AMD Trinity A10-5800K APU Review

[SKYMTL]

GPU Compute Performance: Photoshop CS6 / WinZip

GPU Compute Performance (pg.2)

Photoshop CS6

In our previous Photoshop CS6 benchmark, we deliberately disabled the GPU acceleration features so we could get an apples to apples CPU comparison. However, in this test, we are enabling that acceleration to see what affect it has upon the benchmark numbers. Please remember: Photoshop’s GPU features only improve performance on SOME (rather than all) editing tools so performance will not scale in a linear fashion due to the CPU still playing a role.

WinZip 16.5

As of version 16.5, WinZip now supports OpenCL acceleration for some of its file compression features. Unfortunately, it only works on AMD’s graphics processors for the time being. In this test, we compressed (using the maximum compression routine) a 2.6GB folder with random files and logged the time it took to complete.

RESULTS: Since our Photoshop test doesn’t focus solely on GPU acceleration, the Intel Ivy Bridge processor (which also supports OpenCL) still wins in this case. However, the A10-5800K is able to significantly lower its time to completion by using its graphics processor for certain tasks.

Winzip meanwhile ONLY supports AMD acceleration and represents a huge win for Trinity considering its popularity.

 

[SKYMTL]

IGP Gaming Benchmarks: 3DMark06 / 3DMark11

IGP Gaming Benchmarks

Integrated graphics processors have never been thought of as legitimate gaming devices but that stigma has begun to change. In this section, we test 3DMark06 and 3DMark11 to see how modern IGP systems reach to DX9 and DX11 environments.

RESULTS: Regardless of DX9, DX11, Intel HD3000 or HD4000, Trinity cannot be matched when it comes to integrated graphics performance. The HD 7660D is simply in another league here, regardless of the A10-5800K’s poor CPU performance within games.

 

[SKYMTL]

IGP Gaming Benchmarks: Deus Ex: HR / Dirt 3 / Skyrim

IGP Gaming Benchmarks (pg.2)

As with previous in-game tests, we have selected a number of games for our IGP-only testing suite. As you may expect, these benchmarks are run without a discrete card installed. All applications are tested at moderate detail settings (remember, these aren’t fully fledged discrete cards so we can’t expect miracles) in both DX11 and DX9 environments at a resolution of 1080P.

RESULTS: Once again, these tests tend to make a mockery of Intel’s integrated graphics processors. At times, the A10-5800K’s results nearly double the benchmark numbers set by the HD4000 within the i3 3225.

 

[SKYMTL]

IGP Gaming Benchmarks: Street Fighter IV / Torchlight

IGP Gaming Benchmarks (pg.3)

As with previous in-game tests, we have selected a number of games for our IGP-only testing suite. As you may expect, these benchmarks are run without a discrete card installed. All applications are tested at moderate detail settings (remember, these aren’t fully fledged discrete cards so we can’t expect miracles) in both DX11 and DX9 environments at a resolution of 1080P.

RESULTS: While these two games continue AMD’s dominance, there are some interesting things going on here. Street Fighter is one of the older titles in our gaming suite and it looks like Intel has some optimizations built into their driver stack for it. However, Torchlight was released last month and Intel’s latest drivers (dating all the way back from May) completely lack performance enhancements for it. This goes to show you how much importance driver development is in this day and age.

 

[SKYMTL]

The Effect of Memory Speeds on IGP Performance

The Effect of Memory Speeds on IGP Performance

On AMD’s APUs, the integrated graphics processor is linked directly to the system’s memory. Not only does AMD suggest that you have at least 8GB of memory to ensure there’s enough for Windows, applications and GPU use but they also recommends using 1866MHz memory for Trinity. The main reason for this is quite simple: the HD 7000-series cores within this architecture are based off of desktop parts that typically run memory speeds of between 1800MHz and 4000MHz. As such, anything lower than 1800MHz may cause some bottlenecking and lower than expected in-game performance.

In order to put AMD’s claims to the test, we benchmarked a number of applications at 1600MHz, 1866MHz, 2133MHz and a blistering 2400MHz as well. Timings were left at 9-9-9-25 1T and all other methodologies for game testing were left in place.

As we can see, the HD 7660D within the A10 5800K absolutely adores higher memory speeds with significant performance increases being realized across the board. However, framerates start to level out at 2133MHz and bumping speeds up to 2400MHz yields minimal benefits. In our opinion the best route for Trinity users is to buy a good kit of 1866MHz memory and overclock it as needed.

 

[SKYMTL]

System / IGP Power Consumption

System / IGP Power Consumption

Our power consumption numbers are broken down into two categories: one which simply stresses all of the CPU cores with WPrime and another which puts a high amount of load on both the CPU cores and the IGP. The latter will only be included if a given processor includes a dedicated internal graphics sub-processor.

For the CPU power consumption test, we use the standard testing system (with an NVIDIA GTX 670 installed) and wait until the system and discrete GPU are at idle speeds in order to log the idle power consumption. After this, WPrime 1024M is looped for 15 minutes while the power consumption is logged with a calibrated power meter to determine the peak watts.

IGP power consumption testing follows very much the same route as above but with some changes. First and foremost, the GTX 670 is removed and the video output is run through the processor’s graphics engine. In order to fully load the graphics cores and the primary processing stages within the CPU, we run the Unit Benchmark (in DX9 mode) from Civilization V for exactly 15 minutes.

Please note that after extensive testing, we have found that simply plugging in a power meter to a wall outlet or UPS will NOT give you accurate power consumption numbers due to slight changes in the input voltage. Thus we use a Tripp-Lite 1800W line conditioner between the 120V outlet and the power meter.

When placed in comparison to previous AMD architectures, the A10-5800K represents a good step forward from a performance per watt perspective. It manages to consume less power than an A8-3870K despite its higher clock speeds.

Against Intel processors, things tend to fall apart. AMD’s 32nm manufacturing process and relatively large die size contribute to significantly more power consumption than any of the Intel processors we tested in this review. This causes an issue for AMD since their APU architecture is well suited for small form factor chassis but SFF manufacturers desperately need to retain peak efficiency.

One small glimmer of hope rests in the idle power consumption numbers which are simply astonishing. It looks like AMD’s power gating and memory P-state expansion has served them well.

The IGP-only results follow almost perfectly in the footsteps of the previous CPU test but with some differences. The HD 7660D seems to be quite efficient but its extra performance does come with a price and it consumes nearly as much power as the previous generation’s HD 6550D. Nonetheless, it is great to see that AMD has augmented performance by leaps and bounds without sacrificing efficiency.

 

[SKYMTL]

Overclocking Results

Overclocking Results

With an unlocked A10-5800K APU in-hand and a cooling system backed up by one of the best air coolers available, we were under the assumption that overclocking would be a breeze. Unfortunately, that wasn’t the case this time around.

Typically increasing the clock speeds of a K-series chip is as simple as turning off Turbo Core / power saving features, upping the multiplier, fiddling with bus speeds and adjusting voltages to compensate for any instability. To an extent, this approach worked with our chip and we attained a moderate, nominal overclock of 4.52GHz with compete stability at reference voltages. Unfortunately, that’s where the fun stopped.

After hours of tuning and trial and error, we couldn’t get an ounce of stability above the 4.85GHz mark and benchmarks would only complete with the A10 running at 4.81GHz. For full stability, we finally settled at a clock speed of 4.77GHz which isn’t bad but it is far from the impressive numbers we’ve seen others achieve.

We do however want to say that even without going to the next level, AMD’s architecture does provide a relatively straightforward overclocking experience. The unlocked multipliers of these chips add value to anyone willing to take an hour of their time and put it towards some tweaking. Just make sure an upgraded cooling solution is purchased before setting out on any overclocking adventures.

While it may have been our chip that lacked overclocking headroom, we suspect the culprit was MSI’s A85XA-GD65 motherboar. Not only did it flat out refuse to boot Windows when AMD’s Overdrive was installed but GPU overclocking didn’t carry over into Windows and voltage adjustments were listed it outrageously cryptic terms. Oh, did we mention that its BIOS uses erroneous Intel XMP profiles instead of AMD’s own AMP? Supposedly, all of these items will be (or have already been) fixed in subsequent BIOS updates so we will be revisiting overclocking with this board in the near future. Let’s hope things have improved.

We included some comparative performance testing below:

 

[SKYMTL]

Conclusion

Conclusion

Before we dive further into the challenges facing AMD’s Fusion strategy, let’s take a closer look at what the A10-5800K was able to achieve, and where it fell flat. In terms of raw central processing tasks Trinity is a mixed bag. Its Piledriver CPU cores have allowed for a drastic clock speed improvement over Llano but that ~30% bump hasn’t translated into significantly better performance. The A10-5800K is at best 15% faster than the previous flagship APU and in some cases is notably worse than an A8-3870K due to Trinity’s lack of legacy instruction set support. Newer applications do tend to highlight Trinity’s potential but performance justifications for the removal of legacy instruction sets are few and far between.

One of the main critiques leveled against Llano was its poor single thread performance and Trinity’s Piledriver cores have allowed for improvements in this area. The move towards next gen CPU cores also benefits gaming performance where the A10-5800K is able to stay well ahead of the A8-3870K, and by quite a bit in most cases. Despite better in-game framerates than Llano, Trinity still looses to Intel’s similarly priced Ivy Bridge and Sandy Bridge processors in every title, making the A10-5800K a poor choice for gamers using discrete GPUs.

While CPU performance may see the A10-5800K trading blows with -and in many cases leading- Intel’s offerings, its graphics subsystem is a game changer. Elitists may scoff at anything short of a $150 discrete card but the mere fact that AMD’s integrated solution can play DX11 games at mid to high detail settings (at 1080P no less!) gives credibility to Trinity’s ultimate goals.

The HD 7660D is impressive in other respects as well. Its ability to excel at parallelized workloads is one of the cornerstones of AMD’s Heterogeneous System Architecture strategy and with Trinity that vision is finally becoming reality. As a result, certain programs have their performance boosted by significant margins due to the HD 7660D’s mere presence. APUs from this generation could also make excellent HTPC options with their inclusion of advanced image processing techniques, a robust driver stack and native HD outputs.

Speaking of drivers, this is one area where AMD holds an ace up their sleeves. Intel’s slow software release cycles just can’t keep up with game releases, as evidenced by the HD4000’s poor showing in Torchlight. Meanwhile, AMD is able to effectively leverage the expertise of a strong graphics division in order to stay well ahead of the development curve. As both companies move on to upcoming generations, this fact alone could become a major differentiator between products, with AMD continuing to hold an upper hand unless Intel does something drastic.

The success or failure of AMD’s core HSA fundamentals hinges upon the notion that software support will eventually catch up with the hardware being provided. Up until this point development hasn’t been moving forward all that quickly and we’re sure this causes AMD no small amount of headaches since the entire raison d’être of Fusion is a seamless harmony between CPU and GPU tasks. But new development IS happening and we’ll surely see additional steps towards true integration as upcoming APU generations are released.

This brings us back to Trinity. For the time being, it doesn’t really stand out since software that can properly utilize its forward-looking architecture is either in its infancy, still under development or anything but seamless. In the few instances where programs can natively harness the A10-5800K’s integrated parallel processing abilities (MuseMage, WinZip, Photoshop and a few others) the design pulls ahead by leaps and bounds. Those applications alone will make nearly anyone an APU convert in the blink on an eye.

Despite all of its benefits, we still struggle to understand where the A10-5800K falls into the current market. For HTPC users, small SFF systems and the market’s new “slim and light” desktops, its 100W TDP is far too much, despite spectacular idle power consumption. Those segments would be much better served with one of the 65W Trinity variants. Meanwhile, the 5800K’s $129 price tag may seem appealing to anyone looking for a budget gaming CPU but once again, there are better options from AMD’s own camp not to mention Intel’s offerings. True entry level gamers may be salivating due to great IGP performance but this still leaves us with a largely orphaned product that isn’t well suited for any one situation and is merely “good enough” to not get overwhelmed by the competition. There is however a glimmering ray of light: the A10-5800K is able to stay well ahead of Intel's i3 3225 and its Sandy Bridge equivalent in nearly every test that uses newer instruction sets.

There are those who understand the fundamentals of AMD’s APU push and those that don’t. When looking at the A10-5800K’s benchmark numbers without an open mind, you’ll likely see processor that lacks performance consistency which is paired up with some impressive graphics capabilities. However, there is much more to this equation than a strict separation of CPU and GPU tasks. In many ways Trinity is a step in the right direction and its architecture represents a giant leap ahead for the viability of a truly heterogeneous computing environment. We can’t wait to see what the next generation will bring.