Is the my memory running at 1T?

[Thermopylae_480]

I recently upgraded from 1 to 2 GB of memory and I now have all four slots filled. I have an i925X chipset. CPU-Z is reporting my memory is still running in dual channel mode, My memory speed has not decreased, and my memory timings have remained the same. Is this correct? Usually memory takes some kind of performance hit when all four slots are filled. Is the i925x Chipset just different?

 

[djbbenn]

It's always 1T on Intels systems. The chipset was programmed only to handle 4 memory banks per channel reguardless of how much memory is installed. All unbuffered memory moduals can run 1T, but once you go over the 4 banks on both channel (by putting more modules in, resulting on 8 banks per channel/on both channels) you must run 2T because of the memory controller limitation.

AMD, there is no limitation on the number of banks per channel, so when you install 4 double sided DIMM's (16 banks), you must run 2T. The momory contoller just can't handle 16 banks at 1T, but it can 8 banks.

I believe thats how it works, but I'm not a 100% sure about it. I don't know how the limition of 4 banks per channel on Intel works exactly, I think it works on bank select. I'll look into it.

-Dan

 

[wtf8269]

It's always 1T on Intels systems. The chipset was programmed only to handle 4 memory banks per channel reguardless of how much memory is installed. All unbuffered memory moduals can run 1T, but once you go over the 4 banks on both channel (by putting more modules in, resulting on 8 banks per channel/on both channels) you must run 2T because of the memory controller limitation.

AMD, there is no limitation on the number of banks per channel, so when you install 4 double sided DIMM's (16 banks), you must run 2T. The momory contoller just can't handle 16 banks at 1T, but it can 8 banks.

I believe thats how it works, but I'm not a 100% sure about it. I don't know how the limition of 4 banks per channel on Intel works exactly, I think it works on bank select. I'll look into it.

-Dan

I know the RDX200 and SLI-DR Expert can handle 4 double sided DIMMs at 1T. Don't ask me how though.

 

[djbbenn]

Yeah, thats something to do with bank select I believe. It must be something that over rides the memory controller, or limits the amount of banks that the controller can see at a given time. I'm not sure though, thats just my guess, I don't build motherbaords.

-Dan

 

[Thermopylae_480]

I don't think I understand. what is the difference between 1T and 2T?

 

[djbbenn]

Command rate is the amount of time in cycles (1T or 2T) ("T" just means cycle/time) when a bank can be selected/chip select is executed, then be issued a command. So if it takes one cycle (1T), it will perform faster, 2T will be slower where it takes 2 cycles. 2T is used when there is sability issues, or 4 double ranked DIMM's installed. It gives the chipset more time to properly address the memory. So when there is more memory, it takes longer to address where there more chips/banks.

Does that help?

-Dan

 

[wtf8269]

Actually it helps me a lot lol. My friends always asked me what the difference was and all I could ever tell them was that 1T is faster than 2T but I was never really sure why.

 

[Hypersonik]

Hmm, its doesnt make that much difference 1T or 2T. Obviously 1T is faster, but only by about 2% max! Its the same with Dual and Single channel. You would never notice the difference! Not saying dont do it, though! If its there, you might as well have it

 

[trog100]

"You would never notice the difference! Not saying dont do it, though! If its there, you might as well have it"

with an amd 64 cpu that is probably true in real life system peformance.. however if u run a sandra memory bandwidth benchmark u will see a huge difference.. something like double the memory bandwidth..

when i was playing with my new amd chip i tried the memory at 333.. ran 3dmark2005 and got a score.. set the memory at 400 ran the benchmark and got another... the difference was exactly 19 points.. as a percentage of 6000 odd this definitley fits in the "not notice the difference" category.. less than one frame a second in a game for example..

sandra of course told a diferent story..

memory at 400 2T seems about the same as memory at 333 1T.. memory at 500 2T seems about the same as memory at 400 1T..

the moral of this story seems.. dont run memory bandwidth benchmarks like sandra.. save yourself a fortune and dont buy over-priced super-fast memory cos in real life u "wont notice the difference" tween it and the cheapo stuff..

mind u i can same the same about my overclocked cpu.. in a real life gaming situation i cant tell the difference tween running it at 2.4 gig or 2.9 gig.. he he he..

and most definitely dont ever dream of paying over $1000 for a real fx57 at 2.8 gig.. he he he

the grfx card speeds.. yep.. everything else.. nope..

perfect gaming system.. cheapo memory (lots of it).. cheapo amd cpu.. bloody expensive grfx card..

trog

 

[Hypersonik]

Well, got my CABNE 0545 146 coming next week So i'll see how it all turns out. Sandra! According to sandra my P4 @ 3.6GHz (15*240) runs faster than an FX-55....