditional" Algol-like programming languages. Now granted, there are still some issues to be handled - even though the CPU->GPU->CPU call latency is vastly lower (and data sharing is zero-copy), it's still not zero, and the compilers will have to be clever about a lot of stuff, especially about control flow divergence, at least on some particular hardware implementations. And there's still some work to be done on context switching and overall making the GPU cores behave like CPU cores as OS resources. But if the Oracle people think it's worthy of the effort, it probably is. (Honestly, I certainly like it much more that the latest UltraSparc chips.)

      Also, I mentioned Java mostly because that's a potential major workload for the future Opteron APUs. Me, I will be perfectly with HSAified Gimp and Blender, and with somewhat more conservative languages such as Go or C and the like reaping the benefits for the purposes of text mining, AI, NLP and such.

      --
      Ezekiel 23:20
    30. Re:AMD could do a 24 core desktop chip right now by Carewolf · · Score: 1

      Well, in a way gcc and the linux kernel is also rigged, though in a very fair an honestly good way. Intel is sponsoring a lot of open source development. The only conditition is their open source teams has is to make sure the open source projects support Intel products as good as possible. This means that there a many optimization for Intel than AMD in all central open source projects. The Intel compiler stuff is terrible and Intel needs to be reminded of it and scolded for it all the time. I can't scold them for just making things better though.

  3. Re:news for nerds by ls671 · · Score: 2

    Great! Can you explain to me why "GDDR5 isn't bonded in sticks for easy motherboard socketing" ?

    --
    Everything I write is lies, read between the lines.
  4. Of course, that would miss the point by guacamole · · Score: 4, Interesting

    The whole point of AMD APUs is low cost gaming. That is, lower cost than buying a dedicated GPU plus a processor. Many already argue that you don't save much by buying an APU. A cheap Pentium G3220 with a AMD Radeon 7730 costs the same as the A10 Kaveri APU, and will give better frame rate. Even if the Kaveri APU prices come down, the savings will be small. If you have to buy the GDDR5 memory, there won't be any savings. It's understandable that AMD didn't take that route.

    1. Re:Of course, that would miss the point by Anonymous Coward · · Score: 1

      Also benchmarks have already proven that Kaveri couldn't utilize GDDR5 well enough. memory speeds over 2133 MHz no longer improve benchmarks on it, so while "bog standard" 1600 MHz DDR3 will leave Kaveri's GPU somewhat memory starved, 2133 MHz is already enough for it (upping that to higher frequencies through overclocking the memory alone doesn't help).

      Besides, Kaveri could just go for four DDR3 memory channels. The Chip supposed can do it, it's just that motherboards available right now can't.

      Now in the future APUs may get to a point where they would greatly benefit from on-motherboard block of GDDR5 for the GPU, but Kaveri ain't it - it is off by about factor of 3 or so. 8 CUs is just far too little... would need something like 20-24 CUs to start getting to the range of high end *laptop* GPUs and good level for today's high end gaming. Then it'd need faster CPU cores (or 8 cores instead of 4, if we assume that games really will multithread - which is still bit of hit and miss). That would be a Very Big Chip and probably wouldn't be available anywhere near current price points.

      Kaveri has a good balance - a budget CPU with a reasonable low end GPU tacked on. It could use bit more memory bandwidth, but it isn't terribly RAM-bandwidth-starved. It is hopelessly outclassed by any setup with reasonable midrange gaming card, but the price reflects that.

    2. Re:Of course, that would miss the point by Lonewolf666 · · Score: 1

      Besides, Kaveri could just go for four DDR3 memory channels. The Chip supposed can do it, it's just that motherboards available right now can't.

      It would also require a new and presumably more expensive socket, and motherboards would always need four DDR3 sockets for provideing the full bandwidth - no more cheap mini boards with only two sockets.

      Overall, I'm not sure if it would be much cheaper than GDDR5 on the mainboard.

      --
      C - the footgun of programming languages
    3. Re:Of course, that would miss the point by girlintraining · · Score: 2

      The whole point of AMD APUs is low cost gaming.

      Sigh. You people with your myopic vision. If AMD consigned itself to your view of what it should do, it'll be dead in another 5-7 years. Let's take a look at what Intel offers: Higher performance. Lower energy consumption. Less heat. Smaller die size. In fact, you'd be hard pressed to find anything AMD has in its favor from an engineering standpoint. So what does AMD have that's keeping it in business? Cost. AMD offers a lower price point for economy systems.

      But that's not where the profit is. That's not what's going to take AMD into the mid-21st century. If AMD sticks to that line of thinking, it'll go the way of Cyrix... and for exactly the same reason. AMD can't invest in a new fab plant because its cash reserves are too low, whereas Intel's pile of gold just keeps growing.

      The only way to swing AMD back into the mid and high-end market is parallelization -- multiple chips, embedding everything into everything, and cutting costs everywhere it can... because it can't shrink the CPU and it can't do a damn thing about energy consumption. Their only avenue of escape is a radical re-think of the chipset, integration, and anything it can to boost profits to make that leap forward in minaturization to catch up with Intel. And it will not be easy. Many market analysts have suggested AMD throw in the towel.

      That's the position AMD has to play. Of course, your myopic view only sees an engineering problem to solve. You have absolutely no comprehension of the business considerations behind the move... which is probably why you insist they're missing the boat. They are trying to claw their way into the mid-range market and undercut Intel. This is one of the select few ways they're going to do it. Unfortunately, the cost of commodity components to get this Hindenburg off the ground isn't going down -- AMD's position is shrinking and if they don't leverage everything now, they're going to find themselves in an untenable market position and will simply fold, leaving Intel as the only major player left in the market.

      And if that happens... we're all fucked.

      --
      #fuckbeta #iamslashdot #dicemustdie
    4. Re:Of course, that would miss the point by Daniel+Hoffmann · · Score: 1

      " So what does AMD have that's keeping it in business? "
      A licensing agreement with intel for the x64 instruction set also helps. Every intel chip sold gives AMD a few bucks and every AMD chip sold gives intel a few bucks (because of the licensing of x86 instruction set). I'm not sure on the specifics but I assume this to be a lot of money.

    5. Re:Of course, that would miss the point by mrchaotica · · Score: 1

      motherboards would always need four DDR3 sockets for provideing the full bandwidth - no more cheap mini boards with only two sockets.

      Mini-ITX and smaller boards could just use SO-DIMMs.

      --

      "[Regarding the 'cloud,'] ownership was what made America different than Russia." -- Woz

    6. Re:Of course, that would miss the point by Kartu · · Score: 1

      Uh oh. Yet it is much cheaper than i5 (+ premium on mainboard), but does much better in games.
      http://www.anandtech.com/show/7643/amds-kaveri-prelaunch-information

    7. Re:Of course, that would miss the point by K.+S.+Kyosuke · · Score: 1

      Now in the future APUs may get to a point where they would greatly benefit from on-motherboard block of GDDR5 for the GPU, but Kaveri ain't it - it is off by about factor of 3 or so. 8 CUs is just far too little...

      Why don't you simply get a discrete card and use the APU CUs for what they're supposedly good for, that is, random-access-ish general-purpose computation, as opposed to streamed-access shading and texturing? I mean, the fact that the APU is also reasonably competent as an iGPU is nice, but...

      --
      Ezekiel 23:20
    8. Re:Of course, that would miss the point by edxwelch · · Score: 2

      The thing is the real Kaveri star is the A8 7600, not the A10 models. The A8 7800 is only $119 yet thrashes any Intel APU in gaming.

    9. Re:Of course, that would miss the point by hairyfeet · · Score: 2

      And anyone who went that route would frankly be short sighted and foolish and would regret it later. The simple fact is that more and more games are taking advantage of multicores and that dual core will be a bottleneck pretty quickly on any of the new games coming out. Already many of the bigger games like Bioshock infinite, Far Cry 3, Dirt 3 and the like will run better on a quad than a dual and that number is going nowhere but up.

      As someone who has been selling the APUs since liano the advantages are threefold: One you can buy a nice budget system that will game OOTB and can add a dedicated GPU down the line, thus making the system more affordable to more folks, Two if you stick with AMD and buy one of the dedicated chips that support zerotech like a 7xxx chip or above you can significantly lower your power usage when you aren't gaming as zerotech will let the dedicated GPU drop down to a couple watts when not required and the APU can take care of hardware acceleration normally handled by the dedicated card, and finally Three these APUs also make great HTPC chips, with hardware acceleration for pretty much every major format out there you can get smooth 1080P playback even when using a tuner for recording or while multitasking.

      As for TFA while I can't see a point in embedded 8GB of GDDR5 in a board for one of these I CAN see putting a smaller amount, say 512Mb, to give the APU a nice high bandwidth buffer to improve performance.

      --
      ACs don't waste your time replying, your posts are never seen by me.
    10. Re:Of course, that would miss the point by guacamole · · Score: 1

      Sigh. You people with your myopic vision. If AMD consigned itself to your view of what it should do, it'll be dead in another 5-7 years. Let's take a look at what Intel offers: Higher performance. Lower energy consumption. Less heat. Smaller die size. In fact, you'd be hard pressed to find anything AMD has in its favor from an engineering standpoint. So what does AMD have that's keeping it in business? Cost. AMD offers a lower price point for economy systems.

      I am sorry, but I think you can't realize that at this point AMD can't compete even on the low end. A 300-400 dollar BestBuy special laptop with Intel i3 still beat AMD in speed and power efficiency. How much lower can AMD go in prices? They can't. They don't even have the volume to get there. They can integrate all they want, but that doesn't change the fact the AMD offerings are woefully uncompetitive at almost any price level. Forget Intel. After 6months of hyping their product, the flagship Kaveri A10 can't even decidedly beat the _last generation_ AMD's Richland A10.

    11. Re:Of course, that would miss the point by guacamole · · Score: 1

      Midrange market? HAHA

      Intel Core i5, mobile or desktop, is the midrange market. NO ONE of AMD's processors, APUs or FX, can match that in speed or power efficiency, not even the new Kaveri APU.

    12. Re:Of course, that would miss the point by guacamole · · Score: 1

      NO ONE buys an i5 CPU to use its integrated graphics for games. The Intel HD Graphics are a bonus for people who don't play games. Add a discrete GPU to i5 based PC as well as to AMD PC, and it will be game over to ANY AMD based system, not just Kaveri based. As for Kaveri A10, which costs $170 online, the lowest as of now, you can beat it with a $69 dollar Intel Haswell Pentium G3220 and a discrete graphics card like a $100 dollar Radeon 7730.

    13. Re:Of course, that would miss the point by guacamole · · Score: 1

      The problem is that AMD cores aren't that great. For one, each AMD module has two integer cores but only one FPU. Despite that, the call it "dual core" even though for floating point stuff, the AMD architecture awfully sounds like hyperthreading. And so to me it's not surprise that Intel i3 (2 real cores but 4 logical because of hyper-threading) can challenge 4 and 6-core AMD CPUs.

    14. Re:Of course, that would miss the point by guacamole · · Score: 1

      The A8 makes sense for a very low end system. If I was building a PC, either for playing games or not, I think my might have considered the A8 or A10-6800K (last generation A10) if I was looking for a $130 processor. It looks like they both have compute power somewhere in the neighborhood of Intel i3-4150 but somewhat better graphics.

    15. Re:Of course, that would miss the point by Bengie · · Score: 1

      The whole point of AMD APUs is low cost gaming. That is, lower cost than buying a dedicated GPU plus a processor. Many already argue that you don't save much by buying an APU. A cheap Pentium G3220 with a AMD Radeon 7730 costs the same as the A10 Kaveri APU, and will give better frame rate. Even if the Kaveri APU prices come down, the savings will be small. If you have to buy the GDDR5 memory, there won't be any savings. It's understandable that AMD didn't take that route.

      AMD is aiming for "good enough", and they did a great job. Per thread, AMD is now on par with Intel's Haswell and has an integrated GPU that can cover the 80/20 rule for games. The only issue I personally have is that AMD's current Kaveri offerings are limited to a 2 module(4 core) setup that consumes about 50% more power than 2 core(4 Hyper-thread) Haswell while idle and about 100% more power under pure CPU load. Since I will have a discrete GPU, I see no benefit to consuming that much more power. We're talking about a 20watt difference idle and about a 50watt difference under load. And they're about the same price.

    16. Re:Of course, that would miss the point by guacamole · · Score: 1

      I don't understand what you mean "per thread". AMD claims A10 is a four core CPU, but each module has only one FPU despite presenting two logical cores, which sounds awfully like hyper-threading to me, but Intel was more honest and called the i3 2-core CPU with hyper-threading. Basically, AMD overestimates how many real cores its processors really have, but this strategy seems to work since the web forums are filled with fanboys who think that more cores is always better.

      This is why benchmarks show AMD can't beat Intel CPUs in single threaded applications. If one thread all you can use, then Intel will be faster. If you need to use more threads, you need 8 threads or more before to start beating even Intel's i5.

    17. Re:Of course, that would miss the point by aliquis · · Score: 1

      the fact the AMD offerings are woefully uncompetitive at almost any price level

      They simply aren't.

      If they was then Sony and Microsoft wouldn't had opted to put tens of millions of their solutions in their consoles.

      Also for the older A10s faster memory really helped with graphics performance, and that the consoles have a different memory configuration likely help them perform better than the current desktop parts.

      It's nothing new GPUs like having massive memory bandwidth, sure the one in the APU is pretty small but it would likely still enjoy having better memory access than what a (multiple..) CPUs would have. Also who knows maybe with higher memory bandwidth they would had put a more capable GPU part in the APU too but as is what is worth putting in there may be limited by the memory access.

    18. Re:Of course, that would miss the point by Bengie · · Score: 1

      It has one 256bit FPU that can split to handle any mix of FPU between both cores. It can do 2x 128bit or 128bit and 2x 64bit or 128bit and 4x 32bit or 2x 64bit and 4x 32bit.. etc etc... Right now only AVX instructions can use the full 256bit, so any SSE or less won't "share" the FPU. Each core is actually able to pipeline up to 4x 32bit, even if not using SIMD. No matter how you look at it, it can handle 256bit worth of FPU calculations. The FPU eats up a lot of transitors and is one of the least used parts. Even in heavy FPU core, there are enough memory stalls or branching logic to make it not used often. There is little reason to have two full 256bit FPU units.

      Code the best works with heavy FPU usage typically works well with GPUs. Thus, the APU. Need a mix of high throughput and low latency? The APU has you covered. With about 10ns latency to the IGP compared to 10-100 microseconds for discreet GPUs, many work loads can notice a throughput increase.

    19. Re:Of course, that would miss the point by Bengie · · Score: 1

      AMD has two separate core logics with their own execution units that only share the FPU. They have their own L1 caches and a bunch of other things. The FPU can handle 256bits worth of FPU calculations split it almost any configuration of x87, SSE, AVX. The only real time you see contention for the FPU unit is when doing 2x 128bit SSE instructions on a single core or a single 256bit AVX instruction.

      Intel's hyper threading shared nearly everything, right down to the integer execution units and L1 cache. Intel made a highly pipelined CPU with a lot of execution units, then made it so it presented itself as two logical CPUs, as many of those execution units will be idle because of memory stalls or just can't pipeline enough instructions in parallel to use them all.

      Intel's design has a better best case and a worse worst case, AMD is more middle of the road. Both designs increase throughput per transistor and reducing power per transistor on average.

      A huge benefit Intel's design gets is if you disable HT or the OS puts a one of the logical cores to sleep, you can make use of the entire CPU, making single threaded performance strong as you have a single core with massive L1 cache and a ton of pipelinable execution units.

      But remember, Intel optimized for throughput, so that large L1 cache comes with a 2 cycle latency, but can handle 128bit loads per cycle. If you're going to optimize for single threaded performance, make use of those extra registers in 64bit mode and help hide that latency. when hyper-threading is enabled, it helps hide single threaded latency by interleaving logical threads against the cache, making it mostly seem like 1 cycle latency. Higher throughput, but higher latency.

    20. Re:Of course, that would miss the point by 0123456 · · Score: 1

      Intel have plenty of CPUs that can compete with AMD's prices at the low end. The only thing AMD have is better graphics, which is why they keep pushing APUs, even though there's a tiny market for them... they're basically for people who want to play games, but don't want to play them enough to buy a proper GPU.

      It would be a very viable company if they could just dump the CPU side and concentrate on graphics.

    21. Re:Of course, that would miss the point by Lonewolf666 · · Score: 2

      The 45W Kaveris are interesting, as they show a nice improvement in performance/watt - the new "sweet spot" is not in the top models but in the somewhat slower A8-7600 (3.1-3.3 GHz CPU speed).

      I wonder how a 4 module (8 core) FX on that basis would perform and at which TDP. For software that scales well with many cores, it might be a good buy.

      --
      C - the footgun of programming languages
    22. Re:Of course, that would miss the point by 0123456 · · Score: 1

      If they was then Sony and Microsoft wouldn't had opted to put tens of millions of their solutions in their consoles.

      Intel probably aren't much interested in the tiny margins in the console market, particularly when they'd have to throw a lot of R&D into building a competitive GPU.

      That doesn't make AMD any more competitive in PCs.

    23. Re:Of course, that would miss the point by tlhIngan · · Score: 1

      That's the position AMD has to play. Of course, your myopic view only sees an engineering problem to solve. You have absolutely no comprehension of the business considerations behind the move... which is probably why you insist they're missing the boat. They are trying to claw their way into the mid-range market and undercut Intel. This is one of the select few ways they're going to do it. Unfortunately, the cost of commodity components to get this Hindenburg off the ground isn't going down -- AMD's position is shrinking and if they don't leverage everything now, they're going to find themselves in an untenable market position and will simply fold, leaving Intel as the only major player left in the market.

      And if that happens... we're all fucked.

      I'm sure Intel realizes it too, because they've avoided a lot of anti-trust because of AMD. Heck, Intel probably foisted Sony and Microsoft onto AMD to provide AMD with some steady cashflow for the next 8+ years.

      To Intel, AMD's the perfect competitor - not too big, in just enough trouble to not be an issue. But big enough to matter and to provide competition in the eyes of government, keeping regulators and other stuff off Intel's back.

      And Intel probably has a "AMD rescue" package available which can involve burying millions of AMD processors to boost their sales. .

      Because if AMD fails - all of AMD's patents are most likely NOT going to be given to Intel. Instead, they'll probably be spread out among dozens of entities, and what was once a lucrative zero-cost cross-licensing with AMD becomes an expensive, lawyer filled license negotiation.

      You got to think similar things as well for odd competition, like Apple iAds. Probably supported by Google to justify the purchase of AdMob (DoJ said iAds provides sufficient competition in mobile advertising). Of course, it really isn't and Apple's quit promoting it for some time, yet it lives on. Probably some Google cash paying someone to keep it alive.

    24. Re:Of course, that would miss the point by Kjella · · Score: 1

      But that's not where the profit is. That's not what's going to take AMD into the mid-21st century. If AMD sticks to that line of thinking, it'll go the way of Cyrix... and for exactly the same reason. AMD can't invest in a new fab plant because its cash reserves are too low, whereas Intel's pile of gold just keeps growing.

      Dude, this already happened back in 2009 when they first spun off and later sold out of GlobalFoundries.

      They are trying to claw their way into the mid-range market and undercut Intel.

      Again it sounds like you dropped out of a time machine from 2009, when Thuban was aging and Bulldozer was supposed to be AMDs ace in the hole. Since then AMD has done nothing but dodge Intel selling all-in-one APUs using their graphics division and special case architectures for consoles, supercomputers, ARM servers and everything but going head to head with Intel. Their flagship CPU is still a Bulldozer refresh from 2012 built on 32nm, they've got nothing to compete with for systems that use no graphics (CPU crunching) or dedicated graphics (on board graphics irrelevant).

      AMDs own roadmap shows there's no replacement coming in 2014, the only thing you'll get are Kaveri APUs. And the only thing selling them are the GCN graphics, otherwise they're almost out of the CPU business. Xeons now rule totally supreme over the once so good, but horribly aging Opterons. Sadly you're projecting what you'd like to happen with what is happening, AMD is not turning to fight they're running looking for something else to turn a profit on. It might be the best (or only) remaining choice for AMD, even if it sucks for us.

      --
      Live today, because you never know what tomorrow brings
    25. Re:Of course, that would miss the point by aliquis · · Score: 1

      In the "chips for people who aren't ultra-graphics extreme gamers"-department they seem to be doing pretty ok currently. The APUs doesn't seem terrible. Sure you can get faster processors and sure you can get dedicated graphics cards.

      Maybe chip A is quicker in this and that area vs chip B and prices are whatever but for those consumers I don't know how much it really matters. For me personally if I went without a dedicated graphics card I think game performance would rate pretty high / the highest.

      If more gaming go over to portable machines and more code/game content is shared among the platforms I guess PC gaming will become even less demanding. The best phones and tablets will still not put up much of a challenge against the APU and since similar chips sit in the game consoles neither should they really.

      And if you do get a dedicated graphics card they seem to be about on par with Nvidia. Nvidia will give you better Linux drivers, accelerated PhysX and G-sync while the AMD chips support DX11.2, got their PureAudio (not sure that's worth caring about) and AMD seem more interested in having Mantle a standard rather than something just they use and for obvious reasons think GPU syncing should be done as a command standard too (I seriously don't get Nvidias bragging about having coming up with the idea and the solutions. Like really? No-one have wished for it before? It was all that hard? I imagine the hardest part for anyone with it would be to actually get it implemented but Nvidia is large enough to make that happen.)

      (And earlier I've thought where Nvidia will go since they don't offer their own x86 chip but then they have their ARM stuff instead so I guess they are all fine they too.)

      As a consumer I've got no interest in proprietary standards in the above mentioned areas though. Fuck it all.

  5. A Better Explaination At Anandtech by Anonymous Coward · · Score: 5, Informative

    Anandtech's writeup (which Hothardware seems to be ripping off) has a much better explanation of what's going on and why it matters.

    Let me be very clear here: there's no chance that the recently launched Kaveri will be capable of GDDR5 or 4 x 64-bit memory operation (Socket-FM2+ pin-out alone would be an obvious limitation), but it's very possible that there were plans for one (or both) of those things in an AMD APU. Memory bandwidth can be a huge limit to scaling processor graphics performance, especially since the GPU has to share its limited bandwidth to main memory with a handful of CPU cores. Intel's workaround with Haswell was to pair it with 128MB of on-package eDRAM. AMD has typically shied away from more exotic solutions, leaving the launched Kaveri looking pretty normal on the memory bandwidth front.

    It's also worth noting that the Anandtech article implies that AMD is still on the fence on Kaveri APUs with more memory bandwidth, and that it may be something they do if there's enough interest/feedback about it.

    1. Re:A Better Explaination At Anandtech by basecastula+ · · Score: 1

      Thanks.

    2. Re:A Better Explaination At Anandtech by danknight48 · · Score: 1

      Cheers for the better article.
      It wasn't just me who thought the author of the hothardware article didnt have a clue?

  6. Re:news for nerds by davester666 · · Score: 2

    Motherboard manufacturers see the profit margin's Apple has with RAM that can't be increased in a number of their high-end models and they want in on that action.

    --
    Sleep your way to a whiter smile...date a dentist!
  7. Re:Was tha intentional? by Luckyo · · Score: 1

    Actually "pal". Friend would be "ystävä".

  8. Re:news for nerds by SuricouRaven · · Score: 4, Informative

    Because it's electrically so delicate that you can't keep bit sync when shoving such high frequencies through a slot connector. The price of higher bandwidth, in both the analog and digital senses.

  9. Re:news for nerds by basecastula+ · · Score: 1

    I feel you brother. Its has been nonstop garbage for months.

  10. Re:news for nerds by symbolset · · Score: 1

    /. is a user supported blog. That means that you can submit any link you like. In case you missed the link present on every page, here it is again.

    --
    Help stamp out iliturcy.
  11. Re:news for nerds by hairyfeet · · Score: 2

    Sure can. You see currently all GDDR5 is sold to GPU OEMs and they haven't used sockets on graphics card since the mid 90s.

    As for TFA they sold some HD43xx boards with a small amount of dedicated RAM and since AMD makes their own board chipsets I see no reason why they couldn't sell a board or two with dedicated GDDR5 of 512Mb to say 2Gb. Lets face it guys you can only go so far with an APU before it would make more sense to have a dedicated GPU instead but with AMD having the same chips that are in the new consoles this might be a good option for a budget gaming rig.

    The key will be if they can get the GDDR5 at a price that will still let the board make sense versus just buying say an HD7750 card, but if they could get the boards made so a kit would cost around $250-$270 (current their quad APU kits sell for around $200 sans HDD) this could be a really compelling option for those on a budget that want to game.

    --
    ACs don't waste your time replying, your posts are never seen by me.
  12. This show up every time! by higuita · · Score: 1

    No, its a river in india... AMD uses river names for APU/CPUs families and GPUs are islands (where the type and location of the island is used as family grouping)

    --
    Higuita
  13. Re:news for nerds by unixisc · · Score: 2

    Great! Can you explain to me why "GDDR5 isn't bonded in sticks for easy motherboard socketing" ?

    Reason is that they are used exclusively for the video cards/GPUs, and are not meant to be accessed directly by the CPU. In cases of integrated video on motherboards, they're not used in the first place. In case of video cards, they are soldered right on the video cards - video cards don't have slots b'cos then, you'll have video cards going into PCIe slots, and then the cards would have slots of their own, and then the height of the GDDR5 modules would potentially eliminate other motherboard slots that may be needed. In other words, configuration complications.

    Short answer: GDDR5 is not meant to be socketed on to motherboards

  14. Packaging != modular design by unixisc · · Score: 2

    CPUs - the last that I saw - were PGAs - except for embedded systems, I'm not sure of many BGA based CPUs. DDR3 onwards changed its packaging from TSOP to BGA due to excess pin count, where on TSOP, only a larger package would work, but then, the length of the wire bonds would become a factor in the speed of the sub-system (CPU to DDR3). Also, while TSOP is cheaper than BGA for lower pin counts, when the pin counts become comparable - ~50, the equation flips - BGAs become cheaper than 56 pin TSOPs. As a result, memory started getting packaged in BGA, which were already in use by flash memory, especially in portable applications such as cell phones, PDAs and so on.

    But even those DDR3 modules are available on sticks, even in their BGA packages.

    1. Re:Packaging != modular design by Shirley+Marquez · · Score: 1

      Most laptop CPUs are now coming in BGAs rather than PGAs. The push for thinner laptops is driving the change. It is also causing more laptops to come with memory soldered to the motherboard rather than socketed in SODIMMs.

      GDDR5 is not currently available on DIMM sticks. The high speed of the memory may make it impossible to package effectively that way.

  15. The point is MANTLE by ducomputergeek · · Score: 1

    I'm shopping for a new gaming computer on a budget. And even models shipping with this APU still usually have a R9 270x dedicated card as well, for a price point of about $850 USD.

    Where this gets interesting is if MANTLE gets widely adopted. Suddenly it can treat those 6 or 8 GCN nodes on the APU as additional GPU Processing power to be used in the queue. While maybe not as powerful as a second video card, it should give a boost in performance at no additional cost.

    Of course assuming game developers start using Mantle...

    --
    "The problem with socialism is eventually you run out of other people's money" - Thatcher.
  16. Obvious questions by zigfreed · · Score: 1

    Why couldn't this be a package-on-package chip in the current generation? The raspberry pi does this with it's graphics & system RAM; a redesign of the heat spreader and the chip package wiring would allow older motherboards to support it since the new pins come out the top, directly attaching to the GDDR5. Since AMD already sells DDR3 memory, selling GDDR5 with the chip would be business as usual.

    Regardless, motherboard manufacturers might still want to integrate the GDDR5 to sell the next generation like they did with sideport memory back when the HD3000 series was integrated in the northbridge.

    Third point: since build-your-own computers are commodity devices, traditional layout isn't prohibitively expensive -- splitting the CPU so it's more like a xeon phi and adopting a UMI interconnect is acceptible especially if it could sell on mobile.

  17. Less FUD please by Sycraft-fu · · Score: 1

    I get real tired of this "XBox one cannot do 1080p" crap that Sony/ Nintendo fans keep trotting out. Yes, it can. FIFA Soccer 14, Forza Motorsport 5, NBA 2K14, Need for Speed Rivals, all run at 1080 internally. Yes, a number of games run at less and I imagine that'll only be more common as time goes on, but it doesn't change the fact that the system is perfectly capable of 1080. Heck for that matter, not all PS4 titles run at 1080, BF4 being an example.

    I have no stake in this fight, I don't do console games I'm all PC all the time. I say "lolconsoles" to 1080p30, I do my gaming at 2560x1600 @ 60fps please and thanks. However don't go spreading misinformation, particularly technical information on a tech site.

  18. Chips by phorm · · Score: 1

    I've often wondered if it would be useful to have RAM in socketed chips (similar to a CPU) rather than on a stick.

    1. Re:Chips by SuricouRaven · · Score: 1

      That used to be common practice, back in the Z80 to 286 era. Stick form (SIMM, then DIMM) is just more convenient to work with. Look how small the pins are on a surface-mount chip - if those chips were big enough to socket, they'd be unweildy. Plus the stick form connects the whole bus at once, so it's fewer dust-prone connections to make than each individual chip together would have.