1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Hardware Is 2016 a make or break year for AMD?

Discussion in 'Article Discussion' started by Combatus, 23 Jul 2015.

  1. Taua

    Taua What's a Dremel?

    Joined:
    20 Sep 2014
    Posts:
    87
    Likes Received:
    0
    I'm a huge fan of "good enough", however AMD have fallen so far behind that using an APU over an Intel CPU/discrete card, in form factors which allow for this, lands on the wrong side of the line. Good enough does not mean making huge sacrifices to upgrade prospects and initial performance for the sake of a £20 note. A Pentium anniversary has an upgrade path ending in the hefty 4790k, and a discrete GPU will be changeable over the years. Buy cheap, buy twice. It's a cruel irony that it can be more expensive to occupy a financially limited position than not, but given the differences in price here, there is no excuse to chose the vastly more limited system, in both initial performance and future prospects.

    Perhaps it would be for the best if AMD spun off the GPU division, which is still potent though cash-starved and subject to poor management - the Fury X is a lovely itx sized flagship. The only value I can see in the CPU arm, no pun intended, is the x86 license. Consumer software has not become hugely multi-threaded, as some dreamed it would, nor has the fact all 3 next gen consoles use AMD hardware resulted in PC games performing better with AMD hardware, it's a much more complex equation that that.

    My first CPU was AMD, as was my second, but AMD have been stunted ever since core2duo replaced the ghasty pentium 4 based pentium Ds, and that was years ago. As I said in the first paragraph, being poor is expensive, AMD's CPUs simply can't recover as-is, it is not possible, Intel utterly dominates now, even AMDs trick-shot APU is now in danger of being replaced by an Intel low end CPU with iris pro graphics.

    Intel played dirty and won, but it's not like AMD haven't tried pulling fast ones, they hyped bulldozer up, moar coarz0rz, got people buying the motherboards, and then revealed the abysmal 8150. Ebay told the story of how that was received, and the aging 2500k when juiced up is still better than anything AMD have for general purpose/gaming. There's just no competition.

    And that's partly why we're still looking at 4 core mainstream CPUs from Intel heading into 2016, when the q6600 came out at the start of 2007, which is ironically AMDs only hope, that CPUs just aren't the limiting factor they used to be, and the push is to make them even less limiting via software, not brute CPU hardware. We probably all have more powerful CPUs than we really make use of baring 'content creators', but we all frequently max out our GPUs, pushing them as hard as possible as that workload scales better, in terms of power and the flexibility of programming around the differences of GPU power available in the platform the software is being run on as opposed to hitting real-time limits in CPU performance that to adjust would alter the software in probably unworkable ways.

    I personally think VR is going to shake things up, as if a reset button has been been hit. 3 way monitor setups are going to be quaint relics of how people used to sit down infront of computers, like when we see a mainframe operator gleefully entering code on a 10" monitor as monoliths of silicon and wirelooms moot round them.

    This whole sitting down at a computer has got to change - powerful computing unit hidden somewhere linked wirelessly to a lagless VR headset, sensors for everything, massive information processing to recreate environments, truly natural input/control methods...rambling now, but putting too much stock into how technology is currently used, and just thinking "more of that" is not really seeing how it will be used, how it will innovate to evolve.
     
  2. bionicgeekgrrl

    bionicgeekgrrl Minimodder

    Joined:
    1 Oct 2009
    Posts:
    223
    Likes Received:
    7
    A point to remember is that AMD provide the chips for the Xbox One, PS4 and iirc the Wii U?, so they certainly have a steady contract there for a number of years.

    Top end performance cards/chip isn't where the profits are made either it should be remembered, as the majority of cpus/gpus will be sold to the lower power markets.
     
  3. rollo

    rollo Modder

    Joined:
    16 May 2008
    Posts:
    7,887
    Likes Received:
    131
    The main chip income comes from servers that's where Intel makes the bulk of there money. AMD abandoned this a long with other things.
     
  4. SchizoFrog

    SchizoFrog What's a Dremel?

    Joined:
    5 May 2009
    Posts:
    1,574
    Likes Received:
    8
    Should only buy the best? Yes, you are blind. Go read my posts again... you know, the ones where I suggest people buy a Pentium K and a 750Ti instead of a high-end AMD APU, or where I suggest buying better hardware second hand.

    Yes I buy for my own priorities... you say you are thinking of 'others' and yet you are constantly criticising when people buy low end Intel.

    I'll go right back to the beginning and repeat my very first comment...
    'I really hate it when people bring out the 'Intel will rip us all off if AMD goes' argument as if that somehow makes AMD's situation or their products better.'

    At what point can we have a discussion about AMD and their failings and positives WITHOUT pointing the finger at Intel?
     
  5. rollo

    rollo Modder

    Joined:
    16 May 2008
    Posts:
    7,887
    Likes Received:
    131
    To answer the question

    YES it's a big 12 months for AMD screw it up and they won't get another chance. At least not as the company is today.

    If they survive that long there's fresh buyout rumours doing the rounds.

    They need to release some accurate benchmarks no screwing around give the people the truth and we may buy it.

    The 2 biggest mistakes AMD have made in recent years abandoning servers and abandoning the enthusiast market. They are going to try and correct one of them.

    AMD fans can not really defend those 2 points in any argument I've seen about AMD. Yes they are good enough for most but on tech websites very few give a crap about been just good enough. They want the fastest they can afford and it's not AMD.
     
  6. Corky42

    Corky42 Where's walle?

    Joined:
    30 Oct 2012
    Posts:
    9,648
    Likes Received:
    388
    I'm not jumping to AMD's defense, but did they really have any choice in leaving the servers and enthusiast market? If the microarchitecture that you've built your chips on causes them to run *slow and hot there's not much you can do about that until you do a total re-design.

    *In comparison to your competitors offering
     
  7. exceededgoku

    exceededgoku What's a Dremel?

    Joined:
    19 Oct 2006
    Posts:
    75
    Likes Received:
    0
    Intel is treading carefully at the moment to NOT devastate AMD's business by being too competitive as that would be an absolute disaster for them.

    They'd be hit with anti competitive restrictions, probes in to price fixing, etc.

    AMD needs a good CPU to compete on a level playing field but just doesn't have the cash to achieve this.

    Leaving the server market was a big hit in my opinion. Speaks leagues when all of my servers are Intel powered, I wouldn't even entertain AMD. Infact we do have a NAS that has an AMD (FX8350), but anything greater (e.g. DB, Directory services, etc.) is put on a powerful CPU and in most cases mainly because if I was to buy a cheap CPU my MD would ask why we're not getting the best...
     
  8. Corky42

    Corky42 Where's walle?

    Joined:
    30 Oct 2012
    Posts:
    9,648
    Likes Received:
    388
    Developing a new CPU microarchitecture takes more than just money, even small tweaks to existing microarchitecture takes time.
     
  9. Huw_Dav

    Huw_Dav What's a Dremel?

    Joined:
    17 Jul 2015
    Posts:
    1
    Likes Received:
    0
    The Bit-Tech i7-5700HQ review of 07 July contained comparisons to an FX-8350 and i7-4790K running at 4Ghz. AMD are talking about a 40% better performance, but if you multiply or divide the FX-8350 scores by 1.4 it is still well behind the i7-4790K.
    Where AMD are depends on how efficient and expensive to build their "Almost i5-2500K" is. If a 4Ghz version has a 65W TDP (with 2KGaming/4K 2D Graphics) and is profitable at £80 retail they have a medium term future. If the thing consumes 125W and needs to be sold for over £200 they might as well throw in the towel now.
    Assuming that their new chips are financially viable, they should perhaps concentrate on speeding up data flows over the motherboard . If they were also to develop an inexpensive SAS microprocessor, the better hdd performance/lower cpu overhead should allow their complete computers to compete on many workloads, and (with a Linux partner) get them back into at least the SME Server/Shopfront arenas.
     
  10. rollo

    rollo Modder

    Joined:
    16 May 2008
    Posts:
    7,887
    Likes Received:
    131
    They been using a rehash of the same one since bulldozer launched that's what 6 years or something.
     
  11. rayson

    rayson Damn sure it was legal

    Joined:
    23 Jul 2010
    Posts:
    238
    Likes Received:
    0
    isn't having a x86 license a big thing wouldn't it benefit large companies like Samsung or some other company with decent foundries to buy up amd, I'd imagine it because a lot of companies expect amd's share prices to go even further.
     
  12. Corky42

    Corky42 Where's walle?

    Joined:
    30 Oct 2012
    Posts:
    9,648
    Likes Received:
    388
    By the time Zen hits the shops pretty much yea, although that's nothing unusual or exclusive to AMD.

    Intel have been using the same Core series microarchitecture, all be it with lots of refinements, for almost a decade now, before that the NetBurst microarchitecture was in use for 6 years, as was the P6 microarchitecture before it.
     
  13. rollo

    rollo Modder

    Joined:
    16 May 2008
    Posts:
    7,887
    Likes Received:
    131
    X86 licence is not transferable
     
  14. ZeDestructor

    ZeDestructor Minimodder

    Joined:
    24 Feb 2010
    Posts:
    226
    Likes Received:
    4
    Some very good points and arguments in this thread, but I think everyone here is missing the key point that's been killing AMD: the server market. Sure, the consumer market is importnat and all, but you're comparing $20-$200 margins vs $600-$4000 margins (top-end CPUs cost over $6000 in modern times)

    Fact is, since IVB-EP/EX launched, AMD has had nothing, and I mean NOTHING (It's so bad places like AnandTech who review server CPUs/platforms didn't even bother comparing to AMD for Haswel and instead went all out to source IBM POWER8 machines to compare with) that could compete with Intel.

    Sure, the CPUs are cheaper, but (when new) the full platforms cost near enough the same (RAM and disk being the bigger factors), then you have the power costs. Over the usual 3-6 year cycle, it all balances out to costing about the same, but with Intel either having an edge in performance or power (or both), plus they resell better on the 2nd-hand market, and if you're wise with your timing you can make back a nice gaming PC worth of cash off of an old server.

    Right now though, Intel has more cores, more cache, far more IPC, and you can configure any LGA2011-3 server to start with a miserly 4core non-HT CPU all the way up to an 18-core HT-enabled monster of a CPU, so tell me, how would you in good conscience justify buying anything but Intel for your servers? Thanks to that, Intel pulls in billions in revenuew from servers, while AMD has pretty much nothing.

    So for all our sakes, watch Zen, and pray it can compete with Haswell-EP/EX, because if it can't, we're not at the mercy of Intel, we're at the mercy of Intel and IBM, and they both want their pounds of flesh at the high-end. Should Zen fail, I can genuinely see AMD falling by the wayside as yet another HW company casualty in the face of faster-moving competition.
     
  15. ZeDestructor

    ZeDestructor Minimodder

    Joined:
    24 Feb 2010
    Posts:
    226
    Likes Received:
    4
    Not quite actually... Intel used P6 in various modified forms right up until Nehalem (Core/Core2 is a modified P6 with multicore support added mostly at the Northbridge level and a fancy cache controller allowing two-core shared cache).

    But with Penryn out of the way, Intel decided the platform needed revamping, so they applied a three-part strategy:

    1. Move the memory controller/multiCPU half of the Northbridge into the CPU, which they did with Nehalem and QPI
    2. Move the rest of the northbridge (the fast IO (wide PCIe by then, previously AGP)) into the CPU (mostly through Lynnfield, then iGPU as well with Clarkdale)
    3. Redesign the core and put P6 back to rest, which they did with Sandy Bridge.

    Thus over 3 generations, Intel essentially completely rebuilt their entire platform, both core and IO. This is also why the improvements have essentially stopped overnight: Intel can't find easy ways to improve performance without harming server performance (servers these days care much more about total CPU throughput than clockspeeds, so big high core count CPUs scale better than high-frequency ones) anymore.

    Sorry for double-post, but I felt it was necessary to keep seperation of topics.
     
  16. StoneyMahoney

    StoneyMahoney What's a Dremel?

    Joined:
    10 Jul 2009
    Posts:
    287
    Likes Received:
    13
    Here's a graph of Intel vs AMD x86 processor market share since Q1 2004. Two things worth noting. Firstly, the Core architecture came out in 2006 and AMD's market share works it's way down to the lowest figure so far in Q2 2015. Secondly, the most recent quarterly results show a significant up-tick in AMD's CPU-selling fortunes. There have been a few small upticks in the past, but not one this big. Let's hope it's enough of a boost for AMD to survive long enough to get Zen out the door.
     
  17. SchizoFrog

    SchizoFrog What's a Dremel?

    Joined:
    5 May 2009
    Posts:
    1,574
    Likes Received:
    8
    Market share is well and nice but it is the 'bling factor' added to impress and is ultimately non-informative. What matters is profits and this is where AMD fail big time. You could have a company 3x the size of Intel, 100x... but if they lose money then eventually they will collapse.

    The next most important thing is what you do with those profits to make sure (as best you can) you continue to make profits further down the line. Again this is where AMD fail as for a decade now they have made bad choices over and over.

    Zen may be a good thing, maybe it will be more of the usual over-hyped mediocrity from AMD, time will tell, but it will take a long road map of improvements to turn things around for them and I just don't see that happening.

    It's a like a massive game of Top Trumps. AMD doesn't need the 'perfect' card and it isn't like Intel has perfect cards either, but it does seem that whenever AMD gets a turn, they happen to pick the wrong category each time and fall short of Intel's offerings and the same seems to have been true on the GPU side of things up against nVidia.
     
  18. maverik-sg1

    maverik-sg1 Minimodder

    Joined:
    18 Aug 2010
    Posts:
    371
    Likes Received:
    1
    Reasons I see for some failures at AMD - fabs not being able to keep up with Intel on a manufacturing size (28nm parts versus 14nm parts I mean c'mon!!) - Bulldozer being a turd of a cpu (which might be because most software doesn't take, or rather, isn't optimsied to it's areas of strength).

    GPU business is not far from the mark (drivers withstanding) just some efficiencies needed.

    To succeed Zen needs to be 14nm, fast and efficient, it doesn't need to be the fastest, just close enough.

    If the onboard ZenGPU has good performance then it could compete in the mainstream (read as 1080P 60fps gaming on popular titles).

    Of course, if the cpu overclocks like a biatch, then enthusiasts will love it too.
     
  19. Corky42

    Corky42 Where's walle?

    Joined:
    30 Oct 2012
    Posts:
    9,648
    Likes Received:
    388
    I would disagree, if it's just close enough that won't be a compelling reason for people to buy it.
    It needs to bring something better to the table whether that be faster, lower power draw, better IGPU, or whatever, it needs to have a selling point other than just being cheaper.
     
  20. maverik-sg1

    maverik-sg1 Minimodder

    Joined:
    18 Aug 2010
    Posts:
    371
    Likes Received:
    1
    Peeling one part of my post takes the context and contorts it slightly - That's disappointing, but okay:

    Close enough isn't a compelling reason true, however, if (as you say) it brings superior APU/GPU capabilities (which I also state in my thread above), which I also state greater OC potential (with scalable performance) or some other interesting party trick (which I also state).

    My point is being cheaper with some unique selling points is how the market share comes back to you, price hikes can come over time as confidence returns to the consumer......although there's no denying that true overall speed sells best with greater profit.
     
Tags: Add Tags

Share This Page