AMD Trinity Desktop APU - Graphics Performance Review
AMD Trinity Desktop APU - Graphics Performance Review
In May this year, AMD officially launched its Trinity Accelerated Processing Unit (APU) that features the next generation APU architecture. It is built on a mature 32nm SOI process technology, using a second revision of the Bulldozer architecture. At its launch, AMD only released mobile versions of Trinity APUs that will power ultrathin AMD platform based notebooks.
Now, AMD has launched the desktop versions of the Trinity APU with its flagship A10-5800K leading its charge to bringing its Heterogeneous System Architecture (HSA) to desktop computing this time round. AMD's previous generation desktop Llano series APUs, that were launched last in 2011, failed to make inroads with its touted ability to leverage on both the CPU and GPU portions of the APU to execute tasks in parallel for a balanced system architecture. It wasn't much of a technical failure as it is a logistics failure because its availability to key partners was too limited to make any impact. This time round with this Trinity update, AMD claims that its HSA can run much more efficiently than ever before, as the CPU and GPU of the next generation APU work in tandem once more to run desktop applications with improved compute power and at lower power consumption. Can AMD maintain inventory to satiate its partners and keep users interested? We'll probably get a better idea of this situation over the next few months, but for now, we'll concentrate on its technical capabilities.
In our initial preview last week, we looked into the most important aspect of the new AMD Trinity desktop APU - its graphical performance prowess; in particular, its performance levels for both DirectX 10 and 11 games. We also conducted a quick Blu-ray playback test to see how efficient its new Radeon graphics core can offload its CPU core for this HD audio-visual test.
In this review, we've updated graphics performance results and we will also be delving into its implementation of the AMD Dual Graphics technology that allows users to pair a suitable Radeon HD graphics card with the APU in a CrossFireX configuration that will give a boost to graphics performance. AMD has already listed the recommended graphics processor to pair with each Trinity APU and for our A10-5800K, it is the Radeon HD 6670 GPU. The general rule is that the performance discrepancy between the integrated and the discrete GPUs should be minimized in order to get the best performance out of this technology.
Although we do not have the price point of the flagship A10-5800K (at the point of publication), we do know that it will be priced to compete directly with the third generation Intel Core i3-3220 processor that can be bought from local retailers at ~S$165. Hence, without further ado, we shall put the flagship Trinity desktop APU through its paces in our integrated graphics performance suite. For those interested in the full line-up of Trinity APUs available, here's how they compare:-
(Base / Turbo)
|L2 Cache||AMD Turbo
|Radeon Cores||GPU Clock
|DDR3 Memory Support|
|A10-5800K||HD 7660D||100W||4||3.8GHz / 4.2GHz||4MB||Yes||384||800MHz||
up to 1866MHz
|A10-5700||HD 7660D||65W||4||3.4GHz / 4.0GHz||4MB||Yes||384||760MHz||up to 1866MHz|
|A8-5600K||HD 7560D||100W||4||3.6GHz / 3.9GHz||4MB||Yes||256||760MHz||up to 1866MHz|
|A8-5500||HD 7560D||65W||4||3.2GHz / 3.7GHz||4MB||Yes||256||760MHz||up to 1866MHz|
|A6-5400K||HD 7540D||65W||2||3.6GHz / 3.8GHz||1MB||Yes||192||760MHz||up to 1866MHz|
|A4-5300||HD 7480D||65W||2||3.4GHz / 3.6GHz||1MB||Yes||128||723MHz||up to 1600MHz|
Using AMD's flagship A10-5800K Trinity APU, we are able to put its Radeon HD 7660D graphics core to the test and determine how decent its performance numbers will be. For comparisons, we've put together integrated graphics performance numbers that we've gathered previously from the Intel Core i7-2600K and Core i5-2500 that feature the Intel HD Graphics 3000 engine as well as the top-of-the-line third generation Intel Core i7-3770K processor and its Intel HD 4000 graphics core. We have decided to include our past results with a NVIDIA GeForce GT 220 (512MB) discrete graphics card to ascertain just how close or far off the performance of integrated GPU's capabilities against a low-end graphics card.
Since the GeForce GT 220 is only DirectX 10 capable, we have decided to include a couple of low-end DirectX 11 graphics cards such as NVIDIA GeForce GT 520 (courtesy of an MSI N520GT-MD1GD3/LP with 1GB DDR3) and an NVIDIA GeForce GT 440 (using an ASUS ENGT440/DI/1GD5 with 1GB of GDDR5). The reason for including two cards is due to the anemic performance of the GeForce GT 520 as it grossly under-performs even when it is up against the GeForce GT 220. As such, we brought something a step ahead of the old DX10 card which happened to be a GeForce GT 440.
We also tried downclocking the Intel Core i7-3770K to the level of the Intel Core i7-2600K in order to create a level playing field of comparison for the Intel integrated graphics performance comparison. However as we found out, all of our tests are so GPU constrained that the difference in CPU speeds didn't matter one bit - the differences were too negligible. As such, the results shown are with the default CPU clock speed.
AMD A10-5800K Test System Configuration
- ASUS F2A85-M Pro Desktop Board (AMD A85X FCH), BIOS version 5104
- 2 x 2GB Kingston HyperX DDR3-1600 (CAS 9-9-9-27)
- Radeon HD 7660D (AMD Driver Version 8.982.0.0)
- Western Digital Caviar Black 1TB SATA 6Gbps hard drive (one single NTFS partition)
- Windows 7 Ultimate (64-bit)
Discrete Graphics Card (used for AMD Dual Graphics and Discrete Graphics Testing)
- Sapphire Radeon HD 6670 1GB GDDR5 (Catalyst 12.8)
More Test Details
Given that the AMD A10-5800K Trinity desktop APU supports upto DDR-1866MHz officially (just like Llano), we did want to run it at this speed, but for some reason we couldn't get the board working stable to support this memory speed. From our tinkering, it did feel like a BIOS related issue that should hopefully get better in time to come. As such, for now we've set the clock frequency at 1600MHz in order to proceed with our performance testing. In our earlier foray with the ASUS F2A55-M (AMD A55 chipset) motherboard, we were unable to operate the system with our usual test timings; however, with our new ASUS F2A85-M Pro (AMD A85X chipset) board, we were able to use our usual tighter memory timing settings. As such, we've updated the graphics performance results which did vary slightly.
For further analysis, we also conducted gaming benchmarks on its Dual Graphics performance. According to the reviewers' guide, the recommended discrete GPU to work in tandem with our new Trinity APU is still the AMD Radeon HD 6670. Hence, for our analysis for the latest iteration of the AMD Dual Graphics technology, we included the Sapphire Radeon HD 6670 1GB GDDR5 graphics card to run in tandem with our AMD A10-5800K Trinity desktop APU. In order to ascertain the performance boost provided by the AMD Dual Graphics technology, we conducted the same benchmark tests with our Sapphire Radeon HD 6670 graphics card as a stand-alone discrete card on the same test system (without CrossFireX). Out of curiosity, we attempted to pair our Trinity APU with reference cards from the AMD Radeon HD 7000 series; however, both the Radeon HD 7770 and HD 7750 failed to work in tandem with our A10-5800K under the auspices of the AMD Dual Graphics technology. This is surprising considering that the GPU model within the APU is supposed to be equivalent to the discrete edition, yet they do not cooperate. We'll get more input from AMD on this matter.
As mentioned earlier, the direct competitor of our test APU is the Intel Core i3-3220 processor, an entry-level Ivy Bridge CPU. Its test configuration is the same as that of the third generation Intel Core i7-3770K processor found here. The details of the older comparative test setups for the Intel Core i7-2600K and Core i5-2500 systems are found in our original review of the Sandy Bridge processors. Just for kicks, we've even included results from the first generation Core processor's original Intel HD Graphics engine. Note that for all three generations of Intel HD Graphics engines tested, we've set aside 256MB graphics frame buffer. There is no other granular option available other than to select maximum system memory, but it didn't make much of an impact.
We have included AMD's desktop Llano APU - the A8-3850 processor which has a built-in Radeon HD 6550D GPU engine, to see how this previous generation APU stacks up against the current one. The AMD A8-3850 is definitely doesn't have the same amount of compute power as the AMD A10-5800K even though both are targeted at the mainstream user group - in fact they cater to two different spectrums of mainstream DIY needs. Despite their compute discrepancies, we thought it would be great to see how much stronger discrete-class integrated GPU of the Trinity APU is against the older Llano APU. You can find the test setup of the Llano desktop system here but as a point to note, our testing of Trinity's Radeon HD 7660D graphic core was paired with a 512MB frame buffer, double that of all the integrated graphics engines tested from AMD and Intel prior to today's test (256MB). This shouldn't affect the compatibility of our results because we've tested higher frame buffer sizes for Llano previously, but they've not shown improvements at our test settings. In any case, the results with our test settings will show that even more horsepower is required before we can tax these integrated graphics engines further.
The following is the full list of benchmarks used to test the integrated graphics and the current implementation of the AMD Dual Graphics technology. The Blu-ray playback testing was omitted for our analysis of the AMD Dual Graphics technology:
- 3DMark Vantage
- Far Cry 2
- Battlefield: Bad Company 2
- Blu-ray Playback Testing (Black Snake Moan, Superman Returns) using PowerDVD 10 (ver 2308)
So read on for the performance breakdown of the new Trinity desktop APU in the form of the flagship AMD A10-5800K!
Integrated Graphics Performance
Futuremark 3DMark Vantage
For this DX10 benchmark, the AMD A10-5800K came in second and it was just trailing by a 2% margin behind the discrete GeForce GT 440! We understand that 3DMark Vantage also factors in the CPU's capabilities in computing the overall score; hence, its performance against the Intel Core i7-3770K that was paired with a discrete NVIDIA GeForce GT 440 graphics card was rather impressive as this result seems to point to a decent showing of the general compute performance of the AMD A10-5800K APU. The Trinity APU outperformed the Intel HD Graphics 4000 integrated GPU by slightly over 40% and extends its lead over the previous iteration of Intel's integrated GPU of the Intel Core i7-2600K by over 170%. In terms of improvement over the Llano APU, the new generation Trinity triumphed resounding over the older APU by almost 50%.
When compared to its equivalent-class Intel Core i3-3220 processor, the AMD A10-5800K is over three times faster! That should certainly provide a very tangible improvement in game play without even adding a discrete graphics card.
Far Cry 2
We expect to see strong DX10 performance levels for Far Cry 2, having seen the theoretical graphics performance numbers in 3DMark Vantage. The A10-5800K's internal Radeon HD 7760D took second position again, trailing the leader in this comparison by just a margin of approximately 12% as it churned out an average frame rate of 59 FPS. Its strong performance over is predecessor is evident as the older generation Llano's average frame rate is 47% lower at approximately 40 frames per second. Once more, the new APU is three times speedier than the Intel Core i3-3220 with its old Intel HD Graphics 2500 engine. Even if you were to compare against the fastest Intel CPU and its HD Graphics 4000 engine, the AMD A10-5800K is 50% speedier! That says a lot about the CPU/GPU influence in these games.
Battlefield: Bad Company 2
We put our AMD A10-5800K Trinity APU to the DirectX 11 challenge by using a game that supports it - Battlefield Bad Company 2. As a precautionary note to our readers, the results should be taken in with care - there's a mix of DX10 and DX11 capable GPUs for which the latter will run the game utilizing DX11 routines. As such, the scores aren't exactly comparable. In this test, the A10-5800K has made a marked improvement over its previous score by a large margin of 56% when we previewed it earlier. As a matter of fact, its performance is slightly higher than the discrete GeForce GT 440 by a razor thin margin of less than 1%! That's a very commendable improvement and we decided to up the ante and push the game's overall settings to 'High' with which the integrated Radeon HD 7660D graphics core managed to churn out a pretty respective average frame rate of 36.24.
Video Decoding Performance - Blu-ray Playback
In terms of CPU utilization for our Blu-ray playback, the Trinity APU didn't see any improvement even though we ported the APU over to the higher-end ASUS F2 A85-M Pro motherboard. Its performance of its integrated Universal Video Decoder trailed behind that of the older Llano APU by margins in the range of 37% to 42% (although the actual difference in absolute values is small). From this set of results, the performance of the AMD A10-5800K seems similar to the third generation Intel Core i7-3770K CPU, until the latter is paired with discrete graphics cards. As expected, the AMD A10-5800K also saw marked improvements over the previous generation Intel Core CPUs and their Intel HD Graphics 3000 integrated engine.