Company of Heroes - Performance
Test,
1024 x 768
E6300
P965
E6550
P965
E6300
P35 DDR2
E6550
P35 DDR2
E6300
P35 DDR3
E6550
P35 DDR3
| P35 to P965 - DDR2 | E6300 | E6550 | ||||
| 7x400 | 6x400 | 7x333 | 7x400 | 6x400 | 7x333 | |
| Difference | Difference | Difference | Difference | Difference | Difference | |
| Min | -5.41% | 2.70% | -15.79% | -7.89% | -13.16% | -5.41% |
| Avg | 2.25% | -0.20% | 0.88% | -3.58% | -1.28% | -1.89% |
| Max | -2.93% | -4.37% | -0.49% | 0.47% | -1.90% | -10.73% |
We've decided to give Company of Heroes a try as a benchmark, but we need to profile it first to see whether its more of a graphics benchmark, a system benchmark or both. This means comparing frame rates seen on the performance test and in actual gameplay, plus comparing frame rates reported by the game and FRAPS. Generally, we found frame rates measured with the performance test to be representative of frame rates in gameplay, but with a catch. You'll have to use the default camera angle, not the cinematic one (you can change camera angles by pressing and holding the 'Alt' key and moving your mouse). Cinematic skirmish replays use cinematic cameras and frame rates with such replays are much lower. Frame rates in replays (with FRAPS) also tend to fluctuate more. So, we've decided to use the built in performance test. However, we also saw the frame rates reported by the game to be different from FRAPS - much lower than we actually saw, so we're going to use FRAPS numbers instead.
Back to the article. It looks like Company of Heroes is also more graphically bound like F.E.A.R, Quake 4, Serious Sam and Tomb Raider: Legend. That's good if we were testing graphics cards, but they don't lend anything useful in this article.
| P965 | E6300 | E6550 | ||
| 7x400 | 6x400 | 7x400 | 6x400 | |
| Compared to 7x333 | Compared to 7x333 | Compared to 7x333 | Compared to 7x333 | |
| Min | -2.63% | -2.63% | 2.70% | 2.70% |
| Avg | 0.61% | -0.30% | 2.22% | -0.46% |
| Max | 0.99% | 1.48% | 4.39% | 2.44% |
| P35 | E6300 | E6550 | ||
| 7x400 | 6x400 | 7x400 | 6x400 | |
| Compared to 7x333 | Compared to 7x333 | Compared to 7x333 | Compared to 7x333 | |
| Min | 9.38% | 18.75% | 0.00% | -5.71% |
| Avg | 1.99% | -1.36% | 0.46% | 0.16% |
| Max | -1.49% | -2.48% | 17.49% | 12.57% |
| P35 DDR3 | E6300 | E6550 | ||
| 7x400 | 6x400 | 7x400 | 6x400 | |
| Compared to 7x333 | Compared to 7x333 | Compared to 7x333 | Compared to 7x333 | |
| Min | 0.00% | 0.00% | -2.63% | -5.26% |
| Avg | 1.95% | -0.06% | 2.52% | 1.19% |
| Max | -4.46% | 0.50% | 3.00% | 2.50% |
| P965 | E6550 to E6300 | |
| 7x400 | 6x400 | |
| Difference | Difference | |
| Min | 2.70% | 2.70% |
| Avg | 3.89% | 2.10% |
| Max | 4.39% | 1.94% |
| P35 | E6550 to E6300 | |
| 7x400 | 6x400 | |
| Difference | Difference | |
| Min | 0.00% | -13.16% |
| Avg | -2.03% | 0.99% |
| Max | 8.04% | 4.57% |
| P35 DDR3 | E6550 to E6300 | |
| 7x400 | 6x400 | |
| Difference | Difference | |
| Min | 0.00% | -2.70% |
| Avg | 1.84% | 2.53% |
| Max | 6.74% | 0.99% |
Nothing more need to be said here.
Conclusion:
Simply put, the P35 is a successor of Intel's P965 Express chipset. Many marketing pitch and reviews alike seem conclude that there's three things P35 offer over P965 - official support for 1333 MHz bus, Intel's upcoming 45nm Penryn processor family and DDR3 800 to 1333 MHz support. Of those three, only Penryn support is still up in the air. Hopefully Intel's promise of 'just a BIOS upgrade' is true. One extra we found to the list is the return of upward asynchronous memory settings with DDR2 800 modules when used with 1333 MHz FSB processors. If you're using modules with SPD containing support for higher clocks, you should be able to use limited asynchronous settings, but we think this is a moot point since Intel chipsets generally works best in synchronous mode anyway.There are things we slightly dislike about the P35. First and foremost is heat or more appropriately, heatsink. Intel claims P35 to be more power efficiency and dissipate less heat. Despite this, motherboard manufacturers are putting much larger heatsinks on P35 motherboards. Premium boards like Gigabyte's DQ6 tend to get overboard, making a simple heatsink installation more troublesome than it should be. If the chipset is truly power efficient, why does it seem to require more cooling?
Our second beef with P35 is double bank (double sided) memory modules support. You can use them with no trouble at all, but only if you're using two slots. With Vista and games like Battlefield 2, Company of Heroes and other memory hog games, you'll have to get 1 GB memory modules. If somewhere down the line you feel the need to get more memory, then you'll have to buy new ones. Some people might think this is not a big problem, most gamers and harsdware enthusiast seem to upgrade their rig every 6 months or so. Well, the problem is there is no such problem with P965 - you're free to use all four slots of memory with double bank, double sided memory with virtually no performance penalty compared to P35.
Now, to performance. Depending on application and the load (graphics or system), we can expect around 10 to 15 percent improvement on certain situations. Note that remark - certain situations.. From the benchmarks we've done, it's most likley you will not notice those improvements during gameplay. Performance differences occurs on situations where the system is 'holding up' the graphics card from rendering as fast as it can. As games get more predominantly graphics bound, these situations will be very rare and only on very high performing graphics card. Of course, users who buy / use these cards will likely push their graphics card with high resolutions and graphical details, so again the improvement will likely not be noticed.
From an overclocking perspective, P35 does promise more FSB frequency headroom and it seems to offer more performance at higher FSBs. This will vary from board to board and vendor to vendor, but that seems to be the general consensus. At the very least this should interest overclockers. Even with the coming X38, the P35 will likely still be the platform choice of many because of its price and availability. The next question is of course memory - it's clear that DDR2 is still the best choice (performance and pricewise) for users running at default clocks and up to 400 MHz FSB with synchronous memory settings. If you plan to use FSBs higher than 400 MHz, you can find more alternatives of overclocked DDR2 modules (PC8500, PC9000 etc) much more easily than their DDR3 siblings. If you're not satisfied with synchronous memory settings, the P35 with DDR3 running asynchronously should be interesting.
[Previous Page]
[Go to top]
[Home]