AMD’s not-really-new Radeon RX 500 graphics

And lo on the 18th day of the fourth month, verily did AMD wheel out its latest but not necessarily greatest new graphics cards. I give you the new AMD Radeon RX 580, 570, 560 and 550 – precision engineered to smite the evil Nvidia. Except they’re not really new. Instead, they’re basically dirty old re-badges of existing graphics cards. That doesn’t automatically mean they’re not of interest. But it does mean we’ll have to wait a little longer for something really revolutionary from AMD.

In all candour, you can gauge the novelty of AMD’s new boards by the fact that I haven’t made great efforts to bag one for some hands-on action. There’s not really any point. That’s because they are for the most part only very slightly tweaked versions of the existing Radeon RX 400 family.

Take the new Radeon RX 580. It has exactly the same number of those pixel-prettifying stream processor thingies (precisely 2,304) as well as texture units (144) and render outputs (32) as the old RX 480. Not a huge surprise given they’re both based on the same 14nm AMD Polaris 10 graphics chip. The new RX 580 doesn’t even have faster graphics memory.

Ye olde RX 480, yesterday

What is does get is a slightly faster core clockspeed. The baseclock jump looks reasonably impressive, leaping from 1,120MHz to 1,257MHz. But it’s actually the Boost clock that these things usually run at in games and that has only ticked up from 1,266MHz to 1,340MHz. Yes, that’s six per cent faster, which is the most you can hope for in terms of a performance uptick. In reality, given the memory speed has remained static, you’ll be lucky to get even that.

Indeed, such is the similarity between the two cards that it’s said people are flashing the video bios on RX 480 boards to ‘turn’ them into RX 580s. This, of course, isn’t necessarily a good idea. Firstly, that’s because a standard RX 480 has only a single six-pin supplementary power connector and was already bumping up against the power spec of the PCI-E x16 slot. Ramping up the power consumption could put both your video card and motherboard at risk.

The 580’s higher clocks also make for much higher operating temps on the video board itself. That goes for not only the graphics chips but also the components that regulate the chip’s power supply, known as VRMs. Again, running a board not designed for those hotter temps may end in tears. That said, the fact the video bios update is possible and the board will reportedly function for a time, at least, emphasises just how closely related this new RX 500 family is to the old RX 400 boards.

Single power connector possibly problematic when hacking your 480 into a 580

Anyway, the RX 570 has a similar relationship with the old RX 470, albeit the 570 does get a very small memory clock boost, to boot. Where things get a tiny bit more interesting is the RX 560.

That’s based on the smaller AMD Polaris 11 GPU. As it happens, AMD never shipped a fully enabled version of that chip for the old RX 460. That’s allowed AMD to fully switch on Polaris 11 for the RX 560 and thus the stream processor count jumps from 896 to 1,024. It also gets slightly faster clocks, all of which adds up to (at a guess) 15 per cent or so additional performance – potentially significant in an entry-level card.

The final part of the puzzle is pricing. By the official numbers, the RX 580 and RX 570 are about 10 bucks cheaper at $229 and $169 than the 480 and 470 were at launch last summer. Again, the 560 looks like the most compelling proposition with a $20 price drop to $99. Pu a ‘£’ sign in front of those numbers for UK prices.

That said, I really wasn’t wild about the old RX 460 and honestly I can’t see this revised 560 board changing that. Save up for the 570 if at all possible, is my advice.

The old 460 was a bit of a bummer, the new 560 might be better

In the meantime, we’re left waiting and wondering about AMD’s increasingly fabled Vega graphics chips, which is a properly new family of GPUs which should include a really high performance version as well as generally improving AMD’s graphics performance at ever price point.

Vega is now late enough for legitimate questions to be pondered regards its health. I imagine the mostly likely candidate for the delay will be production problems at Global Foundries. That’s the chip production outfit that used to be part of AMD itself.

GloFo has only recently started to produce PC graphics chips as opposed to CPUs, and the high-end Vega GPU will be a much bigger chip than it is accustomed to making. Maybe it’s having problems producing enough working chips or getting the clock speeds up. Whatever, at this point I’m losing faith that AMD will pull the wraps off a new Vega GPU that can beat everything Nvidia currently offers.

Not that many of us will care about giving the likes of the Nvidia Titan Xp a bloody nose. If Vega is good enough to force prices down at the Nvidia GTX 1070 and 1080 level, that’ll be fine by me. As ever, we are all rather reliant on AMD keeping both Nvidia and Intel honest. So fingers crossed for Vega – it should finally appear in the next few months.

From this site

21 Comments

  1. Booker says:

    “Vega is now late enough for legitimate questions to be pondered regards its health. ”

    What? Weren’t they saying “June at the earliest” for ages now? So how can it be late? Is it August already?

    btw: The RX 580 has significantly lower power consumption during video watching, second monitor and so on. I wouldn’t call that nothing. For people who never bought the 480 (but were thinking about it) this is still really interesting. I’m probably going to get one.

    • Asurmen says:

      Yeah, I don’t get this at all. Vega has been 1st half/2nd quarter since, well, forever, and everything still looks like that’s the case, with a late May/early June launch.

    • Jeremy Laird says:

      Should have been out by now by the timetable AMD itself showed early last year:

      link to images.anandtech.com

      That shows Polaris mid 2016 (and it did indeed launch in June 2016) and Vega at the beginning of 2017. Moreover, I very much doubt the plan was to leave Nvidia with no high end competition for a full year.

      I suspect the original plan was to get Vega out the door in 2016, but that has slipped. Indeed, you’ll find reports of a mooted October 2016 launch circulated last year. It slipped such a long time ago that, yes, it’s been some time since they adjusted the launch window.

      So, it’s gone from October last year, to ‘early’ 2017, to May / June 2017. I think it’s fair to say it’s running fairly late.

      • Sakkura says:

        It isn’t the most super-detailed timetable, but yeah.

        Maybe the HBM2 delay messed with their plans.

        • brucethemoose says:

          HBM don’t work out too well for AMD, did it?

          They invested all that R&D in the HBM1, only to for it to make Fiji expensive to produce and limit its memory capacity more than anything. A little while later, their competitor is raking in the cash from an HBM2 Tesla line their R&D partially paid for (from alot of long-term orders basically), while it screwed over the timetable for their next big consumer launch.

      • Asurmen says:

        Those October reports were only ever rumours.

        So actually based on the only official thing, it’s a slip of one quarter that’s actually been known about for a few months.

        • Jeremy Laird says:

          As above, you only have to look at AMD’s past product schedule to conclude that Vega is quite late. It’ll launch about two years after Fury, a card which itself was hardly early to market.

          Moreover, sometimes the rumours are true. I’d be surprised if October last year, or perhaps even earlier, wasn’t the original plan.

          • Asurmen says:

            Quite late by whose standards though? Not sure if you can class something as quite late, if it’s being delivered at or more or less on time as they intended.

          • hpstg says:

            Since the publicized investor meetings of more than a year ago, and even the original Ryzen Preview, we know that Vega would be 1H 2017. It will most likely be in Computex and fully launched around June, as is tradition.

    • Sakkura says:

      There were once rumors about it launching in Q4 last year or maybe even Q3.

      It’s possible AMD changed their plans and Vega’s going to drop those 6-9 months later than originally intended.

  2. Asurmen says:

    Bring out Vega already.

  3. fish99 says:

    There’s a 580 less than a year after the 480?

    • bananana says:

      Why wouldn’t there be, if the retailers have paid them for the RX4xx cards? It’s not like it’s much more than a slight optimization, at most 10% faster than RX4xx. You can probably get the same results by overclocking.

  4. ravenshrike says:

    The entire purpose of the 500 series was to position themselves to cover every aspect of the low to mid end market so when Vega comes out there are no gaps in their product line.

  5. Pizzzahut says:

    You really gotta wonder about AMD sometimes.

    “Here’s the successor to our four year old 400 series! And the advantage!? Nothing!!! Please buy our stuff.”

Comment on this story

XHTML: Allowed code: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>