AMD FX

This article explains the AMD FX desktop processor . For the processor of the same name from 2003 see AMD Athlon 64 FX .

AMD FX
AMD FX logo
Production:	since 2011
Producers: AMD Globalfoundries
Processor clock:	2.8 GHz to 5.0 GHz
Instruction set :	x86 / AMD64
Microarchitecture :	AMD bulldozer
Base:	Socket AM3 +
Names of the processor cores: Zambezi Vishera

The FX series is a family of x86 - microprocessors with AMD64 extension of the manufacturer AMD . These are multi-core processors for desktop computers . The processors are based on the Bulldozer or Piledriver architecture .

history

After its announcement in Q4 2010, there was initially a lot of fuss about the new architecture, as it was supposed to be competitive with Intel for the first time in several years. Less than a year later, the first four processors were released. The FX-4100, FX-6100, FX-8120 and FX-8150 are the first available processors; the FX-8100 is only supplied to OEM customers and is not available separately. At the end of February 2012, two more models were presented with the FX-4170 and the FX-6200.

Commercially available processor AMD FX 8350, manufactured in Dresden

target group

According to AMD's marketing, the CPUs should serve the high-end segment and enable high overclocking thanks to their open multiplier . In the first tests under Windows, however, the top model also lagged behind its counterpart from Intel, the Core i7 2600K, and consumed significantly more electrical power under load. The enormous power consumption under load is also the limiting factor when overclocking on common mainboards with common cooling. Two patches were subsequently submitted for Windows 7, with which the performance of software with fewer threads is slightly increased, but the multi- thread performance is sometimes somewhat lower . In a good mix of these applications, the average performance of the AMD FX-8150 remained almost the same. Even with Windows 8, performance hardly improves, contrary to previous announcements.

Linux and UNIX systems

This article or the following section is not adequately provided with supporting documents ( e.g. individual evidence ). Information without sufficient evidence could be removed soon. Please help Wikipedia by researching the information and including good evidence.

As is also known with Intel CPUs, the AMD Bulldozers experience a leap in performance under Linux . Especially in use cases in which multi-threading performance is important, such as booting a Linux system with systemd, with the Apache benchmark or when compiling software (on an 8 kernel, software compilation normally causes 100% -Continuous load of all cores with a total of 16 threads), it can be clearly observed here that the leap with AMD is much larger than with Intel. With BSD systems such as FreeBSD, this leap can sometimes be even greater - most likely one of the reasons why an AMD CPU is used in combination with FreeBSD in Sony's PlayStation 4 . However, the processors of the AMD-FX series are promoted preferentially for gamers, which cost the CPU quite a reputation , although their performance can exceed that of an Intel CPU depending on the application.

Software optimization

When using the Open64 compiler under Linux, depending on the benchmarks, more than twice the performance can sometimes be obtained from the bulldozer.

However, the Open64 compiler is not fully equipped with all the functions of a compiler such as GCC , which is why you cannot compile an entire system with Open64 and thus not use the full optimization in all application areas.

In current GCC versions, however, GCC itself also offers optimization for the bulldozer, which can be used with -march = bdver2 (or -march = native for automatic detection). This increases the performance. This optimization in GCC can also be used to compile the entire operating system with all applications, whereby the optimization is applied in all applications. Depending on the application, this optimization runs faster or slower than compiling with AMD's own Open64 compiler.

Block diagram of a 4 module processor

Technical

Each processor consists of several so-called modules. In addition to the cache and everything else that belongs to a typical processor core, a module contains two integer clusters, but only one floating point unit. The latter can either process one thread with 256-bit commands or two threads with 128-bit commands. Only one thread can run on the integer clusters at a time. All other units of the processor such as fetch and decode share all execution units. AMD calls this approach core multithreading. This method takes a similar approach to Simultaneous Multithreading , but goes one step further. While with simultaneous multithreading one essentially only doubles the registers in order to execute two threads on one core, with the Bulldozer architecture several integer clusters are built into one module.

In addition to the special commands for AES encryption and the Advanced Vector Extensions , the FX is the first processor to offer real four-operand commands (Fused-Multiply-Add, FMA4 ).

Model data

Four modules, eight threads (8 integer clusters, 4 floating point units)

Vishera "Centurion Edition" ( Piledriver architecture )

• Sales name AMD FX-9XXX
• Multi-core processor
• Revision: OR-C0
• L1 cache: 16 KiB data per core  + 64 KiB instructions per module
• L2 cache: 2048 KiB per module
• L3 cache: 8 MiB with northbridge clock (2.4 GHz)
• MMX , SSE , SSE2 , SSE3 , SSSE3 , SSE4a , SSE4.2 , AVX , AES , FMA3 , FMA4 , XOP , AMD64 , Cool'n'Quiet 3.0, NX-Bit , AMD-V , Turbo Core 3.0
• Integrated dual-channel DDR3 memory controller: support up to DDR3 -2400
• Socket AM3 + , HyperTransport
• TDP : 220 W
• Manufacturing technology: 32 nm ( SOI ), immersion lithography
• The size: 315 mm² with 1.2 billion transistors
• Clock frequencies: 4.4-4.7 GHz
• Models:

model	Processor clock	Turbo 4 modules	Turbo 2 modules	TDP	introduction
AMD FX-9370	4.4 GHz	N / A	4.7 GHz	220 W	August 28, 2013
AMD FX-9590	4.7 GHz	N / A	5.0 GHz	220 W	August 28, 2013

Vishera ( Piledriver architecture )

• Sales name AMD FX-83XX
• Multi-core processor
• Revision: OR-C0
• L1 cache: 16 KiB data per core  + 64 KiB instructions per module
• L2 cache: 2048 KiB per module with processor clock
• L3 cache: 8 MiB with northbridge clock (2.2 GHz)
• MMX , SSE , SSE2 , SSE3 , SSSE3 , SSE4a , SSE4.2 , AVX , AES , FMA3 , FMA4 , XOP , AMD64 , Cool'n'Quiet 3.0, NX-Bit , AMD-V , Turbo Core 3.0
• Integrated dual-channel DDR3 memory controller: support up to DDR3 -1866
• Socket AM3 + , HyperTransport
• TDP: 95-125 W.
• Manufacturing technology: 32 nm ( SOI ), immersion lithography
• The size: 315 mm² with 1.2 billion transistors
• Clock frequencies: 3.5-4.0 GHz
• Models:

model	Processor clock	Turbo 4 modules	Turbo 2 modules	TDP	introduction
AMD FX-8300	3.3 GHz	3.5 GHz	4.2 GHz	95 W	October 23, 2012
AMD FX-8310	3.4 GHz	3.7 GHz	4.3 GHz	95 W	2nd September 2014
AMD FX-8320	3.5 GHz	3.8 GHz	4.0 GHz	125 W	October 23, 2012
AMD FX-8350	4.0 GHz	4.1 GHz	4.2 GHz	125 W	October 23, 2012
AMD FX-8370	4.0 GHz	4.1 GHz	4.3 GHz	125 W	2nd September 2014
AMD FX-8320E	3.2 GHz	3.5 GHz	4.0 GHz	95 W	2nd September 2014
AMD FX-8370E	3.3 GHz	3.6 GHz	4.3 GHz	95 W	2nd September 2014

Zambezi (" Bulldozer Architecture ")

• Sales name: AMD FX-81XX
• Multi-core processor
• Revision: OR-B2
• L1 cache: 16 KiB data per core  + 64 KiB instructions per module
• L2 cache: 2048 KiB per module with processor clock
• L3 cache: 8 MiB with northbridge speed (2.0 to 2.2 GHz)
• MMX , SSE , SSE2 , SSE3 , SSSE3 , SSE4a , SSE4.2 , AVX , AES , FMA4 , XOP , AMD64 , Cool'n'Quiet 3.0, NX-Bit , AMD-V , Turbo Core 2.0
• Integrated dual-channel DDR3 memory controller: support up to DDR3 -1866
• Socket AM3 + , HyperTransport with 5.2 GT / s
• TDP: 95-125 W.
• Manufacturing technology: 32 nm ( SOI ), immersion lithography
• The size: 315 mm² with 1.2 billion transistors
• Clock frequencies: 2.8-3.6 GHz
• Models:

model	Processor clock	Turbo 4 modules	Turbo 2 modules	TDP	introduction
AMD FX-8100	2.8 GHz	3.1 GHz	3.7 GHz	95 W	October 12, 2011
AMD FX-8120	3.1 GHz	3.4 GHz	4.0 GHz	95-125 W.	October 12, 2011 / Q2 2012
AMD FX-8150	3.6 GHz	3.9 GHz	4.2 GHz	125 W	October 12, 2011

Three modules, six threads (6 integer clusters, 3 floating point units)

Vishera ( Piledriver architecture )

• Sales name: AMD FX-63XX
• Multi-core processor
• Revision: OR-C0
• L1 cache: 16 KiB data per core  + 64 KiB instructions per module
• L2 cache: 2048 KiB per module with processor clock
• L3 cache: 8 MiB with Northbridge clock (2.0 GHz for CPUs with 95 W TDP and 2.2 GHz for CPUs with 125 W TDP)
• MMX , SSE , SSE2 , SSE3 , SSSE3 , SSE4a , SSE4.2 , AVX , AES , FMA3 , FMA4 , XOP , AMD64 , Cool'n'Quiet 3.0, NX-Bit , AMD-V , Turbo Core 2.0
• Integrated dual-channel DDR3 memory controller: support up to DDR3 -1866
• Socket AM3 + , HyperTransport with 5.2 GT / s
• TDP: 95-125 W.
• Manufacturing technology: 32 nm ( SOI ), immersion lithography
• Clock frequencies: 3.5-3.9 GHz
• Models:

model	Processor clock	Turbo 3 module	Turbo 1 module	TDP	introduction
AMD FX-6300	3.5 GHz	3.8 GHz	4.1 GHz	95-125 W.	October 23, 2012
AMD FX-6330	3.6 GHz	3.9 GHz	4.1 GHz	95-125 W.	December 2015
AMD FX-6350	3.9 GHz	4.0 GHz	4.2 GHz	125 W	May 2013

Zambezi (" Bulldozer Architecture ")

• Sales name: AMD FX-6XXX
• Multi-core processor
• Revision: OR-B2
• L1 cache: 16 KiB data per core  + 64 KiB instructions per module
• L2 cache: 2048 KiB per module with processor clock
• L3 cache: 8 MiB with northbridge clock (2.0 GHz)
• MMX , SSE , SSE2 , SSE3 , SSSE3 , SSE4a , SSE4.2 , AVX , AES , FMA4 , XOP , AMD64 , Cool'n'Quiet 3.0, NX-Bit , AMD-V , Turbo Core 2.0
• Integrated dual-channel DDR3 memory controller: support up to DDR3 -1866
• Socket AM3 + , HyperTransport with 5.2 GT / s
• TDP: 95-125 W.
• Manufacturing technology: 32 nm ( SOI ), immersion lithography
• Clock frequencies: 3.3-3.8 GHz
• Models:

model	Processor clock	Turbo 3 modules	Turbo 1 module	TDP	introduction
AMD FX-6100	3.3 GHz	3.6 GHz	3.9 GHz	95 W	October 12, 2011
AMD FX-6200	3.8 GHz	3.9 GHz	4.1 GHz	125 W	February 2012

Two modules, four threads (4 integer clusters, 2 floating point units)

Vishera ( Piledriver architecture )

• Sales name: AMD FX-43XX
• Multi-core processor
• Revision: OR-C0
• L1 cache: 16 KiB data per core  + 64 KiB instructions per module
• L2 cache: 2048 KiB per module with processor clock
• L3 cache: 4–8 MiB with Northbridge clock (2.0 GHz for CPUs with 95 W TDP and 2.2 GHz for CPUs with 125 W TDP)
• MMX , SSE , SSE2 , SSE3 , SSSE3 , SSE4a , SSE4.2 , AVX , AES , FMA3 , FMA4 , XOP , AMD64 , Cool'n'Quiet 3.0, NX-Bit , AMD-V , Turbo Core 2.0
• Integrated dual-channel DDR3 memory controller: support up to DDR3 -1866
• Socket AM3 + , HyperTransport with 5.2 GT / s
• TDP: 95-125 W.
• Manufacturing technology: 32 nm ( SOI ), immersion lithography
• Clock frequencies: 3.8-4.2 GHz
• Models:

model	Processor clock	Turbo 2 modules	Turbo 1 module	L3 cache	TDP	introduction
AMD FX-4300	3.8 GHz	3.9 GHz	4.0 GHz	4 MiB	95 W	October 23, 2012
AMD FX-4350	4.2 GHz	4.2 GHz	4.3 GHz	8 MiB	125 W	May 2013

Zambezi (" Bulldozer Architecture ")

• Sales name: AMD FX-4XXX
• Multi-core processor
• Revision: OR-B2
• L1 cache: 16 KiB data per core  + 64 KiB instructions per module
• L2 cache: 2048 KiB per module with processor clock
• L3 cache: 8 MiB with Northbridge clock (2.0 GHz), with FX-4130 only 4 MiB
• MMX , SSE , SSE2 , SSE3 , SSSE3 , SSE4a , SSE4.2 , AVX , AES , FMA4 , XOP , AMD64 , Cool'n'Quiet 3.0, NX-Bit , AMD-V , Turbo Core 2.0
• Integrated dual-channel DDR3 memory controller: support up to DDR3 -1866
• Socket AM3 + , HyperTransport with 5.2 GT / s
• TDP: 95-125 W.
• Manufacturing technology: 32 nm ( SOI ), immersion lithography
• Clock frequencies: 3.6-4.2 GHz
• Models:

model	Processor clock	Turbo 2 modules	Turbo 1 module	TDP	introduction
AMD FX-4100	3.6 GHz	3.7 GHz	3.8 GHz	95 W	October 12, 2011
AMD FX-4130	3.8 GHz	3.9 GHz	4.0 GHz	125 W	August 27, 2012
AMD FX-4170	4.2 GHz	4.3 GHz	4.3 GHz	125 W	February 2012

Web links

Commons : AMD FX  - collection of pictures, videos and audio files

References and comments

1. ↑ AMD cracks the 4 GHz mark with a four-core bulldozer. Hot
2. ↑ Test: AMD "Bulldozer". Computer base
3. ↑ AMD FX-8150 - Bulldozer in the detailed test. HT4U
4. ↑ The overclocked AMD FX-8150 is hungry - 193 watts at 4.4 GHz. ht4u.net
5. ↑ AMD Bulldozer Performance Patch. TweakPC
6. ↑ Windows 7, the “Bulldozer Patch” and Windows 8. Planet3dnow
7. ↑ Performance check: does Windows 8 give the Bulldozer architecture a boost? Tomshardware.de
8. ↑ AMD Bulldozer With GCC, Open64, LLVM / Clang Compilers
9. ↑ Open64 Release Notes ( Memento of the original from November 25, 2011 in the Internet Archive ) Info: The archive link was automatically inserted and not yet checked. Please check the original and archive link according to the instructions and then remove this notice.  
10. ↑ AMD Bulldozer "bdver1" Compiler Performance
11. ↑ Overview of the bulldozer structure using the example of a module.
12. ↑ More detailed picture to illustrate the bulldozer architecture
13. ^ Presentation c't
14. ↑ AMD FX-Series FX-9590 - FD9590FHW8KHK / FD9590FHHKWOF . cpu-world.com. Retrieved January 19, 2014.
15. ↑ AMD FX-Series microprocessor family. In: cpu-world.com. Retrieved October 6, 2014 . 
16. ↑ Nero24: AMD FX-8310 introduced - the best for last? In: Planet3DNow! October 6, 2014, accessed October 6, 2014 . 
17. ↑ Processors with this part number / OPN are only available for certain major customers / OEMs .