FX-8350 CPU Review; AMD's Vishera Arrives

[SKYMTL]

A Quick Refresher:
Bulldozer: 1st generation architecture
Zambezi: Code name for FX-series CPUs with Bulldozer cores
Piledriver: 2nd generation Enhanced Bulldozer architecture
Vishera: Core name for FX-series CPUs with Piledriver cores

Hot on the heels of a successful launch for their desktop Trinity APUs, AMD is back again with yet another refreshed architecture: the Vishera FX-series processors. Based off of AMD’s Piledriver architecture, these new CPUs are supposed to be an evolutionary step forward rather than an revolutionary leap into the unknown.

Bulldozer and its associated Zambezi brand name have been around for over a year now so it was high time AMD began updating what could be construed as an outdated product stack. Intel on the other hand has been busy marching to their own “tick / tock” drummer by continuing onwards to Ivy Bridge and an efficient 22nm manufacturing process. Unfortunately, AMD’s Piledriver is still at 32nm but that hasn’t stopped them from improving in other areas to reduce some of Zambezi's shortcomings.

AMD’s current lineup is quite straightforward with very little segment overlap. At the top end is the FX-series of unlocked CPUs –the subject of today’s review- which don’t include onboard graphics processors and are aimed at systems with discrete graphics capabilities. The A-series of Trinity APUs hit the mid tiers and are lumped in with AMD’ “Fusion”, an initiative which promotes a heterogeneous environment where the central processing cores and integrated GPU work together to accelerate overall performance. The APUs are closely followed but a trio of Athlon X4 and X2 processors which, like the FX-series, make do without an integrated graphics subsystem. They target budget-conscious entry level users who want a CPU but don’t want to be tied at the hip to lower end gaming performance. Finally, AMD’s E-Series of energy efficient APUs brings up the rear.

All in all, this is very much the same product stack as last year’s but drilling down towards the individual FX-series processors shows us a slightly different story.

When Bulldozer / Zambezi was first introduced, the FX-8150 was the only true 8-core processor in the mainstream desktop market. Now, one year later that still holds true for its spiritual successor; the FX-8350. While similarities abound, there are some telltale changes here as well. Piledriver’s first and foremost goal is to improve upon Bulldozer’s performance per watt and that carries over into the FX-8350 which has 10% higher clock speeds than an FX-8150 but it still retains a TDP of 125W. Naturally, due to the design equivalencies between the Piledriver and Bulldozer architectures, transistor count remains at 1.2 billion and die size is still 315 mm². AMD’s Turbo Core makes a comeback as well and pushes this new processor’s frequencies up to 4.2GHz where situations permit.

Other than the flagship FX-8350, AMD is rounding out their lineup with a number of lower end offerings, all of which feature unlocked multipliers for easy overclocking and Turbo Core. Replacing the FX-8100 is the 8-core, 125W FX-8320. It follows closely in its big brother’s footsteps but comes with much lower clock speeds. Meanwhile, the 6-core FX-6300 and quad core FX-4300 take over from the FX-6100 and FX-4100 respectively. These may look like lower-end parts but they serve to complete a top to bottom lineup for AMD that will be in place until the second half of 2013.

One of the most important aspects of Vishera is pricing. AMD may be talking about eight core processors that operate at 4GHz but they aren’t meant to compete against the $1000 Extreme Edition CPUs of this world. The most expensive FX-8350 will be priced at a mere $195 with the FX-8320 and FX-6300 coming in at approximately $30 intervals below that. Other than the $122 FX-4300, this strategy represents a full court price decrease from the last generation even though Vishera’s performance could be substantially higher.

Clock speeds and introductory pricing may be changing this time around to account for the market’s new realities but past that, not much has changed from one evolutionary step to another. All of the distinguishing features from Zambezi CPUs have carried forward into the refreshed generation. L2 cache, L3 cache, memory support and even the venerable AM3+ motherboard compatibility have returned.

AMD is aiming Vishera at a market segment that wants more processing power for less money and that mantra seems to hold true when the new FX-series is placed up against similarly priced solutions from Intel. In addition to high clock speeds, more cores and every processor’s unlocked multiplier, AMD believes their architecture has a leg up with additional instruction sets which aren’t supported by lower end Intel SKUs.

If performance could be easily distilled down to a combination of the above-mentioned points, then AMD should have a winner on their hands. Unfortunately, CPUs don’t work with such a straightforward dynamic and because of this; Vishera may be facing an uphill battle. While the core / thread count seems like a cakewalk for the FX-8350 and FX-8320, we can’t forget these are 32nm 125W processors going up against 22nm parts that have TDPs of just 77W or less. In addition, the Ivy Bridge and Sandy Bridge architectures are known for their strong single thread performance while Piledriver continues to lag behind in this respect.

Going further down market, AMD is once again throwing core counts and high clock speeds at the same issues they faced with Zambezi. Intel has traditionally (for the last two years at least) retained a lead in per-thread horsepower so both the FX-6300 and FX-4300 go with the same focuses as their higher end FX siblings: price, frequencies and cores.

One feature everyone should give AMD credit for is their willingness to sell every one of their FX-series products as unlocked. This makes overclocking a breeze, even if you don’t have a serious background in system tweaking. Intel on the other hand has separated enthusiasts into neat little segments with their premium K and X-series processors. If you are looking for a budget-friendly overclocking experience, AMD is currently the only route to go.

From an intergenerational perspective, AMD seems to be moving at about the same pace as Intel’s Sandy Bridge to Ivy Bridge transition with an approximate 15% (or more) performance bump between Zambezi and Vishera. Of that fifteen percent, roughly 10% comes from clock speed improvements while the other 5% has been gained through some architectural optimizations.

Bulldozer represented the first step in a long trek for AMD and some hard lessons were learned throughout the course of its design process. The Piledriver cores within Vishera processors are the first step towards refining these basic principles in an effort to create a CPU architecture that is both adaptable and highly scalable as future generations are rolled out.

AMD are still firm believers that there’s a market for CPUs without integrated graphics processors and we tend to like this approach. The day of truly heterogeneous CPU / GPU computing is coming but it isn’t here yet. Not everyone wants an IGP or the cost and increased TDP that's associated with them. However, in this day and age, selling mid-level processors with TDPs of 125W may be tough for AMD regardless of overclocking capabilities and core counts.

Even though Vishera can be considered an incremental update rather than a brand new ground up redesign, the principles used in its architecture will be used by AMD for years to come. These FX-series CPUs along with the Trinity APUs are AMD’s current answer to Intel’s Ivy Bridge and the hope is they will tide the company over until 28nm products are introduced to battle Haswell. But will the FX-8350 and its ilk be enough and do they even represent enough of a performance increase to remain viable alternatives against Intel’s juggernaut?

 

[SKYMTL]

Revising Bulldozer; Say Hello to Piledriver & Vishera

Revising Bulldozer; Say Hello to Piledriver & Vishera

The entire raison d’être behind Vishera is to provide the next logical step in the Bulldozer architecture’s ongoing maturation process. This time around, an Enhanced Bulldozer design has been implemented in order to create the Piledriver cores found within these new CPUs.

Piledriver doesn’t represent a brand new core design for AMD. Rather, it is an evolutionary step that aims to implement the gradual changes without the costs associated with ramping up a brand new architecture. As such, we can’t expect miracles but Piledriver should bring some noteworthy benefits to the table for the new FX-series.

Based around a 32nm manufacturing process, equipped with a pair of cores and up to 2MB of L2 cache, the basic Bulldozer module hasn’t changed all that much in its Piledriver guise. There have been some minor changes like the addition of the FMA3 and F16C instruction sets but AMD’s major focus here was to increase the instruction per clock (IPC) rate and generally improve upon the operational frequencies of the previous generation. The relative maturity of GlobalFoundries’ 32nm node also led to a substantial leakage reduction when compared against Bulldozer.

Diving a bit further into the Piledriver, the enhancements seem to be everywhere. Most are supposed to home in on streamlining branch scheduling throughout the architecture and optimize certain elements for quicker communication.

However, there have been some sacrifices here since –as with the previous Bulldozer design- space has been given over for the aforementioned new instruction sets rather than retain a strict adherence to legacy standards. As a result, programmers using compilers like Visual Studio 2008 and older versions of Visual Basic could see their applications run slower on Bulldozer-based cores due to a lack of x87 and other optimizations.

While this may not be an optimal solution for every situation an emphasis upon SSE, AVX, XOP and other new, emerging instruction sets has allowed AMD to maximize their die space for current (and future) computing needs. Plus, the number of programs that use the legacy instruction sets is diminishing at a rapid rate.

AMD has also put some focus into single thread performance, which was a major stumbling point for the original Bulldozer design. Improvements have been instituted but AMD admits their architecture is still less than optimal for these situations.

The internal workings of Vishera processors are arranged in a similar way to Zambezi, with the only major difference being the replacement of typical Bulldozer cores with the updated Piledriver version. As with the previous generation, the top-end eight core Vishera chips are split into four distinct modules. Each of these modules consists of two x86 processing cores, a set of shared L1 cache and 2MB of L2 cache. This design can be scaled from the high end FX-8350’s eight cores and 8MB of L2 cache downwards to lower end processors like the quad core FX-4300 with its 4MB of cache.

Meanwhile, the unified L3 cache is shared between all of the modules, thus creating an adaptable pipeline between processing stages which can be dynamically assigned depending on how many cores are engaged.

 

[SKYMTL]

990FX & AM3+; More Lives Than Disco

990FX & AM3+; More Lives Than Disco

Welcome to the future....and the past. AMD has staunchly adhered to their AM3 platform for the last few CPU generations and not much has changed here. This time around, the AM3+ 900-series chipsets (970, 990X and 990FX) are being carried forward en masse and they all boast forwards compatibility with the new Vishera FX CPUs. Due to the architectural similarities between Zambezi and Vishera this shouldn’t come as any surprise and users with an eye to upgrade from slightly older CPUs will count this as a blessing.

While the 900 chipsets were derived from previous 800-series designs, these older chipsets don’t typically support the new AM3+ Vishera processors since there was a power spec revision right before 990FX was released. However, according to AMD, some motherboard vendors were able to include this updated design into their late 800-series products so there may be some older boards that actually do support Vishera. We suggest checking with your motherboard vendor prior to assuming compatibility with these new processors.

Since the market is going to be using this platform for the next year or so, let’s go through a quick crash course on the 990FX and the features it brings to the table.

At its heart the 990FX is simply an 890FX with updated microcode and expanded power modes to support the latest Bulldozer-based processors (Zambezi and now Vishera) while the 950 Southbridge is simply a rebranded SB850 chip. Is this a bad thing? Well, we don’t think so because the 890/850 combination used on previous boards proved to be forward thinking enough that many of its features (like SATA 6G) are still very much legitimate today.

While Intel has transferred most of their Southbridge functionality to their processor die, AMD’s AM3 chipset remains a two-part configuration consisting of a Northbridge and a Southbridge. We have seen this layout persist through the 790FX days on to the 890FX and the 990FX as well. This may change in the future but for the time being all current AM3+ products feature on-chip DDR3 memory controllers while all the I/O and expansion slot functions are handled through the chipset. We also get native support for DDR3 1333MHz to 1866MHz memory while the processor and 990FX Northbridge communicate through a Hypertransport 3.1 link at 6.4 GT/s.

The Northbridge of AMD’s two-chip solution acts as a controller hub for most of the board’s PCI-E 2.0 lanes and facilitates communication between the SB950 and the processor. PCI-E 3.0 isn’t supported but as we have seen in the past, the benefits the new protocol brings to the table are virtually nonexistent, even with today’s high-end dual GPU cards.

The 32 dedicated graphics card lanes are split up into either two 16x slots or can be evenly dispersed for up to a quartet of 8x slots which gives the boards incredible flexibility for multi GPU setups. Naturally, SLI and Crossfire are fully supported.

Meanwhile, the remaining ten PCI-E 2.0 lanes are divided up into one grouping of four lanes while an additional six lanes 1x lanes can be dispersed as needed for integrated components like audio and networking controllers.

Moving on down to the SB950, we see that AMD has once again foregone any updating as the layout is absolutely identical to that of the SB850. The chipset interconnect still uses a 2GB/s interface dubbed “Alink Express III” which essentially uses four PCI-E 2.0 lanes to speed up on-board communications.

Speaking of the Southbridge, we see that AMD has native 6Gbps SATA support but integrated USB 3.0 support was never included on these chipsets or the processors themselves. Third party USB 3.0 controllers can be interfaced to the Southbridge or Northbridge chipsets using the 1x PCI-Express 2.0 lanes for a maximum theoretical throughput of 500MB/s.

So there you have it. What was old is new again and while motherboard vendors aren’t too pleased about AMD’s decision to keep 990FX around for yet another generation, end users will end up benefitting. These motherboards are typically quite affordable and anyone that already has one will only need a $195 processor to complete a Vishera-based system.

 

[SKYMTL]

Test Setups & Methodology

Test Setups & Methodology

For this review, we have prepared a number of different test setups, representing many of the popular platforms at the moment. As much as possible, the test setups feature identical components, memory timings, drivers, etc. Aside from manually selecting memory frequencies and timings, every option in the BIOS was at its default setting.

For all of the benchmarks, appropriate lengths are taken to ensure an equal comparison through methodical setup, installation, and testing. The following outlines our testing methodology:

A) Windows is installed using a full format.

B) Chipset drivers and accessory hardware drivers (audio, network, GPU) are installed.

C)To ensure consistent results, a few tweaks are applied to Windows 7 and the NVIDIA control panel:

• UAC – Disabled
• Indexing – Disabled
• Superfetch – Disabled
• System Protection/Restore – Disabled
• Problem & Error Reporting – Disabled
• Remote Desktop/Assistance - Disabled
• Windows Security Center Alerts – Disabled
• Windows Defender – Disabled
• Screensaver – Disabled
• Power Plan – High Performance
• V-Sync – Off

D) Windows updates are then completed installing all available updates

E) All programs are installed and then updated

F) Benchmarks are each run three to eight times, and unless otherwise stated, the results are then averaged.

G) All processors had their energy saving options / c-states enabled

 

[SKYMTL]

Clock for Clock: Vishera vs Zambezi

Clock for Clock: Vishera vs Zambezi

As AMD transitions from one architecture to another, they have augmented a number of aspects within the Bulldozer architecture in order to create Piledriver. As we saw on the previous pages, average clock speeds have been increased somewhere in the neighborhood of 10% which will count for some performance differentiation between the FX-8350 and its predecessor. However, the improvements are supposed to go beyond mere frequency bumps and AMD has supposedly massaged other areas to squeeze some additional processing power from their latest CPUs. Just how much refinement did AMD build into their new CPUs? That’s what we want to find out.

In order to level the playing field forefront between Zambezi’s first generation Bulldozer design and the Piledriver cores within Vishera, we used an stock clocked FX-8350 and an overclocked FX-8150. Both processors were set to run at 4.0GHz and AMD’s Turbo Core was turned off in order to ensure clock speeds weren’t impacted by TDP constraints. All other settings remained identical to our standard testing setup but power saving features were disabled as well. This allowed us to highlight actual architectural improvements rather than let clock speeds get in the way of results.

It looks like AMD has actually managed to improve Piledriver in a number of ways that aren’t directly impacted by clock speeds. While the design optimizations haven’t necessarily yielded earth-shattering changes in every instance, there are a few noteworthy areas where this new architecture seems to shine. On average, we see a 5-10% performance bump which surely isn’t anything to ignore.

 

[SKYMTL]

Clock for Clock: Vishera vs Zambezi (Single Thread)

Clock for Clock: Zambezi vs. Vishera (Single Thread)

This section closely follows the previous page but we’re now going to take a closer look at one area where Bulldozer initially received a ton of negative press: single thread performance. This is one of the areas AMD admittedly needed to work on and Piledriver should have given them ample opportunity.

Unfortunately, while there are some improvements it doesn’t look like Piledriver is able to offer the single thread performance increases many were hoping for. In most of these situations, the actual difference between the FX-8350 and FX-8150 is less than 10%.

 

[SKYMTL]

System Benchmarks: AIDA64 / Cinebench r11.5

System Benchmarks

In this section, we will be using a combination of synthetic benchmarks which stress the CPU and system in a number of different domains. Most of these tests are easy to acquire or are completely free to use so anyone reading this article can easily repeat our tests on their own systems.

To vary the results as much as possible, we have chosen a selection of benchmarks which focus upon varied instruction sets (SSE, SSE3, 3DNow!, AVX, etc.) and different internal CPU components like the floating point units and general processing stages.

AIDA64 Extreme Edition

AIDA64 uses a suite of benchmarks to determine general performance and has quickly become one of the de facto standards among end users for component comparisons. While it may include a great many tests, we used it for general CPU testing (CPU ZLib / CPU Hash) and floating point benchmarks (FPU VP8 / FPU SinJulia).


CPU ZLib Benchmark

This integer benchmark measures combined CPU and memory subsystem performance through the public ZLib compression library. CPU ZLib test uses only the basic x86 instructions but is nonetheless a good indicator of general system performance.

CPU Hash Benchmark

This benchmark measures CPU performance using the SHA1 hashing algorithm defined in the Federal Information Processing Standards Publication 180-3. The code behind this benchmark method is written in Assembly. More importantly, it uses MMX, MMX+/SSE, SSE2, SSSE3, AVX instruction sets, allowing for increased performance on supporting processors.

RESULTS: The benchmark suite gets off to a roaring start for AMD’s FX-8350 with a number of tests that obviously love the eight core CPU. In one benchmark this $195 product actually runs circles around Intel’s $1000 3960X.



FPU VP8 / SinJulia Benchmarks

AIDA’s FPU VP8 benchmark measures video compression performance using the Google VP8 (WebM) video codec Version 0.9.5 and stresses the floating point unit. The test encodes 1280x720 resolution video frames in 1-pass mode at a bitrate of 8192 kbps with best quality settings. The content of the frames are then generated by the FPU Julia fractal module. The code behind this benchmark method utilizes MMX, SSE2 or SSSE3 instruction set extensions.

Meanwhile, SinJulia measures the extended precision (also known as 80-bit) floating-point performance through the computation of a single frame of a modified "Julia" fractal. The code behind this benchmark method is written in Assembly, and utilizes trigonometric and exponential x87 instructions.

RESULTS: Unfortunately, the VP8 and SinJulia benchmarks bring AMD’s new processor crashing back down to earth. In the FPU VP8 benchmark, the FX-8350 does post some respectable results and ties Intel’s 2600K but the x87 instruction set used by the SinJulia test proves to be its undoing.

CineBench r11.5 64-bit

The latest benchmark from MAXON, Cinebench R11.5 makes use of all your system's processing power to render a photorealistic 3D scene using various different algorithms to stress all available processor cores. The test scene contains approximately 2,000 objects containing more than 300,000 total polygons and uses sharp and blurred reflections, area lights and shadows, procedural shaders, antialiasing, and much more. This particular benchmarking can measure systems with up to 64 processor threads. The result is given in points (pts). The higher the number, the faster your processor.

RESULTS: Even though CineBench 11.5 isn’t fully optimized for AMD’s Piledriver cores, the FX-8350 still hangs with the 2600K and the Sandy Bridge E 3820. It also exhibits about 15% better performance than the FX-8150 which proves AMD’s claims in this regard.

 

[SKYMTL]

System Benchmarks: Civ V / PCMark 7

System Benchmarks (pg.2)

Civilization V: Gods & Kings Unit Benchmark

Civilization V includes a number of benchmarks which run on the CPU, GPU or a combination thereof. The Unit Benchmark simulates thousands of units and actions being generated at the same time, stresses multi core CPUs, system memory and GPU We give the non-rendered score below as it is more pertinent to overall CPU performance within the application.

RESULTS: These results aren’t all that impressive but AMD is still able to narrowly beat Intel’s similarly price quad core 3570K and once again there’s a respectable gain over the FX-8150.

PCMark 7

PCMark 7 is the latest iteration of Futuremark’s system benchmark franchise. It generates an overall score based upon system performance with all components being stressed in one way or another. The result is posted as a generalized score. We also give the Computation Suite score as it isolates the CPU and memory within a single test, without the influence of other components.

RESULTS: These results had us scratching our heads since the FX-8350 should shine in PCMark. This may be more of a platform issue since we noticed our SSD results (a major contributor to the overall score) weren’t anywhere close to those from the Intel platform despite using our motherboard’s SATA 6Gbps port and the latest AMD drivers. Luckily, the Computation suite pulls AMD out of the fire and their FX-8350 is able to post very respectable results.

 

[SKYMTL]

System Benchmarks: 3DMark (CPU) / WPrime

System Benchmarks (pg.3)

3DMark06 CPU

While 3DMark06 may be a slightly older synthetic benchmark, its CPU test still allows for multi threaded performance evaluations within a gaming environment. It effectively removes the CPU from the equation, generating a CPU-centric score.

WPrime

wPrime is a leading multithreaded benchmark for x86 processors that tests your processor performance by calculating square roots with a recursive call of Newton's method for estimating functions, with f(x)=x2-k, where k is the number we're sqrting, until Sgn(f(x)/f'(x)) does not equal that of the previous iteration, starting with an estimation of k/2. It then uses an iterative calling of the estimation method a set amount of times to increase the accuracy of the results. It then confirms that n(k)2=k to ensure the calculation was correct. It repeats this for all numbers from 1 to the requested maximum. This is a highly multi-threaded workload. Below are the scores for the 32M and 1024M benchmarks.

RESULTS: Once again we see the FX-8350 competing against the quad core 3750K but this is nowhere near where we would have expected an eight threaded processor to be. Nonetheless, it is great to see substantial improvements made over Bulldozer.

 

[SKYMTL]

System Benchmarks: Single Thread Performance

System Benchmarks: Single Thread Performance

Even though most modern applications have the capability to utilize more than one CPU thread, single threaded performance is still a cornerstone of modern CPU IPC improvements. In this section, we take a number of synthetic applications and run them in single thread mode.

RESULTS: The source of many of AMD’s Bulldozer failures doesn’t seem to have been addressed to the extent some were hoping. Single thread performance is still far behind Intel’s processors and while there is some improvement, it doesn’t look like it was anywhere near enough.