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

News Dev releases Radeon R9 Fury CU unlock tool

Discussion in 'Article Discussion' started by Gareth Halfacree, 6 Aug 2015.

  1. Gareth Halfacree

    Gareth Halfacree WIIGII! Staff Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    12,818
    Likes Received:
    2,036
  2. Jimbob

    Jimbob Member

    Joined:
    2 Jul 2009
    Posts:
    191
    Likes Received:
    1
    Nice, reminds me of the BIOS tools to unlock AMDs old triple core CPU parts.

    At the moment I should imagine yields of perfect chips are low on Fury parts as Fury X availability is still pretty low.
     
  3. Corky42

    Corky42 Where's walle?

    Joined:
    30 Oct 2012
    Posts:
    9,395
    Likes Received:
    334
    Hasn't most of the GPU being used in the Fury/X been in production for a long time?
    I thought the Fiji was an unscaled Tonga, that AMD's 28nm node was a fairly mature process.
     
  4. jb0

    jb0 Active Member

    Joined:
    8 Apr 2012
    Posts:
    502
    Likes Received:
    71
    Read as: "scarcity of high-end parts is artificially enforced to enable shameless price-gouging."

    Because let's be honest here... if the parts are the same, the costs for those parts are the same.
    The only reason to charge more for the fully-unlocked ones is that fewer fully-functional parts make it off the line, so they ARE actually harder to get. This is not just untrue if you're crippling fully-functional parts to make lower-spec hardware, it is COMICALLY obvious what's actually going on.
     
  5. Gareth Halfacree

    Gareth Halfacree WIIGII! Staff Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    12,818
    Likes Received:
    2,036
    That's one way to view it; the other way, and the way AMD et al would prefer, is that availability of low-end parts is artificially increased to enable those who can't afford to buy the top-end models to play. In truth, it's somewhere between the two: higher profits on the real high-end parts makes up for lower profits - or, in extreme cases, outright losses - on the low-end parts. You don't think that AMD - or any other semiconductor company - is selling you the parts for what they cost to make, do you? (Although, having seen AMD's most recent financial filings, it's not far off...)

    You may want to re-read the article, as I attempted to address this. Let's imagine a new semiconductor company, called MAD. MAD makes Redaon graphics chips, and it's just launched a new 22nm process node. As with all node shrinks, yields are god-awful: MAD is lucky if 10 per cent of the chips that make it out of the fab work as-expected. So, it launches a trio of products: the Redaon UltraMax, using those 10%; the Redaon Max, using the next 40; and the Redaon, using the final 50%. Everyone's happy: MAD doesn't have to thrown any of the chips away, and customers get a range of products suitable for all budgets.

    It's a year down the line. MAD's 22nm process node has had the wrinkles ironed out. Yields have vastly improved: now it's more like 10 per cent of chips that don't meet specs. Trouble is, demand for the Redaon Max and Redaon hasn't lessened - people still want 'em. So, what do you do? You tighten the binning process, 'rejecting' chips on an ever-tighter set of tolerances, and artificially limit them in order to meet Redaon Max and Redaon specifications - thus meeting demand for those parts.

    Now, you might think that's unfair - but let's look at MAD's alternatives. It could ditch the Redaon and Redaon Max lines altogether, in favour of selling the Redaon UltraMax at the same price as its entry-level stablemates. Whoops, MAD's profit margins just took a nose-dive; it's not getting those low-volume but oh-so-deliciously high-margin sales any more. Worse, its enthusiast customers just took their bats and balls home in a sulk: the card they paid £300 for a year ago is now available for £50.

    MAD could instead design an entirely new chip which lacks the extra features of the Redaon UltraMax - but that involves R&D costs and running a separate production line, which means the cost of the Redaon Max and Redaon just doubled. Adding insult to injury, it's still got those 10 per cent of parts that don't meet specs - but now it throws 'em away instead.

    The final option on MAD's plate is to ditch the Redaon and Redaon Max lines and keep selling the Redaon UltraMax at the high-margin, full-price RRP. Trouble is, it never really sells much of those; sure, enthusiasts like to think that they're the be-all and end-all of the market, but sales of low-end parts to budget-conscious customers and OEMs is where the real money is. With no entry-level and mid-range parts any more, these customers switch to MAD's rival Nveedeea - whoopsie.

    When viewing all these scenarios, you need to remember that producing the chip itself is one of the smaller costs involved in building a new GPU: the bulk of the costs were in research and development. Building one chip then crippling it to meet the demand for lower-spec and lower-cost parts is cheaper than doing the R&D on three separate chip designs - and, as covered above, neatly handles what would otherwise be 'wastage' from the production process.

    It's nothing new, either: back in the 60s and 70s IBM sold a bunch of mainframes with a hard drive in two size choices. You could pay a few tens of thousands of dollars, post-facto, to double the size: IBM would send an engineer out, he would undo a couple of security bolts, flip a switch, and redo the bolts. See, the drive already had two platters - but one was disabled via a physical lock-out switching inside the housing.

    Blimey, that was a bit of an epic, wasn't it?

    TL;DR: If you have a better way of doing business, AMD would love to hear it!
     

Share This Page