While CES is generally not a big show for the GPU divisions of the major chip makers – those firms preferring to make major GPU announcements at their own events – due to the fact that CES itself aligns so closely with OEM refresh cycles, we do see some quieter GPU announcements come out of the show every year. In recent years this has especially been the case for laptop/mobile graphics thanks to the larger volume and higher levels of integration as found on those parts, though this does have the unfortunate side effect of meaning GPU announcements get caught up in the deluge of announcements coming out of CES. At any rate with our first 2014 discrete GPU announcements in hand, we’ll be starting our look at CES 2014 dGPU news with AMD’s first mobile product unveilings.

AMD Radeon M200 Series

For their 2014 mobile GPU series, AMD’s mobile GPU branding is being caught up with their desktop GPU branding. On the desktop AMD launched their new Radeon 200 series of products in Q4 of last year, and now at CES 2014 we’re seeing the launch of the mobile counterpart of the Radeon 200 brand, the Radeon M200 brand.

Meanwhile in what has quickly become a CES tradition if there ever was one, the lack of alignment between GPU development cycles and OEM refresh cycles means that we’re going to be starting off the year with a combination of new SKUs and rebadges of existing SKUs. For their product refreshes OEMs need new products to show off – whether or not new SKUs are truly available – leading to component suppliers such as AMD and NVIDIA playing the rebadge game to satisfy their OEM customers. The end result is that we get some “new” products that aren’t really new, a situation we’re not big fans of, but something that we have come to terms with for the OEM market.

Unlike CES 2013, AMD hasn’t held a press briefing for these mobile parts, and accordingly this is a far more tempered rollout than the nearly top-to-bottom 8000M rollout last year. Still, we already have published information for 2 M200 parts, and partial information for a third. If history is any indication there will be some more SKUs (and more rebadges) to further flesh out this lineup, but this is what AMD is officially starting with for the moment.

AMD Radeon M200 Series GPU Specification Comparison
  AMD Radeon R9 M290X AMD Radeon R7 M265 AMD Radeon R5 M230
Was 8970M ? New
Stream Processors 1280 320
Texture Units 80 20
ROPs 32 4
Core Clock 850MHz 855MHz
Boost Clock 900MHz N/A
Memory Clock 4.8GHz GDDR5 2GHz DDR3
Memory Bus Width 256-bit 64-bit
FP64 1/16 ?
Transistor Count 2.8B ?
GPU Neptune (Pitcairn) Sun
Manufacturing Process TSMC 28nm TSMC 28nm TSMC 28nm
Architecture GCN 1.0 GCN 1.? GCN 1.0

Starting at the top we have the flagship of the group, the Radeon R9 M290X. The M290X is a straight-up rebadge of the 8970M (itself a minor variant of the 7970M), offering the same specifications and performance as the 8970M that came before it. This means we’re looking at a fully enabled Neptune (Pitcairn) GPU, with its 20 CUs and 32 ROPs running at a base clock of 850MHz with a boost clock of 900MHz. This is attached to 2GB or 4GB of GDDR5, running at 4.8GHz.

Since we’ll be starting 2014 on the same 28nm process node as 2013 (and 2012), there’s little surprise that AMD’s flagship hasn’t changed. AMD’s Tahiti GPU was too hot and power hungry for laptops and Hawaii is the same, so Neptune (Pitcairn) remains the largest GPU AMD can offer in this space.

Moving on to our second confirmed product, we have a new low end SKU called the R5 M230. The details on this one are a bit sketchier, as we know the functional unit counts and clockspeeds, but not the underlying GPU. In terms of performance this is a very low end part, featuring just 5 CUs (320 SPs), 4 ROPs, and a tiny 64bit DDR3 memory bus. AMD’s specs call for it to ship at clockspeeds up to 855MHz for the GPU and 2GHz for the DDR3 RAM, so shader performance and memory bandwidth should be competitive with the wider but lower clocked 8570M (Mars/Oland), but the ROP count is only half as much, which will hurt performance in pixel-heavy scenarios and higher resolutions. Functionally speaking this is a lower end part than any of the 8000M parts, and based on what we know about desktop Kaveri, unlikely to offer better performance than the iGPU in many Kaveri SKUs.

Meanwhile we also don’t have official confirmation on what GPU is in the R5 M230. Our initial assumption would be that this is another Mars (Oland) part, utilizing a cut down die. But AMD never offered such a product in the 8000M series. Furthermore Mars was 8 ROPs attached to a 64bit memory bus; while it’s easy enough to independently disable a CU, disabling ROPs without disabling part of the memory bus is an unusual endeavor. Sticking our arm into the rumor pool, we’ve heard rumors that AMD has developed another GCN GPU, mobile codename Sun, that’s even lower end than Mars. Seeing the number of functional units available on M230 this would certainly fit the description of Sun. For the moment we’re listing this as a Sun part, but we’re trying to chase down AMD to get confirmation of this.

Update: AMD has confirmed that the GPU is indeed Sun

Technical specifications aside, it’s not immediately clear what M230 is targeted at. With such a low performance profile, even in the best case it’s not going to be much faster than AMD’s iGPUs. At the same time it’s going to be rather big (relatively speaking) to pair with something like Kabini. It may be getting pitched as a cheap upgrade for Intel’s iGPUs, but that’s just a guess on our part.

AMD Roadmap Slide, Source: ComputerBase.de

Finally, we have a small bit of information on a 3rd AMD mobile SKU, whose full details have not been published yet. ComputerBase.de has a slide from an (unknown to us) AMD presentation that lists the R7 M265, which has yet to be posted by AMD. Since we don’t have specifications on this one we won’t spend too much time speculating on it, but there are really two ways this can go. Considering this replaces the 8700M series, this is either another Mars part, or we may finally be seeing a Bonaire part reach mobile. Heathrow (Cape Verde) was retired from AMD’s desktop lineup with the 200 series, and we’d expect the same to happen on the M200 series. In any case we’ll have more details on this SKU once AMD publishes some additional information on it.

Source: AMD Notebook Graphics Page

Comments Locked


View All Comments

  • ToTTenTranz - Wednesday, January 8, 2014 - link

    Ryan, I think you missed a possible purpose for the R5 M230.
    It's a contender to Crossfire with Kaveri and get a substantial performance boost in 3D.
    Even more now that Crossfire is exclusively made through PCI-Express and Mantle should give it increased control over frame pacing.

    Perhaps a performance boost that is substantial enough to hold its own against Broadwell GT3e at a much lower price point.
  • YuLeven - Wednesday, January 8, 2014 - link

    While frame pacing did improve in symmetric crossfire, APU + dGPU dual graphics is still so broken that instead of an improvement you take a performance penalty by enabling it in many titles.

    This substantial boost was of -18% on Tomb Raider 2013, last time I checked. Let's pray that Mantle change this. I've been hearing people saying "Drivers will fix it" since Llano.
  • ToTTenTranz - Wednesday, January 8, 2014 - link

    Yet this isn't new drivers trying to mix a VLIW4 GPU with a GCN GPU.
    This is new hardware, with both GPUs using GCN architecture, built with the new Crossfire in mind.
  • YuLeven - Wednesday, January 8, 2014 - link

    Dual graphics poor performance isn't about mixing VLIW4 and GCN. It performs far from great with a pair of VLIW4 graphic engines. It is a driver issue. New symmetrical bridgless crossfire is one thing, asymmetrical dual graphics is another beast.

    I won't expect much since llano, trinity and richland poor performance in this area. But, I truly want it fixed and operational. It would kill that unpleasant scenario that is enduring since 2011: Ins most cases, a cheap Intel CPU paired with cheap dGPU performs better in games for the same money of an AMD APU.
  • silverblue - Thursday, January 9, 2014 - link

    I imagine Kaveri might actually fix this, at least partly, considering AMD is spending most of its time supporting GCN rather than VLIW4/5.
  • Alexvrb - Friday, January 10, 2014 - link

    Err, I've seen quite a few benches that show it provides a big boost to FPS in many cases these days using Richland. Was Llano the last time you looked? Anyway, where they need work is frame pacing - and that remains primarily a driver issue. Unfortunately they're spending most of their frame pacing efforts right now on traditional crossfire, so it will probably remain an issue to some degree for the foreseeable future of hybrid crossfire.

    However, they've been making steady improvements to hybrid crossfire performance, and frame pacing is now very much on their radar across the board. I'm not saying that this is the generation that fixes everything, but I suspect there will be some improvement, and it will continue to get better with time. I would agree that one major use of this GPU would be crossfire with mobile Kaveri.
  • Khenglish - Wednesday, January 8, 2014 - link

    Really sad. I was hoping that they'd put a low voltage tahiti on a card with a 256bit memory bus, but they're doing a rebadge just like the 880m gtx:


    At least AMD isn't wasting a ton of money putting a ludicrous 8GB on the card though, but that will probably cost them sales to the uneducated.
  • extide - Wednesday, January 8, 2014 - link

    It's really Clevo putting 8GB on there, not nVidia...
  • Khenglish - Wednesday, January 8, 2014 - link

    I suppose, but still a waste of money that AMD is not doing (although 4GB is only useful over 2GB to run more than 1 game at a time, so you can say even 4GB is wasteful).
  • Alexvrb - Friday, January 10, 2014 - link

    At 1080p, maybe. But as more of these 4K laptops hit the market, 4GB becomes more useful. Especially in newer titles.

Log in

Don't have an account? Sign up now