NAG Online > Technology > AMD launches three new Kaveri APUs

AMD launches three new Kaveri APUs

AMD Kaveri analysis header 800x450

Although AMD’s Kaveri family has been available to the public for some time now, that’s only been in the form of the A10-7850K and the A10-7700K, two unlocked quad-core chips that pack a lot of processing power, but locally don’t have the kind of price-competitiveness that is required to make any inroads against Intel. Internationally it’s a different story, as AMD’s APUs with or without discrete graphics do very well in the budget markets, with the FX CPU lineup doing particularly well as budget workhorses and tweakable rigs. Today AMD finally announced the launch of the A10-7800, A8-7600 and the A6-7400K. Hit the jump to see what the fuss is about.

The APUs on the table today are a bit different to the launch products introduced earlier this year. The A10-7800, A8-7600 and A6-7400K are all what AMD calls “configurable APUs.” What this means, really, is that you can plop these into any socket FM2+ motherboard and enable whichever thermal profile best suits your use-case. If you’re using a thin ITX chassis, switching to 45W mode for any of these processors will limit power draw to accommodate for extra heat into the chassis, while 65W mode is better for larger chassis and better coolers as well as power supplies which are rated for 150W and higher power draws.

AMD Kaveri family lineup 2014

The most impressive of the lot, arguably, is the A8-7600. It consumes half the power of the older A10-5800K but pushes out marginally better performance. In the 45W mode it’s also faster than the older A10-6700T, but pushes out more performance for the same power draw. AMD still has to work around thermal limits and this is why stock A10-7850K chips aren’t much faster than the outgoing A10-6800K. But there is the potential for better performance with lower power consumption when overclocked and that, together with the HSA-compliant hardware and GCN graphics, makes it a worthwhile consideration for a budget rig. It even can be a reliable workhorse if need be.

The A6-7400K is also interesting because it launches in the US and Canada today at just $77, very close to Intel’s new Pentium G3258 which is at $75. Both chips are fully unlocked, both can be overclocked to extreme levels and both can do this on boards that most people would consider “budget.” The A6-7400K scores because it’s not just limited to expensive Z87 or Z97 motherboards and it can deliver playable game performance at 720p.

Although it’s good news that AMD’s finally releasing these chips, I’m not sure what kind of price point they’ll reach when they reach our shores. With the A10-7700K already hitting the R2200 mark, the A8-7600 needs to be priced around R1400 if AMD needs it to be properly effective. pricing it at the same level as some of the more expensive Core i3 Haswell processors, or even at the same level as a Core i5 processor would be a foolish move.

Source: Anandtech

Tags:  ,
  • ShockG

    AMD needs to stop trolling us with dated technology and offerings. There’s nothing valuable in the APUs anymore. The poor CPU performance is ever more evident the better the GPU inside becomes. No amount of price cutting is going to save these products. Even from a TDP POV, not only does AMD calculate TDP based on average use as opposed to peak power draw, the platform that serves the APU is old and lacking in features. Global Foundries is years behind in process technology, trailing not only INTEL, SAMSUNG, TSMC, but UMC as well.

    • Wesley Fick

      I’ve been wondering how much the silicon process they’re making Kaveri on is holding them back. With 28nm SHP the GPU performance is pretty solid, but the CPU transistor switching is so much slower than 32nm SOI.

      And I really like the kinds of boards that are pushed out at the low-end on FM2+. If only the APUs themselves were more price-competitive and/or performant. There are times when I’m not sad about being stuck on AM3+ (still saving up for a FX-6300) and times when I am and I’d like to try out Kaveri instead.

      • ShockG

        INTEL houses the IRIS PRO GPU on a different die (same package) on the 4770R and all IRIS PRO enabled parts. That’s because process geometry needs are very different for GPUs and CPUs. There’s no sense in sacrificing CPU performance, TDP and performance per mm so you can pack in more GPU transistors. If you choose to do so, the given process must be specifically tuned for that kind of logic and in this case it is not. There’s no significant if any re-tooling that has taken place at GloFo since AMD owned the fabs. Being able to claim 28nm or 32nm doesn’t say anything these days as gate contact pitch, metal pitch, drawn pitch etc can all be different. Each company quoting the most favorable number. INTEL does this allot, but then again nobody else manufactures fin-fet products at such densities or volumes. Regardless, a GPU isn’t as sensitive to latencies like a CPU, with such deep execution pipelines, they are well hidden in it’s parallel nature. This is in direct contrast to CPUs that are very sensitive to latencies. Putting the two on a single die is no doubt the future, but I”m not sure if AMD did it because they wereare ready or something else was the motivation. Heat is a big issue for many and power draw. AMD is unfavorable in these two matrices because there isn’t much more performance sacrifice you can give to reduce it’s thermal footprint, where as you can with the INTEL products. cutting the 88W TDP of a 4790K for instance still gives you better performance than what AMD can deliver at 100W.


Login / Search

Latest games

Latest opinions