||Half-Life 2 CPU Shootout: AMD versus Intel
November 18, 2004 Chris Crazipper Angelini
Summary: Wondering which processor is best for Half-Life 2? Or perhaps you're curious to see how older Athlon XP processors perform with this game in comparison to Pentium 4 and Athlon 64 processors? In this article we've rounded up 17 different processors ranging from the Athlon XP 2100+ all the way up to the Athlon 64 FX-55 and Pentium 4 3.46GHz Extreme Edition. See how the CPUs compare to one another inside!
| Introduction||Page:: ( 1 / 8 )|
And now it’s out, available to the gaming enthusiasts who’ve waited so long for Half-Life 2 to be finished. It should come as no surprise that we’re revisiting CPU scaling; after all, if processor performance takes a backseat to graphics, there’s still time to fine-tune the upgrades on your holiday wish list.
We received plenty of feedback on the last processor scaling piece and we tried to consider it all for this story. Thus, we’ve added an intermediate resolution, presenting 800x600, 1280x1024, and 1600x1200. Each setting corresponds to a different set of quality options, reflecting a wide range of visual experiences.
The tests at 800x600 employ low model detail, low texture detail, no anti-aliasing, no bilinear filtering, simple water reflections, low shader detail, low shadow detail, and the DirectX 9.0 code path.
At 1280x1024, those settings rise to medium model detail, medium texture detail, 2x anti-aliasing, 2x anisotropic filtering, world reflective water detail, high shader detail, high shadow detail, and again, the DirectX 9.0 code path.
Then, 1600x1200 steps in to represent Half-Life 2 at its pinnacle. Here you’ll find high model detail, high texture detail, 4x anti-aliasing, 16x anisotropic filtering, world reflective water detail, high shader detail, high shadow detail, and naturally, DirectX 9.0.
We’ve also added more processors to cast more light on the various systems out there. There are 17 total this time around. And before you start wondering who’d pair an Athlon XP 2100+ with NVIDIA’s GeForce 6800 Ultra, anyway, bear in mind that we’re looking at processor performance here. By using a high-end graphics card, we minimize that component’s affect on the overall scores to the best of our ability. We used NVIDIA graphics cards because ATI still doesn’t have matching PCI Express and AGP products, which we need in order to normalize graphics performance between the Intel and AMD platforms.
| System Setup||Page:: ( 2 / 8 )|
AMD Athlon 64 FX-55
AMD Athlon 64 FX-53
AMD Athlon 64 4000+
AMD Athlon 64 3800+
AMD Athlon 64 3500+
AMD Athlon 64 3200+
AMD Athlon 64 3000+
AMD Athlon XP 3200+
AMD Athlon XP 2800+
AMD Athlon XP 2100+
Intel Pentium 4 3.46GHz Extreme Edition
Intel Pentium 4 3.4GHz Extreme Edition
Intel Pentium 4 570J 3.8GHz
Intel Pentium 4 560J 3.6GHz
Intel Pentium 4 550J 3.4GHz
Intel Pentium 4 540J 3.2GHz
Intel Pentium 4 520J 2.8GHz
MSI 7025 Socket 939 nForce3 Ultra Motherboard
ASUS A7N8X Socket A nForce2 Motherboard
Intel D925XECV2 LGA775 925X Motherboard
1GB Corsair 3-3-3-8 533MHz DDR2 Memory (2x512MB)
1GB Corsair 2-2-2-5 400MHz DDR Memory (2x512MB)
NVIDIA GeForce 6800 Ultra 256MB (AGP)
NVIDIA GeForce 6800 Ultra 256MB (PCI Express)
34GB Western Digital Raptor (10,000RPM, 8MB cache)
Windows XP Professional with Service Pack 2
Desktop resolution 1024x768, 32-bit color, 85Hz refresh
All power saving options were turned off, as were the Automatic Update and System Restore services. Graphics options under the ‘Performance’ tab were all disabled for maximum performance.
This demo is recorded outside the train station in City 17. There are relatively few actors and little to tax graphics hardware. Consequentially, the demo tends to be bound by host processor performance – ideal for testing high-end and mid-range CPUs
| Juggernauts: EE vs. FX||Page:: ( 3 / 8 )|
Intel Pentium 4 Extreme Edition versus AMD Athlon 64 FX
At 800x600, with all of the graphical detail minimized, there’s a notable difference between AMD’s Athlon 64 FX series and the Pentium 4 Extreme Edition. As we’ve seen through countless reviews, the FX is simply a better performer in games. The difference between AMD’s FX-55 and Intel’s own flagship, the 3.46GHz Extreme Edition, is greater than 18 percent.
That lead persists into 1280x1024, where the Athlon 64 FX-55 leads the 3.46GHz Extreme Edition by a margin of 14 percent. Then again, with both processors delivering in excess of 70 frames per second, they’re both considered playable still.
The upper echelon of processors is powerful indeed. So much so, in fact, that graphics horsepower starts to lag and all of the chips begin piling up on one another. The FX-55 still leads, but only by four percentage points. And at 60+ frames per second, the lead is hardly decisive. Fortunately, AMD still holds the crown for value (how’s that for a relative term, especially when we’re talking thousand dollar processors?). Its Athlon 64 FX-55 costs just over $800 online, while the 3.46GHz chip isn’t even registering yet.
| Mainstream Performance: Athlon 64 vs. Pentium 4||Page:: ( 4 / 8 )|
Intel Pentium 4 versus AMD Athlon 64
There’s a lot more to see here, especially at 800x600, where Athlon 64 processors, right down to the 3200+, outperform Intel’s Pentium 4 at up to 3.8GHz. The fastest Athlon 64 bests the speediest Pentium 4 by about 23 percent, and if you’re looking at prices, the Athlon 64 3500+ and Pentium 4 550J at 3.4GHz, both about $260, are separated by 15 percent in favor of AMD’s chip.
At 1280x1024 the 3.8GHz Pentium 4 improves its standings, though not considerably. And already, it’s looking like we’re somewhat limited by processor performance, especially with the slower chips.
Finally, there’s some equalization at 1600x1200, where the Athlon 64 3800+ outperforms Intel’s 3.8GHz Pentium 4 by about three percent. And once again, those are scores in excess of 60 frames per second with maximized graphical details. Thus, almost any of them will yield an enjoyable gaming experience.
| Value Gaming: Athlon XP versus Pentium 4||Page:: ( 5 / 8 )|
Intel Pentium 4 versus AMD Athlon XP
We’re seeing plenty of spread at 800x600 as the Athlon 64 3000+ puts more than 25 frames between itself and the older Athlon XP 2100+. At the same time though clearly it’s still possible to get respectable frame rates in Half-Life 2, even if you don’t own the latest processor technology.
Moving up to 1280x1024, the scores don’t change much and the order of placement remains the same, indicating that we’re indeed bound by processor performance in this test. Right down to the Athlon XP 2800+, it’s still possible to get more than 50 frames per second with minimal anti-aliasing and anisotropic filtering enabled.
Another small hit at 1600x1200 with all of the details maxed out confirms the processor limitation of our lower-end chips. You’re probably pushing it if you try to use an Athlon XP at that resolution, but the other processors all turn in surprisingly impressive results considering the game’s sheer detail at those settings and the age of that hardware (the Athlon XP 2800+ is more than two years old!).
| Low-Resolution Summary||Page:: ( 6 / 8 )|
All 17 CPU’s At 800x600
It’s hardly surprising that at 800x600 we see such diverse numbers. From the aging Athlon XP 2100+ at 40 frames per second to the high-end Athlon 64 FX-55 at more than twice that number, Half-Life 2 is, by and large, playable on all of these platforms. Of course, not all areas of the game are going to be purely limited by processor performance – we very specifically chose a scene light on shading and with relatively little character interaction. So, if you’re already sitting pretty with a 2.8GHz Pentium 4 and are debating an upgrade, a video card would probably be a smart choice. But that’s if you intend on playing at 800x600…
| High-Resolution Summary||Page:: ( 7 / 8 )|
All 17 CPUs at 1600x1200
The graph rounds out at 1600x1200, where you can see graphics performance factor into the equation a little more. AMD’s Athlon 64 FX-55 still takes its victory, but the margin is slim indeed. We’re still seeing processor-bound play here, though, as right down to the Athlon XP 2800+, you can still play the game. And that’s with all the goodies turned on! So, even at 1600x1200, we’re seeing the effects of processor-bound game play, except for the very fastest chips, which are limited by graphics. Your best bet here is to get a DirectX 9 card as long as you’re already using a processor in the 3GHz or equivalent range.
| Conclusion||Page:: ( 8 / 8 )|
There are areas that employ much more demanding shader detail, so expect graphics hardware to play a more significant role in the game than these tests suggest. For example, the canal level, where you pilot the jet-ski-like boat, relies heavily on water. That water shader will have a significant impact on performance, making it particularly important that you use the best video card available to you.
If you find yourself in a situation where your processor is older than any of those tested, it’s a great time to think about upgrading, especially if Half-Life 2 is a priority. NVIDIA’s nForce4 chipset is slowly materializing with support for PCI Express and some advanced storage options. Moreover, high-end PCI Express graphics cards, though still rare, will be rolling out prior to the end of 2004. And best of all, there are some remarkably inexpensive processors available, thanks in part to AMD’s transition to 90nm manufacturing. You can pick up a Socket 939 Athlon 64 3200+, a great performer in our tests, for less than $200. The 3.2GHz Pentium 4 is priced similarly, though it doesn’t benchmark quite as well.
But to revisit the question posed at the beginning of our story, does Half-Life 2 more closely resemble the scoring of Valve’s Video Stress Test or Counter-Strike: Source? Not to cop out here, but it actually looks like a balance between both. There are clearly processor-limited environments in the game, as demonstrated by our benchmarking here. And even at resolutions up to 1600x1200, there remains variance between different platforms. However, there are also much more graphically intensive scenes that tax video hardware.
If you’d be interested in looking at scaling characteristics for more graphically intensive levels, let us know. In the meantime, check out Brandon’s high-end graphics performance analysis for a better idea of how the latest 3D architectures handle Half-Life 2. Also keep an eye out for his mainstream graphics piece, which will tie all of these scaling stories together nicely.