If you read our AMD Trinity Preview, you’ve already seen the integrated GPU performance in the new A-Series APU from AMD. Today, we take a look at the CPU side of the APU. In addition, we are able to see just how the new chip will perform at various applications and workloads, including media transcoding, data encryption, and OpenCL.
Let’s face it, Bulldozer was a bit of a disappointment. While the radical modular design of a single floating point cluster paired with two integer clusters has a lot of potential, it ultimately failed to move AMD forward, as the processor delivered worse performance than AMD’s own several-year old Phenom II X6 processor. When we reviewed Bulldzoer, we felt that it had potential and with a a bit of fine-tuning, it would possible for AMD to deliver a more competent processor to the market.
Fast-forward a year, and we are here today with the official launch of the desktop Trinity: the successor to the Llano, and AMD’s new A-series APU for mainstream and budget systems. Trinity, released back in May 2012, has finally brought us the processor we hoped AMD would put forth when it launched Bulldozer. Trinity is built based on AMD Piledriver, which stems from Bulldozer; and the Northern Islands GPU family found on the last generation HD 6000 graphic cards. Built on the same 32nm SOI die, Trinity features two to four Piledriver x86 core with up to 384 VLIW Radeon cores. This results the Trinity to have 1.303 billion of transistor (up from 1.178 billions on Llano) and 246mm2 die area (vs 228 on Llano).
Desktop Trinity APU: Piledriver + HD 7000
While “Piledriver” is still based on the same architecture and module design of Bulldozer, it has improved significantly over Bulldozer. The picture above shows what improvement and enhancements AMD has done on Piledriver. AMD has improved on the branch prediction, scheduling, and the hardware pre-fetcher. Together, these will help Piledriver to improve on its instruction per clock. By keeping it more streamline, Piledriver offers approximately 10 to 15% improvement over Bulldozer.
With Piledriver, AMD also adds two new ISA instruction sets: FMA3 atnd F16C. Bulldozer already supported FMA4 and with the addition to the FMA3, Piledriver is first CPU with such support as Intel won’t add FMA3 until Haswell.
The Trinity Turbo Core has been upgraded over Llano. Unlike Llano where only the CPU could turbo up to a higher speed, the Trinity APU will feature Turbo Core for both CPU and GPU. Both sides will be able to turbo up when there is thermal headroom available. Now if CPU is under heavy load while GPU is not, the CPU is able to run at higher speed. The converse is also true: when the CPU is idle but the GPU is under heavy load, it is also able to run at higher clockspeed. With Trinity there are two modes of Turbo clockspeed, full turbo and half turbo. Take the A10-5800K, for example. Its base clock is 3.8GHz and it is able to turbo to 4.0 GHz (half turbo) or 4.2 GHz (full turbo). The turbo speed seems to be either a 150MHz or 200MHz stepping depending on the model.
On the GPU front, it can turbo up to 800MHz. Unlike Intel’s processor where the turbo speed depends on the number of active cores, both modules on the APU will turbo up to the same clockspeed. We asked AMD whether there was a preference for GPU vs CPU turbo if a particular workload was both CPU and GPU intensive. We were told that even under these scenarios, both CPU and GPU would still turbo up. So it appears that the turbo is based on the workload and the thermal envelope.
While Turbo Core (and Intel’s Turbo) makes it hard to quantify absolute performance due to the dynamic clockspeed, it is nonetheless good for consumers to gain extra performance.
Piledriver’s memory controller has been updated so it now supports DDR3-1866 speeds (up from 1600 on Llano). The controller supports DDR3 up to 64GB for the desktop.
Trinity GPU borrows the AMD Northern Islands family’s VLIW4 design. The GPU features 6 SIMD engines with a 16 VLIW4 array for up to 384 Radeon cores. There are 24 texture units and 8 ROPs. The A10 models will come with all 384 cores enabled while the A8 will get 256 and the A6 will get 192 cores. The GPU will be clocked at 800MHz for the A10 an 760 MHz for the A8.
As expected, the features on the GPU is exactly the same as the Radeon HD 6000. We get DirectX 11, OpenGL 4.1, OpenCL 11, Eyefinity support. In fact, this is the first processor that is supporting 3+1 displays when using DisplayPort 1.2 port. The AMD Universal Video Decoder (UVD3) is here as well for helping video decode. What is new here is the addition of the hardware encode component it has borrowed from the Graphics Core Next’s Video Codec Engine (VCE). This should help out with transcoding multimedia where hopefully it is capable of delivering comparable performance like QuickSync on the Intel’s Sandy or Ivy Bridge
Today’s launch is not just an APU launch but a platform launch. We get the A-Series APU and a new A85X chipset.
Three different chipsets will support the A-series APU: A55, A75, or A85X. The A55 and A75 are not new as they were introduced when Llano was launched. No new features have been added to ether one of these two chipsets. They still support only one PCI Express 2.0 x16. The A55 only comes with SATA 3 Gbps and USB 2.0, while the A75 will come with SATA 6 Gbps and four USB 3.0
The new addition, the A85X, is not much different from the A75 except it now is able to split the PCI-E bandwidth into dual x8 for CrossFire. Additionally, we gain two extra SATA 6 Gbps ports, bringing a total of 8.
Trinity will require a new socket, FM2, that is not backward compatible with previous generation APU (socket FM1). The socket FM1 board will not work with FM2 APU and neither would the FM2 APU will fit the FM1 board. This is a little bit unfortunate since the FM1 socket was introduced only a year ago and it is no longer being supported. According to AMD, the FM2 socket should stay a little bit longer as future APUs (ie. the die shrink of Trinity) will also be FM2 socket so hopefully we will see a longer upgrade path.
The good news is that the older heatsink that is compatible with the Socket FM1 will be compatble with the new Socket FM2. We are actually using an old Thermalright cooler from way back and it is still functioning without any issue.
During our meeting with AMD, we were informed that the new platform brings a new feature: Radeon RAMDisk. The idea of RAMDisk is not new but it is the first time that AMD is putting it on its own platform. When you pair the AMD A10 APU with AMD Memory, users are able to use RAM as a caching drive for faster data accessing and hence better performance With the A10 APU, AMD allows up to 64GB of storage for caching. AMD’s implementation is similar to Intel’s Smart Response (SSD Caching) except Intel uses an SSD for caching while AMD chooses cheaper alternative–DDR RAM. Either way, the goal is to let the end-user to have a much more responsive system. We could not get much information from AMD as to whether the RAMDisk cache would be backward compatible to the older generation A-Series APU. We also could not get hold of the software needed to setup the RAMDisk so we cannot comment on it more than what we present here. We would have to wait until we can get a hold of the software to test this feature in the near future.
AMD shipped us the the GIGABYTE F2A85X-UD4P board with the A10-5800K and the A8-5600K.
The board is an ATX form factor that features GIGABYTE’s Digital Power, Ultra Durable 5, 2x copper PCB, and GIGABYTE’s 3D BIOS. Around the CPU socket, we can see that the CPU has been moved closer toward the center of the board as opposed to close to the rear IO. This leaves more room for big horizontal heatsinks, but it also puts the CPU closer to the DIMM slots, which can potentially be an issue if you choose a large heatsink that extends out out of the CPU. The board features four DIMM slots that supports 64GB DDR3 up to 1866 MHz.
It is very nice to see that GIGABYTE is not skipping the onboard Power On/Off button, Clear CMOS button, and the reset button on the board. We absolutely love these onboard buttons as it makes our job much easier. The only thing we probably wish the board comes is the rear Clear CMOS button. Another neat feature is the dual 64Mbit BIOS with a dedicated BIOS switch.
For storage, the board comes with 7 SATA 6Gbps courtesy of the A85X chipset. The board supports RAID 0/1/5/10 and JBOD.There are four fan headers sprinkles across the board in addition other CPU fan header. All five fan headers are 4-pin design. GIGABYTE put the APU fan header toward to the center of the board, between the PWM and the PCI-E expansion slot, as oppose to the edge of the board. The placement makes it a little bit hard to reach if the board is installed inside the case. The four additional fans are all placed close to the edge of the board where one is located on the top, two are next to the 24-pin power connector, and the last one is on the bottom next to the front panel header. Three of the four system fans support speed control.
For expansion, the board comes with three PCI Express 2.0 x16 slot. The first slot runs at x16 bandwidth and if the second slots are populated, both slots will run at x8 fr CrossFire. The third slot is always running at x4. There are three PCI-E x1 slots and a PCI slot for other add-on cards. The spacing of the first two x16 PCI-E slots is far enough for a three slot cooler, however, if you install two cards with dual slot cooler, you will only be left with two single PCI-E x1 and the last PCI-E x16 for expansion. The PCI slot will be blocked by the second PCI-E x16.
Along the bottom edge of the board is where the usual internal headers are located. We got the front panel header, the front audio header, the S/PDIF out header, one USB 3.0 header, four USB 2.0 header, one serial port header, one TPM header. The USB 3.0 and USB 2.0 headers are all controlled by the A85X chipset. GIGABYTE seem to be one of the few motherboard vendors to keep supporting TPM on their boards.
Flip to the rear of the board and we see the board comes with one PS/2 keyboard/mouse port, one D-sub pot, one DVI-D port, one HDMI port, one DisplayPort, one optical S/PDIF out connector, four USB 3.0 ports, two USB 2.0 ports, one eSATA 3 Gbps port, one Ethernet port, and the standard audio jack. The network is controlled by the ubiquitous Realtek Network Controller and the audio is the equally ubiquitous Realtek ALC892. We were a little surprised to find that the eSATA port is 6Gbps as oppose to 3Gbps since the AMD A85X chipset supports 8 SATA 6 Gbps ports and seven of them are used for internal SATA, so we presumed GIGABYTE would simply route the remaining SATA 6 Gbps to the eSATA port. Two of the four USB 3.0 headers are controlled by the A85X while the other two are controlled by Eton EJ168.
Trinity supports 3+1 display. A note with three display configuration where if you connect both the D-sub and the DisplayPort, only the DisplayPort will be functional. Thus, you can mix and match any of the ports, provided that you do not use the D-sub and the DisplayPort simultaneously.
GIGABYTE ships the board with the manual, driver CD, six SATA data cable, and the rear IO plate.
AMD launched the mobile Trinity SKUs back in May and we will not repeat them here. Today we will just focus on the desktop Trinity. The table below shows the specification and the models:
Desktop Trinity will either be 65W or 100W TDP. Unlike what Intel has done with Ivy Bridge by lowering the maximum TDP, AMD chose to deliver higher performance while maintaining the same TDP as Llano. Given the size of the die and the amount of transistors, we can understand why AMD is unable to lower the TDP, especially since both Trinity and Llano are based on the same 32nm die. Hopefully we will see lower TDP once AMD shifts to the 28 nm fabrication process.
AMD will have A10, A6, A8, and A4 models available. It will also have the K series with unlocked multiplier for overclockers. The desktop APU will fit in the 65W or 100W TDP envelope. The A10 and A8 will come with 4 cores (2 modules) and the former will feature HD 7660D GPU with 384 Radeon cores clocked at 800 MHz while the later will come with HD 7560D with 256 Radeon Cores clocked at 760 MHz. The A6 and A4 series be dual cores (or one module) where the A6 will have the HD 7540D GPU with 192 Radeon Cores clocked at 760MHz and the A4 will have HD 7480D with 128 Radeon Cores clocked at 723 MHz.
In addition to the A series APU, AMD also brings back the Athlon branding. The difference between the Athlon and the A series APU is appreantly the lack of the integrated GPU. Other than that, the turbo core is still supported and it still comes with either four cores or two cores configuration. There is even an K model for overclockers. One thing to note here though because the Athlon models lack the integrated GPU, they will not have the VCE and the UVD3 support.
The most expensive model A10-5800K retails at $122 which is the same as the Intel Core i3 2120/2100. So it is clear that AMD is targeting the high-end A series APU at the Intel Core i3 CPU and the low-end APU at the Intel Celeron.
For consumers, AMD’s nomenclature created a little bit confusion. First, although the GPU on the Trinity is based on the HD 6000 GPU, it is branded as 7000. Making the matter a little worse, these APUs will also able to run AMD Dual GPU with the 6000 graphic cards. Additionally, the same chipset naming can be slightly confusing when shipping for a new board. This is not as big of an issue if you just watch out for a board that is socket FM2 compatible as opposed to FM1 when shopping for a board that will work with Trinity. Obviously, non-tech savvy individuals may not care much about the numbering scheme and branding, so we do wish AMD could make it easier for consumers to identify their various products and their compatibility.
Synthetic: AIDA64 AND SANDRA 2012
AMD positioned the A10-5800K against the Intel Core i3 processor and based on the AIDA64 CPU Queen benchmark, the competition is justified. Notice that the older Llano actually performs faster here despite both APUs featuring four integer cores. This test involves the branch prediction capabilities of the CPU and looks like despite the improvement on the Piledriver’s branch prediction, it may not be as fast as the old offering.
Things turns around for Trinity in the Phtoworxx benchmark where the new APU not only out-performs the older models but also is faster than the Core i3 2100. The test involves not just the CPU integer arithmetic and multiplication execution unit but also the memory subsystem.
The A10-5800K does very well in this test where it is even faster than the Core i7 920.
All new AMD APUs will feature a dedicated hardware AES acceleration and thus it is no surprise to find it performs very well here. Some of the Intel Core i3 processors (like the Core i3 2100) lack a dedicated hardware AES support. While the APU is still not as fast as the more expensive Intel CPUs with hardware AES support, it is nonetheless good to see it is significantly faster than those CPUs without. If AES is important for you, then be sure to choose one with hardware support.
The new entry to the AIDA64 supports many of the new instructions such as AVX or XOP extensions. Here again the A10-5800K does very well and we can see that the A8-5600K is actually just as fast as the fastest Llano, the A8-3870K.
Despite only having two float-point units, the A10-5800K is not doing such a bad job here. It is almost as fast as the A8-3850 and is faster than the Intel Core i3 2100. For both single (32-bit) and double (64-bit) precision, the A10-5800K is trailing behind the A8-3850 with about 7% and 14%, respectively. It is faster than the Core i3 2100 by 30%.
The extended float point calculation (80-bit) is a weak spot for Trinity, most likely because it only has two float-point units.
VP8 is yet another new entry to the AIDA64 where it tests the FPU’s performance at video compression. Here the A10-5800K edges out the Core i3 2100 but is still lagging behind the A8-3850.
The memory subsystem has been improved significantly on Trinity where its memory performance is much better not only over Llano but also over Bulldozer. The write speed on A10-5800K is almost as fast as the Sandy Bridge and is more than double the K8 and about 70% faster than the Bulldozer.
The read is not as big of a jump as the write. Trinity is still faster than the the K8 by about 50% but it is slower than Bulldozer. Intel still dominates here, and even the older Nehalem memory controller is still faster than Piledriver.
The copy is faster than Llano but still trailing behind the Intel platform.
Interestingly, the latency on the Trinity is worse than that of either Llano or Bulldozer. Intel once again dominates the chart here.
The A10-5800K is 67% faster than the A8-3850 and 30% faster than the Core i3 2100.
Here, the A10-5800K is just marginally faster than the A8-3850 and is actually a tad slower than the Core i3 2100.
Not much of a surprise here as with four dedicated integer calculation core, the A10-5800K comes ahead of the Core i3. Notice that Trinity is actually much faster than the K8 CPUs (Phenom II and A8-3850) and is about as fast as the Nehalem (Core i7 920).
A10-5800K performs pretty good here where it is about as fast as the A8-3850 and faster than the Core i3 2100.
When comes to Cryptography, we can see that Intel still dominates here but the dedicated hardware assistance on the Trinity definitely helps a lot with the workload. Keep in mind that that chart above only shows the performance difference between CPUs with dedicated hardware AES encryption and not those without (Llano and some Core i3 CPU models).
Oddly enough, the memory performance on Trinity seemed to be worse here.
If we use PCMark 07 as our tool to test the overall performance, we can see that A10-5800K is about 6% faster than the A8-3850 and 27% faster than the Core i3 2100.
The Productivity test is also not particularly heavily threaded enough to really show the power of multiple cores. Here the A10-5800K performance is impressive where it is about 15% faster than the A8-3850 and Core i3 2100.
The Creativity test involves video transcoding and image manipulation. AMD traditionally does really well in this test, even beating out the Core i5-2500K.
Computation again involves in the video transcoding. With a dedicated hardware assisted VCE, Trinity does rather well here. The score from the Core i5 2500K onboard was 13572, and was out of the range of the graph.
Entertainment involves both streaming video and gaming. A10-5800K does well here.
Cinebench 11.5 x64
Single threaded performance for the A10-5800K has improved over the A8-3870K slightly. The APU is in fact just as fast as the FX-8150 but still lags behind the Core i3 2100.
The multi-thread performance, on the other hand is better for the A10-5800K. However, the A8-3870K is still a tad faster than the A10-5800K.
We include the Cinebench OpenGL benchmark to show the performance of the integrated GPU. The A8-5600K is about 15% faster than the A8-3870K and the A10-5800K is about 30% faster.
x264 HD v4.0
When comes to media transcoding, Trinity is about 10-20% faster with the x264 benchmark. The x264 benchmark do not yet support the dedicated hardware encoder so the performance here shows us only the CPU side.
We are glad to see that AMD did not cripple Trinity like Intel did with the Core i3 when it comes to hardware-assisted AES support. Granted, AMD’s implementation is still not on par with the Intel’s offering, but we can see that there is a big performance jump having the dedicated block doing the work. Just give you some idea, we can see that the A8-5800K is twice as fast as the six-core Phenom II X6 1100T. Granted, AES may not be as important for Trinity’s target audience, it is nonetheless good to see even a budget CPU is able to offer some sort of hardware encryption. This also makes buying decision much simpler as you know that no matter which APU you choose, all of the features and special instruction are present no matter what.
We already know that Intel’s latest HD 4000 GPU is almost as fast as the HD 6550D found on the Llano APU, so it should not come as a big surprise that Trinity’s GPU performance takes AMD even farther ahead of the game in the GPU front. Much like what we observed with the transition from the HD 5000 (VLIW5) to the HD 6000 (VLIW4), the GPU improvement is significant on the Trinity despite having fewer cores than Llano.
The HD 7560D on the A8-5600K delivers about a 10-20% gain over the HD 6550D on the A8-3870. AMD expects the A10-5800K’s HD 7660D to deliver about 20% more over the HD 7560D, or about 40% gain over the Llano. We also put the HD 6670 as a comparison here and we can see that the A8-5800K is actually able to deliver about the same performance as the HD 6670 or $50 budget GPU performance. With Trinity, AMD is getting really close to delivering a great integrated GPU for us. For most part, the GPU on the Trinity is decent enough, playing games at 1680 x 1050 with mid to high level, where it is capable of delivering 30 FPS. The APU is almost good enough to play Metro 2033 at 1680 x 1050 low setting, just a tad shy of the 30 FPS needed for smooth game play.
Compare it against the Intel HD 4000, the HD 7660D is about twice as fast. This puts the AMD APU further ahead of Intel’s integrated GPU. Obviously, AMD would want to highlight the importance of a balanced computing experience on both GPU and CPU. On this note, AMD is clearly moving closer to what it is trying to achieve with Trinity. AMD is not alone here. Intel is also pushing GPU performance as well. Ivy Bridge’s graphics doubled GPU performance over Sandy Bridge, so it is not hard to see just how serious both companies are trying to provide a good enough integrated GPU.
Discreet Card Performance
Paired with the GTX 670, we can see that at high-end gaming, it really depends on if the game is more CPU bound or GPU bound. If it is more GPU bound the A8-5800K has absolutely no problem in keeping up with the Core i7 3700K. However, if it is more CPU bound, then the A8-5800K still lags behind the Intel’s offering. The Core i7 3770K is selling at more than double of the price that A8-5800K will be selling so its performance is not too shabby considering its pricing. We put the A10-5800K against the Core i7 3770K because AMD markets the APU to be a quad core and we want to see just how much performance difference would it be between the two quad core CPUs. AMD targets the A8-5800K against the Core i3 from Intel and we will need run that benchmark to make a fair comparison.
Media Encoding and OpenCL
Cyberlink MediaExpresso 6.5 and Arcsoft MediaConverter 7.5 are two softwares that are currently supporting the Intel Quick Sync and AMD VCE. When comes to video transcoding, nothing can touch Intel’s QuickSync at the moment. The A10-5800K is just as fast as the Core i5 2500K without any hardware acceleration and is faster than the Llano. With VCE enabled, we shed 3 seconds off compared to the fast mode and 27 seconds off compare to the quality mode.
HandBrake: OpenCL Performance
HandBrake is an open source media converter that supports OpenCL. The software is available online free to download but the current version 0.9.8 does not yet support the OpenCL and we have gottena beta version that allows us to test the performance with and without OpenCL. We did not have much time to run the test through a list of CPUs so we only include a small data set.
Clearly, A10-5800K is slower than the Sandy Bridge CPU when transcoding video through the x86. With OpenCL enabled, we got 54% gain in performance. We will have to do more test and compare the performance with the GPU acceleration on Sandy Bridge and OpenCL performance of the Ivy Bridge. So far, we can see that if the software is capable to tapping into the GPU power via OpenCL, Trinity can gain significant gain where it goes from being underperforming be a very competitive at its price range.
Ultimately, this all lies on the software support. Currently, AMD lists about 10 applications that supports OpenCL that includes some of commonly used software titles: WinZip 16.5, GIMP 2.8, Adobe Flash 11.2, Adobe Photoshop CS6, ArcSoft MediaConverter 7.5, and CyberLink MediaExpresso 6.5, Handbrake. Obviously, software programmers would need to take their time and implement such support and we hope more software will adopt OpenCL and similar technology to fully take advantage of modern processing power.
Compute Power: OpenCL and OpenGL
Let’s take a look the theoretical GPU compute power of the HD 4000. We ran Sandra’s GP Computing benchmarks. Keep in mind that these numbers are theoretical performance.
OpenCl GP Processor
|OpenCl GP||Native Float Shader (Gbpix/s)||Native Double Shader (Mpix/s)|
|Core i5 2500K (HD 3000)||0.10388||43.3|
|Core i7 3770K (HD 4000)||0.15791||72.75|
|A10-5800K (GPU only)||0.6425||42.8|
|A8-5600K (GPU only)||0.4102||57.3|
The theoretical OpenCL compute power on the A10-5800K is about six times greater than the Core i7 3770K in the native float shader (single precision) benchmark. However, the A10-5800K falls slightly behind the Core i5 2500K in the Native double shader (double precision) benchmark.
Sandra also provides the ability to measure the CPU and GPU combination performance and with the
|Compute Shader||Native Float Shader (Gbpix/s)||Native Double Shader (Mpix/s)|
|Core i7 2500K (HD 3000)||0.05881||5|
|Core i7 3770K (HD 4000)||0.27276||57.2|
Looking at the Computer Shader, we see the A10-5800 is twice as fast as the Core i7 3770K in the single precision float point calculation but it is only half as fast in the double precision calculation. The A10-5800K is, however, much faster than the Core i7 2500K.
|Compute Shader||Cryptography Bandwidth (GB/s)||Encryption/Decryption Bandwidth (GB/s)||Hashing Bandwidth (GB/s)|
|Core i7 2500K (HD 3000)||0.363||0.129||1.019|
|Core i7 3770K (HD 4000)||1||0.902||1.15|
In the Cryptography benchmark, the Trinity trumps the fastest Ivy Bridge
|Compute Shader||Internal (GB/s)||Transfer (GB/s)|
|Core i5 2500K (HD 3000)||8||3.29|
|Core i7 3770K (HD 4000)||14.77||6.7|
Looking at the compute shader bandwidth, we an see the A10-5800K is faster than the Core i5 2500K by 50% in the internal and 100% in the transfer. However, comparing it against the Core i7 3770K, it falls behind in the internal but is able to keep up with the transfer.
Power Consumption, Temperature and Overclocking
Keep in mind that the power consumption is dependent on the platform as opposed to just the CPU alone. With both Intel and AMD producing not just the CPU but also the chipset, both companies are able to more tightly regulate the chip in order to minimize power leakage.
The idle power consumption for Trinity has been improved significantly. The A8-5600K is 14 watts less and the A10-5800K is 6 watts less than the A8-3870K. In fact, the idle power consumption for the Trinity is on par with the Ivy Bridge CPUs.
The load power consumption for the A8-5600K is about 10 watts higher than the Core i3 system. With the A10-5800K, we see 7 more watts of power consumption. While the Intel Core i3 has a lower power consumption, we an see that the power consumption for these two APU is about the same as the Core i5 2500K (aka quad core Sandy Bridge).
First of all, do not read too much on the absolute value of the temperature reading as it depends on the cooler used and the architecture. We use the same cooler for all of our AMD CPUs but we use a different cooler for the Intel CPU. One thing we can notice is that AMD has done some nice work in reduction the idle heat-output. We believe this is partly due to fewer transistor on the Trinity compare to Llano and possibly with internal power regulation in Trinity.
The load temperature is slightly higher on Trinity where the A10-5800K is 3 degrees hotter than the A8-3870K. If your old cooler is able to keep your Llano cool, it should not have any issue at keeping the Trinity cool as well.
At default voltage, we were able to overclock the CPU to 4.4 GHz and the PC is able to boot but when we overclock it to 4.5GHz, we were not able to get our system to post. After raising the voltage to 1.5 V we were able to push the CPU to 4.6 GHz but it would not pass Prime95, and two of the four workers would give us errors and not complete the test. We had to lower the speed to 4.5GHz in order for us to run Prime 95 stable.
At 4.5 GHz, we ran Prime 95 and saw the power consumption jumped to 193 W, or 46 watts more than the sotkc speed. The CPU temperature also reached 85ºC and core is 68ºC.
Obviously, overclocking is not Trinity’s best feature if you plan to do it with air-cooler. We were told that it can reach higher speeds but ultimately that would require more extreme cooling like liquid nitrogen.
When we reviewed Bulldozer, we felt that the new architecture had potential despite its luckluster performancer. Trinity is the first APU from AMD that features the much improved Piledriver core and the result is that compared to Llano, Trinity is a step forward for AMD. The single thread performance for Trinity is either same or faster than Llano but its multi-thread application is definitely faster than Llano (provided the load is not heavy on float-point computations). Compared against the Core i3, the Intel still has the same advantage on the single thread as it is always been, but multi-thread favors AMD due to the extra integer calculation unit.
While the float point calculation of Trinity falls short of Llano because of reduction in the FPU (2 vs 4), it’s actually more efficient than the Bulldozer. As a result, despite lower theoretical numbers, its performance per unit is actually higher than both Llano and Trinity.
As expected, AMD continues to move forward on the GPU front. Ivy Bridge is almost as fast as Llano and now Trinity moves AMD even further ahead of the competition. Not to mention that Trinity also supports 3+1 display configuration and have the ability to pair with another Radeon card for added performance. Trinity’s power consumption is also looking good. Its idle power is lower than Intel platform but its load power is higher than Intel’s offering due to its 100W TDP vs Intel’s 95W (Sandy Bridge) or 77W (Ivy Bridge).
We are happy to see Trinity brings all updated instructions that includes AVX and FM3 which helps out with the floating point calculation. In addition, the hardware assisted AES encryption and dedicated VCE and UVD3 engine for media encoding and encoding makes Trinity an even more competitive processor at performing these special tasks not only against the older CPU from AMD but also against the Intel Core i3. It also helps to alleviate some of its weaker floating-point performance. We like that AMD did not cripple their APU line up by disabling some of these features. No matter which APU you choose, you will get all of these ISAs. The only difference is that you would be choosing an APU with different clockspeeds for CPU and GPU; or if you choose Athlon models, it will not come with the integrated GPU (note that Athlon models will also lack the UVD3 and the VCE as well).
With Trinity, it is not hard to see AMD is marching forward with the idea of heterogeneous computing. Trinity improves on CPU and GPU performance. While its floating point is still weak compared to Llano due to the reduction in the floating-point units, it is more efficient than its predecessor. As we see with HandBrake, if software is able to tap into the OpenCL, then Trinity can definitely keep up and even performs better than Llano in every aspect. Ultimately, the CPU industry is going through the same evolution as when the industry moved to multi-core where the hardware is here and ready but the software still need to catch up in order to fully take its advantage. The current list of software title is much better than a year ago and we do hope that it will grow with more software companies take advantage of the new technology.
Retailing at around $100, Trinity is a good all around APU. For day to day computers, it is fast enough for the average user. As a platform, Trinity and A85X chipset offer a good combination of chipset (SATA 6Gbps, USB 3.0), CPU, and GPU power for anyone who are looking for a mainstream and budget PC. In fact, with a more powerful integrated GPU than Intel’s offering and VCE for assisting encoding, Trinity would make a great HTPC.