 | |||
|
|
|
|
|
 |
Half-Life 2 VST/Counter-Strike Source Beta CPU Shootout: AMD vs Intel September 28, 2004 Chris Crazipper Angelini |
|
 |
|
| |
Summary: Now that we've discussed Half-Life 2 video stress test and Counter-Strike: Source beta performance with dozens of graphics cards, it's now time to evaluate the CPUs. In this article we've rounded up Athlon 64 FX, Athlon 64, Athlon XP, Pentium 4, and Pentium 4 Extreme Edition processors. See which CPUs come out on top at various clock speeds, including clock-for-clock comparisons with Prescott and Northwood in this article!
 Introduction | Page:: ( 1 / 13 ) |
There’s plenty of debate online if you’re looking for somewhere to speculate on whether ATI or NVIDIA will perform better in Half-Life 2, once it’s released. We’ve run our own benchmarks, drawn some early conclusions, and seen Valve make various code changes to throw off some of our predictions. It’s still beta software; go figure. But while it’s not a secret that game performance varies wildly across different video card configurations, we also wanted to explore the effect your processor has on gaming speed, specifically in the Half-Life 2 Video Stress Test and Valve’s Counter-Strike: Source Beta.
We took a handful of the fastest processors currently available to determine if and when the Video Stress Test, included with the Counter-Strike: Source Beta test, and CS beta itself were hampered by processor performance. The contenders include Intel’s Pentium 4 Extreme Edition at 3.4 GHz, the 3.6 GHz Pentium 4 560, both 130nm and 90nm 3.4 GHz Northwood and Prescott chips, an older 3.2 GHz Pentium 4, AMD’s Athlon 64 FX-53 on Socket 939, the Athlon 64 3800+, the 3400+, and an Athlon XP 3200+ for good measure.
In order to make the distinction between processor and graphics performance limitations, we ran one set of tests at 800x600, using trilinear filtering, no anti-aliasing, model, texture, and shadow detail set to high, vsync disabled, and water detail set to world. Then we ran the same settings, only with 6x anti-aliasing and 8x anisotropic filtering at 1600x1200 to see if there were any changes in scaling.
Of course, all of that was after downloading the freshly updated beta several times at 889MB per transfer. While there’s economical genius to the way Valve plans to sell Half-Life 2 through Stream, it’s little more than a pain in the neck for loading several times on freshly formatted hard drives.
Also, in order to normalize for graphics performance, the PCI Express RADEON X800 XT we used was overclocked to 520/560MHz to match the AGP RADEON X800 XT PE using RadLinker. Any performance differences that spring up as a result of the 925X chipset’s DDR2 memory support simply have to be attributed to the platform itself
| System Setup | Page:: ( 2 / 13 ) |
System Setup
AMD Athlon 64 FX-53 (S939)
AMD Athlon 64 3800+ (S939)
AMD Athlon 64 3400+ (S754)
AMD Athlon XP 3200+
Intel Pentium 4 3.4GHz Extreme Edition (LGA775)
Intel Pentium 4 560 (3.6GHz; LGA775)
Intel Pentium 4 3.4GHz (90nm, S478)
Intel Pentium 4 3.4GHz (130nm, S478)
Intel Pentium 4 3.2GHz
MSI 6702E Socket 939 K8T800 Pro Motherboard
Gigabyte K8VNXP Socket 754 K8T800 Motherboard
ASUS P4C800-E Deluxe Socket 478 875P Motherboard
Intel D925XCV LGA775 925X Motherboard
1GB Kingston 4-4-4-10 675MHz DDR2 Memory (2x512MB)
1GB Corsair 2-2-2-5 400MHz DDR Memory (2x512MB)
ATI RADEON X800 XT PE (AGP 8x)
ATI RADEON X800 XT (PCI Express)
Catalyst 4.9
34GB Western Digital Raptor (10,000RPM, 8MB cache)
Windows XP Professional with Service Pack 2
DirectX 9.0c
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.
Benchmarks
Half-Life 2 Video Stress Test
Counter-Strike: Source Beta custom timedemo
| FX-53 vs. 3.4 EE: VST | Page:: ( 3 / 13 ) |
Athlon 64 FX-53 vs. Pentium 4 3.4GHz Extreme Edition – Video Stress Test
Notes
At low-resolutions, the Athlon 64 FX-53 is significantly faster than Intel’s flagship, the Pentium 4 3.4 GHz Extreme Edition. Granted, you wouldn’t expect a gamer to play such a visually engaging game at 800x600, nor would they play at 180 frames per second without either anti-aliasing or anisotropic filtering, which is why the test at 1600x1200 is more realistic.
Not surprisingly, the high-resolution test shows the FX-53 and Extreme Edition within one frame per second of each other. As you’ll see in some of the following graphs, all of the processors hit a similar wall between around 80 frames per second, indicating that something other than CPU performance is inhibiting frame rate, at least in the Video Stress Test.
Then again, the Stress Test isn’t necessarily representative of actual Half-Life game play, which is why we’ll also check out the Counter-Strike: Source beta.
| FX-53 vs. 3.4 EE: CS: Source | Page:: ( 4 / 13 ) |
Athlon 64 FX-53 vs. Pentium 4 3.4GHz Extreme Edition – Counter-Strike: Source
Notes
Surprisingly, the Counter-Strike benchmark is more demanding than the Video Stress Test, at least when it comes to processor scaling, with significantly lower frame rates. Even still, the FX-53 establishes an 11 percent lead over Intel’s Extreme Edition at 800x600. The move up to 1600x1200 sees that gap shrink substantially; however, the relatively small drop in performance moving from low resolution to high indicates that, indeed, we’re dealing with a processor-bound test.
| 3800+ vs. 3.6GHz: VST | Page:: ( 5 / 13 ) |
Athlon 64 3800+ vs. Pentium 4 560 3.6GHz: Video Stress Test
Notes
Once again, the Video Stress Test exposes a notable difference between AMD’s Athlon 64 and Intel’s Pentium 4 at low resolutions – right around 15 percent. The move to 1600x1200 eliminates that gap, suggesting a limitation of the graphics hardware.
| 3800+ vs. 3.6GHz: CS: Source | Page:: ( 6 / 13 ) |
Athlon 64 3800+ vs. Pentium 4 560 3.6GHz: Counter-Strike: Source
Notes
There’s certainly a pattern developing here. While both the Athlon 64 3800+ and Pentium 4 560 trail slightly behind their flagship counterparts, there’s a 12 percent gap between the two processor architectures at 800x600, which shrinks to just four percent at 1600x1200 and with all of the graphical details enabled.
| 3400+ vs. 3.4GHz: VST | Page:: ( 7 / 13 ) |
Athlon 64 3400+ vs. Pentium 4 3.4C: Video Stress Test
Notes
The lead established by AMD’s Athlon 64 3400+ at 800x600 disappears at 1600x1200 in the Video Stress Test. This serves to reaffirm that the benchmark test is bound by processor performance down low and more constrained by video processing at high resolutions.
| 3400+ vs. 3.4GHz: CS: Source | Page:: ( 8 / 13 ) |
Athlon 64 3400+ vs. Pentium 4 3.4C: Counter-Strike: Source
Notes
At 800x600, the Athlon 64 3400+ is eight percent faster than Intel’s Pentium 4 3.4C. Moving to 1600x1200, that lead shrinks to roughly six percent in favor of the Athlon 64. We’re once again looking the Counter-Strike benchmark being limited by processor performance, even as the Athlon 64 carves out a minimal victory.
| Northwood versus Prescott versus EE: VST | Page:: ( 9 / 13 ) |
Pentium 4 3.4GHz EE, 3.4C GHz, Pentium 4 3.4E GHz: VST
Notes
With three 3.4GHz processors in the Pentium 4 lineup, it’s easy to line them all up against the wall for evaluation. Clearly, at 800x600, the Extreme Edition and its 2MB L3 cache are able to usurp the other two 3.4GHz contenders by an astounding 18 percent margin. As we move to 1600x1200, though, the EE’s lead evaporates in the face of heightened graphical processing demands, and all three chips fall within a frame of each other.
| Pentium 4 3.4GHz Summary: CS: Source | Page:: ( 10 / 13 ) |
Pentium 4 3.4GHz EE, 3.4C GHz, Pentium 4 3.4E GHz: Counter-Strike: Source
Notes
The Counter-Strike: Source benchmarks tell a different tale, of course. Intel’s Pentium 4 3.4GHz Extreme Edition claims its victory once again, this time by 13 percent at 800x600. Then, at 1600x1200, the lead actually expands to 19 percent over the other 3.4GHz chips.
| XP 3200+ vs. 3.2C: VST | Page:: ( 11 / 13 ) |
Athlon XP 3200+ vs. Pentium 4 3.2C: VST
Notes
The Video Stress Test once again pits one processor against another at 800x600, while effectively forcing a draw at 1600x1200. Only this time, it’s the Pentium 4 on top, while the Athlon XP trails in the low-res test. The difference is only six percent, but it’s certainly a change of pace from the Athlon 64 scores seen previously.
| XP 3200+ vs. 3.2C: CS: Source | Page:: ( 12 / 13 ) |
Athlon XP 3200+ vs. Pentium 4 3.4C: Counter-Strike: Source
Notes
As we start moving down to slower processors (if you can consider 3.2GHz and the fastest Athlon XP available slow), there’s less separation in the Counter-Strike: Source test, even at 800x600. Naturally, performance at 1600x1200 with these chips is dead even. With that said, CPU-bound or not, 53 frames per second is still very much playable, and that’s with full anti-aliasing and anisotropic filtering.
| Conclusion | Page:: ( 13 / 13 ) |
The good news for all of those Athlon 64 owners is that, without having to go out and buy new DDR2 memory modules and a PCI Express video card, you’ll likely enjoy the best of Half-Life 2, once it’s released. In those tests that weren’t constrained by graphics performance, all variants of the Athlon 64 architecture dominated. Even tests that were graphics-limited often favored AMD’s Athlon 64 to a small extent.
Intel’s Pentium 4 doesn’t do as well in the low-resolution tests. But it mostly catches up in the high-resolution metrics, which are largely held back by graphics performance (in the VST) or processor speed (in CS: S). And while the Extreme Edition is severely overpriced for the performance it delivers in these tests (never mind that it’s L3 cache does prove advantageous in many of the benchmarks), it’s worth noting that the 560 (3.6GHz) is officially priced $226 below AMD’s Athlon 64 3800+; $100 less that the A64 3800+ when looking at current street prices. You’ll need to buy a new motherboard, graphics card, and possibly DDR2 memory modules to support it, of course. That’s just par for the course when it comes to upgrading though, right?
Falling in just below the Pentium 4 is AMD’s Athlon XP. It’s still a potent piece of silicon – the 3.2GHz Pentium 4 only slightly beats it – but it doesn’t deliver the same compelling frame rates as the Athlon 64 3400+, which runs at the exact same clock frequency.
When you look at the big picture, Counter-Strike: Source runs well at 1600x1200, even with 6x AA and 8x AF enabled on the latest graphics hardware and an Athlon XP processor. And while scores at 800x600 might give you the impression an Athlon 64 will yield a major performance advantage, those high-resolution scores show it just isn’t so. Even still, the fact that there’s variance between the Pentium 4 lineup at 1600x1200 indicates the Counter-Strike test is still CPU-bound.
On the other hand, the Video Stress Test, known to be shader-heavy, is very reliant on graphics performance. No matter which processor you’re looking at, they all score the same at 1600x1200, while varying at 800x600. That naturally won’t reflect actual game play, so it remains to be seen if Half-Life 2 incorporates more of the graphically intensive detail of VST or the processor-bound game play of CS.