Zenoth
Diamond Member
- Jan 29, 2005
- 5,202
- 216
- 106
The Crucial Ballistix are very sensible to changes made to the timings outside of the main ones (tCL, tRCD, tRP and tRAS). Making changes to timings such as tRFC (ROW Refresh Cycle Time), tWR (Write Recovery Time) and other such "advanced" timings will inevitably result in memory errors and will cause all sorts of problems such as the infamous Blue Screen of Death (although a BSOD can also be related to so many other things, but it is just more frequent to see them occurring due to memory-related issues).
Usually when a consumer buys memory he/she should immediately, in my opinion, run a program like Memtest at default settings and voltages (just installing them, and then in the BIOS doing zero manual changes, let it boot to the Desktop and start testing from there or just let it start with Memtest right away even before it goes into Windows for the first time), but how many even do that? Probably not much. When we make Memtest analysis it's usually because we have been getting errors and that we're suspecting the memory being the culprit. In my case I am pleading guilty. I should have made tests with Memtest right away, but I didn't due to impatience and excitation because I just got that new system, one that has been faster than anything I've ever built before, and I just took everything for granted, it seems.
I experienced a lot of issues, and I thought that my Ballistix was faulty, but it's not, because I've changed a lot of timings and played a lot around the BIOS memory settings during the past 24 hours or so, and after a lot of Memtest analysis and Orthos Blend testing I finally figured out what was the problem. The problem was simple: a bad BIOS memory timings configuration, and nothing else. I had put too much voltage, first, and secondo I ended up mixing some timings for others (due to the strange timings labeling/naming by Gigabyte, for example naming "tRAS" as "Precharge Delay", instead of the more standard naming of it being "RAS Active Time"). There were just a few timings which I had to reduce (some a little, others greatly) to achieve so far what looks like absolute stability.
When I read the users reviews in web-sites such as Newegg and others I am amazed by the number of them complaining and worrying that their memory just doesn't work and that because they are getting BSODs "from the start" that it automatically means that their memory modules are faulty and that they need to RMA them immediately. Well yes of course there are such faulty memories being shipped and very unlucky consumers ending up with them in their system, it does happen yes, but it has to be more rare than it looks like. Many consumers just don't investigate enough themselves. I mean... when you buy a motherboard like the GA-X38/X48-DQ6 or the ASUS Maximus Formula you've GOT TO act like an enthusiast yourself and actually DARE around the BIOS and play with the settings, or else things just won't stabilize themselves alone. You can't just go out and buy some nice expensive Mobo just because you've got the money and not move a single finger in the BIOS - and alone without asking any help - when you got errors in Windows, and then expect the problems to be caused by the actual Memory itself and somewhat also expect the problems to fix themselves if you buy the exact same thing again later on (or if you RMA it), because if it's the same thing it'll have the same problems (pretending it wasn't faulty, like the majority of times) since the BIOS settings are still set to their default values which very often tend to cause problems.
My Ballistix, for example, is rated at 5-5-5-15 by Crucial, at 2.2v. That's cool, but do you think Crucial is testing all of their modules in Gigabyte motherboards? Certainly not, and it's not Crucial's job to ensure that their memory does exactly what it's been engineered as in every single motherboards out there that "supports" 1066Mhz modules of DDR2 memory. My current latest BIOS, namely F7 is telling me that, by default, my memory should be running at 5-7-7-24, at 1.8v. At those settings I get errors. If I bump the voltage to 1.9v I can run 5-7-7-24 no problems. Then if I change the timings to 5-5-5-15, still at 1.9v, I get errors... but if I change the voltage to 2.0v, I can run at 5-5-5-18 (yes, 18), because at 5-5-5-15 I get errors. At 2.1v and 2.2v I get errors at any timings I chose, from 6-6-6-18 to 8-8-8-24, along with all sorts of very low "advanced" timings, I always get errors. So at 2.0v, 5-5-5-18, along with some very carefully chosen timings in the advanced category, so far everything has been entirely stable in Memtest (6+ hours, no errors whatsoever).
It IS true however that some brands (or very specific models/speeds) will react VERY differently to timings changes. It's very similar to how that 'x' CPU will react to voltage increases when compared to that 'y' CPU of the exact same model (two E8400's won't ever react at the exact same extent and way when their voltages are being augmented). My E8400 just can't get passed 3.48Ghz at stock, while others will go as high as 3.7Ghz with a tiny, very shy voltage increase in perhaps a 0.002v range. Some people need 1.45v to get to 4.0Ghz, others will need 1.38v for the same goal and will reach it stable. Well, memory and timings are the same story. My own PC2-8500 Ballistix will do 'x' and 'y' timings under certain voltage conditions (and under main timings conditions), and another one such memory kit will do better or worse. It's very unpredictable, but the best way to figure out anything is to try it yourself and to not be afraid of the results. Just don't exaggerate anything (don't put 2.4v in a DDR2 module rated at 2.0v, and don't go higher than that without expecting it to react well, don't forger that memory sticks are more sensible to voltages than CPUs, it's almost as sensible as increasing the PCI-E voltages).
By the way all this wasn't directed to anyone in particular, just generally speaking to the mass.
Usually when a consumer buys memory he/she should immediately, in my opinion, run a program like Memtest at default settings and voltages (just installing them, and then in the BIOS doing zero manual changes, let it boot to the Desktop and start testing from there or just let it start with Memtest right away even before it goes into Windows for the first time), but how many even do that? Probably not much. When we make Memtest analysis it's usually because we have been getting errors and that we're suspecting the memory being the culprit. In my case I am pleading guilty. I should have made tests with Memtest right away, but I didn't due to impatience and excitation because I just got that new system, one that has been faster than anything I've ever built before, and I just took everything for granted, it seems.
I experienced a lot of issues, and I thought that my Ballistix was faulty, but it's not, because I've changed a lot of timings and played a lot around the BIOS memory settings during the past 24 hours or so, and after a lot of Memtest analysis and Orthos Blend testing I finally figured out what was the problem. The problem was simple: a bad BIOS memory timings configuration, and nothing else. I had put too much voltage, first, and secondo I ended up mixing some timings for others (due to the strange timings labeling/naming by Gigabyte, for example naming "tRAS" as "Precharge Delay", instead of the more standard naming of it being "RAS Active Time"). There were just a few timings which I had to reduce (some a little, others greatly) to achieve so far what looks like absolute stability.
When I read the users reviews in web-sites such as Newegg and others I am amazed by the number of them complaining and worrying that their memory just doesn't work and that because they are getting BSODs "from the start" that it automatically means that their memory modules are faulty and that they need to RMA them immediately. Well yes of course there are such faulty memories being shipped and very unlucky consumers ending up with them in their system, it does happen yes, but it has to be more rare than it looks like. Many consumers just don't investigate enough themselves. I mean... when you buy a motherboard like the GA-X38/X48-DQ6 or the ASUS Maximus Formula you've GOT TO act like an enthusiast yourself and actually DARE around the BIOS and play with the settings, or else things just won't stabilize themselves alone. You can't just go out and buy some nice expensive Mobo just because you've got the money and not move a single finger in the BIOS - and alone without asking any help - when you got errors in Windows, and then expect the problems to be caused by the actual Memory itself and somewhat also expect the problems to fix themselves if you buy the exact same thing again later on (or if you RMA it), because if it's the same thing it'll have the same problems (pretending it wasn't faulty, like the majority of times) since the BIOS settings are still set to their default values which very often tend to cause problems.
My Ballistix, for example, is rated at 5-5-5-15 by Crucial, at 2.2v. That's cool, but do you think Crucial is testing all of their modules in Gigabyte motherboards? Certainly not, and it's not Crucial's job to ensure that their memory does exactly what it's been engineered as in every single motherboards out there that "supports" 1066Mhz modules of DDR2 memory. My current latest BIOS, namely F7 is telling me that, by default, my memory should be running at 5-7-7-24, at 1.8v. At those settings I get errors. If I bump the voltage to 1.9v I can run 5-7-7-24 no problems. Then if I change the timings to 5-5-5-15, still at 1.9v, I get errors... but if I change the voltage to 2.0v, I can run at 5-5-5-18 (yes, 18), because at 5-5-5-15 I get errors. At 2.1v and 2.2v I get errors at any timings I chose, from 6-6-6-18 to 8-8-8-24, along with all sorts of very low "advanced" timings, I always get errors. So at 2.0v, 5-5-5-18, along with some very carefully chosen timings in the advanced category, so far everything has been entirely stable in Memtest (6+ hours, no errors whatsoever).
It IS true however that some brands (or very specific models/speeds) will react VERY differently to timings changes. It's very similar to how that 'x' CPU will react to voltage increases when compared to that 'y' CPU of the exact same model (two E8400's won't ever react at the exact same extent and way when their voltages are being augmented). My E8400 just can't get passed 3.48Ghz at stock, while others will go as high as 3.7Ghz with a tiny, very shy voltage increase in perhaps a 0.002v range. Some people need 1.45v to get to 4.0Ghz, others will need 1.38v for the same goal and will reach it stable. Well, memory and timings are the same story. My own PC2-8500 Ballistix will do 'x' and 'y' timings under certain voltage conditions (and under main timings conditions), and another one such memory kit will do better or worse. It's very unpredictable, but the best way to figure out anything is to try it yourself and to not be afraid of the results. Just don't exaggerate anything (don't put 2.4v in a DDR2 module rated at 2.0v, and don't go higher than that without expecting it to react well, don't forger that memory sticks are more sensible to voltages than CPUs, it's almost as sensible as increasing the PCI-E voltages).
By the way all this wasn't directed to anyone in particular, just generally speaking to the mass.
Facebook
Twitter