The good news: You can set the timings manually
The bad news: You might have to set them to the speed of your slowest module
if you want them to work together.
Set the timings to 3-4-4-8 Voltage at 2.6..
if no post, try 2.7...up to 2.8 should be ok, but only if needed. Your mileage may vary.
Once they post together you can try speeding them up (3-3-3-8) etc while adding more voltage,
but do so at your own risk (usually ok, but you never know)
Doesn't work? Try tRAS of 2T instead of 1T or AUTO
Still doesn't work? How old is the power supply? Going from 1 stick to 3 could be a big draw on a weak/cheap power supply
This is good information. Thanks!
I have a 460W PS. It should be sufficient. I downloaded Crucial's mem scan utility and it proposed CT12864Z40B saying it would be compatible with the system and the current memory configuration. We'll soon see how accurate that statement is as I get the chips tomorrow. Hopefully I won't have to play with the settings.
I'm disappointed. I installed the new memory and the board dropped the DDR to 325, down from 400. It reports it did this because the new chips are not matchable to the old ones. Apparently, Crucial's idea of compatibility is without regard to performance and only relavent in terms of the chips being able to coexist on a functional level. That's not what I thought I was buying.
In terms of access speeds:
- The old configuration 2 X 512MB PC3200 DDR400 2-2-3-6 (SPDTool says 2-3-3-6) runs fastest at DDR400.
- The new configuration 2 X 1024MB PC3200 DDR400 3-3-3-8 (SPDTool says 4-3-3-7) runs a little slower but still at DDR400.
- The mixed configuration, 3GB total, split between the channels, runs slowest at DDR325.
The new configuration by itself may have slower access times than the old configuration by itself, but the extra GB seems to compensate for the loss in access time by increasing the performance of Windows and various applications, especially things like Photoshop.
I never expected to have this kind of difficulty upgrading the RAM on this machine. Staying with a single manufacturer, I haven't been able to find an OCZ module that matches the original OCZ configuration. Beyond that, I haven't been able to find any module to match the configuration such that the board doesn't degrade DDR. I don't really like the idea of pushing the timings. I bought this board for stability and it's been the most stable board I've ever used. I would like to stay within all published performance specs, but if I can't find matching chips I don't know what other solution I have available to me if I'm interested in maximum performance.
If I was in your situation, I'd:
1) try to manually override the speed, making them DDR 400, or
2) try to raise the RAM to the highest it will go, eg 333, etc.
3) lower the timings + raise the voltage when needed (should not need to much at lower speeds); test with memtest86+ until I find a setting with no errors for at least 2 full run cycles - really tight timings at 333 will be nearly fast as looser timings as 400
4) update the BIOS
Apparently there were some fixes to memory issues in various BIOS revisions. Short of that, I'm scratching my head.
I updated the BIOS, I guess really just to be up to date. But it did alter its auto configuration to DDR320, 2.5 - 3 - 3 - 7 with refresh 12, delay 2, timing 6 and voltage of 2.625. It has an agressive mode of DDR320, 2.5 - 2 - 2 - 5 with refresh 10, delay 1, timing 5 and voltage of 2.625, yet I'm not sure I want to try it after reading about some of latent hard to find potential problems. I'm not a gammer or a speed freak. My primary application in terms of memory demands is editing and manipulating larger graphics files. I also do a lot of digital audio processing, which is one place I can't really afford mistakes to show up, or worse, to not be aware mistakes are being written as data. I want most of all a stable board.
I downloaded memtest86+ and have determined that you really need to dedicate some long runs in order to be reasonably sure your configuration is stable after pushing it. Given the incremental nature of pushing it, this process could take days of dedicated configuration and testing.
I'm curious, in terms of some kind of actual speed test, what my system is losing in dropping from DDR400 2 - 3 - 3 - 6 to DDR320 2.5 - 3 - 3 - 7. It seems slower, but am I capable of actually discerning this loss through normal use of my applications. Do you know of any utility that does such testing such that it reports a start and end time for a predefined series of tests? It doesn't appear as if memtest86+ does this.
I'm still mulling things over. Am I better off with the slowest speed with 3GB (mixed chips), or the mid range of 2GB (new chips only), or the fastestest 1GB (original)? Or perhaps I should go with 4 X 1GB or 2 X 2GB, all with the same timings and manufacturer with hopefully faster timings than the 2GB new chips?
I find it interesting that nowhere on Crucial's site do they give timings for the chips I bought, and even the chips are not labeled. It appears the only way to really know what they are is to buy them and inspect them under software control. Is that just Crucial? Maybe because they're mass consumer oriented? It doesn't seem reasonable to not publish these things, especially when it has the potential to degrade your system. Worse yet, they give you a utility in place of published specs that simply reports what chips you need to buy to upgrade to whatever total RAM configuraiton you desire saying that the proposed chips are "compatible." But degrading a system's performance is not my definition of compatible.
You have some options for benchmarking memory.
(This list is not exhaustive)
The Advanced Memory Test is part of the PerformanceTest application,
You can also try Sisoft Sandra Lite
(There are many others)
In either case, run the Memory benchmarks (or all of them if you like)
and see what numbers you get. These are synthetic benchmarks so
they are really only useful for looking at the behaviour of changes
on systems under the conditions the benchmarks create. (IE, it may say it
is faster with one setting, but you may not notice the difference, or you may
find that some other factor the benchmark does not take into account
changes your experience of the change/tweak)
Anyhow, I find that doing a couple runs gives you an idea of the performance
of a particular setting. Make note of the scores it gives you. (esp. the relevant
sub-system one - ie. memory)
Then, make the changes. Run the benchmarks again and see what happens.
Compare your results.
Another option will be to try manipulating a huge file (rendering, encoding, etc)
and timing how long it takes.
Then, do the same with changes to the memory and see whatcha get.
make changes to setting, repeat.