AMD new processor shootout

AMD recently refreshed their Phenom II and Athlon II lines with six new processors.  Of course it wouldn’t be any fun to just tell you about new processors without showing off what they can do, now would it?  We didn’t think so either, so today we’re bringing you three of the new processors: The Phenom II x6 1075T, Phenom II x4 970 Black, and Athlon II x4 645.

The Phenom II x6 1075T, as its name implies, is a six-core processor is positioned right between the flagship 1090T and 1055T in both price and performance.  Like its siblings, the 1075T uses the Thuban core and includes AMD’s TurboCORE technology (similar to Intel’s Turbo Boost).

The Phenom II 970 Black is a speed-bumped revision of the 965 Black.  It boasts a stock speed of 3.5GHz and, as with all of the Black Edition processors, is not multiplier locked.

Finally the Athlon II 645 is the new top end of AMD’s value quad core processors, thanks to a bump to 3.1GHz.  The main differentiating feature between the Athlon II’s and Phenom II’s lines is that the former lacks any L3 cache.  As you’ll see in the benchmark results, this makes a difference in a few real-world applications.

Test Setup

Benchmark Results

Synthetic benchmarks

Synthetic benchmarks are a great way to isolate specific parts of a system to see how it could potentially perform.  It is also useful to keep in mind that these tests don’t always reflect real world performance.

The more cores and threads available to 3DMark’s CPU tests, the better the score.  All three of the AMD processors scale almost perfectly along core count and speed lines.  Extrapolate the Phenom II data out to the equivalent of an eight-core 1075T, and you’ve got the i7 870.

SiSoft Sandra’s CPU Arithmetic benchmark tests integer and floating point performance.  Whetstone is the integer component, and Dhrystone is the FPU component.  Both are multithreaded and test in 32- and 64-bit modes.   Whetstone shows pretty much the same performance scaling as 3DMark.  Dhrystone, on the other hand, seems to prefer core speed to core count, with the Phenom II x6 falling behind the faster-clocked x4.  The Intel processor is disproportionately ahead of AMD’s offerings here.

The Multimedia tests really play to AMD’s strength as can be seen in the charts.  Again, core/speed scaling is the same, but this time the Phenom II x6 is able to at least edge out the Core i7 870 in all the tests.  Even the x4 970 is able to keep up with the i7 processor in the iSSE4 test.

Moving on to Sandra’s Memory Bandwidth and Latency testing, we see that AMD’s integrated memory controller performs very uniformly as far as bandwidth is concerned.  Latency is a bit of a different story with the Athlon II for some reason.  Maybe it’s the lack of an L3 cache, but it lags pretty far behind the Phenoms.

wPrime32M takes advantage of a CPU’s multithreading capabilities.  Generally, the more threads a CPU has, the better it will perform.  That being said, the Phenom II x6 holds the slightest of leads over the hyperthreaded quad core i7 870 despite the Intel processor handling two threads more than the x6 1075T.  And just like before, the numbers scale to perfection for all three AMD processors.

Applications

WinRAR is coded to take advantage of multicore processors, and it definitely shows here.  The six-core Phenom II completes our test faster than the other AMD processors, but is no match for the hyperthreaded Core i7 which can handle eight threads.  The Athlon II lags far behind the other processors, likely due to a combination of slower speeds and lack of L3 cache.

The MP3 Encoding test shows single-core performance, and it’s a close race for the performance-class processors.  The similarly clocked x6 1075T and i7 870 are in a dead heat, while the faster-clocked x4 970 edges everyone out.  The Athlon II x4 645 clearly lags behind the other processors.

H.264 encoding shows a much tighter race between the x6 1075T and i7 870.  In fact, it’s close enough to call a tie.  The x4 970 Black performs quite well, scaling right in line with the speed/core difference of the x6 1075T in the DVD and 1080p tests.  720p encoding shows a much different story though, with the x4 970 Black taking a very slight lead over everything else.  Surprisingly the Athlon II x4 645 follows this pattern as well, following the exact same scaling for speed and core count.

Games

The CPU doesn’t matter as much for gaming as the GPU.  This has been the case for some time, and it’s definitely reflected in the results below.  As you’ll see, if a game produces playable frame rates with one CPU, all of them will produce playable results.  Our GPU for this test is a Radeon 4850, an average graphics card by today’s standards but still good enough for demonstration purposes.

A little background information is needed for our Crysis: Warhead (simply referred to as Crysis from here on out) testing.  Crysis has three major settings that can be tweaked for overall performance: resolution, antialiasing, and shader mode.  The shader modes offered are, from lowest to highest, Performance, Mainstream, Gamer, and Enthusiast.  Our testing has shown that independent of resolution and antialiasing levels, moving from Performance to Mainstream mode results in a 45% drop in frame rate, while moving from Gamer to Enthusiast mode provides a 33% reduction in frame rate.  Since those reductions were so predictable, only the Mainstream and Enthusiast modes are provided to show the best and worst case scenarios.  The other modes can be reliably calculated from the results provided.

Crysis is able to demonstrate what happens to a system when the GPU is pushed to its limits.  As you can see, the game is quite playable in Performance mode regardless of resolution or AA level (Mainstream is fine as well).  Gamer and Enthusiast modes aren’t anything you’d want to try with the test system’s GPU, but it does still reinforce the fact that all of the CPUs tested perform about the same.

Tom Clancy’s H.A.W.X. produces some of the largest frame rate differences between CPUs (ten percent at its worst), but even the slowest of the bunch is more than sufficient to enjoy this game with the eye candy turned all the way up.

Warhammer 40,000: Dawn of War II shows some differences between CPUs, but again the results are all playable and the performance delta is 15% or less in every case.  Again, even with our older video card, this game is playable at 1920×1080 with all the settings cranked up regardless of the CPU used.

Heat and Power

All of the AMD processors we’ve tested are built on a 45nm process.  The Phenom II x4 970 and x6 1075T both have a maximum TDP of 125W, while the Athlon II x4 645 is rated at95W.  Our Intel Core i7 870 is also built on a 45nm process and has a max TDP of 95W.  Despite this, the only processor with a higher idle temperature than the Core i7 is the faster clocked Phenom II 970 Black.  When it comes to load temps though, AMD’s processors run far cooler.

Also of interest is the power draw required by both platforms.  Both of our Phenom IIs draw more power at idle, but once a load is applied to the CPU the story becomes different.  Only the Phenom II x6 1075T doesn’t draw less power.

Overclocking

The Phenom II line, especially the Black Editions, are known to be very overclockable.  The new batch of CPUs is no different.  The Athlon II doesn’t do so badly either, but it probably could have done much better with a voltage bump.

As always, every CPU is different, so your results may vary.

Pricing and Competition

Phenom II x6 1075T ($239.99).  Its closest-priced Intel CPU is the i7 860 at $279.99, which is still $40 above the 1075T.  Dropping down in price we have the i5 760 at $208.99.  Unfortunately neither one of these processors was available for testing – our only available Intel processor is the i7 870 ($289.99) from the Socket 1156 test bench, but it’s close enough that a comparison is not unfair.

Phenom II x4 970 Black ($180.99).  The closest quad core Intel CPU here is the i5 750 at $194.99.  If you really want a price equivalent from Intel, you’ll have to drop to the dual core i5 655K ($184.99) or i5 650 ($179.99).

Athlon II x4 645 ($120.99). Intel’s offerings here seem to be a bit underpowered in that they are both dual core processors.  The i3 550 is $129.99, and the i3 540 is $114.99.

Conclusion

As for recommendations, the Athlon II x4 645 would make an excellent choice for an HTPC.  Either of the Phenom II processors would work well for high-performance and/or multithreaded tasks such as video encoding.  All three of the tested CPUs were great in our gaming applications as well.  The overclocking crowd will also be happy with the potential of the new Phenom II processors.  A 20-33% speed boost on air cooling is no small feat.

AMD’s goals are to deliver choice, compatibility and a great price/performance ratio.  The recent CPU refresh certainly fits those goals. The newest top-end Phenom II x6 CPU is able to keep pace with a much more expensive Core i7.  The Phenom II x4 970 Black holds its own pretty well too.  The Athlon II, while not performing as well in real-world applications, still presents an excellent choice in a budget performance system.  On top of price/performance, all of AMD’s offerings run at very low temperatures even under load.

Bottom line? AMD has put out a solid product refresh.  They continue to provide an excellent value and exceptional performance at their various price points, and are deserving of the Icrontic Stamp of Approval.