Test Systems
Test System |
---|
Motherboard: | AMD: ASUS Crosshair V Formula MSI 990FXA-GD80
Intel LGA1366: ASUS P6X58D-E Intel LGA1156: GIGABYTE H55-UD3H Intel LGA1155: ASUS P8P67 EVO |
---|
Cooler: | Scythe Katana III |
---|
Memory: | AM3+ 2 x 2048 MB G.Skill RipjawsX F3 PC3-17066 DDR3 @ 1866 MHz 9-9-9-28
AM3/LGA1156/LGA1155 2 x 2048 MB G.Skill RipjawsX F3 PC3-17066 DDR3 @ 1333 MHz 7-7-7-21
LGA1366 3 x 2048 MB MUSHKIN BlackLine FrostByte PC3-12800 DDR3 @ 1333 MHz 7-7-7-21 (limited to 4GB) |
---|
Graphic card | AMD Radeon HD 5850 |
---|
Harddisk: | Samsung F3 HD502HJ 500GB |
---|
Power Supply: | CORSAIR TX 650 W |
---|
Software: | Windows 7 Ultimate 64-bit SP1 |
---|
Drivers: | Catalyst 10.11 |
---|
As many of you know, AMD claimed the Windows 7 thread scheduler was not optimized for Bulldozer architecture, and its wrong core workload allocation could hinder performance. Windows 8 by their claims, in certain situations will bring up to 10% more performance compared to Windows 7, and in the meantime Microsoft published two hotfixes for Windows 7 correcting this problem.
The problem with Windows 7 was that it could not distinguish modules from cores, and since in Bulldozer each module has two cores with shared resources it is more performance effective to assign threads first per module, then per core if modules are populated. If module one has Cores 0 and 1, module two Cores 2 and 3, with this hotfix Windows 7 will assign first thread to Core 0 in first module and second thread to Core 2 in second module, so both CORES in separate modules will have full control of module resources not having to share them with the other core idling, or in CC6 state. Only after all first cores in every module are busy working, Windows 7 will assign the next thread to the second core in a module and thus making cores in that module share resources.
This patch is highly recommended for all FX processor owners, although you should not expect more than marginal improvements. The update will be most beneficial with lightly threaded workloads, where 2-4 cores are active. In the following graphs, results
with and
without the two hotfixes applied are shown.
Below you can see the entire list of programs used to test our processors. Most of the tests make good use of multithreading, but just like real-life situations, there are some tests that prefer higher clock frequencies over number of cores or size of L2/L3 cache buffer. Since Windows 7 64-bit is used from now on as base for processor benchmarks, some of benchmarks are used in 64-bit versions as well.
Benchmark Suite
- Aida64 Extreme - Read/Write/Copy/Latency
- PC Mark Vantage 64-bit
- 3D Mark 11 Performance preset (Physics and Combined score)
- wPrime v2.04 1024M
- SuperPI 1M
- I'm Too Audio Encoder v6.1.2 - FLAC to LAME MP3 320kbs
- dBpoweramp R14 - FLAC to LAME MP3 320kbs
- Windows Media Player 12 -Audio CD to MP3 320kbs
- x264 benchmark HD V3.0
- Im Too HD Video Converter 6 - 1080p trailer encode via Matroska profi
- AVS Video Converter 7.1 - 1080p trailer encode via Video for Facebook profile
- Photoshop CS5 64bit Retouch Artists Speed Test
- Cinebench R11 64-bit
- POV Ray 3.7b 64-bit
- 7zip v9.20 64-bit - 32 MB mutithreading
- WinRAR Benchmark v4.0 64-bit
- Real life compression ~ 1 GB folder (139 files)
- TrueCrypt - 100 MB encryption/decryption benchmark
- Batman Arkham Asylum
- Warhammer 40k: Dawn of War II Chaos Rising
- Formula 1 2010
- Mafia II
- Metro 2033
- Prime95 (Small FTT-s) for maximum heat and power consumption
Specifications
In the table below you can review detailed specifications of FX-8150 and other models that will show up later in the performance graphs.
Intel Processors | Core i5 760 | Core i7 875K | Core i7 950 | Core i3 2120K | Core i5 2500K | Core i7 2600K |
---|
Number of cores | Quad | Quad | Quad | Dual | Quad | Quad |
---|
Core codename | Lynnfield | Lynnfield | Bloomfield | Sandy Bridge | Sandy Bridge | Sandy Bridge |
---|
Socket | LGA 1156 | LGA 1156 | LGA 1366 | LGA 1155 | LGA 1155 | LGA 1155 |
---|
Core speed | 2800 MHz | 2933 MHz | 3066 MHz | 3300 MHz | 3300 MHz | 3400 MHz |
---|
Turbo Boost | 3333 MHz | 3600 MHz | 3333 MHz | - | 3700 MHz | 3800 MHz |
---|
L2 Cache | 256 KB/core | 256 KB/core | 256 KB/core | 256 KB/core | 256 KB/core | 256 KB/core |
---|
L3 Cache | 8 MB | 8 MB | 8 MB | 3 MB | 6 MB | 8 MB |
---|
Process Node | 45 nm | 45 nm | 45 nm | 32 nm | 32 nm | 32 nm |
---|
Core die size | 296 mm² | 296 mm² | 263 mm² | 131 mm² | 216 mm² | 216 mm² |
---|
TDP | 95 W | 95 W | 130 W | 65 W | 95 W | 95 W |
---|
Integrated graphics | - | - | - | Intel HD 2000 | Intel HD 3000 | Intel HD 3000 |
---|
Price | EOL | EOL | $ 270 | $ 120 | $ 230 | $ 330 |
---|
AMD Processors | A8-3850 | Phenom II X4 980 BE | Phenom II X6 1055T | Phenom II X6 1075T | Phenom II X6 1090T | Phenom II X6 1100T | FX-8150 |
---|
Number of cores | Quad | Quad | Hexa | Hexa | Hexa | Hexa | Eight |
---|
Core codename | Llano | Deneb | Thuban | Thuban | Thuban | Thuban | Zambezi |
---|
Socket | FM1 | AM2+/AM3 | AM2+/AM3+ | AM2+/AM3+ | AM2+/AM3+ | AM2+/AM3+ | AM3+ |
---|
Core speed | 2900 MHz | 3700 MHz | 2800 MHz | 3000 MHz | 3200 MHz | 3300 MHz | 3600 MHz |
---|
Turbo Core | - | - | 3300 MHz | 3500 MHz | 3600 MHz | 3700 MHz | 4200 MHz |
---|
L2 Cache | 1 MB/core | 512 KB/core | 512 KB/core | 512 KB/core | 512 KB/core | 512 KB/core | 2 MB/module |
---|
L3 Cache | - MB | 6 MB | 6 MB | 6 MB | 6 MB | 6 MB | 8 MB |
---|
Process Node | 32 nm | 45 nm | 45 nm | 45 nm | 45 nm | 45 nm | 32 nm |
---|
Core die size | 228 mm² | 258 mm² | 258 mm² | 346 mm² | 346 mm² | 346 mm² | 315 mm² |
---|
TDP | 100 W | 125 W | 125 W | 125 W | 125 W | 125 W | 125 W |
---|
Price | $ 120 | $ 160 | $ 150 | $ 165 | $ 180 | $ 200 | $ 269 |
---|