1t vs 2t timing on A64. Does it effect 3dmark 05?

[menlatin]

So i've been trying to figure out my low 3dm 05 scores and was thinking it maybe my new ram and timing. with my 7800gt @ 485/1275 i was able to get about 8200 with my x2 @ 2.7ghz. With my x1900xt @ 732/801 and x2 @ 2.75ghz i've gotten a high of 11281. Seems kinda low compared to others. i'd say about 1000 pts low. The only other thing that changed was my ram. with 7800gt i had OCZ plat 3200 @ 227mhz 3-3-3-6 1t. My new ram is gskill xtreme @ 275mhz 3-4-4-8 2t. Does it sound like the 2t timing is killing me? or is it the loose overall timings? Or maybe this needs to be in the AMD section?

[menlatin]

that sucks. i hope the EVGA ppl can hurry up and come out with their "1t allowing" bios. but is it mostly the 1t, or does th 3-4-4-8 have any significant part?

[equalizr]

Ive heard and seen the oposite here.

IVe done tight timings and lower fsb's (220) and got say 30k in '01
then gone really slack timings and 240fsb and got almost 32k

for A64's i think all that matters is that you get there, not how you do it.

[biohead]

it's (frequency)/(sum of latencies * XT) = bandwidth.
300mhz 3-3-3-8 equals (+/-) 200 2-2-2-5 when CPU frequency is the same.

[ahmad]

Sometimes 3DMark05 gets a speed bump from 2T. You just gotta test both.

[menlatin]

1t @ 210mhz 2.5-3-4-7 = ~same 3dmark score. this gskill does NOT like tight timings AT ALL.

[menlatin]

or do you think me running windows 2000 would have something to do with it?

[Kaffebord]

it's (frequency)/(sum of latencies * XT) = bandwidth.
300mhz 3-3-3-8 equals (+/-) 200 2-2-2-5 when CPU frequency is the same.

Wow, I didn't know that. Thank you, that explains a lot for me.

[WeStSiDePLaYa]

it's (frequency)/(sum of latencies * XT) = bandwidth.
300mhz 3-3-3-8 equals (+/-) 200 2-2-2-5 when CPU frequency is the same.

ummm..... dont think that is right at all.


you say 300mhz 3-3-3-8 will have same bandwidth as 200mhz 2-2-2-5?

i say BS.

especially since tras has been shown to have little effect on bandwidth AND latencys. in fact often super low tras LOWERS bandwidth.

[God_Lx]

Well, i have some Single-Channel comparisons on 260Mhz 2-2-2-5 and 3-4-4-7... One BH-5 and one UCCC...

All tests done with ~ CPU Speed, and same GFX clocks, 6800GS @ 500/1200 that's my day to day overclock... Results are weak but don't forget the SINGLE-Channel...

3DMark01: @ 260Mhz 3-4-3-7 1T



3DMark01: @ 260Mhz 2-2-2-5 1T




And the difference is...? Not quite 2K but not far... You can easily see where the difference comes from...

The first stick is 1x512Mb Samsung UCCC that was about 43€ here in Portugal... Same chips as the Gskill HZ but only 512Mb stick...

[menlatin]

Well, i have some Single-Channel comparisons on 260Mhz 2-2-2-5 and 3-4-4-7... One BH-5 and one UCCC...

All tests done with ~ CPU Speed, and same GFX clocks, 6800GS @ 500/1200 that's my day to day overclock... Results are weak but don't forget the SINGLE-Channel...

3DMark01: @ 260Mhz 3-4-3-7 1T
.
.
.
3DMark01: @ 260Mhz 2-2-2-5 1T
.
.
.
And the difference is...? Not quite 2K but not far... You can easily see where the difference comes from...

The first stick is 1x512Mb Samsung UCCC that was about 43€ here in Portugal... Same chips as the Gskill HZ but only 512Mb stick...

Can you run 290 1:1 ? can we see a score of that?

[God_Lx]

This ram doesnt do more than 270 stable... So 290 1:1 is impossible

[Fission]

ummm..... dont think that is right at all.


you say 300mhz 3-3-3-8 will have same bandwidth as 200mhz 2-2-2-5?

i say BS.

I gotta agree. Total BS. Inspite of DRAM being held back by the processor's clocks, the RAM refreshing faster will put out more bandwidth than having a tighter strobe time, and tighter isn't always better, because it is still a matter of timing, not just time.
Like those old fighter planes where the machine gun would fire bewteen the propeller blades, you don't want to just crank one up to it's fastest if the others are co-dependant on another setting.

And a command rate 1T is the same or faster than 2T, period. The only reason 2T is available on newer systems is that memory controller on legacy chipsets wasn't as stable at the native 1T, so 2T was an option to give compatability/stability, especially for systems using more physical banks of memory.

[ahmad]

I won't comment on anything else, but I can say I agree with biohead. With A64s, CPU speed is worth more than anything.