Oblivion Athlon 64 CPU Performance April 25, 2006

Introduction

Page:: ( 1 / 15 )

That’s a quick summary of what we’ve learned so far. Up for today we’re going to be taking a look at how big of a role the CPU plays on Oblivion performance. Bethesda has claimed that Oblivion takes advantage of today’s latest dual-core processors, making it one of the first games on the market to take advantage of this feature out-of-the-box, no patches are necessary. We were curious to see how much of an impact dual-core CPUs have on performance, if any. In addition, we’ve also run CPU scaling tests over a wide range of clock speeds, as well as L2 cache tests, to see how much of a performance improvement processors with a larger 1MB L2 cache has on performance.

One area where this article is a little different than the others is in its scope. Whereas the previous two articles were more like shootouts with ATI and NVIDIA going head-to-head, in this article we’re going to attempt to provide more of a guide to help you determine what type of processor you should be looking for. To keep this simple, we’ve decided to stick solely with AMD’s Athlon 64 line of CPUs. Why AMD over Intel? Since their introduction, AMD’s Athlon 64 CPUs have been a huge hit with gamers due to their superior performance and competitive pricing; AMD’s Athlon 64 CPUs also tend to run cooler than competing Intel processors. As a result, most gamers in the market to upgrade have been plucking up AMD CPUs: for most it’s not a question of which CPU they should get AMD or Intel, but which AMD CPU to buy? Hopefully this article will help answer some of those questions.

Before we get started on the benchmarks though, we’d like to provide a quick refresher for those who may not have the feature set of AMD’s current CPU lineup memorized. As you look over the benchmarks on the following pages, those of you who aren’t as well-versed in AMD’s model number methodology may want to flip back to this chart for reference:

AMD Socket 939 CPU Comparison
Model Number	Frequency	L2 Cache Size
X2 4800+	2.4GHz	1MB per core
X2 4600+	2.4GHz	512KB per core
X2 4400+	2.2GHz	1MB per core
X2 4200+	2.2GHz	512KB per core
X2 3800+	2.0Ghz	512KB per core
FX-60	2.6GHz	1MB per core
FX-57	2.8GHz	1MB
Opteron 165	1.8GHz	1MB per core
A64 4000+	2.4GHz	1MB
A64 3800+	2.4GHz	512KB
A64 3700+	2.2GHz	1MB
A64 3500+	2.2GHz	512KB
A64 3200+	2.0GHz	512KB
A64 3000+	1.8GHz	512KB

Test Systems

Page:: ( 2 / 15 )

System Setup

AMD Athlon 64 FX-57
AMD Athlon 64 FX-60
AMD Athlon 64 4000+
AMD Athlon 64 3800+
AMD Athlon 64 3500+
AMD Athlon 64 X2 4200+

ASUS A8R32-MVP Deluxe (CrossFire Xpress 3200)

2GB OCZ DDR400 SDRAM

ATI Radeon X1900 XTX
Catalyst 6.4

250GB Maxtor Hard Drive Maxline III SATA Hard Drive w/16MB Cache

Windows XP Professional SP1

DirectX 9.0c

Benchmarks

Elder Scrolls IV: Oblivion

Notes

Testing CPUs for their performance is entirely different than testing video cards. Lower resolutions with less visual detail (settings) are required in order to stress the CPU’s performance as much as possible – this directly contrasts the way most people game nowadays: with the highest screen resolution and graphics settings as possible, often mixed in with a little bit of anti-aliasing and/or anisotropic filtering. By cranking up these eye candy settings, the burden shifts from the processor to the graphics card, thus making it less of an ideal CPU test.

Because of this we faced an interesting dilemma, since this article is focused on CPU performance, should we just test at low resolutions like 640x480 and 800x600 with the Oblivion’s graphics settings turned down to their minimums? Or should we test in a way that’s similar to how people will actually play the game?

We decided to settle on a compromise, running one test at 800x600 (no AA/AF or HDR lighting) with the same graphics settings we used in Part 2 of our Oblivion performance articles, while also testing at 1280x1024 and 1600x1200 with HDR and much higher quality graphics settings, but not quite the maximum settings we used in Part 1. We’ve provided a copy of our Oblivion.ini file here.

Dual-core Performance Mountains

Page:: ( 3 / 15 )

The #1 question we’ve received when it comes to CPU performance with Oblivion is without a doubt if dual-core processors are capable of outperforming older single-core processors in this game. And if so, by how much? Bethesda has already gone on record stating that the game would take advantage of dual-core while it was still in development, and we’ve all seen by now how much Quake 4 benefits from dual-core processors, so everyone’s been eager to see if Oblivion yields similar performance improvements with multi-core CPUs.

One thing a lot of dual-core users have been reporting are performance enhancements with the following tweaks:

bUseThreadedBlood=1
bUseThreadedMorpher=1
bUseThreadedTempEffects=1
bUseThreadedParticleSystem=1
bUseMultiThreadedTrees=1
bUseMultiThreadedFaceGen=1
iNumHavokThreads=5
iThreads=9
iOpenMPLevel=10

We couldn’t find one case where the aforementioned settings improved performance for dual-core CPUs, the performance impact in our testing by turning on/off these settings was negligible for the most part. One thing we noticed is that a couple of those tweaks, bUseMultiThreadedTrees=1 and bUseMultiThreadedFaceGen=1 are already turned on by default, regardless of whether you have a dual-core processor or not. It’s possible that our demos just aren’t capable of taking advantage of these settings, or perhaps they have nothing to do with multi-threading at all, as we certainly saw a performance increase with dual-core CPUs:

Oblivion – Direct3D

Oblivion Performance 800x600x32
CPU	Min FPS	Max FPS
X2 4800+	96	140
A64 4000+	81	120
X2 4200+	88	127
A64 3500+	75	111

Notes

The dual-core Athlon 64 X2 4800+ outperformed the similarly configured Athlon 64 4000+ by a whopping 16% at 800x600. We also ran tests at the slightly lower clock speed of 2.2GHz to see how much of an impact the slower clock speed and smaller cache plays a role on performance. Here the X2 4200+ outran the A64 3500+ CPU by a slightly slimmer margin of 13%. However, once Oblivion’s image quality settings are cranked up, and the screen resolution increases the performance difference between the various systems is insignificant.

Foliage Dual-core Performance

Page:: ( 4 / 15 )

Oblivion – Direct3D

Oblivion Performance 800x600x32
CPU	Min FPS	Max FPS
X2 4800+	88	98
A64 4000+	79	86
X2 4200+	82	89
A64 3500+	74	81

Note

Under the more stressful foliage demo the performance impact of dual-core isn’t as large, as the GPU is stressed much more than in our mountains area. Here the X2 4800+ and A64 4000+ are separated by 11%, while the margin between the X2 4200+ and A64 3500+ is trimmed to 10%.

Indoors Dual-core Performance

Page:: ( 5 / 15 )

Oblivion – Direct3D

Oblivion Performance 800x600x32
CPU	Min FPS	Max FPS
X2 4800+	108	245
A64 4000+	92	212
X2 4200+	95	224
A64 3500+	87	207

L2 Cache Size Performance - Mountains

Page:: ( 6 / 15 )

Oblivion – Direct3D

Oblivion Performance 800x600x32
CPU	Min FPS	Max FPS
A64 3800+	76	113
A64 4000+	81	120
X2 4200+	88	127
X2 4400+	90	132

Notes

With the 3800+ and 4000+ running at 2.4GHz and the X2 4200+ and 4400+ both clocked at 2.2GHz, we were curious to see how much of an improvement a larger L2 cache has on performance. As you can see, the difference is pretty profound at 800x600 with the Athlon 64 4000+ running about 5% faster than the 3800+, and the Athlon 64 X2 4400 running 4% faster than the X2 4200+ (it’s possible the slightly lower clock speed of the X2 chips may prevent that margin from being a little larger). Once you crank up the res to 1280x1024 and crank up the IQ settings though, performance is limited by the X1900 XTX video card, the system’s perform the same regardless of the amount of L2 cache inside the CPU.

Foliage L2 Cache Results

Page:: ( 7 / 15 )

Oblivion – Direct3D

Oblivion Performance 800x600x32
CPU	Min FPS	Max FPS
A64 3800+	76	82
A64 4000+	79	86
X2 4200+	82	89
X2 4400+	83	92

Notes

Under the more stressful demands of our foliage demo, the margins are a little slimmer between the two X2 chips at just 3% in favor of the X2 4400+, which has 1MB L2 cache per core, while the Athlon 64 CPUs remain separated by a margin of 5%.

Indoors Area L2 Cache Results

Page:: ( 8 / 15 )

Oblivion – Direct3D

Oblivion Performance 800x600x32
CPU	Min FPS	Max FPS
A64 3800+	85	206
A64 4000+	92	212
X2 4200+	95	224
X2 4400+	102	234

Notes

The X2 4400+ sees a nice performance improvement in indoors testing thanks to its larger L2 cache, it’s running 5% faster than the X2 4200+ CPU. Both single-core processors remain separated by a margin of 5%.

Clock Speed Scaling – Mountains

Page:: ( 9 / 15 )

Oblivion – Direct3D

Oblivion Performance 800x600x32
CPU	Min FPS	Max FPS
A64 2.0GHz	72	109
A64 2.2GHz	77	117
A64 2.4GHz	81	120
A64 2.6GHz	88	130
A64 2.8GHz	91	137
A64 X2 2.2GHz	90	132
A64 X2 2.4GHz	96	140
A64 X2 2.6GHz	101	147

Notes

At 800x600, clock speed definitely plays a large role in Oblivion’s performance, with the Athlon 64 series really showing improved performance once you make the jump from 2.4GHz to 2.6GHz, here we saw an improvement of nearly 9% (although the transition from 2.0GHz to 2.2GHz also shows a nice performance 5% spike).

For the X2 CPUs, we saw the greatest improvement when going from 2.2GHz to 2.4GHz.

Foliage MHz Scaling

Page:: ( 10 / 15 )

Oblivion – Direct3D

Oblivion Performance 800x600x32
CPU	Min FPS	Max FPS
A64 2.0GHz	72	79
A64 2.2GHz	78	85
A64 2.4GHz	79	86
A64 2.6GHz	84	94
A64 2.8GHz	89	98
A64 X2 2.2GHz	83	92
A64 X2 2.4GHz	88	98
A64 X2 2.6GHz	89	102

Notes

Under the more intensive foliage area, the results fall in line with what we just saw in mountains, with the biggest improvement in terms of percentage occurring during the transition from 2.4GHz to 2.6GHz for the Athlon 64 line of CPUs, while the X2 chips saw the highest improvement when going from 2.0GHz to 2.2GHz.

Indoors Clock Speed Scaling

Page:: ( 11 / 15 )

Oblivion – Direct3D

Oblivion Performance 800x600x32
CPU	Min FPS	Max FPS
A64 2.0GHz	83	196
A64 2.2GHz	89	212
A64 2.4GHz	92	212
A64 2.6GHz	101	233
A64 2.8GHz	107	245
A64 X2 2.2GHz	102	234
A64 X2 2.4GHz	108	245
A64 X2 2.6GHz	114	260

Oblivion CPU Performance – Mountains

Page:: ( 12 / 15 )

Oblivion – Direct3D

CPU Performance Roundup – Foliage

Page:: ( 13 / 15 )

Oblivion – Direct3D

CPU Performance Roundup – Indoors

Page:: ( 14 / 15 )

Oblivion – Direct3D

Final thoughts

Page:: ( 15 / 15 )

Oblivion also plays nice with processors that run at higher clock speeds as long as you keep the screen resolution and graphics settings low: again once you’re running at high-res with HDR and all the graphics settings are turned up, it doesn’t make one bit of a difference what kind of processor you have inside your system.

In this regard, picking the right graphics card for Oblivion is going to be very important. In case you didn’t know, we’ve already got a pair of articles on the topic that should hopefully be helpful in helping you choose which graphics card is right for your budget.

Before we go, we want to make one thing clear: the CPU still plays a critical role in the overall performance of your system. While you may not see much of a performance impact from having one of AMD’s new dual-core CPUs if you’re gaming at 16x12 with all the eye candy turned on you will see tangible performance benefits if you find you’re into audio/video encoding or lots of multitasking in general. Game developers will also increasingly take advantage of multi-core processing to handle tasks like more advanced AI and sound, as well as physics calculations. As dual-core processors become more prevalent you’ll begin to wonder how you got by on just one processor core. That day isn’t here quite yet though.

If you’re in the market to upgrade your CPU for Oblivion but don’t want to spend a lot of money, obviously the 3500+ delivered an awesome price/performance ratio, particularly in light of the high-res results. If you can afford to spend a little more, around $330 (a little over $100 more than the 3500+), AMD’s Opteron 165 CPU is a great value. The Opteron 165 is a dual-core CPU and runs at just 1.8GHz, 400MHz slower than the 3500+, but ships with 1MB of L2 cache per core and is known for being an excellent overclocker, often running at speeds in excess of 2.3/2.4GHz on air cooling. The most remarkable part is that these chips used to sell for just under $300, making them an absolute steal! AMD has wizened up and raised the price on these parts, but with Opteron 165 CPUs selling online at right around $320-$330 it’s still a terrific value, especially once you factor in its overclocking potential and consider the fact that X2 3800+ CPUs ship at higher clock speeds (2.0GHz) but with only 512KB of L2 cache per core. You can easily make up the 200MHz clock speed difference with the Opteron 165, but there’s no way you can ever drop 1MB of cache into an X2 3800+.

Hopefully once you combine this article with our previous video articles you should have a pretty good idea of the performance potential you can get from these upgrades. But what about the AGP market? Sounds like another article idea to me!