With recent in-OS fixes, game patches, and a great many games being designed around their architecture, AMD haven't been doing too badly when it comes to their CPUs when compared to how they were before. This is great for all end users, as a stagnant market with little to no competition stunts progress and hitches prices up.
Though their standalone CPUs have been dwindling in popularity, their Accelerated Processing Units (APUs) are way ahead of the curve. Pairing a relatively low powered processor with a decent on-die graphics chip is perfect for a HTPC on a budget. The iGPU can deal with HD video playback with no problem, and it's even able to play low powered games at decent framerates, albeit with lowered settings.
And so we get to today, with AMD's new APU range, 'Richland', successor of Trinity. We get to take a look at Richland's top dog, the A10-6800k, and see how it compares to it's predecessor, the A10-5800K, and also Intel's best match, the i3-3220. This APU comes with four fully overclockable 'Piledriver' cores at 4.1GHz, a memory controller capable of handling 2133MHz memory, and an AMD Radeon HD 8670D for its graphics solution. The updated architecture is said to excel in applications which utilise OpenCL, so we'll be factoring that in with our tests.
We've been praising APUs for a while now, so it's time to see how the new kid on the block settles in.
Richland is still based on the same 32nm manufacturing process as Trinity, but with a beefed up core and a stronger graphics solution there doesn't seem to be anyway for this not to be a great release. The high end chips, the A10s, are treated to a stronger memory controller which, when paired with high speed RAM, should help boost the capabilities of the iGPU.
This latest range is to be released with many different APU models, but as ever with AMD's aggressive pricing it's not going to break your bank to purchase the model that we have with us today.
Intel i3 3220 Gigabyte Z77 UP4TH
4GB Kingston HyperX 1600MHz @ 1333MHz
AMD Stock Cooler
Mushkin Joule 1000W
Windows 7 x64
It should be evident that we're testing a realistic scenario in terms of the motherboard, RAM, and cooling solutions used for our setup. This is to showcase the real world performance and characteristics of the hardware in the spotlight, to benefit those wanting to invest in such a solution.
As this review shall be looking at the stock results of the A10-6800K, as well as the alternative solutions being used for comparison, all hardware on show will be tested as if it were just taken out of its box and turned on. No overclocking, no XMP profiles, no other optimisations that would deviate from the stock results.
The frequency being shown is one of the 6800K's Turbo steps, which goes up to a maximum of 4.4GHz. When fully stressed, the chip operates at 4.1GHz over all cores.
Power Consumption at the Wall
Power consumption is something often overlooked when talking about custom systems, as performance is typically paramount. However, when we're talking about an all in one CPU/GPU solution to be used in media PCs as well as low-power gaming systems, power consumption means a lot.
If the system's going to be on for extended periods of time, such as a media PC used to transcode video files for future playback, having an efficient system can save you a lot in the long run.
It is clearly evident that Intel's i3 processor is more frugal at the mains, sipping less than 60% of the power used by the 6800K when being stressed by OCCT's Linpack AVX stress test. This is of course a worst case scenario as the CPU is unlikely to face such extreme stresses in every day use, but even when running Unigine: Valley's Extreme Preset, a benchmark not unlike many modern games in terms of stress levels, the Intel system uses 50% of the power of Richland's top dog.
When playing back a HD MKV format video file with VLC, the AMD chips are still using 10w more power than the i3.
It should also be noted that the Richland processor is consistently using more power than its Trinity predecessor - a product perhaps of the higher performing cores on the 6800K.
As well as overall power draw, the temperatures reached by these CPUs are important to note. Both stock coolers used for the testing were ran at a constant 12 volts from the power supply, and both were at a noise level which would be unacceptable for a living room area where HTPCs are often used. As such, the temperatures yielded by our testing should give an insight over whether fan control could be used to keep the noise down, or if an aftermarket solution would be required for a peaceful experience.
The AMD processors are far, far hotter than their Intel counterpart. Even with the heatsink's fan running at 12v the AMD CPUs were nearing 100°C when the CPU was put under heavy load, while the i3 was sitting at a respectable 54°C. During HD Video playback the temperatures became a lot more manageable, as the 6800K maxed out at 68°C.
With these results it should be clear that if CPU intensive tasks are commonplace on your HTPC you would need to purchase an aftermarket cooling solution to keep both A10s in check, or else put up with the loud noise from the stock fan.
AIDA64 is a benchmark suite containing a plethora of tests for the whole system. For our needs we singled out the tests that represent the performance of the CPU and the memory.
A win for the 6800K as the new chip surpasses the performance of both the 5800K and the i3 in most tests. We're not quite sure why our i3 reacted so poorly to the AES tests, but we've double checked our figures and that is what it scored.
For the memory tests we can see that the 6800K is faster than the 5800K in every way, albeit by a small margin. However, the run away winner in terms of memory is clearly the Intel i3, whose results show a stronger memory controller used on the chip.
Maxxon's Cinebench provides an insight into how well the CPUs would fare when tasked to render complex images, as well as how the GPU deals with rendering a scene using OpenGL.
The results clearly show that the graphics solution of the AMD chips is far superior than the i3's in this scenario, while the Intel chip does well to keep up with its opponents in the CPU test despite lower clock speeds.
We see that the 6800K loses marginally to the 5800K in the OpenGL render test, which is strange seeing as the 8760D should trump the 7660D.
POV-Ray is another benchmark focused on how well the CPU deals with complex image renders, and we can see that the 6800K does pull ahead of the Trinity chip, as well as just beating the i3.
Media Transcoding Using Handbrake
A common task for media PCs to undertake is the transcoding of various video files, whether to change the format to one more desirable or as a means of compression. For our tests we have employed the popular software known as Handbrake to transcode three separate files. As previously mentioned, Richland boasts an impressive boost to applications which employ OpenCL, which Handbrake is included. As such, we're expecting good things from the 6800K.
The small file comes in at just under 100MB and is 1080p. This test is to show how the CPUs will deal with small video clips that can be taken by most modern smart phones.
The larger files come in two distinct flavours - a 2GB standard definition video to demonstrate how non-HD footage is processed, and a 1080p MKV file weighing in at almost 5GB.
Together we feel we have covered the most common usage scenarios.
The 6800K comes out triumphant in all three tests, which is a promising result for the CPU. When processing the large MKV file the Richland chip was more than 1,000 seconds in front of the i3-3220, and 650 seconds in front of the 5800K - an impressive result by and large.
Video conversion is an extremely strenuous ordeal for a processor, but it's something that a HTPC may very well be tasked with. As such, the x264 Benchmark gives a great insight into how each processor would deal with such a work load.
The Richland processor trades blows with the i3; AMD's chip edging ahead in the second pass average while the 3220 just wins with the first pass. The results also show the clear improvement over the 5800K, though there's still not all that much in it.
wPrime, a hugely popular benchmark of a CPU's ability, shows us much of what we've seen already. The 6800K beats the 5800K pretty handily, but it can't quite catch up with the i3. With more physical cores and a higher clock speed, this proves that architecture plays a big role too, and that Intel's is superior when tasked with wPrime's workload.
We ran the three processors through a plethora of Sandra's tests, ranging from the raw arithmetical performance of the CPUs to how they deal with media transcoding.
Throughout the arithmetic tests we can see that the i3 and the Richland A10 constantly compete for first place, while Trinity's offering trails a little behind.
The Multimedia and Cryptography results show the i3 struggling with many of the test's aspects, resulting in the 6800K and the 5800K coming out on top.
For this part of Sandra's test we focus on the processor's ability with Financial Analysis, as well as the overall power management of the chips.
Where the i3 triumphs with the Binomial results, it falls behind to Richland when faced with the Black-Scholes portion of the test. The way the results are weighted mean that Intel pulls ahead as the overall winner of this test.
We see familiarly mixed results with the processor efficiency tests, with AMD's offering miles ahead with overall Power Management and the efficiency of the ALU, but are then dwarfed by Intel's multi-core efficiency.
Processor Multi-Core Efficiency and Power Management
We can see with the Video Shader Compute results that AMD leave Intel in the dust. The integrated graphics of the APUs thwart the HD2500 by 400% - a result which shall be shown consistently throughout the rest of the review.
Moving on to the two other results on show and we see a complete reversal, with the i3 demolishing the APU's results. We believe that the A/V transcoding favours the Ivy Bridge chip because of its QuickSync support, which rockets the i3 into the lead. With the memory bandwidth tests we see that the 3220's memory controller is superior to the two A10s, with the same RAM producing radically different results. This should be an area for AMD to focus more on, as greater memory bandwidth should do nothing but boost the APU's graphical performance.
Video Shader Compute
Media Audio/Video Transcode
Futuremark's PCMark benchmarks are great for testing the overall performance of a system. Tests include processor performance, web browsing, spreadsheets and image manipulation to name a few, and the results should indicate how well the processors are affecting the overall system.
PCMark Vantage gives us a mixed bag of results. Though the overall PCMark score follows the same trend as other benchmarks, with the Richland APU just beating the 5800K, and the i3 trailing behind a little more. However, when we look at the Gaming and Productivity results we see that the i3 trumps the other chips.
The Richland and the Ivy Bridge processors are pretty much neck and neck here, with the 6800K just taking the lead. We also see one of the biggest improvements for the 6800K over the 5800K, with the Trinity chip trailing a good 450 points behind its successor.
No real surprises here as we move onto the GPU-weighted section of our review. Both APUs are showing the massive difference in performance between the Radeon graphics solution and Intel's HD2500. The results also show a clear increase in performance from Trinity to Richland, which is a good thing for the guys in red.
Similar results with Vantage as the Intel continues to fall behind. Interestingly enough the 3220 refused to complete anything above Vantage's Performance test, leaving the Higher and Extreme tests to the AMD team. Between the two APUs we again see the 6800K edge in front, though not by as substantial a margin as before.
And the same again for 3DMark 11's results. Intel's onboard graphics solution is leaving much to be desired here.
Alien vs Predator
Moving on to Alien vs Predator, a benchmark based from the real game, we see far less difference between the two A10 APUs. Intel once again is far behind, showing that gaming on the HD2500 would be nigh on impossible.
The benchmark runs at the same resolution as your system, or 1080p in our case.
Counter-Strike: Source is one of the most popular games out there, and it uses one of the most common game engines too. This should mean that our results from this benchmark can be extrapolated to other Source based games, such as Team Fortress 2 and the Half Life series.
With the in-game benchmark completely maxed out at 1080p we see great results from the APUs, with the Richland offering yielding an impressive 65FPS. So if you're into your Source based games then the APUs may be just the thing for you.
One of the most popular and highly rated games to be released this year, the new instalment to the Bioshock series is heralded for its graphics. As such, it only made sense to include it in our testing.
We used two different quality presets to do our run-throughs; Low and Normal, both at 1080p. In our results we can see that, same as ever, the AMD's are fighting each other while the i3 limps on behind.
One of the newest arrivals to the world of benchmarks, ALLBenchmark's Catzilla is already a popular option for those looking to test their systems. With a dubstep sound track and mutant cats there's not much that can beat Catzilla in terms of entertainment.
Moving on to the results and we're only able to compare Kitty and Cat for the APUs, while the i3 point blank refused to run the benchmarks at all.
From what we can see the Trinity processor is once again being beaten by its successor, but there's not much in it.
Showdown is one of the games often bundled with an AMD graphics card, so it makes sense for the APUs to perform well. Running at 1080p with Medium settings, the two APUs perform well no matter the MSAA settings being used, with only a ~5FPS drop in frames by going from 0xMSAA to 8xMSAA. To some, playing games at around the 35FPS range would be perfectly fine, but for those that require a solid 60FPS you'll need to further turn down settings and reduce the resolution.
It must be noted that the i3 is still massively behind the APUs, and it could not run the game with 8xMSAA.
SimCity has been with us since March this year and is often bundled in with AMD's A-Series of APU. With this in mind we should be seeing great performance from the APUs . . . which we do. With the game maxed out at 1080p, with Lights & Shadows set to Low, both APUs sit at just under the 25FPS range. Unlike fast paced first person shooters, SimCity should not need to be played at a constant 60FPS to be enjoyable. If more frames are needed, settings are to be turned down and resolutions reduced.
The 6800K only just beats the 5800K here, with the Trinity APU having a slightly higher minimum FPS count. Once again, the i3 could not open SimCity whatsoever at these settings.
Another game from one of AMD's bundles and another clear win for the APUs. We completed six separate runthroughs to include a variety of settings and resolutions - 1080p, 720p, Lowest preset, Medium preset, 0xAA, and 8xAA.
As is evident from the results, the i3's HD2500 really is lacking in gaming performance, though it's nice that the benchmarks ran at all.
Once again we see a marginal performance increase throughout the board for the 6800K when compared to its predecessor. The gain's there, but it's not as big as we feel it should be.
In the above results we can see that the 6800K's maximum FPS is inordinately high, and we're classing it as an outlier to our results.
Yet another game which is affiliated with AMD through their bundle deals. We ran two instances of the game's built in benchmark; one using the Normal Preset, and the other using the High Preset with both at 1080p.
Looking at the average FPS we see a small improvement from Trinity to Richland, with both being miles ahead of the i3.
The results to the Sleeping Dogs benchmark, built into the game itself and ran at 1080p with medium settings, show us what we've been seeing already. The i3 is falling behind, the APUs fighting for first place with the Richland iteration slipping ever so slightly ahead.
Resident Evil 5 DX10
We've been using the Resident Evil 5 fixed benchmark for a while now and so it only made sense to include it in these testings. We used Medium settings at 1080p and took the results for both 0xAA and 8xAA.
Nothing we haven't seen before here, with the new kid taking the lead from the 5800K. Slightly more of a difference than we've been used to seeing with the other game benchmarks, but it's good to see that Richland is beating Trinity.
Nothing much to talk about with the i3 which, as ever, performs dismally. It is worth noting that the i3 seems to be affected most by Anti Aliasing being turned on.
Resident Evil 6
Resident Evil 6's benchmark utility is DX9 only, which is strange when its predecessor has DX10 support. We ran the benchmark with the Medium preset over 1080p and 720p, both showing familiar trends. Another victory for Richland, another commiseration for the i3.
Unigine's Heaven benchmark is a favourite for anyone wanting to test their graphics solution.
With the Basic preset we see an interesting result - the Trinity APU takes first place, with a 6 point lead over Richland. In saying that, the 5800K's minimum FPS leave a lot to be desired, getting close to the result given by the i3.
With the Extreme preset the 6800K takes a two point lead over the Trinity. The difference is there, but it should be more than that when you consider the claimed improvement to Richland's architecture.
Another benchmark offering from Unigine, with this one being new to the party. Much like the Basic preset on Heaven's runthrough, we again see a performance decrease when going from Trinity to Richland. It's an odd one, but the fact that the result's been repeated twice shows that there's something that needs to be addressed here, as the 6800K should not be losing to its predecessor.
Going to the Extreme preset we see Richland take the lead once more, managing to score 7 points more than the 5800K for a completely underwhelming 2.86% improvement. Really nothing to get excited about, unfortunately.
AMD came to us in 2011 with their APU lineup to fill a gap in the market that not many realised was there. Fast forward to present day and, on the performance front, they're still going reasonably strong. For the price that this chip is being released at the performance has been pleasantly surprising, beating the Ivy Bridge i3-3220 in most scenarios. Not too shabby at all for a product whose main focus is not on the CPU, but instead the integrated graphics solution.
In saying that, it does seem as though AMD have focused more on improving the CPU side of their new range, with a 300MHz boost to stock clocks, Piledriver cores, an improved memory controller, and a stronger focus on OpenCL. Both the Handbrake media encoding and the x264 benchmark have seen the benefits from the OpenCL support, which is great for those of you who transcode video files. AMD's Piledriver cores employ two modules containing two cores each, with each core sharing some resources with its module pair. Unlike Intel's i3, this makes the A10 a fully fledged quad-core processor. Many dismiss the architecture as being weak, but there's no denying that the 6800K does beat the other processors in the majority of our CPU oriented tests.
On to the iGPU and we see that the 8670D on board the Richland APU beats both of its opponents throughout our tests, though we don't see as much as an improvement as we'd like over last generation's Trinity solution. Synthetic benchmarks seem to show more of an improvement than those sourced from actual games, with results from the relatively new Hitman Absolution game, an AMD title nonetheless, showing barely any improvement over Trinity's 5800K. This trend is found frequently throughout our results, with next to no increase in frame rates over last year's tech. It is possible, of course, that future updates will prove beneficial for Richland, but we must report on the results that we're getting at present. From the get go it was obvious that the i3's HD2500 offering was no where near the performance of the APUs, which is a great spot for AMD to find themselves. Intel simply have no real answer to this section of the market.
Following on from this it must be said how impressive the gaming performance of both APUs are. Being able to play games at decent framerates on an integrated solution is remarkable, and it's often easy to forget that there's no dedicated card being used. For those that care less about graphical settings and more about the gameplay itself, the AMD chips may be the perfect match for you. This is such a far cry from how things were just a few years ago that it's not too far fetched to believe that the casual gamer has little need for a video card anymore.
Though the general performance of the 6800K is good, we think it's prime time for AMD to focus more on the power efficiency of their APUs. The difference in performance between Intel's i3 and the 6800K in terms of the CPU isn't all that much, and yet the i3 uses significantly less power. This is not great for a HTPC, as such systems are often left on for long periods of time, sometimes 24/7. With this kind of usage expected the extra power consumption is definitely a bad thing, though there will be many of you who won't care that much.
The power efficiency does lead into another issue, and that's the temperatures. The extra power used by the 6800K is creating heat - a lot of heat. When OCCT is used to stress the CPU AMD's stock cooling solution doesn't stand a chance, with the core reaching temperatures of 97°C. Though such stresses aren't going to be experienced to such a degree in real life, applications such as Handbrake which transcode media files can certainly turn up the heat. We ran our stock cooler at a constant 12v to take any variations in motherboard fan profiles from effecting results. Even with the cooler manually set to run at maximum we still experienced temps up to nearly 100°C, showing that if you want your system for CPU intensive tasks OR low noise you will have to invest in a good aftermarket cooler to tame the heat that can be produced.
All in all we find that AMD's biggest competition for an on-die gaming solution with the 6800K is the 5800K. Tiny improvements in gaming scenarios lead us to recommend those already owning a Trinity APU to stay put. For those of you who are thinking about buying an APU for a low powered gaming system for the first time, it'd be worth checking the price differences between the two A10s in today's tests. If the 5800K undergoes a serious price drop then, to us, it still seems to be the best option.
For those who're not interested in gaming on the APU, but instead more interested in media playback, the two A10s may be a little more than you need. There are other less power hungry options which can play movies just as well, and these should also have less issues with heat.
After much deliberation, we give this Richland processor the OC3D Value for Money award. For the price that it's being released the gaming performance is great, and the CPU performance is steadily improving, though not by much over last year. For the people who can still enjoy gaming at lower settings and resolutions this APU is a fantastic option.
Thanks to AMD for supplying the A10-6800K. Discuss your thoughts on the OC3D Forums.