If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below. |
|
|
Thread Tools | Display Modes |
#12
|
|||
|
|||
In article , Lou
wrote: Sure enough - and it's only $79.99 with free shipping straight from their web site! I notice that it is only CAS-3; I'd be more comfortable with 2 or 2.5. Even though I have a 166MHz. FSB CPU (Athlon 2600+ Barton) I can run the memory at 200 MHz. async. for testing with memtest86 during the return period. Thanks, Louis If you buy two and run it dual channel 3-3-3-8, tell us whether it passes Prime95 (mixed) torture test or not, at DDR400. That is the "acid test". The CT6464Z40B PC3200 module is likely based on a MT46V32M8-5B chip. On page 59, CL=3 at 5ns, Trcd is 15ns, and that equals 3 ticks of 5ns each, Trp is also 15ns, so that is 3 ticks. Tras is 40ns min, and that is 8 ticks of 5ns each. That is where I came up with the guess of 3-3-3-8 above. (When you do the conversion math, you round up to the nearest whole tick for all the parameters except CL, as CL is rated in half tick increments.) http://download.micron.com/pdf/datas...DDRx4x8x16.pdf You can use an auto setting in the BIOS, and then verify the memory timing in Windows using http://www.cpuid.com/cpuz.php . The BIOS should find the 3-3-3-8 parameters in the SPD EEPROM chip (the tiny eight pin device) on the DIMM. Set Vdimm to = 2.6V, as 2.6V is the minimum for operation at PC3200 speeds. Paul On Tue, 11 Jan 2005 21:32:08 -0000, "Matt" wrote: "Paul Busby" wrote in message ... Thus spake Paul: I use Ballistix PC3200 at DDR400 2-2-2-6 and find it is error free in dual channel mode. But it is a little more expensive than the Mushkin. If you only plan to ever run the memory at DDR333 (166MHz), you may not need to bother with the more expensive memories. Just what I wanted to know Paul, so thanks for that. I'm toying with the idea of replacing my XP1800+ with an XP2400 m & 2x512 PC3200 to replace my Crucial 2x256 sticks. My concern was that the Crucial RAM isn't recommended by Asus but do they update these recommendations - probably not for an "old" m/b such as this one. Crucial test and guarantee there own RAMS on each motherboard, so if you went through their RAM selector and selected the correct RAM for your motherboard and it did not work then you would be able to get it replaced. Matt |
#13
|
|||
|
|||
Thanks a lot Paul, I really appreciate it. I will read through some
of the info you provided links to and try to figure out what you said (you're over my head here). In the end then, you think it should run at 200 MHz. DDR in dual channel mode on my A7N8X-E Deluxe, right? Thanks again, Louis On Wed, 12 Jan 2005 06:07:10 -0500, (Paul) wrote: In article , Lou wrote: Are memory timings typically referred to in a specific order? You refer to running Balistix at 2-2-2-6. To which parameter does each number refer? In my BIOS I get the following timings: Column "A" = "AUTO" (166 MHz.) Column "B" = "BySPD" (200 MHz.) A B SDRAM Active Precharge Delay 7 8 SDRAM RAS to CAS Delay 3 3 SDRAM RAS Precharge Delay 3 3 SDRAM CAS Latency 2.5T 3T This is obviously in a different order than the way you state it. (BTW this is with two sticks of 256MB Kingston PC3200 memory and it seems to run fine at 400 MHz. async with an Athlon 2600+). I was planning to upgrade the RAM on my A7N8X Deluxe to two 512MB modules and pass my two 256MB modules down to my kids (I'm upgrading their PC with a new mobo & CPU). If I do that, I want capability to run the new 1GB @ DDR400 in case I upgrade to a 400 FSB processor or an Athlon64. Thanks everyone for all of the input. Louis Here is a Micron datasheet for a PC2700 DIMM. (Micron makes memory chips and they also make memory modules. The modules are sold by Crucial, their retail arm, and normal Crucial modules have a Micron sticker and a Crucial sticker on them. Trouble is, there is no web page that maps Crucial part number, to Micron datasheet, so you cannot easily take advantage of this fact.) http://download.micron.com/pdf/datas...32_64x64AG.pdf On page 2 of the datasheet, you can see they use "CL - Trcd - Trp" as their first three numbers. Not the same order as your BIOS screen. CL = CAS Latency, Trcd = RAS to CAS Delay, Trp = RAS Precharge Delay. The final parameter is Tras, which in the Micron datasheet is defined as "ACTIVE to PRECHARGE command". The minimum time is what is set in the BIOS, but larger times can be used, up to the point that the dynamic RAM cell gets discharged (70000 nanoseconds). The 8 chips used on that module (64Mx8 chips) are here. Page 78 shows how Tras is the sum of Trcd+2+Twr, meaning it is not a simple timing parameter, but a composite number, and on page 78, the write cycle seems to make Tras longer than on a read. http://download.micron.com/pdf/datas...DDRx4x8x16.pdf This slide set defines some of this stuff. See slide 24: http://www.corsairmemory.com/memory_basics/153707/ Terms are listed here as well: http://forums.techpowerup.com/showthread.php?t=28 HTH, Paul On Mon, 10 Jan 2005 13:41:25 -0500, (Paul) wrote: In article , Lou wrote: I'm building a new PC with an ASUS A7N8X-E Deluxe and an Athlon XP 2800+ (Barton). I want to install two 512MB PC3200 modules to run in dual channel mode at 200MHz. (or maybe 166 MHz. to match the bus frequency - I think I've read that with the Atholn XP, synchronous at the lower speed is actually faster). The problem is that I just can't find memory with the exact same part number as listed on the ASUS web site as having been tested. (I can't find a list for the -E Deluxe but there is one for the A7N8X Deluxe and I assume they are the same memory-wise). www.newegg.com has a two module kit (2 X 512) of TWINMOS for $140 and TWINMOS is on the list, but with no specific part number. www.ms4me.com has a two module kit (2 X 512) of Samsung memory for about $150 and htey have a 100% positive feedback rating on PriceWatch but they don't show the manufacturers full part number either. Has anybody here used either of these memories on an A7N8X / A7N8X-E Deluxe or have anyother advice for me. I appreciate it. Louis There is a thread here on TwinMOS: http://www.nforcershq.com/forum/viewtopic.php?t=49071 There are also plenty of other threads on nforcershq about memory: http://www.nforcershq.com/forum/viewtopic.php?t=51516 My suspicion, is that you'll find few problems running a couple of generic CAS3 sticks at 166MHz (DDR333). But, if you want to run at DDR400 (say when you are overclocking the FSB of your new processor, and running FSB and memory in sync), then expect trouble. If you want to play at DDR400, pick up some CAS2 (low latency) PC3200 memory. That seems to work better than buying CAS3 PC4000 type memory. Looking in the Newegg memory lists, this is the cheapest CAS2 product I can find. Mushkin PC3200 2x512 Blue Dual pack 2-3-3 memory for $182. On one of the threads on nforcershq.com, about memory, one user reported brands that worked and didn't work. It will take you a while to sift through all the threads, but that is the only way to find memory that actually works. When judging the threads, the users reporting their results should have used memtest86/prime95 to test the memory, and list the memory settings used. They should have used enough Vdimm to get the memory to work well (like 2.6 to 2.8V). You'll find many reports from people who are not using taxing configurations for their hardware (like only 166MHz), or from people who are unaware of the need to test the memory, and their experiences are best ignored. Start here, and click "search for all terms" to get logical AND in your searches: http://www.nforcershq.com/forum/search.php I use Ballistix PC3200 at DDR400 2-2-2-6 and find it is error free in dual channel mode. But it is a little more expensive than the Mushkin. If you only plan to ever run the memory at DDR333 (166MHz), you may not need to bother with the more expensive memories. I took a look at the Newegg customer reviews for some of their cheapest memories, and unfortunately there are very few reports from people who test their memories properly. Since the cheapest memories involve a mixture of stick designs (can change from lot to lot), there is really no way of knowing what you'll get. If you want to play "bargain memory" games, it is best to do that with a vendor having a generous return policy, or use a local vendor who will accept returns or allow upgrading. Make sure the memory has a warranty of longer than one year, as I've been burnt twice by memory failures (on cheap generic "sale" memory) just after the one year warranty expired. Dammit, I want to own my memory, not rent it :-) HTH, Paul |
#14
|
|||
|
|||
In article , Lou
wrote: Thanks a lot Paul, I really appreciate it. I will read through some of the info you provided links to and try to figure out what you said (you're over my head here). In the end then, you think it should run at 200 MHz. DDR in dual channel mode on my A7N8X-E Deluxe, right? Thanks again, Louis I've tried to pass on as much info as I've got, and that is, start with extensive reading on nforcershq.com, as to what works and doesn't. If you wish to take up Crucial on their CAS3 memory, and give it a try, you are welcome to try. I've used Crucial products with zero problems for other computers. Crucial may guarantee their memory, but in my _opinion_, CAS3 DDR400 memory is a _higher risk_ solution than finding some CAS2. I would either want some independent confirmation from someone who has tried some, and knows what they are doing, in terms of motherboard setup and testing. You'll find a lot of posts from people who are not taxing the hardware setup on the machine, and use their OS stability as a measure of how well the memory works, and that is not good methodology. I guess, stated in other terms, I'm still looking for _anyone_ using CAS3 DDR400 (PC3200 or higher) memory, who is able to make it work error free at 1:1 CPU and FSB400 setting for the processor (memtest86/prime95/superpi stable). It is my belief that the memory itself is not at fault, and the nforce2 Northbridge has specs that place a premium on memory performance. The nforce2 Northbridge seems to need better quality memory than some other chips (like some Intel dual channel solutions). Now, maybe some nforce2 chips leave the factory working better than mine does, so you never know. Somebody must have tested some CAS3 for the board, and I would start sifting through nforcershq.com search engine, to find those experiences. My choice of Ballistix was done, to get the job over and done with. (My generic RAM wouldn't run error free, and I wasn't about to waste another week doing more experiments.) As you can tell from the above description, I get uncomfortable recommending anything other than what I've tested, or can find a quality posting about testing experience. Memory is such a crapshoot (with the industry taking product with no markings on it, and stamping some bogus manufacturing info on it), that the lack of traceability to the original chip maker, makes it hard to get to the bottom of why there are so many problems. Some chipsets are quite tolerant of all this nonsense, and then I don't feel like I'm going out on a limb - with the A7N8X family, at FSB400/DDR400, that tolerance to crap ram is not there. YMMV. You may have noticed that in all of my posts so far, no CAS2.5 memory got mentioned. When I was doing my testing, I noticed that a setting of CAS2.5 and a setting of CAS3, were giving the same measured memory bandwidth numbers. Based on that, I cannot really be sure that buying a CAS2.5 DDR400 memory would necessarily function any better than CAS3. But again, only sifting through user reports will tell us whether nforce2 Northbridge can actually run at CAS2.5, or there is some issue that causes the BIOS to use some other setting. (Maybe I can try some testing later today, and reproduce that.) Paul On Wed, 12 Jan 2005 06:07:10 -0500, (Paul) wrote: In article , Lou wrote: Are memory timings typically referred to in a specific order? You refer to running Balistix at 2-2-2-6. To which parameter does each number refer? In my BIOS I get the following timings: Column "A" = "AUTO" (166 MHz.) Column "B" = "BySPD" (200 MHz.) A B SDRAM Active Precharge Delay 7 8 SDRAM RAS to CAS Delay 3 3 SDRAM RAS Precharge Delay 3 3 SDRAM CAS Latency 2.5T 3T This is obviously in a different order than the way you state it. (BTW this is with two sticks of 256MB Kingston PC3200 memory and it seems to run fine at 400 MHz. async with an Athlon 2600+). I was planning to upgrade the RAM on my A7N8X Deluxe to two 512MB modules and pass my two 256MB modules down to my kids (I'm upgrading their PC with a new mobo & CPU). If I do that, I want capability to run the new 1GB @ DDR400 in case I upgrade to a 400 FSB processor or an Athlon64. Thanks everyone for all of the input. Louis Here is a Micron datasheet for a PC2700 DIMM. (Micron makes memory chips and they also make memory modules. The modules are sold by Crucial, their retail arm, and normal Crucial modules have a Micron sticker and a Crucial sticker on them. Trouble is, there is no web page that maps Crucial part number, to Micron datasheet, so you cannot easily take advantage of this fact.) http://download.micron.com/pdf/datas...32_64x64AG.pdf On page 2 of the datasheet, you can see they use "CL - Trcd - Trp" as their first three numbers. Not the same order as your BIOS screen. CL = CAS Latency, Trcd = RAS to CAS Delay, Trp = RAS Precharge Delay. The final parameter is Tras, which in the Micron datasheet is defined as "ACTIVE to PRECHARGE command". The minimum time is what is set in the BIOS, but larger times can be used, up to the point that the dynamic RAM cell gets discharged (70000 nanoseconds). The 8 chips used on that module (64Mx8 chips) are here. Page 78 shows how Tras is the sum of Trcd+2+Twr, meaning it is not a simple timing parameter, but a composite number, and on page 78, the write cycle seems to make Tras longer than on a read. http://download.micron.com/pdf/datas...DDRx4x8x16.pdf This slide set defines some of this stuff. See slide 24: http://www.corsairmemory.com/memory_basics/153707/ Terms are listed here as well: http://forums.techpowerup.com/showthread.php?t=28 HTH, Paul On Mon, 10 Jan 2005 13:41:25 -0500, (Paul) wrote: In article , Lou wrote: I'm building a new PC with an ASUS A7N8X-E Deluxe and an Athlon XP 2800+ (Barton). I want to install two 512MB PC3200 modules to run in dual channel mode at 200MHz. (or maybe 166 MHz. to match the bus frequency - I think I've read that with the Atholn XP, synchronous at the lower speed is actually faster). The problem is that I just can't find memory with the exact same part number as listed on the ASUS web site as having been tested. (I can't find a list for the -E Deluxe but there is one for the A7N8X Deluxe and I assume they are the same memory-wise). www.newegg.com has a two module kit (2 X 512) of TWINMOS for $140 and TWINMOS is on the list, but with no specific part number. www.ms4me.com has a two module kit (2 X 512) of Samsung memory for about $150 and htey have a 100% positive feedback rating on PriceWatch but they don't show the manufacturers full part number either. Has anybody here used either of these memories on an A7N8X / A7N8X-E Deluxe or have anyother advice for me. I appreciate it. Louis There is a thread here on TwinMOS: http://www.nforcershq.com/forum/viewtopic.php?t=49071 There are also plenty of other threads on nforcershq about memory: http://www.nforcershq.com/forum/viewtopic.php?t=51516 My suspicion, is that you'll find few problems running a couple of generic CAS3 sticks at 166MHz (DDR333). But, if you want to run at DDR400 (say when you are overclocking the FSB of your new processor, and running FSB and memory in sync), then expect trouble. If you want to play at DDR400, pick up some CAS2 (low latency) PC3200 memory. That seems to work better than buying CAS3 PC4000 type memory. Looking in the Newegg memory lists, this is the cheapest CAS2 product I can find. Mushkin PC3200 2x512 Blue Dual pack 2-3-3 memory for $182. On one of the threads on nforcershq.com, about memory, one user reported brands that worked and didn't work. It will take you a while to sift through all the threads, but that is the only way to find memory that actually works. When judging the threads, the users reporting their results should have used memtest86/prime95 to test the memory, and list the memory settings used. They should have used enough Vdimm to get the memory to work well (like 2.6 to 2.8V). You'll find many reports from people who are not using taxing configurations for their hardware (like only 166MHz), or from people who are unaware of the need to test the memory, and their experiences are best ignored. Start here, and click "search for all terms" to get logical AND in your searches: http://www.nforcershq.com/forum/search.php I use Ballistix PC3200 at DDR400 2-2-2-6 and find it is error free in dual channel mode. But it is a little more expensive than the Mushkin. If you only plan to ever run the memory at DDR333 (166MHz), you may not need to bother with the more expensive memories. I took a look at the Newegg customer reviews for some of their cheapest memories, and unfortunately there are very few reports from people who test their memories properly. Since the cheapest memories involve a mixture of stick designs (can change from lot to lot), there is really no way of knowing what you'll get. If you want to play "bargain memory" games, it is best to do that with a vendor having a generous return policy, or use a local vendor who will accept returns or allow upgrading. Make sure the memory has a warranty of longer than one year, as I've been burnt twice by memory failures (on cheap generic "sale" memory) just after the one year warranty expired. Dammit, I want to own my memory, not rent it :-) HTH, Paul |
#15
|
|||
|
|||
Thanks again Paul. I understand your position and certainly am not
looking for "your word" that some particular RAM will work flawlessly. I went ahead and bit the bullet with Crucial today because of their money back guarantee and because I cannot afford premium memory right now. I talked to a sales guy about the compatibility guarantee and he said that if I set the FNS to run at 800 MHz. the RAM would run at 400 MHz. I said that I thought he was thinking of a Pentium IV mobo and that an Athlon XP mobo ran at 400 MHz. max. and he said "that will work fine - it's backward compatible so at that FSB speed it will automatically back down to PC2100 speed." OK, so he doesn't know any more than I do, but I chalked that up to him being a salesman and not a technician. Since I have a 333 MHz. FSB CPU (Athlon XP 2800+) I understand that I may get better performance running the memory synchronously at 333, but for testint the memory, will it be a valid stress test to run it asynchronously at 400 MHz. while running memtest86 and the other tests? Thanks (I'll try to leave tyou alone now!). Louis On Wed, 12 Jan 2005 14:57:21 -0500, (Paul) wrote: In article , Lou wrote: Thanks a lot Paul, I really appreciate it. I will read through some of the info you provided links to and try to figure out what you said (you're over my head here). In the end then, you think it should run at 200 MHz. DDR in dual channel mode on my A7N8X-E Deluxe, right? Thanks again, Louis I've tried to pass on as much info as I've got, and that is, start with extensive reading on nforcershq.com, as to what works and doesn't. If you wish to take up Crucial on their CAS3 memory, and give it a try, you are welcome to try. I've used Crucial products with zero problems for other computers. Crucial may guarantee their memory, but in my _opinion_, CAS3 DDR400 memory is a _higher risk_ solution than finding some CAS2. I would either want some independent confirmation from someone who has tried some, and knows what they are doing, in terms of motherboard setup and testing. You'll find a lot of posts from people who are not taxing the hardware setup on the machine, and use their OS stability as a measure of how well the memory works, and that is not good methodology. I guess, stated in other terms, I'm still looking for _anyone_ using CAS3 DDR400 (PC3200 or higher) memory, who is able to make it work error free at 1:1 CPU and FSB400 setting for the processor (memtest86/prime95/superpi stable). It is my belief that the memory itself is not at fault, and the nforce2 Northbridge has specs that place a premium on memory performance. The nforce2 Northbridge seems to need better quality memory than some other chips (like some Intel dual channel solutions). Now, maybe some nforce2 chips leave the factory working better than mine does, so you never know. Somebody must have tested some CAS3 for the board, and I would start sifting through nforcershq.com search engine, to find those experiences. My choice of Ballistix was done, to get the job over and done with. (My generic RAM wouldn't run error free, and I wasn't about to waste another week doing more experiments.) As you can tell from the above description, I get uncomfortable recommending anything other than what I've tested, or can find a quality posting about testing experience. Memory is such a crapshoot (with the industry taking product with no markings on it, and stamping some bogus manufacturing info on it), that the lack of traceability to the original chip maker, makes it hard to get to the bottom of why there are so many problems. Some chipsets are quite tolerant of all this nonsense, and then I don't feel like I'm going out on a limb - with the A7N8X family, at FSB400/DDR400, that tolerance to crap ram is not there. YMMV. You may have noticed that in all of my posts so far, no CAS2.5 memory got mentioned. When I was doing my testing, I noticed that a setting of CAS2.5 and a setting of CAS3, were giving the same measured memory bandwidth numbers. Based on that, I cannot really be sure that buying a CAS2.5 DDR400 memory would necessarily function any better than CAS3. But again, only sifting through user reports will tell us whether nforce2 Northbridge can actually run at CAS2.5, or there is some issue that causes the BIOS to use some other setting. (Maybe I can try some testing later today, and reproduce that.) Paul On Wed, 12 Jan 2005 06:07:10 -0500, (Paul) wrote: In article , Lou wrote: Are memory timings typically referred to in a specific order? You refer to running Balistix at 2-2-2-6. To which parameter does each number refer? In my BIOS I get the following timings: Column "A" = "AUTO" (166 MHz.) Column "B" = "BySPD" (200 MHz.) A B SDRAM Active Precharge Delay 7 8 SDRAM RAS to CAS Delay 3 3 SDRAM RAS Precharge Delay 3 3 SDRAM CAS Latency 2.5T 3T This is obviously in a different order than the way you state it. (BTW this is with two sticks of 256MB Kingston PC3200 memory and it seems to run fine at 400 MHz. async with an Athlon 2600+). I was planning to upgrade the RAM on my A7N8X Deluxe to two 512MB modules and pass my two 256MB modules down to my kids (I'm upgrading their PC with a new mobo & CPU). If I do that, I want capability to run the new 1GB @ DDR400 in case I upgrade to a 400 FSB processor or an Athlon64. Thanks everyone for all of the input. Louis Here is a Micron datasheet for a PC2700 DIMM. (Micron makes memory chips and they also make memory modules. The modules are sold by Crucial, their retail arm, and normal Crucial modules have a Micron sticker and a Crucial sticker on them. Trouble is, there is no web page that maps Crucial part number, to Micron datasheet, so you cannot easily take advantage of this fact.) http://download.micron.com/pdf/datas...32_64x64AG.pdf On page 2 of the datasheet, you can see they use "CL - Trcd - Trp" as their first three numbers. Not the same order as your BIOS screen. CL = CAS Latency, Trcd = RAS to CAS Delay, Trp = RAS Precharge Delay. The final parameter is Tras, which in the Micron datasheet is defined as "ACTIVE to PRECHARGE command". The minimum time is what is set in the BIOS, but larger times can be used, up to the point that the dynamic RAM cell gets discharged (70000 nanoseconds). The 8 chips used on that module (64Mx8 chips) are here. Page 78 shows how Tras is the sum of Trcd+2+Twr, meaning it is not a simple timing parameter, but a composite number, and on page 78, the write cycle seems to make Tras longer than on a read. http://download.micron.com/pdf/datas...DDRx4x8x16.pdf This slide set defines some of this stuff. See slide 24: http://www.corsairmemory.com/memory_basics/153707/ Terms are listed here as well: http://forums.techpowerup.com/showthread.php?t=28 HTH, Paul On Mon, 10 Jan 2005 13:41:25 -0500, (Paul) wrote: In article , Lou wrote: I'm building a new PC with an ASUS A7N8X-E Deluxe and an Athlon XP 2800+ (Barton). I want to install two 512MB PC3200 modules to run in dual channel mode at 200MHz. (or maybe 166 MHz. to match the bus frequency - I think I've read that with the Atholn XP, synchronous at the lower speed is actually faster). The problem is that I just can't find memory with the exact same part number as listed on the ASUS web site as having been tested. (I can't find a list for the -E Deluxe but there is one for the A7N8X Deluxe and I assume they are the same memory-wise). www.newegg.com has a two module kit (2 X 512) of TWINMOS for $140 and TWINMOS is on the list, but with no specific part number. www.ms4me.com has a two module kit (2 X 512) of Samsung memory for about $150 and htey have a 100% positive feedback rating on PriceWatch but they don't show the manufacturers full part number either. Has anybody here used either of these memories on an A7N8X / A7N8X-E Deluxe or have anyother advice for me. I appreciate it. Louis There is a thread here on TwinMOS: http://www.nforcershq.com/forum/viewtopic.php?t=49071 There are also plenty of other threads on nforcershq about memory: http://www.nforcershq.com/forum/viewtopic.php?t=51516 My suspicion, is that you'll find few problems running a couple of generic CAS3 sticks at 166MHz (DDR333). But, if you want to run at DDR400 (say when you are overclocking the FSB of your new processor, and running FSB and memory in sync), then expect trouble. If you want to play at DDR400, pick up some CAS2 (low latency) PC3200 memory. That seems to work better than buying CAS3 PC4000 type memory. Looking in the Newegg memory lists, this is the cheapest CAS2 product I can find. Mushkin PC3200 2x512 Blue Dual pack 2-3-3 memory for $182. On one of the threads on nforcershq.com, about memory, one user reported brands that worked and didn't work. It will take you a while to sift through all the threads, but that is the only way to find memory that actually works. When judging the threads, the users reporting their results should have used memtest86/prime95 to test the memory, and list the memory settings used. They should have used enough Vdimm to get the memory to work well (like 2.6 to 2.8V). You'll find many reports from people who are not using taxing configurations for their hardware (like only 166MHz), or from people who are unaware of the need to test the memory, and their experiences are best ignored. Start here, and click "search for all terms" to get logical AND in your searches: http://www.nforcershq.com/forum/search.php I use Ballistix PC3200 at DDR400 2-2-2-6 and find it is error free in dual channel mode. But it is a little more expensive than the Mushkin. If you only plan to ever run the memory at DDR333 (166MHz), you may not need to bother with the more expensive memories. I took a look at the Newegg customer reviews for some of their cheapest memories, and unfortunately there are very few reports from people who test their memories properly. Since the cheapest memories involve a mixture of stick designs (can change from lot to lot), there is really no way of knowing what you'll get. If you want to play "bargain memory" games, it is best to do that with a vendor having a generous return policy, or use a local vendor who will accept returns or allow upgrading. Make sure the memory has a warranty of longer than one year, as I've been burnt twice by memory failures (on cheap generic "sale" memory) just after the one year warranty expired. Dammit, I want to own my memory, not rent it :-) HTH, Paul |
#16
|
|||
|
|||
In article , Lou
wrote: Thanks again Paul. I understand your position and certainly am not looking for "your word" that some particular RAM will work flawlessly. I went ahead and bit the bullet with Crucial today because of their money back guarantee and because I cannot afford premium memory right now. I talked to a sales guy about the compatibility guarantee and he said that if I set the FNS to run at 800 MHz. the RAM would run at 400 MHz. I said that I thought he was thinking of a Pentium IV mobo and that an Athlon XP mobo ran at 400 MHz. max. and he said "that will work fine - it's backward compatible so at that FSB speed it will automatically back down to PC2100 speed." OK, so he doesn't know any more than I do, but I chalked that up to him being a salesman and not a technician. Since I have a 333 MHz. FSB CPU (Athlon XP 2800+) I understand that I may get better performance running the memory synchronously at 333, but for testint the memory, will it be a valid stress test to run it asynchronously at 400 MHz. while running memtest86 and the other tests? Thanks (I'll try to leave tyou alone now!). Louis snip Here are some results I gathered: Timings, where CAS Latency is the first number, followed by memory bandwidth, as detected by memtest86 version 1.4. The CPU conditions are given after that. 2-2-2-8 1511MB/sec (CPU=200x11, DDR400) 2.5-2-2-8 1485MB/sec (CPU=200x11, DDR400) 3-2-2-8 1485MB/sec (CPU=200x11, DDR400) 2.5-3-3-8 1435MB/sec (CPU=200x11, DDR400) 3-3-3-8 1435MB/sec (CPU=200x11, DDR400) What this seems to be showing me, is the Nvidia memory controller cannot start the memory transfer on a half cycle, or something like that. It certainly gives the appearance of running at CAS3, when CAS2.5 is selected. Even CPUZ reports CAS2.5, when the BIOS is set to CAS2.5, so I don't know exactly how this works. But the exactly equal memory bandwidths don't seem to be an accident. To duplicate your conditions, I tried the following with my memory set back to 2-2-2-8. The CPU was set to 166x12.5, which is as near as I can get to 2200MHz without hacking something. Memtest86 version 1.4 reports: CPU=166x12.5 Memory=100% setting 2-2-2-8 == 1254MB/sec CPU=166x12.5 Memory=120% setting 2-2-2-8 == 1236MB/sec So, in fact, running memory synchronous at 100%, is more efficient than running the memory higher. I suspect if the memory was lifted a lot higher, that could change to seeing an improvement. Now, the conventional wisdom is, you should run synchronous to avoid memory errors. I'm running four copies of Prime95 under Knoppix linux, and so far (10 minutes) there are no signs of error. So maybe the reports of problems running async were actually due to the speed problem with the nforce2, rather than a problem caused by running async in and of itself. It would appear that using 100% is the way you should go for long term performance, while if you want to test the memory at 120%, there should be no harm done. Running the memory at 166MHz should allow use of CAS2.5, while 200MHz takes you to CAS3 (i.e. the rated 3-3-3-8 performance). But as I observed in the memtest86 results above, I doubt the difference in the CAS setting will make a visible difference, as CAS2.5 seems to work exactly like CAS3. (On an Intel chipset, it might make a difference.) If Prime95 throws any errors, I'll post a followup. (I'm not posting from the test machine, so whatever happens to it doesn't matter :-) HTH, Paul |
#17
|
|||
|
|||
I guess, stated in other terms, I'm still looking for _anyone_ using CAS3 DDR400 (PC3200 or higher) memory, who is able to make it work error free at 1:1 CPU and FSB400 setting for the processor (memtest86/prime95/superpi stable). It is my belief that the memory itself is not at fault, and the nforce2 Northbridge has specs that place a premium on memory performance. The nforce2 Northbridge seems to need better quality memory than some other chips (like some Intel dual channel solutions). Now, maybe some nforce2 chips leave the factory working better than mine does, so you never know. Somebody must have tested some CAS3 for the board, and I would start sifting through nforcershq.com search engine, to find those experiences. My choice of Ballistix was done, to get the job over and done with. (My generic RAM wouldn't run error free, and I wasn't about to waste another week doing more experiments.) I am using :- 2*256 MB DDR400 (PC3200) Crucial 1*512 MB DDR400 (PC3200) Crucial in dual channel mode and overclocked to 210Mhz without any problems on my AN78X-E Dlx with my Barton XP3200+ Matt |
#18
|
|||
|
|||
I am using :- 2*256 MB DDR400 (PC3200) Crucial 1*512 MB DDR400 (PC3200) Crucial in dual channel mode and overclocked to 210Mhz without any problems on my AN78X-E Dlx with my Barton XP3200+ Matt Oh and I had to set RAM timings to SPD in the BIOS which reports 9,4,4,4,3T. If I run RAM at 200Mhz then BIOS reads 8,3,3,3,2.5T Matt |
#19
|
|||
|
|||
In article , "Matt"
wrote: I am using :- 2*256 MB DDR400 (PC3200) Crucial 1*512 MB DDR400 (PC3200) Crucial in dual channel mode and overclocked to 210Mhz without any problems on my AN78X-E Dlx with my Barton XP3200+ Matt Oh and I had to set RAM timings to SPD in the BIOS which reports 9,4,4,4,3T. If I run RAM at 200Mhz then BIOS reads 8,3,3,3,2.5T Matt Would that be 8,3,3,3T at 200MHz ? I thought that is what vanilla PC3200 Crucial runs at. The chips on a PC3200 DIMM could be Micron 46V32M8-5 , and the -5 refers to the 5ns for a 200MHz clock, or DDR400. (At least the 512MB stick is likely to use the 32Mx8 chip, but the 256MB DIMMs can be made at least two ways, so no guarantees on which chips are used on those. If the 256MB modules are single sided, that raises the odds of finding 46V32M8-5 on them. On page 14 here, the allowed values for CAS are 2, 2.5, and 3. I don't think CAS will go any higher than 3. Or at least, higher settings are not documented. http://download.micron.com/pdf/datas...DDRx4x8x16.pdf It is good to hear of at least one report of a CAS3 memory working in dual channel on the Nforce2. Louis can sleep easy tonight :-) I'm curious. Have you tested with Prime95 (mersenne.org) ? I'd like to know if it passes Prime95 or not. My first (generic, cheap) memory passed memtest86, but failed Prime95 after less than an hour of testing. The Prime95 test I was running earlier today didn't have any errors, using a memory setting of 120%, and a CPU clock of 166MHs, so running sync doesn't seem to be an absolute requirement for dual channel. (I normally run at 100% and 200Mhz clock, so won't be leaving it that way.) Paul |
#20
|
|||
|
|||
|
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Asus A7N8X Deluxe and Serial ATA drive churning | Paul D. Motzenbecker, Jr. | Asus Motherboards | 1 | September 23rd 04 02:39 AM |
A7N8X Deluxe...Asus says it DOES support 150Mhz with it's onboard serial ATA..are they lying? | John Q. Public | Asus Motherboards | 3 | September 1st 04 01:08 AM |
Asus a7n8x deluxe gold 1.04 with XP 3200+ | VP | Asus Motherboards | 2 | June 13th 04 05:26 PM |
help folder question ASUS A7n8X deluxe | Sbarel | Asus Motherboards | 0 | October 26th 03 08:40 PM |
ASUS A7N8X Deluxe USB Questions (long) | Space--Coyote | Overclocking AMD Processors | 0 | July 25th 03 01:18 PM |