Crucial PC3200 - top stuff!

[RichL]

I've always regarded memory from Crucial to be high quality, reliable stuff and have used it in my PCs for years, but I've never considered it overclockable, not compared to the Corsair 3200LL sticks I have that run 220@5-2-2-2.

However for my new A-64 rig, I wanted to bump up from 512mb to the full Gig, and as this was to be a workstation rather than a speed demon (I'm running a G550 after all) I plumped for 2 sticks of Crucial 512mb.

Naturally I couldnt resist the lure of free HertzPower for long and the FSB started creeping up...as I write, I'm running 1:1 FSB/DDR ratio and 230@8-3-3-3. It's just ran a SS Sandra benchmark after an evening with BOINC/CPDN whilst browsing, downloading and watching movies, so I think it's stable.

And I know my CPU will run 240fsb...
edit
...and so will the memory.
Now running 240.7@8-3-3-3.

PiFast completes in 51.5 seconds

[spadnos]

I've always regarded memory from Crucial to be high quality, reliable stuff and have used it in my PCs for years, but I've never considered it overclockable, not compared to the Corsair 3200LL sticks I have that run 220@5-2-2-2.

However for my new A-64 rig, I wanted to bump up from 512mb to the full Gig, and as this was to be a workstation rather than a speed demon (I'm running a G550 after all) I plumped for 2 sticks of Crucial 512mb.

Naturally I couldnt resist the lure of free HertzPower for long and the FSB started creeping up...as I write, I'm running 1:1 FSB/DDR ratio and 230@8-3-3-3. It's just ran a SS Sandra benchmark after an evening with BOINC/CPDN whilst browsing, downloading and watching movies, so I think it's stable.

And I know my CPU will run 240fsb...
edit
...and so will the memory.
Now running 240.7@8-3-3-3.

PiFast completes in 51.5 seconds

Two questions:
1) What is the normal clock on that board (speed and timings)?
2) Do you note significant increases in Sandra memory benchmarks?

I only ask because it seems to me that the slower timings would have more of a negative effect than the clock speedup has positive effect.

Thanks
- Steve

[RichL]

Two questions:
1) What is the normal clock on that board (speed and timings)?
2) Do you note significant increases in Sandra memory benchmarks?

I only ask because it seems to me that the slower timings would have more of a negative effect than the clock speedup has positive effect.

The normal clockspeed for the board/CPU/mem is 200fsb, the memory SPD defaults to 8-3-3-3 timings at 200fsb (according to CPU-Z)
Sisoft Sandra benchmarks increased by appromimately the same percentage as the FSB overclocks.

A year or so ago I ran a series of Sandra and PiFast comparison tests with different FSB and Timings on my Barton 2500/Corsair3200ll system. Unfortunately the forum I posted the results on did a prune after a couple of months and the post got deleted.
Good excuse to run them again I guess

[borat]

really depends on how much you can bump up the speed by relaxing the timings. i think im right in saying a latency of 2 at 200Mhz is equal in terms of the actual time involved to a latency of 3 at 300Mhz, but you get your extra bandwidth through the clock increase. Of course, we are not talking of a speed bump of that magnitude.

[spadnos]

really depends on how much you can bump up the speed by relaxing the timings. i think im right in saying a latency of 2 at 200Mhz is equal in terms of the actual time involved to a latency of 3 at 300Mhz, but you get your extra bandwidth through the clock increase. Of course, we are not talking of a speed bump of that magnitude.

Right - that's why I was wondering.

Taking 2 or 3 clocks out of 8+3+3+3 (roughly the time needed for a single burst transfer), you can get pretty massive speed improvements: DDR333 at 6-2-2-2 is 12 clocks = 36 ns; DDR400 @ 8-3-3-3 is 17 clocks = 42.5 ns. (these may not be correct - if anyone has a link to the bus timings for the Athlon64 or Opteron, I'd be interested) You'd have to overclock the DDR400 to 472 MHz to get the same single burst speed as DDR333 at 6-2-2-2.

- Steve