borab
Čuven
- Učlanjen(a)
- 29.06.2000
- Poruke
- 8,578
- Poena
- 860
Na sajtu http://www.legionhardware.com/document.php?id=468, objavljen je interesantan clanak o uticaju kolicine memorije na igre. Najveci je dobitak prelazak sa 256Mb na 512Mb. Test je uradjen za 4 straije igre. Tu su i neka sporedna poredjenja.
Conlusion:
There are a lot of different results to discuss here so I am going to break it down. First of all SiSoft2005 shows absolutely no gains in memory bandwidth when increasing the memory capacity. Clearly the sample data is actually smaller than 256MB as there was no difference in bandwidth between 256MB and 2GB of memory. What did have a significant impact on performance was how the memory was configured for either dual or single-channel operation. The single-channel 64-bit mode produced roughly 2.7GB/s where as dual-channel operation allowed for almost 5GB/s. Surprisingly when moving into the real-world gaming there was for the most part very little difference in performance between single and dual-channel modes.
Given dual-channel DDR400 offers so much more bandwidth I was expecting to see huge frame rate gains when using this configuration. However, moving into Doom 3 this was not the case as dual-channel appeared to be no faster than single-channel. Despite the average frame rate being so similar the game did seem to play much smoother in dual-channel mode, which really is no surprise. Nevertheless, it was surprising to find that the 512MB, 1GB and 2GB capacities all rendered 60fps in Doom 3 with “No AA/AF” enabled. The only weakness here was the 256MB configuration which was extremely painful to test. The game just seemed to take forever to load and I honestly find it hard to believe the frame rate sustained an average of 53.2fps. There was no point in the demo where the game did not look extremely laggy.
Unlike the Doom 3 results Far Cry does favor the larger memory capacities. Though there was little difference between 1GB and 2GB’s of memory, 512MB was noticeably slower. The 256MB single-channel configuration was again extremely choppy. However, once again 1GB appears to be the sweet spot, matching the 2GB setup very well. The UT2004 results were much like those of Far Cry, showing very little difference between the 1GB and 2GB configurations. Even the 512MB setup produced much of the same performance we saw out of the larger capacities. Battlefield 2 was the big test here and again there was nothing separating the 1GB and 2GB configurations. The 512MB capacity was noticeably slower and made the game unplayable in my opinion.
The latest test looked at load times; this was something I did not include in the 1GB vs. 2GB article as I was purely interested in frame rate performance. Here we see the dual-channel memory technology come into play. The 512MB dual-channel setup was much faster than 2GB of single-channel memory. However, I was surprised to find that Battlefield 2 did not improve its load times when paired with 2GB’s of memory. Both the single and dual-channel modes showed negligible difference between the 1GB and 2GB load times. Even the 512MB dual-channel memory was only 2 seconds slower in completing the entire demo when compared to the 1GB setup.
These results are quite interesting as they are certainly not what I expected. After completing the 1GB vs. 2GB article I was disappointed that I was unable to show any benefits to using 2GB of memory. The reason I was so disappointed is because I know the benefits are there. Playing Battlefield 2 with a 2GB memory capacity is just so much better. The game runs smoothly from start to finish, where as 1GB seems to lag a lot in the beginning and then pause momentarily through-out the game. For this reason I was hoping the load time results would strongly favor the 2GB memory capacity. Sadly I have again failed to capture what makes 2GB memory capacity so much better in games such as Battlefield 2. Although the benefits are not evident through frame rate improvements or faster load times, I assure you they are there.
Reviewed By Steven Walton
Conlusion:
There are a lot of different results to discuss here so I am going to break it down. First of all SiSoft2005 shows absolutely no gains in memory bandwidth when increasing the memory capacity. Clearly the sample data is actually smaller than 256MB as there was no difference in bandwidth between 256MB and 2GB of memory. What did have a significant impact on performance was how the memory was configured for either dual or single-channel operation. The single-channel 64-bit mode produced roughly 2.7GB/s where as dual-channel operation allowed for almost 5GB/s. Surprisingly when moving into the real-world gaming there was for the most part very little difference in performance between single and dual-channel modes.
Given dual-channel DDR400 offers so much more bandwidth I was expecting to see huge frame rate gains when using this configuration. However, moving into Doom 3 this was not the case as dual-channel appeared to be no faster than single-channel. Despite the average frame rate being so similar the game did seem to play much smoother in dual-channel mode, which really is no surprise. Nevertheless, it was surprising to find that the 512MB, 1GB and 2GB capacities all rendered 60fps in Doom 3 with “No AA/AF” enabled. The only weakness here was the 256MB configuration which was extremely painful to test. The game just seemed to take forever to load and I honestly find it hard to believe the frame rate sustained an average of 53.2fps. There was no point in the demo where the game did not look extremely laggy.
Unlike the Doom 3 results Far Cry does favor the larger memory capacities. Though there was little difference between 1GB and 2GB’s of memory, 512MB was noticeably slower. The 256MB single-channel configuration was again extremely choppy. However, once again 1GB appears to be the sweet spot, matching the 2GB setup very well. The UT2004 results were much like those of Far Cry, showing very little difference between the 1GB and 2GB configurations. Even the 512MB setup produced much of the same performance we saw out of the larger capacities. Battlefield 2 was the big test here and again there was nothing separating the 1GB and 2GB configurations. The 512MB capacity was noticeably slower and made the game unplayable in my opinion.
The latest test looked at load times; this was something I did not include in the 1GB vs. 2GB article as I was purely interested in frame rate performance. Here we see the dual-channel memory technology come into play. The 512MB dual-channel setup was much faster than 2GB of single-channel memory. However, I was surprised to find that Battlefield 2 did not improve its load times when paired with 2GB’s of memory. Both the single and dual-channel modes showed negligible difference between the 1GB and 2GB load times. Even the 512MB dual-channel memory was only 2 seconds slower in completing the entire demo when compared to the 1GB setup.
These results are quite interesting as they are certainly not what I expected. After completing the 1GB vs. 2GB article I was disappointed that I was unable to show any benefits to using 2GB of memory. The reason I was so disappointed is because I know the benefits are there. Playing Battlefield 2 with a 2GB memory capacity is just so much better. The game runs smoothly from start to finish, where as 1GB seems to lag a lot in the beginning and then pause momentarily through-out the game. For this reason I was hoping the load time results would strongly favor the 2GB memory capacity. Sadly I have again failed to capture what makes 2GB memory capacity so much better in games such as Battlefield 2. Although the benefits are not evident through frame rate improvements or faster load times, I assure you they are there.
Reviewed By Steven Walton
Poslednja izmena:
