Menu

Bulldozer Arrives: AMD FX-8150 Review

October 12, 2011 - Hardware
Bulldozer Arrives: AMD FX-8150 Review

AMD is kick starting the new FX lineup with seven processors, including 8-, 6- and 4-core models. The flagship processor is the FX-8150, which features a base frequency of 3.6GHz with a Turbo Core clock of 3.9GHz and a Max Turbo speed of 4.2GHz. Before delving any deeper, let us explain what the Turbo Core and Max Turbo frequencies mean.

 

AMD’s Turbo Core technology has been enhanced for FX processors to include a new mode that boosts allCores when there’s enough thermal headroom. This allows new highly threaded scenarios to take advantage of the extra frequency. AMD Power Manager inside the CPU monitors the processor states as seen below.The Max Turbo mode is activated on lightly threaded applications by increasing frequency on half the cores. AMD has enhanced the highest clock level of its Turbo Core Technology to remain in a higher frequency state than in previous AMD Phenom II Processors. The result is better performance in single and lightly threaded applications.

 

Turbo Core and Max Turbo are present on all core configurations (again, the FX series is comprised of 8-, 6- and 4-core chips). The 3.6GHz FX-8150 carries 8-cores, has a 125-watt TDP, and an 8MB L2 + 8MB L3 cache, which is standard across all 8-core FX parts. The flagship octo-core processor is accompanied by the 3.1GHz FX-8120 and 2.8GHz FX-8100.

 

AMD has set the FX-8150’s MSRP at $245, which is only slightly more than the Core i5-2500K’s asking price, while it is considerably lower than the Core i7-2600K. The FX-8120 is even more affordable than the i5-2500K at $205.

 

 

Along with the three 8-core processors, AMD has launched a single 6-core chip known as the FX-6100, which comes clocked at 3.3GHz with a Turbo Core frequency of 3.6GHz and a Max Turbo of 3.9GHz. While the L3 cache remains the same at 8MB, the L2 cache has been reduced to 6MB, as 1MB is allocated per core. The FX-6100 is only $175, which seems like an incredible bargain considering the Phenom II X6 1100T is currently $190.

Lastly, there are three 4-core models: the FX-4170, FX-B4150 and FX-4100. All of the quad-core CPUs have 4MB L2 cache with 8MB L3 cache. The FX-4170 is the highest clocked Bulldozer chip with a standard frequency of 4.2GHz, however, Turbo Core is disabled and the Max Turbo frequency is just 100MHz above the base clock.

 

Under Bulldozer’s hood

AMD’s new microarchitecture is designed to provide the perfect balance between performance, cost and power consumption for multithreaded applications. It focuses on high frequencies and resource sharing to achieve optimal throughput. As mentioned previously, the AMD FX processors offer up to eight power-efficient cores. These represent the first generation of a new execution-core family (15h) from AMD.

The Bulldozer concept is based on a 2-core design that shares latency-tolerant functionality, smoothes bursty/inefficient usage and provides dynamic resource allocation between threads. Each core has its own 16KB L1 cache with a 1MB L2 cache, while the L3 cache is shared. The other units are now effectively shared between two cores and include: Fetch, Decode, Floating-point pipelines, and the L2 cache.

 

This design allows two Cores to use a larger, higher-performance function unit (ex: floating-point unit) as they need it with less total die area than having separate, smaller function units for each Core. It also means that there shouldn’t be Bulldozer-based CPUs with an uneven number of cores like the Phenom X3 series.

The Zambezi Bulldozer-based processors have a die size of 315mm², which is smaller than the Phenom II x6’s 346mm² die, while it’s bigger than the Phenom II X4’s 258mm² die. The 6-core “Gulftown” Intel Core i7 processors are also smaller at 240mm2, and the complex Sandy Bridge chips such as the i7-2600K are 216 mm².

A large 32nm die means a lot of resistors and AMD tells us that the Zambezi architecture has roughly two billion of them. That’s pretty incredible given the Intel Core i7-990X Gulftown (32nm) features 1.17 billion while the Core i7-2600K has just 995 million. The older Phenom II X6 processors have 904 million and the Phenom II X4 chips just 758 million. Those numbers help convey just how complex these Bulldozer CPUs really are.

 

The floating-point unit has also undergone a complete redesign. It has been improved to support many new instructions and it now allows resource sharing between cores. There are two 128-bit FMACs shared per module, allowing for two 128-bit instructions per core or one 256-bit instruction per dual-core module.

 

AMD has also designed a shared front-end which is responsible for driving the processing pipeline and will ensure that the cores are constantly fed with information. It has been designed to work with each dual-core unit and allocate threads to individual cores themselves. AMD has made heavy changes that include decoupled predict and fetch pipelines as well as prediction-directed instruction prefetchers.

A Prediction Queue can manage direct and indirect branches that are now fed with a L1 and L2 Branch Target Buffer, which stores destination addresses. The Bulldozer modules can decode up to four instructions per cycle, which is one more than the Phenom II processors. The prediction pipeline produces a sequence of fetch addresses. The Fetch pipeline performs a look up in the instruction cache and pulls 32 bytes per cycle into the fetch queue to feed the decoders.

 

AMD has also built new instructions into the Bulldozer architecture. While AMD and Intel share SSE3, SSE4.1/4.2, AES, and AVX, there are two new instruction sets called FMA4 and XOP that are now unique to AMD. The former is designed for HPC applications while the latter is used for numeric and multimedia applications as well as algorithms used for audio and radio.

 

Unlike Sandy Bridge, which features an on-die GPU with the System Agent (aka northbridge), AMD has taken a more traditional approach with the Bulldozer architecture. The company is avoiding an IGP (Integrated Graphics Platform) all together with AM3+, leaving that functionality for its 32nm Llano processors, which feature a speedy Radeon core.

The northbridge is also separate from the processor. Even though AMD claims to include an integrated northbridge, it’s really just a memory controller. In fact, AMD pioneered this technology back in the Athlon64 days. Bulldozer’s northbridge features two 72-bit wide DDR3 memory channels and four 16-bit receive/16-bit transmit HyperTransport links.

 

Sticking with Chipsets

As mentioned on the previous page, AMD is sticking with a north and southbridge chipset design for the AM3+ platform and this makes sense in our opinion, particularly for such a high-end platform. By including separate chips to handle PCI Express lanes and connectivity, AMD has not only been able to simplify the processor design, allowing it to focus on performance, but it also grants the company a great deal of flexibility.

The AM3+ platform is supported by the 9-series chipsets, which currently include the 990FX, 990X and 970 northbridge with the SB950 southbridge. This chipset series features the same 65nm silicon used by AMD’s previous-generation 8-series chipsets.

In fact, the SB950 and SB850 are the same chip and therefore offer the exact same features. The reason for the name change is simply to help the user identify that one is designed for the AM3+ platform while the other supports AM3 processors.

The SB950 southbridge supports six SATA 6Gb/s ports with AHCI 1.2 as well as RAID 0, 1, 5 and 10. Additionally there is support for 14 USB 2.0 ports along with two USB 1.1 ports and gigabit Ethernet. When compared to the Intel Z68 PCH (Platform Controller Hub), there is nothing missing from the SB950 southbridge.

Meanwhile, the 990FX (codenamed RD990) is essentially the same chip as the 890FX, though there is one minor change. The 990FX has been updated with HyperTransport 3.1 to support up to 51.2GB/s bandwidth between it and the CPU. The 890FX is equipped with HT 3.0 which features a bandwidth of 41.6GB/s.

The 990X is a cut down version of the 990FX, as it reduces Crossfire support from dual x16 or quad x8 lanes to a pair of x8 lanes with no quad-fire support. Continuing with the theme, AMD’s 970 chipset is basically a rebadged 870 with HT 3.1 support added. It lacks Crossfire support as it only includes a single PCIe 2.0 x16 lane.

 

990FX Partner Boards

AMD 990FX boards have been on the loose for some time now — we’ve had the Asrock Fatal1ty 990FX Professional for about two months. Along with its Fatal1ty- branded board, Asrock offers two other 990FX-based motherboards: the 990FX Extreme4 and 990FX Extreme3. The company also has a couple of AMD 970 motherboards available.The Fatal1ty 990FX Professional is currently retailing for $190, which is very affordable for a high-end motherboard. The price is especially impressive considering the included features, such as USB 3.0, SATA 6Gb/s, 8-channel audio, dual Gigabit LAN with Teaming function, 12 + 2 power phase design as well as Quad Crossfire and SLI support.

Despite being a Fatal1ty-branded product, we like the look and design of Asrock’s premium board. It has an excellent layout and UEFI BIOS interface.

 

We also included the Asus Crosshair V Formula in our review kit. This board costs considerably more at $230, but doesn’t seem to justify that premium.

Features such as USB 3.0 remain the same, while the Crosshair V Formula drops one of the two gigabit Ethernet jacks and eSATA ports. The audio has been upgraded to the SupremeFX X-Fi 2 codec, while the board also features an extra full-length PCIe 2.0 x16 slot. Overall, we prefer Asus’ aesthetics, but its board about equal in terms of functional design.

 

Asus also has the Sabertooth 990FX available, as well as various 990X and 970 motherboards. The Sabertooth 990FX is a cheaper alternative to the Crosshair V Formula, as it costs just $190 yet it is still very well equipped.

Although we’ve only received boards from Asrock and Asus so far, other manufacturers such as MSI and Gigabyte already have their flagship 990FX boards in stores, including the $145 MSI 990FXA-GD65 and $250 Gigabyte GA-990FXA-UD7.

How We Tested & Bandwidth
AMD has supplied us with its top-end FX-8150 processor for testing. Thankfully, the FX range is unlocked and the Asrock Fatal1ty 990FX motherboard allowed us to enable just six, four or two cores. This board also allows customization of the Turbo Core and Max Core multiplier. This allowed us to mimic the performance from the lesser three FX processors launching today: the FX-8120, FX-6100 and FX-4170.AMD AM3+ Test System Specs
– AMD Phenom II X6 1100T (3.30GHz)
– AMD Phenom II X4 980 (3.70GHz)
– AMD FX-8150 (3.60GHz)
– AMD FX-8120 (3.10GHz)
– AMD FX-6100 (3.30GHz)
– AMD FX-4170 (4.20GHz)
– x2 4GB G.Skill DDR3 PC3-12800 (CAS 8-8-8-20)
– Asrock Fatal1ty 990FX Professional (AMD 990FX)
– OCZ ZX Series 1250w
– Crucial m4 256GB (SATA 6Gb/s)
– Gigabyte GeForce GTX 580 SOC (1536MB)
Software
– Microsoft Windows 7 Ultimate SP1 64-bit
– Nvidia Forceware 285.38Intel LGA1155 Test System Specs
– Intel Core i7-2600K
– Intel Core i5-2500K
– x2 4GB G.Skill DDR3 PC3-12800 (CAS 8-8-8-20)
– Gigabyte G1.Sniper2 (Intel Z68)
– OCZ ZX Series 1250w
– Crucial m4 256GB (SATA 6Gb/s)
– Gigabyte GeForce GTX 580 SOC (1536MB)
– Microsoft Windows 7 Ultimate SP1 64-bit
– Nvidia Forceware 285.38Intel LGA1366 Test System Specs
– Intel Core i7-975 Extreme Edition (3.33GHz)
– Intel Core i7-920 (2.66GHz)
– x3 2GB G.Skill DDR3 PC3-12800 (CAS 8-8-8-20)
– Gigabyte G1.Sniper (Intel X58)
– OCZ ZX Series 1250w
– Crucial m4 256GB (SATA 6Gb/s)
– Gigabyte GeForce GTX 580 SOC (1536MB)
Software
– Microsoft Windows 7 Ultimate SP1 64-bit
– Nvidia Forceware 285.38Intel LGA1156 Test System Specs
– Intel Core i5-750
– x2 4GB G.Skill DDR3 PC3-12800 (CAS 8-8-8-20)
– Gigabyte P55A-UD7 (Intel P55)
– OCZ ZX Series 1250w
– Crucial m4 256GB (SATA 6Gb/s)
– Gigabyte GeForce GTX 580 SOC (1536MB)
– Microsoft Windows 7 Ultimate SP1 64-bit
– Nvidia Forceware 285.38 
Compared to the Phenom II, AMD’s FX processors provide significantly more memory bandwidth. Read throughput has been increased by roughly 60%, while write performance is almost 40% better. Although the bandwidth performance has been improved greatly, it’s still considerably slower than Intel’s Sandy Bridge processors. 

The L2 cache performance has also been improved — at least when looking at the Phenom II range. The FX-8150 delivered 30% more read throughput than the Phenom II X6 1100T, though the write performance was slightly slower. As with the previous chart, Bulldozer just can’t match Sandy Bridge’s L2 cache performance.

 

Synthetic Performance
SPECviewperf v11 is clearly not optimized for processors with more than four cores, as the highly clocked 4.2GHz AMD FX-4170 came away with the win. Rendering an average of 10.56fps in the SolidWorks test, it was slightly faster than the FX-8120 and FX-8150 processors. The FX-8150 still managed to outpace the Phenom II X6 1100T by 39% and it held a 17% margin over the Core i7-2600K.

The Maya benchmark revealed a similar trend. AMD’s FX-4170 seized the throne, while the FX-8150 trailed with a 55% lead on the Phenom II X6 110T and a 7% jump on the Core i7-2600K.

 

CINEBENCH R11.5 broke Bulldozer’s momentum as the FX-8150 rendered an average of 48.96fps, which was 9% slower than the i5-2500K and i7-2600K. Nonetheless, it was 4% faster than the Phenom II X4 980 and 6% quicker than the Phenom II X6 1100T.

 

The FX-8150 performed exceptionally in the WinRAR benchmark, delivering a multithreading result of 4287KB/s — 18% faster than the i7-2600K and 32% faster than the i5-2500K. Moreover, it landed 52% ahead of the Phenom II X6 1100T and 66% higher than the Phenom II X4 980.

 

Application Performance
AMD’s flagship Bulldozer processor delivered roughly the same performance as the Core i5-2500K and Core i7-920 in Excel 2010, which is a giant leap forward compared to the Phenom II. AMD’s chips rarely do well in Excel compared to Intel’s. For example, the Phenom II X6 1100T was 30% slower than the FX-8150.

WinRAR’s built-in benchmark pegged the FX-8150 as the fastest processor by far, but the results are quite different in our own custom WinRAR compression test. The FX-8150 took 118 seconds to complete the 700MB file compression test, which was on par with the dated Core i7-920, but 17% slower than the i5-2500K and 22% behind the i7-2600K.

 

Compared to AMD’s last-gen hexa-core processor, the FX-8150 offers a hefty 26% boost in Adobe Photoshop CS5, placing it in the league of Intel’s i5-2500K — though the i7-2600K was still 32% faster.

The FX-8120 was only a fraction of a second slower than its snappier sibling, and although the FX-6100 saw a noticeable performance drop, it was still 8% faster than the Phenom II X6 1100T. Oddly. the FX-4170 was 4% slower than the Phenom II X4 980, which is surprising considering they’re both quad-core processors and the FX-4170 has a significant clock advantage.

 

When testing with Fritz Chess 12’s built-in benchmark, the FX-8150 delivered 11682 kilo nodes per second, which was only 1% faster than the Phenom II X6 1100T and 10% quicker than the i5-2500K, while being 12% slower than the i7-2600K.

Clock for clock, the FX processors are likely no faster than the Phenom II in this test, as the FX-8120 was 6% slower than the Phenom II X6 1100T. Moreover, the FX-4170 was 24% slower than the Phenom II X4 980 despite having a 14% higher frequency.

 

Encoding Performance
The FX-8150 was 15% quicker than the X6 1100T with 113.8fps in HandBrake, while scoring only 2% lower than the i5-2500K and 5% behind the i7-2600K.The lesser-clocked FX-8120 performed as well as the Core i7-975 Extreme Edition, while the FX-6100 was 3% slower than the X6 1100T. Clock for clock, AMD’s dated processor won the hexa-core battle. Likewise, the X4 980 was much faster than the FX-4170.

The x264 HD Benchmark 4.0 produced another interesting result. In the Pass 1 test, the FX-8150’s 121.3fps showing was 2% slower than the X6 1100T, 15% slower than the i5-2500K, and 17% slower than the i7-2600K. At the same time, the dated X4 980 was only 1% slower than the FX-8210 and it fared 11% better than the FX-4170.

Things were quite different in the Pass 2 test, as the FX-8150 scored the best with 37.1fps, beating the i7-2600K by half a percent. The FX-6100 and FX-4170 still lagged behind their Phenom II counterparts.

 

The last encoding benchmark uses TMPGEnc 4.0 Xpress to measure the AVI to MPG conversion time. AMD’s latest offering slipped behind Sandy Bridge again as the i5-2500K and i7-2600K outmatched the FX-8150 by 27% and 31%. The FX-8150, FX-8120 and FX-6100 all delivered roughly the same performance. The FX-4170 fell to the bottom of the chart, sitting 2% lower than the X6 1100T and 5% behind the X4 980.

 

Gaming Performance
Dirt 3’s results appear to be dominated by Intel as AMD occupies the bottom six slots. However, if you look closer there is a difference between 1680×1050 and 1920×1200 performance. Intel clearly controlled the 1680×1050 results (where you could say there was less of a GPU bottleneck) the AMD processors performed better at 1920×1200.At 1920×1200, the FX-8150 matched the i7-2600K and narrowly defeated the i5-2500K. The FX-8120 was in line with the i7-975 EE, while the FX-6100 kept pace with the i7-920. Despite its 4.2GHz clock rate, the FX-4170 was the slowest processor tested.

Besides the Phenom II X4 980, AMD again filled the bottom of our graph when measuring 1680×1050 performance. However, at 1920×1200 the FX-4170 was actually the fastest processor tested, delivering 88fps. The rest of the FX lineup rendered between 82 and 84fps, allowing them to deliver similar results to the Core i5 and Core i7 products.

 

We saw mixed results in The Witcher 2 also. At 1680×1050 the FX processors performed poorly, with the exception of the FX-8150 which managed 77.6fps, placing it in the i7-920’s territory.

The FX-8150 delivered roughly 1fps more than the FX-8120 and FX-6100 when testing at 1920×1200, and although that’s in line with the i7-975 EE and 920, it was slower than the Phenom II X4 980 and X6 1100T.

 

Crysis 2’s performance was also disappointing for the FX processors as it doesn’t benefit from having more than four threads available. This allowed the FX-4170 to dominate at 1680×1050, while the FX-8150, FX-8120 and FX-6100 were all slower than the Phenom II X4 980.

 

Power Consumption
Power consumption is still an issue for AMD, as we found the FX-8150 chugged 252 watts under full load. Although that’s 5% less than the i7-920, it was 52% (86 watts) more than the i7-2600K. Idle consumption wasn’t much better. At 100 watts, the FX-8150 gobbled 32% more power (24 watts) than Intel’s enthusiast chip.The FX-6100 was slightly better than the Phenom II X6 1100T with consumption, though we should stress only slightly. It’s a similar situation when comparing the FX-4170 and Phenom II X4 980, but there is an improvement nonetheless.

Overclocking Performance
Overclocking the FX-8150 from its base 3.6GHz clock to 4.4GHz boosted Dirt 3’s performance at 1680×1050 by 4%. The 1920×1200 performance didn’t budge a single frame.

The 4.4GHz overclock delivered just under 3% more performance in the Pass 1 test and 12% more in Pass 2. This allowed the FX-8150 to overtake the Phenom II X6 1100T, but it was still much slower than the i5-2500K and i7-2600K.

 

CINEBENCH R11.5 OpenGL saw no performance gain when overclocking the FX-8150, though the CPU score increased 17%.

 

Final Thoughts

Breaking down our benchmark results we find that the AMD FX-8150 offers huge performance improvements over the Phenom II range when testing with Excel 2010, while it matched the Core i5-2500K and Core i7 920 processors. Our custom WinRAR benchmark also heavily favored the FX-8150 over the Phenom II, matched the Core i7 920 and trailed behind Sandy Bridge processors in this test.

The Adobe Photoshop CS5 benchmark also saw the FX processors provide decent performance gains over the Phenom II. Although the six-core FX-6100 was only slightly faster than the Phenom II X6 1100T, the eight core FX-8150 and FX-8120 processors provided significant gains and were able to match the Core i5-2500K.

The encoding performance was far less impressive as we found that clock for clock the FX processors were slower than the current Phenom II processors. The FX-6100 for example was slower than the Phenom II X6 1100T in our HandBrake and x264 HD Benchmark 4.0 tests. The FX-6100 did pull ahead by a decent margin when testing with TMPGEnc 4.0 XPress however. For the most part the FX-8150 was still considerably slower than the Core i5-2500K in our encoding benchmarks.

Finally, when it came time to play games the FX range was really no better than the Phenom II. To be completely honest, gaming on these high-end processors is so similar it’s hardly worth worrying about. The FX-8150 was never more than a few frames per second slower than the Core i7-2600K at 1920×1200.

Given that today’s latest game releases are only starting to adopt quad-core processors, having six or even eight threads available is of little consequence.

As for performance vs. power efficiency, the AMD FX processors are really not much better than the Phenom II range either, which is disappointing. When compared to Sandy Bridge CPUs, such as the Core i7-2600K and Core i5-2500K, the new FX processors stack up very poorly.

Despite the unlocked nature of the FX processors, overclocking is not fantastic. We were able to push the FX-8150 to just 4.4GHz on air (from the stock 3.6GHz). Compared to the 4.1GHz of our Phenom II X6 1100T it’s not bad, but if you consider the 5.2GHz possible with a Core i5-2500K or Core i7-2600K it’s definitely not great. Granted we were only able to achieve this extreme overclock using the Asus Maximus IV Extreme-Z, but all other P67 and Z68 motherboards reach at least 4.7 – 4.8GHz.

Then there is the question of value. At $245 the FX-8150 is pretty good, as is the FX-8120 at $205, and the FX-6100 at $165. The FX-8150 is 22% cheaper than the Core i7-2600K and this works to AMD’s favor as the FX-8150 was often less than 20% slower.

However, is the FX-8150 a better buy than the Core i5-2500K? In terms of performance the AMD CPU was more often than not slower, if only by a small margin, while it does consume considerably more power and will not provide the same overclocking results. It’s no secret who wins this round.

Considering that the FX-8120 is essentially the same processor as the FX-8150, we will look to it for the FX vs. Core i5 comparison. The FX-8120 costs $205 and it’s unlocked — all FX processors are — so it can be easily modified to match or exceed the operating specifications of the FX-8150. Therefore we feel the cheaper FX-8120 gives the Core i5-2500K a serious run for its money and it’s a worthy alternative. Meanwhile the FX-6100 is also great value at $165, as it undercuts both the Phenom II X6s but it wasn’t always faster.

We won’t deny it, we really were hoping for a lot more from Bulldozer and AMD’s eight-core processors. It’s disappointing to find these newly launched processors do little to improve AMD’s situation. The FX processors come short of competing hand to hand with the now 9-months old Sandy Bridge processors, and in certain instances surpass their own Phenom II range. Still, this is just the start for Bulldozer, and there’s much more to be seen from the FX range, or so AMD says.

 

Taken From http://www.techspot.com/review/452-amd-bulldozer-fx-cpus

Tags:

6 thoughts on “Bulldozer Arrives: AMD FX-8150 Review

cyberloner

Buy bulldozer or not?

nod32

not unless you’re using it for encryption or utilizing avx/fma instructions

They F’d up the L1 cache big time and L2/L3 is slow as hell, too. It’s bottlenecked. Idiots.

nod32

Power consumption isn’t accurate. Every review is different and the BIOSes seem to be the blame. Also, intel CPUs draw uncore power from the 3.3 and 5v rails. They’re using more than they show (since ppl measure from the 12v). With a proper bios it uses about 10-15W more under load for total system draw and less at idle than i5/i7.

cyberloner

I sold my 1090T and change board to AM3+ , i think bulldozer release just like the 1st dual core cpu era when a lot benchmark is very sucks. But now everybody is using more cores. No software is utilizing it yet i think. I wait and see.

cyberloner

a lot number of the benchmark can be different and high with other brand board. Asus UEFI bios is not matured enough to handle bulldozer when they are reviewing and benchmarking.

Leave a Reply