Today, AMD has completed the transition of their entire line of CPUs/APUs to the Piledriver design with the release of the Vishera, the mainstream/performance FX series CPU. How much performance can we expect from Vishera over Bulldozer? Let’s find out.[review_ad]
AMD chose a rather interesting day to launch Vishera, the Piledriver-based desktop FX CPU, as Apple also has a press event today. We think the timing of the two tech events is purely a coincidence since AMD scheduled the release date long before Apple sent out its invitation. Unlike Apple, which is trying to steal some thunder away from the imminent launch of Windows 8, AMD wants to ride along with the new OS release and hope that it can help to boost its CPU sales as people will be shopping for a new system/CPU to go with the new OS.
Compared to Zambezi (Bulldozer FX CPU), Vishera is more like a refresh than a new launch. The Trinity APU that we reviewed a couple of weeks ago features a greater architectural change in both the CPU and the GPU. However, with Vishera we get a minor tweak of the CPU architecture (Piledriver) in conjunction with a higher clockspeed. This is not a bad thing because Piledriver has improved upon some of the weakness that Bulldozer showed. This, coupled with a higher operating clockspeed, means the Vishera CPU should offer some good performance enhancement over Zambezi.
Piledriver: Enhanced Bulldozer
Major microarchitectural change is always a challenge in the in the PC industry. Something that looks great on paper may not always translate into real-world performance. Part of problem is that while the hardware maybe capable of doing great work, software that can really take advantage of that potential may not exist in the mass market. AMD’s Bulldozer suffered due to the lack of proper software support. Many desktop applications like games are simply not threaded heavily enough to fully take advantage of the eight integer cores. Bulldozer suffered a bit of split personality: it excelled in the heavily threaded tasks like video transcoding, 3D Modeling, and data encryption, but at the same time it failed in the lightly threaded tasks.
Unfortunately for AMD, waiting for the software manufacturers to catch up is impractical: it could take a long time for software makers to release updates to fully utilize all of Bulldozer’s cores. Such a delay would be costly, especially since Intel already has better performance with the existing software. Thus, AMD has to fix what is not working properly on Bulldozer and find a way to improve its performance. What we get as a result is Piledriver.
By now, we’re sure most of you already heard about Piledriver, the enhanced Bulldozer. Piledriver has improved many of Bulldozer’s weakness. Inside the Vishera is the same Piledriver module that is found on the mobile and desktop APU that we reviewed. It is still built based on the same 32nm SOI with 1.2 billion transistors on 315 mm2 die area. Like Buldozer, Piledriver has the same module design with two integer units plus one float unit unit share other resources.
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 instructions per clock. By keeping it more streamlined, Piledriver offers approximately a 10 to 15% improvement over Bulldozer.
As you may recall, one of the problems with Bulldozer is that its single-threaded performance was not only slower compared to Sandy Bridge but also is slower than AMD’s outgoing K10.5 Phenom II CPUs. As a result, in lightly threaded applications, Bulldozer is simply slower than AMD’s older CPU despite its better multi-thread performance. The enhancement that AMD has done with the Piledriver and the higher operational clockspeed should help AMD to deliver at least the same single-thread performance as its older CPU.
The memory controller has not been changed on Vishera: it still supports DDR3-1866 speeds. The controller supports DDR3 up to 64GB. In addition to the performance enhancement, AMD also adds two new ISA instruction sets: FMA3 and 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.
Same AM3+ Socket and the 900 Chipset/SB950
The Vishera CPU will still fit the AMD AM3+ socket motherboard with the same AMD 900 series chipsets that were launched last year. We are glad to see the same chipset and socket support. As you may recall there are some motherboards out like the ASUS Crosshair IV, that support Bulldozer but using an older chipset. AMD has informed us the Vishera will only be officially supported on the 900 series chipsets and it is up to the motherboard manufacturers to decide if their board will support Vishera. We highly suggest you contact your board vendor or check their website before putting the new CPU in the board. If you have a board that uses 900 Series chipset then you are probably fine, but for anything older, you would definitely want to check its compatibility.
As always, before installing a new CPU, we always suggest upgrading the BIOS first. With our ASUS Crosshair V Formula that uses the 990FX chipset, we had no issue getting the board to recognize and run the CPU even without updating to the latest BIOS. Still, it never hurts to check with the board vendor to see if your particular board needs the BIOS update before you install one of the new CPUs in the board.
Three Northbridge chipsets to choose from are the 990FX, 990X, or 970. The Southbridge will be the same, SB950 in this case regardless of which Northbridge you choose. Note that none of the AMD CPU/APU support PCI Express 3.0 yet so we get up to 32 lanes of PCIE Express 2.0 on the 990FX that is capable of split into dual x16 or quad x8 configuration for CrossFire. The 990FX will also support SLI natively for NVIDIA graphic card. It will come with six SATA 6 Gbps capable of RAID 0, 1, 5, 10 and 12 USB 2.0. The USB 3.0, however, would require third party chip.
The top of the line Vishera model is the FX-8350. Like its predecessor the CPU is an 8 core model but now it is clocked at 4.0 GHz as opposed to 3.6 GHz found on the FX-8150. The Turbo speed has not changed as both FX-8350 and FX-8150 will able to turbo up to 4.2 GHz. The official memory speed supported for both CPUs is still up to DDR3-1866. The mount of L2 and L3 cache are still 8MB and so is the 125W TDP.
In addition to the FX-8350, AMD also have the FX-8320, FX-6300, and FX-4300 with 8, 6, or 4 Piledriver cores, respectively. The eight-core variant will all be 125W TDP while the fewer core models will be 95W. All four models will come with unlocked multiplier for easy overclocking.
The picture above shows us how AMD prices these chips against Intel’s offering. The top of the line FX-8350 is positioned against the Core i5 3570K, fastest Ivy Bridge CPU without Hyperthreading support. While the FX-8320 is against the Core i5-3450 and the six-core variant FX-6300 is against the Core i5-2300. The only quad-core variant, FX-4300 is against the Core i3-2120. This model is a rather interesting since it is in direct competition also against AMD’s own A10-5800K. Choosing between these two chip is probably going to be depending on the platform and need. If you need more PCI Express lanes, then you would go with the FX-4300+990FX combinationbut if you want to use the integrated GPU or need native USB 3.0 and/or two additional SATA 6 Gbps ports, then you probably would want to choose the A10-5800K+A85X combination.
Current CPU Landscape
Before we take a look at the benchmark results, we would like to take a bit of time commenting on current CPU landscape. AMD has always been known as an underdog against Intel but this is not always this case. AMD had great success back in the Athlon64 era when it was able to put some pressure on Intel. Since then, however, AMD has been unable to recapture its formal glory. While Intel keeps marching forward with their “tick-tock” cycle that keeps technical advances going year after year, AMD has to play catch up. The last few years have been particularly tough for AMD as it has been unable to deliver a chip that can match Intel’s performance at the high end market. As a result, AMD can only compete against Intel with its aggressive pricing in the mainstream market.
Adding salt to the wound is the lackluster performance from Bulldozer, the successor to the K10.5, where the chip is unable to deliver performance equivalence to Sandy Bridge. Even worse, it also performed worse than AMD’s own outgoing Phenom processors under certain workloads. By now, we are not expecting AMD to release anything that is going to rock the CPU world and claim the fastest processor title. CPU architectures take years to develop and as we can see with mobile and desktop APUs and CPUs, AMD is going to use Bulldozer’s two integer unit plus one floating point unit design for the foreseeable future until the next architecture design.
What AMD can do right now is improve on its architecture and hopefully it is able to keep up and deliver a good performance gain (especially in the single thread department) so that it can at least put some pressure on Intel. After all, healthy competition always drives innovation. And for consumers, having AMD putting pressure on Intel would also help to drive the prices down.
In terms of absolute raw performance, AMD currently is more of a distant runner-up compared to the Ivy Bridge. Being the runner-up is never great, especially if you are in the tech industry where the talk of the town is about who is the fastest. The reality is that for average users it would be hard to detect any difference using either Intel or AMD in simple day to day work. Either camps should be able to deliver a good performance. While we can talk about the benchmark and scores, it is ultimately depending on your own needs when choosing a processor that will fit best into your own usage model and the budget.
TESTING & METHODOLOGY
We did a fresh load of Windows 7 64 bit on the test rig, then once we updated the software and drivers to the latest versions, we made a clone copy of the hard drive because we’re going to run this beast like we stole it. We ran each test a total of three times and the average of each test is reported here.
Intel: Core i5 2500K, Core i7 2600K, and Core i7 920
AMD: Phenom II X6 1100T , and FX-8150, FX-8350, A10-5800K
Intel: Asus P8Z77-V Pro
AMD:Asus Crosshair V Formula (AM3+) and Gigabyte F2-A85X-UD4
|Ram||Kingstone HyperX Genesis Special Edition at 1866Mhz (9-9-9-27)|
|CPU Cooler||Thermalright SI-128|
|Hard Drives||Seagate 7200.11 1.5 TB|
|GPU||Gigabyte GTX 580 SOC|
|Testing PSU||Cooler Master UCP 900W|
We will use the following applications to test the performance of the Motherboard. Benchmarks
|Lost Planet 2|
|Crysis 2 – DX11|
Sandra is a synthetic benchmark that may not always translates into real-world application performance. However, we still use it to give us an idea about a chip’s raw performance. As always, take the results here with a grain of salt.
In the Dhrystone benchmark, we see a healthy 23% improvement over the FX-8150, putting the FX-8350 above the Intel Core i5 2500K and just below the Hyperthreading enabled Core i7 2600K.
In the Whestone benchmark, we see a 12% improvement over the FX-8150.
Armed with eight integer cores, the FX processors perform quite well here compare to Intel’s quad core offering. Here we can see a 23% improvement over the FX-8150. Only the six-core Sandy Bridge E is able to out-perform the FX CPUs.
When comes to floating point calculation, we see the FX-8350 is about 18% faster than the FX-8150. Intel’s Core i5/i7 still dominates here.
Sandra’s Cryptology benchmark puts Intel ahead of the AMD’s CPUs.
When comes to the multi-core efficiency, we see the FX-8350 scored 11.59 vs 9.11 on the Core i5 2500K. However, Intel has a much better Inter-core latency of 44.7 ns vs the 134.8 ns on the FX-8350.
AMD has not made much change to the memory subsystem of the Piledriver and as a result, the memory bandwidth remains the same on the FX-8350.
Like Sandra, AIDA64 is also another synthetic benchmark that tests CPU and system’s performance.
The FX-8350 shows 10% improvement over the FX-8150, putting it ahead of the Core i5 2500K.
In Photoworxx benchmark, we only see a 5% gain.
ZLib shows an impressive 26% gain in performance, putting the FX-8350 faster than the Core i7 3770K.
When comes to data encryption, we see there is a 4% gain in the AES encryption benchmark and 10% gain in the Hash benchmark.
When comes to floating point calculations, we see the FX-8350 shows significant improvement over the FX-8150 in both Julia, Mendel, and VP8 benchmarks. While the FX-8150 failed to keep up with the Phenom II X6 1100T, the FX-8350 is able to deliver slightly better performance over the six-core Phenom CPU. The Intel Core i5 still dominates the float point calculation but at least AMD’s latest flagship desktop CPU is fast enough to replace the Phenom six cores.
AMD has not done much improvement on the Piledriver’s memory controller so we did not expect much of the performance difference here. There is a slightly higher bandwidth in memory read and write on the FX-8350 but the latency is a bit slower.
If we use PCMark 07 as our sole benchmark to assess a CPU’s performance, we can see that the FX-8350 is about 7% faster than the FX-8150. It is still falling behind the Core i5 2500K by 6%.
Looking at the lightweight task, we can see that the single thread performance for the CPU has improved, and the FX-8350 is now faster than the older Phenom CPU but AMD still lags significantly behind Intel’s Sandy Bridge.
The rest of the 3DMark 07 benchmark suit mimics the overall score where we see the FX-8350 is about 5~8% faster than the FX-8150. The Core i5 2500K is faster in every test where it is either slightly higher or up to 10% depending on the workload.
x264 HD v4.0
The single pass x264 transcoding favors Intel’s CPU but the FX-8350 shows a significant gain here by 8% over the FX-8150 and the Phenom II X6 1100T.
The FX-8350 performs much better here, and is about 13% faster than the FX-8150. It even narrowly beats out the Core i7 3770K and only is slower than the six-core Intel Core i7 3960X.
Without an integrated GPU, the FX CPU will not get any help from the VCE engine that is found on the APU. Here the transcoding is purely done through the x86. When comes to video transcoding, the FX-8350 certainly performs quite well. The FX-8350 is able to transcode video as fast as Intel Core i5 2500K transcode with Quick Sync and the fast setting in the MediaExpresso.
With quality setting, the FX-8350 takes about 50% more time to finish the task. When compare to the Ivy Bridge transcoding video via x86, we can see the latest Ivy Bridge is able to finish the workload about 12 seconds faster. Notice how well AMD CPUs and APUs scale up with the number of cores when we compare the A10-5800K (quad-core) to the FX-8350 (eight cores): doubling the number of cores cuts our transcoding time by half.
Cinebench single thread performance shows the combination of the higher clockspeed and the architectual tweak, and the FX-8350 is about 10% faster than the FX-8150. This brings the FX-8350 to the same level as the older Phenom II CPUs. One of the issues we had with the FX-8150 was its lower single thread performance and here we can see that AMD is at least able to deliver the same results as the older CPU.
The multi-thread performance shows a 15% improvement over the FX-8150 and it is faster than the Core i7 2600K.
When comes to data encryption, FX-8350 shows about 8% improvement over the FX-8150. This puts the FX-8350 neck and neck against the Intel Core i7 3770K.
Given the right workload, the FX-8350 actually can be a very powerful chip as it shown with our TrueCrypt benchmark. Here we can see that the chip delivers about 10~15% higher performance over the Core i3 3770K.
Lets start with the 3Dmark11 where we notice a 10% improvement in the overall score.
Warhammer 4000: Dawn of War is a CPU bound game where we can see the Intel CPUs take the top spot due to its better single thread performance. Notice that all of the AMD’s CPUs perform relative close to each other regardless if it is the older Phenom II, FX series (Bulldozer or Vishera), or the Piledriver based APUs. Compare the FX-8350 to the FX-8150, we see a 5% improvement.
Alien vs Predator performs better with AMD CPU than Intel. This game in particular also performs better with higher CPU clockspeed and as a result, we see the FX-8350 takes the top spot with at 4.0GHz where it is able to deliver 7% gain over the Phenom II X4 980.
In Lost Planet 2, the FX-8350 shows 9% improvement over the FX-8150, putting it on the same level as the Phenom II X6 1100T. However, the Intel Core processors are about 13% faster over the FX-8350.
Just Cause 2 is also another CPU bound game where we can see that single thread performance of the Piledriver based CPU/APU are performing about the same at lower resolution. Up the resolution to 1080p and the FX-8350 starts to pull ahead.
With more GPU demanding game titles, the FX-8350 is able to keep up with the Intel’s offering where we can see the FX-8350 narrowly edges out the Core i5 2500K.
Unigine Heaven 3.0 favors the FX-8350 where the CPU edges out the Core i7 3770K by 2%.
Ultimately, what we can conclude from our benchmarks is that in terms of gaming, the latest FX-8350 should be able to perform at least the same level as the older Phenom II based CPUs (light threads), an area where the FX-8150 sometimes fail to accomplish. Under heavy threaded games, the FX-8350 can delivers up to 10% improvements over the FX-8150. In addition, if the games are not particularly CPU bound, then the FX-8350 can certainly keep up with what Intel has to offer at the same price bracket but if it is CPU bound, then the Intel still offers better option with up to 30% improvements.
Power Consumption, Temperature
The idle temperature for the FX-8350 shows a slight improvement over the FX-8150. However, under load, it is tad hotter due to faster clockspeed.
For power consumption, we see the FX-8350 sheds off 20 watts of power or 17% idle. However, under load, we noticed that the system consumes 9 watts more than the FX-8150. The move to a lower idle system power consumption is a right step from AMD. While the FX-8350 is still not up to the same level as Intel’s offering, it is significantly improved over the Bulldozer. We do hope that load power can be further reduced in the next iteration.
With FX-8350, we were able to overclock the CPU to 4.4GHz pretty easily with air cooler, which is similar result to what we had seen with the Bulldozer. We kind of expected to see the same result since both chips are based on the same fabrication nodes. However, we did notice that with the FX-8350, overclocking it appears to be slightly more challenging as the system often will hard-lock whenever the temperature reached 65ºC. Thus, if you wish to push the CPU to its limits, it would require some extreme cooling. We are currently testing the overclocking result with water-cooling and/or liquid nitrogen result as well, so stay tuned to see just how far you can push it.
Piledriver is what Bulldozer should have been in the first place. The FX-8350 puts AMD back on track, but not yet back in the game. If it were not for Intel’s exceptional high IPC and performance per watts ratio on Sandy Bridge/Ivy Bridge, AMD’s Vishera/Piledriver would have been a good successor to Bulldozer as it delivers performance gain and reduced power consumption. Judged by itself, Vishera is certainly a good step for AMD and is a capable chip.
Unlike Bulldozer, where the single threaded performance falls behind the older Phenom CPUs, the FX-8350 (Vishera) is able to deliver a comparable performance while improving upon the multi-threaded performance. On average, we noticed that the processor has about a 10% improvement over the FX-8150. Obviously, the performance is partly due to the combination of the higher clockspeed: FX-8150 is clocked at 3.6GHz, whereas the FX-8350 is clocked 10% higher at 4.0GHz. The second major factor is Piledriver’s improved architecture over Bulldozer. The higher clockspeed seemed to help out with its single-thread performance more while the architectural design seemed to further improve upon the multi-thread performance.
Ideally, what we really want is a chip that is not only good at heavy threaded tasks but also is able to deliver a tangible gain in the lightly threaded tasks. The reality of current computing is that many desktop applications and games are still not fully optimized to truly take the advantage of the CPU’s multiple-core computing power. Some games don’t even use two cores, let alone eight. Because of this, the eight core FX-8350 may not show a huge gain over the dual or quad core variant of the same Piledriver derivatives at the same clockspeed (ie. A10-5800K). However, if media transcoding, 3D Modeling (Cinebench), and data encryption are your primarily workloads, then the FX-8350 would be a great choice for you.
We are glad to see the drop in the overall idle system power consumption. The CPU stays in idle mode for a considerable amount of time and only revs up the clockspeed when needed so anything that can help to reduce the idle power consumption is good. The lower power consumption also translates into lower idle temperature.
Compared to what Intel has to offer, AMD is still lagging behind in the single threaded performance and AMD has conceded that in our meeting with the company. However, the multi-threaded performance for the chip is competitive enough against the Intel’s offering. The chip is able to perform somewhere between the Core i5 and i7 depending on the workload. In most cases, we see the FX-8350 performs about the same as the Core i5 2500K or faster. In fact, given the right workload (like data encryption) at requires higher clockspeed and heavy integer calculation, the FX-8350 may even come ahead of what Intel has to offer.
Considering the performance of the chip, AMD was smart enough to price the FX-8350 at $195 so that it is in direct competition against Intel’s Core i5 processors. The price seems fair to us. In gaming, Core i5 still has advantage over the FX-8350 unless the game titles you play are either GPU bound or heavily threaded enough for you to use all of the eight cores. For common tasks, either Intel or AMD should be sufficient for average users. If you are running older Phenom processors, AMD has finally released a good enough chip that is worth upgrading to. If you are shopping for a new PC this holiday season, we would suggest you think about what kind of task and software you would be doing primarily with your PC. The FX CPU has its strengths and would be well-suited if you can take its advantage. Additionally, as a platform, AMD often offers more bang for the buck for those with a budget.
Looking forward, AMD still needs to address its weakness in order to fully compete against Intel. With the Ivy Bridge and Sandy Bridge, we can see that desktop CPUs may not be clocked at higher than 4.5GHz even with a die-shrink. AMD probably won’t be able to crank up the clockspeed much higher than the 4GHz that is on the FX-8350 when the company transitions to the 28nm. If AMD does not improve upon its instruction per clock, it may have a hard time keeping up with Intel in the next round.
Another issue that AMD also needs to address is the power consumption for the chip. With Intel actively reducing the power consumption of the CPU (and the whole platform), AMD needs something that is high in performance per watts, and not just raw performance. The PC shift into more portable and low power consumption devices means that AMD needs something that can deliver a good performance per wattage in order to fit their chip into the changing market. We can see AMD has already begun working on reducing the idle power consumption. Hopefully Steamroller, the successor to the Piledriver, can make some considerable gains in the single-thread performance and improves upon its performance per watts ratio.