Dracas
06-20-2003, 12:11 AM
Has anyone else noticed this lately? How just about every major memory manufacturer has been selling 'overmarked' memory that really doesn't "stack up" to its specifications?
For instance, memory modules labeled for PC3200 but only run stably at PC2700 speeds, or Modules rated for 2-2-2-5T timings but only work (boot) if set to 2-3-3-5T, there has even been a report (multiple that I've heard) of Corsairs high-end memory rated for 2-2-2-5T having to be set at 2-4-4-10T just to function correctly.
But notice, you only find these mass errors in the "High Performance" arena. A cheap stick of Crucial seems to work every time, doesn't it? But Crucial always sells a bit 'under the buck' with flavors like PC2100/PC2700. Why?
I have my suspicions that (currently) the memory industry is undergoing a kind of gridlock stasis where the technology makes it sanely to PC2700, then anything over that seems to be a roll of the dice because they have to 'hand pick' chips capable of reaching the so called "high performance rating".
However, I doubt its as clear as that. Memory chip makers sell chips to Module manufacturers in "batches", if the module maker sees that a few of the memory chips from this 'batch' works at a higher speed then they thought, they take the entire 'batch' and run them as high performance PC3200/PC3500/PC3700, saying that they have 'hand picked' the chips being used; when in reality they only labeled a 'batch' that had a few good tests.
That said, not every chip in that batch is going to perform well, this is most apparent in PC3700 Memory Modules, where the memory modules are run at 466MHz FSB, but you have to use CAS3.0, and a slower command rate to get them there - sometimes, even if you do, they still won't work correctly. They need to use the higher CAS to force that 'potential batch of PC2700 chips" up to PC3700, try to run some PC3700 DDR SDRAM at CAS2.0 or 2.5 and it won't work. The same tactic is used in PC3200 and PC3500 modules, but is less noticeable the closer you get to the standard PC2700. Less stress on the chips, basically.
Why? There hasn't been any real 'break-throughs' in Memory chip manufacturing, the one exception is 3.3ns BGA ram, which actually does perform better more often then not, so long as its not being forced up the scale like traditional DIP memory chips. So now we're stretching it a bit, we're pulling out the speeds of our old PC2100/PC2700 Memory chip batches like Taffee, and relabeling them. But the effects of doing that are unfortuneate and apparent, even just browsing through the reviews on Newegg.com
This was never a problem until now, hence why I have my suspicions that the above is true, its open for debate, so feel free to comment, its a personal 'thought' on the matter, but maybe it would be better just to aim low, you might get better performance for your buck (when it comes to memory anyhow)
Happy Posting :t
For instance, memory modules labeled for PC3200 but only run stably at PC2700 speeds, or Modules rated for 2-2-2-5T timings but only work (boot) if set to 2-3-3-5T, there has even been a report (multiple that I've heard) of Corsairs high-end memory rated for 2-2-2-5T having to be set at 2-4-4-10T just to function correctly.
But notice, you only find these mass errors in the "High Performance" arena. A cheap stick of Crucial seems to work every time, doesn't it? But Crucial always sells a bit 'under the buck' with flavors like PC2100/PC2700. Why?
I have my suspicions that (currently) the memory industry is undergoing a kind of gridlock stasis where the technology makes it sanely to PC2700, then anything over that seems to be a roll of the dice because they have to 'hand pick' chips capable of reaching the so called "high performance rating".
However, I doubt its as clear as that. Memory chip makers sell chips to Module manufacturers in "batches", if the module maker sees that a few of the memory chips from this 'batch' works at a higher speed then they thought, they take the entire 'batch' and run them as high performance PC3200/PC3500/PC3700, saying that they have 'hand picked' the chips being used; when in reality they only labeled a 'batch' that had a few good tests.
That said, not every chip in that batch is going to perform well, this is most apparent in PC3700 Memory Modules, where the memory modules are run at 466MHz FSB, but you have to use CAS3.0, and a slower command rate to get them there - sometimes, even if you do, they still won't work correctly. They need to use the higher CAS to force that 'potential batch of PC2700 chips" up to PC3700, try to run some PC3700 DDR SDRAM at CAS2.0 or 2.5 and it won't work. The same tactic is used in PC3200 and PC3500 modules, but is less noticeable the closer you get to the standard PC2700. Less stress on the chips, basically.
Why? There hasn't been any real 'break-throughs' in Memory chip manufacturing, the one exception is 3.3ns BGA ram, which actually does perform better more often then not, so long as its not being forced up the scale like traditional DIP memory chips. So now we're stretching it a bit, we're pulling out the speeds of our old PC2100/PC2700 Memory chip batches like Taffee, and relabeling them. But the effects of doing that are unfortuneate and apparent, even just browsing through the reviews on Newegg.com
This was never a problem until now, hence why I have my suspicions that the above is true, its open for debate, so feel free to comment, its a personal 'thought' on the matter, but maybe it would be better just to aim low, you might get better performance for your buck (when it comes to memory anyhow)
Happy Posting :t