DavidNYC Posted November 18, 2007 Share Posted November 18, 2007 I have an Asus M2N (Green) motherboard with 4x1GB of Corsair memory (QUAD2X4096-6400C5DHX) filling all 4 memory slots on the board. I noticed in the BIOS (the very latest version) that the auto-detect settings come up with 400MHz and 5-6-6-18 2T for timing. In Memtest v1.70, it shows the following: Athlon 64 x2 2612 MHz L1 cache: 128K 21,410 MB/s L2 cache: 1024K 5,319 MB/s Memory: 4095M 2,687 MB/s 435 MHz DDR870 I tried to manually set the the timings as stated in one of the posts for the same memory on an Asus M2N32-SLI Premium/Deluxe board. I set them to 800MHz manually, 5-5-5-15 as listed in the post (and I also tried 5-5-5-18 as listed in the memory's specifications). This causes the memory to show up as 400MHz in the BIOS, and it still shows as 435 MHz DDR870 in Memtest. Upon further research, I have read things saying that motherboard doesn't do well with all 4 slots filled. I have read things saying that if you're going to use more than 1 GB of memory or if you are going to fill up all 4 slots, you should be using registered memory. I checked the Compatibility Checker and the Asus M2N board isn't even listed. The other variants of M2N are listed though. I have used Memtest86+ v1.60 and v1.70 for days straight and there are 0 errors reported when tested at 5-6-6-18 or 5-5-5-18 or 5-5-5-15. Since there are 0 errors being reported, should I be happy? Link to comment Share on other sites More sharing options...
DerekT Posted November 18, 2007 Share Posted November 18, 2007 Since there are 0 errors being reported, should I be happy? I would be happy and when you decide to move to Core 2, your DRAM will then, with the right motherboard be able to run at a higher throughput. Link to comment Share on other sites More sharing options...
DavidNYC Posted November 18, 2007 Author Share Posted November 18, 2007 I'm not looking to overclock the memory. It shows up as 400MHz in the BIOS. So if it's dual-channel memory then that's 400MHz x 2 = 800 MHz. Memtest shows it as 435MHz / DDR 870. Why the difference? Link to comment Share on other sites More sharing options...
Corsair Employees RAM GUY Posted November 20, 2007 Corsair Employees Share Posted November 20, 2007 Most likely because of the multiplier in your CPU with odd multipliers the memory frequency will not work out exactly right as long as its stable I would not worry about it. Link to comment Share on other sites More sharing options...
DavidNYC Posted November 20, 2007 Author Share Posted November 20, 2007 It's definitely running stable, and I'm content with that. But I'm just wondering about the speed, for my own curiosity. Memory: 4095M 2,687 MB/s 435 MHz DDR870 Is a DDR800 supposed to show up as 400 MHz or 800 MHz on Memtest? Also, at 2,687 MB/s am I running at just a fraction of the potential speed of DDR2? When I make my next purchase of system parts, I'll be sure to use the Memory Configurator! Link to comment Share on other sites More sharing options...
Corsair Employees RAM GUY Posted November 20, 2007 Corsair Employees Share Posted November 20, 2007 Depending on the version of memory test it will show the DDR frequency DDR800 or the actual clock frequency 400 MHz, can you post a screen shot? Link to comment Share on other sites More sharing options...
DavidNYC Posted November 20, 2007 Author Share Posted November 20, 2007 I'll try to grab a screen shot with my camera phone, but I'm running Memtest+ v1.70. It's showing: Athlon 64 x2 2612 MHz L1 cache: 128K 21,410 MB/s L2 cache: 1024K 5,319 MB/s Memory: 4095M 2,687 MB/s 435 MHz DDR870 Link to comment Share on other sites More sharing options...
DerekT Posted November 20, 2007 Share Posted November 20, 2007 Memtest will show both DDR native speed and Dual Channel. Native = 435Mhz and Dual Channel = 870Mhz. Theoretical is the issue here. With your CPU, the difference will be in theoretical bandwidth, not actual throughput as your CPU<--> HT DRAM bus is nowhere even close to saturation. So benchmarks will show a difference but if you were to test your system vs 1066MHz DDR2 with a program load and run (with a digital stopwatch), you would find little if any difference. DDR2 was created by Intel for their legacy Northbridge Memory Controller which was bus starved. It is also why they installed large memory caches on the CPU. AMD installed their Memory Controller on the CPU and so uses the CPU HT and is not close to being starved even with DDR400Mhz DRAM. The issue of the move to DDR2 was that AMD was being informed by the Memory Manufacturing companies that they did NOT wish to run with two standards, that being DDR and DDR2 so AMD updated their on CPU Memory Controller to accept DDR2. It really was not necessary with regards to processor performance. One can not look the two architectures (Intel/AMD) the same as they access memory functions far differently. You are fine where you are. Link to comment Share on other sites More sharing options...
DavidNYC Posted November 20, 2007 Author Share Posted November 20, 2007 Here's a screenshot with a camera phone. Looks like the memory is running even slower than I claimed. Memtest has been running for 3 days straight. Link to comment Share on other sites More sharing options...
DavidNYC Posted November 20, 2007 Author Share Posted November 20, 2007 Thanks Derek. I know you can't just look at one performance statistic on its own. You need to understand the whole system and where the bottlenecks are in the whole picture. And as long as the memory bandwidth isn't a bottleneck, it's a non-issue. I just want to make sure the memory isn't mismatched with my processor/motherboard combination. Link to comment Share on other sites More sharing options...
DerekT Posted November 20, 2007 Share Posted November 20, 2007 Totally understandable. When you decide to migrate to an Intel Core 2 system, you will be able to migrate that DRAM and then you will get some real performance out of it. As it is, you are just fine until the move. 48 passes of memtest? I think that you are performing memtest far too long. :p: There are two reasons to run memtest. The first is to test the bandwidth for errata that is hardware based. It has been my personal opinion that a few passes are enough. If errors do not occur in the first two passes but do later on, then the issue is external to the DRAM ie. other hardware or thermal issues and removal of the heat in the system. The second is to test the thermal output of the dram and the ability of the dram/system to radiate the thermal output. I would perform 5 passes to test for the systems ability to deal with the thermal radiation. It has been my personal knowledge that errata produced after one pass is due to the DIMM's and system's inability to radiate and remove the heat thus producing errata due to thermal issues. Thus I run one pass and continue bandwith tightening until errata are produced on the first pass. Then I back off the tightening. I do this for bandwidth speed as well. Once I have found the sweet spot then I run 10+ runs of Memtest to see how well my system removes the heat that the DRAM produces. I run Prime95 in the long term (12+ hours) to test the final stablity of my machine. Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.