Intel I7 Gaming Performance AnalysisRam and Uncore performance explored
Page 1 / 13 11/12/10
Introduction, Test Methodology and Set Up
NOTE : If you are not familiar with Intel's I7 platform, it is recommended that you read an introductory I7 article explaining the basic features of an I7 system, its capabilities and overclocking methods. One such article can be read here.
Intel's I7 platform has been around for more than two years now. Since its release in 2008, we have seen numerous reviews looking into the performance and capabilities of the platform. Many reviewers have looked at how CPU frequencies impact various applications. Similarly, RAM frequencies and timings have also been looked at in various articles that can be found all over the internet. However, one thing that seems to have been under investigated is to look at certain aspects of an i7 platform in isolation. By isolation, we mean that one aspect (such as the CPU frequency) is tweaked while all other aspects remain unchanged. Isolation is easy to accomplish with CPU frequency, by just changing the CPU multiplier. However, RAM and Uncore frequencies are a different story all together. Before going further, lets make a list of what aspects of this platform play a role in determining the overall capabilities of the system:
1) CPU Frequency
2) RAM Multiplier
3) RAM Frequency
4) RAM Timings
5) Uncore Multiplier
6) Uncore Frequency
Notice that, multipliers have been listed separate from actual frequencies, since both can have individual affects on the system performance. For a given RAM frequency, two different multipliers (and hence different Base clocks) could result in vast variations in performance. The Uncore frequency consists of the memory controller and L3 Cache and intel strongly suggests that it must be at least twice the RAM frequency. The Uncore multiplier can be raised to give a higher uncore frequency, but this should not necessarily result in a faster set up.
This article is by no means exhaustive in terms of performance investigation. In fact, we will only be looking at gaming performance under some very basic scenarios, yet trying to gain a decent insight into various aspects of core i7 performance. The two things that remained constant through out the tests are the Base Clock at 181 MHZ and the CPU multiplier at 21X to give a CPU frequency of 3.81 GHZ. The tweaking revolved around changing the RAM multiplier, Uncore multiplier and Ram timings. The results that follow will be presented in three graphs per benchmark test:
Combined effect of RAM and Uncore frequency: This was accomplished by using RAM multipliers of 2:10, 2:8, and 2:6. In each case, the Uncore multiplier and hence frequency was double the RAM multiplier and Frequency.
Uncore frequency effect: Here, the RAM multiplier was kept at 2:8 to give a RAM frequency of 1450 MHZ. in addition to the 2:16 Uncore divider, a test was run at a divider of 2:20.
Ram timings: Changing the RAM timings is the only way to isolate RAM performance from other aspects. RAM was kept at 2:10, ~ 1810 MHZ and timings were changed from 7-7-7-18 to 10-10-10-18. The test setup is as follows:
I7 920 @ ~ 3.80 GHZ
3 X 2 GB Patriot Viper 2000 MHZ 9-9-9-18
EVGA X58 Motherboard
3 X Sapphire ATI Radeon 5870 2GB GDDR5 (Catalyst 10.10)
150 GB 10,000 RPM Velociraptor
Dell 3008wfp 30" LCD
Antec 1200 Watt PSU
Windows 7 64 bit
The RAM utilized for this test is rated at 2000 MHZ. However achieving 2000 MHZ on a highly overclocked I7 requires some serious uncore frequency voltage boosts. Going to extreme uncore voltages without water cooling is not a good idea. So instead, a compromise was reached between the CPU / RAM frequencies. Another thing worth mentioning at this point is that the gaming benchmarks were performed in crossfire and Tri fire configurations at resolutions of 1920 X 1200 and 2560 X 1600 respectively. And lastly, the results that follow are RAW frames per second. In most cases they will not have any significant effect on actual gameplay, but are interesting to observe nonetheless.
Software Used
The following benchmarks were utilized for this article:
Batman Arkham Asylum
Crysis Warhead
Dirt 2
Far Cry 2
Just Cause 2
Lost Planet 2
Mafia 2
Metro 2033
Resident Evil 5
World in Conflict
Let us move on to some results now.
| NEXT |
|---|