Prior to writing AMD FX-8350 Review: Does Piledriver Fix Bulldozer's Flaws?, I approached AMD to gauge the importance of testing its new FX processor under Windows 8. Naturally, if it'd change the chip's performance profile, I wanted to run those benchmarks. The company made it clear that Windows 8 and its scheduler should behave like a properly-patched installation of Windows 7. As a result, I didn't prioritize those numbers.
In light of Microsoft's recent note that Windows 8 would immediately receive a number of post-RTM updates that might affect performance, though, I'm going to try to run some new numbers on AMD's latest in the days to come. Until then, Thomas has a look at FX-8150 in a baseline install of Windows 7, a patched Windows 7, and the Windows 8 RTM.
--Chris Angelini
I find the courtship between hardware and software vendors to be particularly interesting. On one hand, you have Microsoft Windows, which was originally written to support Intel's x86 architecture. On the other, you have Microsoft's DirectX API, which graphics vendors design hardware to support. Along the way, there are meetings, committees, and then conferences to discuss what needs to show up in the next generation of hardware, how that'll affect software, and what developers need to do to better exploit the former with the latter.

Before AMD's Bulldozer architecture was even made public, our editor-in-chief was in Austin, TX asking AMD's engineers how Microsoft's Windows 7 would react to this module concept, which clearly needed smart scheduling in order to utilize on-die resources in the most effective way possible. After all, it'd be fairly easy for a "dumb" scheduler to have two threads run on one module, tying up shared resources as other modules say idle. AMD didn't have a good answer at the time, replying only that it was working with Microsoft to address the software side of its hardware dilemma. And at launch, we still had no solution.
Not long after, though, Microsoft introduced a pair of patches that, first, properly recognized Bulldozer-based FX and Opteron CPUs, spreading one thread to each module before back-filling a second thread to already-utilized modules. The second patch selectively disabled Core Parking in Windows 7 and Server 2008 R2, keeping the modules from entering a C6 sleep state.
Once those patches were made public, we revisited the Bulldozer architecture in AMD's FX-8150 After Two Windows 7 Hotfixes And UEFI Updates with the hope that Microsoft's adjustments would let the hardware really sing. Unfortunately, they really didn't. But in the conclusion of that story, we reminded you that developers high up in Microsoft's ranks were saying Windows 8 would perform differently than Windows 7, even fully patched.
Shared resources make it more difficult for FX-8150 to scale performance linearly.
It was a bummer, then, when AMD told us not to expect much from Windows 8 when it introduced FX-8350. But of course we wanted to go back and run benchmarks to follow up on our Bulldozer-based coverage. Can Microsoft's latest help make up some of the performance we were expecting to see back when FX-8150 launched, or are any possible operating system-related benefits already baked in?
| Test System Configuration | |
|---|---|
| CPU | AMD FX-8150 (Bulldozer): 3.60 GHz, 2 MB L2 Cache Per Module, 8 MB Shared L3 Cache, Socket AM3+ |
| CPU Cooler | Sunbeamtech Core-Contact Freezer w/Zalman ZM-STG1 Paste |
| Motherboard | Asus Sabertooth 990FX, BIOS 1304 (07/20/2012) |
| RAM | Kingston KHX1600C9D3K2/8GX: 8 GB DDR3-1600 CAS 9-9-9-27 |
| Graphics | AMD Radeon HD 6950 2 GB: 800 MHz GPU, GDDR5-5000 |
| Hard Drive | Samsung 470 Series MZ5PA256HMDR, 256 GB SSD |
| Sound | Integrated HD Audio |
| Network | Integrated Gigabit Networking |
| Power | Seasonic X760 SS-760KM: ATX12V v2.3, EPS12V, 80 PLUS Gold |
| Software | |
| OS | Microsoft Windows 7 Ultimate x64 SP1, Windows 8 RTM |
| Graphics | AMD Catalyst 12.8 |
| Chipset | AMD Platform Driver 3.0.825.0 |
We tested Windows 7 in two ways: first with Windows Update patches applied through August 2012, and then after adding Microsoft's hotfixes released shortly after Bulldozer's introduction. Those patches have to be installed manually, so our combination of configurations should demonstrate whether any automatic updates affect performance.
Carried over from our previous exploration of post-patch performance, Asus’ award-winning Sabertooth 990FX gets yet another firmware upgrade before participating in today’s tests.
We're throwing our full content creation and productivity benchmark suites, plus a few games, into testing the FX-8150. Heavily-threaded applications should demonstrate fairly similar performance between all three Windows configurations, since full utilization doesn't leave much room to improve scheduling. Programs that use fewer than eight threads are expected to enjoy the biggest boost from Microsoft's optimizations.
| Benchmark Configuration | |
|---|---|
| 3D Games | |
| Battlefield 3 | Campaign Mode, "Going Hunting" 90-second Fraps Test Set 1: Medium Quality Defaults (No AA, 4x AF) Test Set 2: Ultra Quality Defaults (4x AA, 16x AF) |
| DiRT Showdown | V1.0.0.0, In-Game Benchmark Test Set 1: Medium Quality Preset, No AA Test Set 2: Ultra Quality Preset, 8x AA |
| Skyrim | Update 1.5.26, Celedon Aethirborn Level 6, 25-seconds Fraps Test Set 1: DX11, Medium Details Defaults Test Set 2: DX11, Ultra-High Details Defaults |
| Adoby Creative Suite | |
| Adobe Photoshop CS6 (64-bit) | Version 13 x64: Filter 15.7 MB TIF Image: Radial Blur, Shape Blur, Median, Polar Coordinates |
| Adobe After Effects CS 6 | Version: CS5.5: Tom's Hardware Workload, SD project with three picture-in-picture streams, source video at 720p, Render Multiple Frames Simultaneously |
| Adobe Premiere Pro CS 5.5 | Video length 2 m 21s, Export to H.264 Blu-ray Source 960x720, Output 1280x720 |
| Adobe Acrobat X Professional | V10.0.0, PDF Creation from PowerPoint 2010 Presentation (3.6 MB) |
| Audio/Video Encoding | |
| iTunes | Version 10.4.1.10 x64: Audio CD (Terminator II SE), 53 minutes, default AAC format |
| Lame MP3 | Version 3.98.3: Audio CD "Terminator II SE", 53 min, convert WAV to MP3 audio format, Command: -b 160 --nores (160 Kb/s) |
| HandBrake CLI | Version: 0.9.8, Video: Video from Canon EOS 7D (1920x1080, 25 frames) 1 Minutes 22 Seconds, Audio: PCM-S16, 48 000 Hz, Two-Channel, to Video: AVC1 Audio: AAC (High Profile) |
| MainConcept Reference | Version: 2.2.0.5440: MPEG-2 to H.264, MainConcept H.264/AVC Codec, 28 sec HDTV 1920x1080 (MPEG-2), Audio: MPEG-2 (44.1 kHz, Two-Channel, 16-Bit, 224 Kb/s), Codec: H.264 Pro, Mode: PAL 50i (25 FPS), Profile: H.264 BD HDMV |
| Productivity | |
| Blender | Version: 2.62: Syntax blender -b thg.blend -f 1, Resolution: 1920x1080, Anti-Aliasing: 8x, Render: THG.blend frame 1, Cycles renderer and internal tile renderer (9x9) |
| Visual Studio 2010 | Compile Chrome project (1/31/2012) with devenv.com /build Release |
| Autodesk 3ds Max 2012 | V14.1.0.328 x64: Space Flyby Mentalray, Frame 248, 1440x1080 |
| WinZip | Version 16.5 Pro: THG-Workload (1.30 GB) to ZIP, command line switches "-a -ez -p -r" |
| WinRAR | Version 4.20.0.0: THG-Workload (1.30 GB) to RAR, command line switches "winrar a -r -m3" |
| 7-Zip | Version 9.28: THG-Workload (1.30 GB) to .7z, command line switches "a -t7z -r -m0=LZMA2 -mx=5" |
| ABBYY FineReader | Version 10.0.102.82: Read PDF save to Doc, Source: Political Economy (J. Broadhurst 1842) 111 Pages |
Because Battlefield 3's single-player campaign is so graphics-bound, we don't expect optimizations for processor performance to have much effect. And as suspected, changes attributable to Windows 8 (or even a fully-patched Windows 7) don't yield any benefit.


In contrast to Battlefield 3, the hotfixes for Windows 7 appear to hurt performance in DiRT Showdown, which wasn't around when we took our first look at those Bulldozer architecture-specific operating system updates. Fortunately, Windows 8 ameliorates those issues, yielding just over 4% compared to Windows after a round of automatic updates.


The Elder Scrolls V: Skyrim is sensitive to processor performance. However, Windows 8 doesn't really affect its performance running on AMD's FX-8150.


Shorter run times mean better performance, and Adobe Photoshop appears to enjoy a small boost from the switch to Windows 8. And, as we saw in DiRT Showdown, the FX patches can actually send performance back the other way compared to the automatic updates installed through August of this year.

After Effects has demonstrated light threading in the past, so we might expect to see the Core Parking hotfix help performance a little bit if Windows 7 was bouncing that one thread around AMD's FX-8150. But we certainly didn't think we'd see the updates hurt performance.

Creating a PDF out of a PowerPoint presentation is also a single-threaded task, so it's interesting to see Windows 7 with the FX-optimized patches and Windows 8 do better than a standard Windows 7 install, which is the opposite of how After Effects behaved.

Windows 8 falls slightly behind in Premiere, but the difference is so small that we're not prepared to call any of the three configurations better than the others. This workload is very well-threaded, leaving little room for a scheduler- or idle resource-oriented speed-up.

Remember when we saw Acrobat X demonstrate the best performance under Windows 8 and the Bulldozer-optimized Windows 7 setup? Well, this time around, the single-threaded iTunes and Lame workloads favor Windows 8 exclusively. In searching for an explanation, one possibility springs to mind. Previous versions of Windows had a habit of bouncing threads from one core to another. If Windows 8 more effectively keeps threads on one core, alleviating the need to wake powered-down resources, perhaps it's able to affect single-threaded tasks more than we thought.

HandBrake and MainConcept fully-tax the FX-8150's four modules. So, we wouldn't expect either application to serve up a speed-up attributable to better scheduling.


Most of the tests in our benchmark suite are well-threaded, so there's very little to report from 3ds Max, FineReader, Blender, or Visual Studio 2010.




With OpenCL enabled in WinZip 16.5, the Windows 8-based configuration enjoys a nice performance improvement. WinRAR goes the other direction, and the setup with Microsoft's FX-optimized hotfixes does worse than the Windows 7 install current as of August 2012. The 7-Zip benchmark scores don't change much at all.


One of Microsoft's Windows 7 hotfixes alters the behavior of Core Parking, preventing AMD's Bulldozer modules from entering a C6 sleep state as often. Applying that patch has a quantifiable impact on power consumption, which, in turn, negatively affects the efficiency of FX-8150. In Windows 8, that efficiency conundrum is largely resolved, bringing power use back down to the level of Windows 7 with all automatic updates applied.
But how does the performance story shake out? After all, everyone who bought a Bulldozer-based CPU (and anyone now in the market for a Piledriver-based FX) is hoping for sizable gains.

Our testing shows that the FX-8150’s performance doesn't change much at all in the shift from Windows 7 to Windows 8.
The last time we looked at the impact of Microsoft's hotfixes was almost a year ago. In today's comparison, our baseline Windows 7 machine is loaded with a lot more patches from Windows Update, and they cumulatively seem to have a larger impact than the two manually-installed tweaks specific to AMD's Bulldozer architecture. In fact, overall performance is better without the hotfixes applied.
Installing Windows 8 does translate to slightly faster benchmark numbers, and without the power spike. But Microsoft's latest certainly cannot be expected to uncork results that many enthusiasts were hoping might have been bottled up by a poorly-optimized operating system. The onus for fixing Bulldozer was clearly on AMD, and we saw the company take a first step toward that goal with its Piledriver-based FX parts.
AMD told us not to expect any additional performance from FX-8350 prior to our review. But now that we know Microsoft plans to roll out performance- and power-altering updates to Windows 8 right away, rather than waiting for a service pack, there's renewed hope for even a small nudge forward.
Then again, software fixes for hardware problems are only viable when software was the problem originally. I remember once telling a programmer that his computer had a bad memory module. Rather than swapping it out, he charged in with determination to create a software-based solution. Had he identified the bad memory cells and kept his system from accessing them, he might have enjoyed about as much success as AMD waiting for Windows 8.
I eventually talked the programmer into fixing the hardware problem, rather than doggedly looking for a never-quite-finished software solution. AMD, do you see where I’m going with this?
