no benefit for low cl timings

i have a p4 3.2, asus p4c800e deluxe, ocz 3200 pl le 2x512 setup and have tweaked and overclocked for the past 2 months and have come to the conclusion that a 230 fsb (3.7)outperforms any other lower fsb setting no matter the cl settings.

i can get 3.4 stable at 2-2-2-5(all volts are 2.85 mem,1.625 cpu
or 3.55 stable at 2-2-3-5
or 3.6 at 2-4-4-8
or 3.7 at 3-4-4-8

and no matter what benchmark(sandra, 3dmark03, 3dmark01, pcmark04) the 3.7 outperforms

at gaming, 3.7 outperforms (AmericasArmy diehard)
at photoshop 6, 3.7 outperforms

my conclusion is:

screw the low latencies cause the higher fsb always seems to outperform on my system

if some1 here has a link or evidence to disprove my conclusion, please post so I can learn.

LOL!!! i like a man that takes a stand... actually u are comparing 3 things, fsb latencies and cpu speed. with everthing being equal Ie: fsb, cpu, speed and memory speed, lower latencies will perform much better than loose timings.for example, a 2.4 running at a250 fsb @1-1 with memory @2-2-2-5, will perform better than a 2.4 running 250 1-1 @3-4-4-8.

i agree, i noticed a 200 point increase on 3dmark03 with all the same except cl settings. however i see a lot of posts saying to go with pc3200 vs pc3700 or higher based solely on cl timings r better at pc3200 and have found that fsb is more of a performance boost than cl timings r.

Your confused as to what people were saying. If you are at the same FSB then 5:4 with 2-2-2-5 timings will beat out 3-4-4-8. But it's pretty obvious that running a higher FSB no matter what the timings are would be faster than a slower fsb.

And remember for timings on intel they go in this order for amount of performance gained most> trcd-trp-tras and cas>least are about the same. If you buy ram based on it being Cas 2 over cas 3 you've had the sheets pulled over your eyes

Steve

voodoothenoob said:

i agree, i noticed a 200 point increase on 3dmark03 with all the same except cl settings. however i see a lot of posts saying to go with pc3200 vs pc3700 or higher based solely on cl timings r better at pc3200 and have found that fsb is more of a performance boost than cl timings r.

Click to expand...

All this pc3200 3700 etc can be very misleading and confusing, memory speed and timings is a relative number . for example, a guy running his fsb at 300and running some dynamite bh5 memory at a 5/4 ratio(250) at 2-2-2-5 , is going to kill guy number 2 in mem bandwidth running a 265 fsb 1-1 with memory @3-4-4-8. likewise a guy running his fsb and bh5 1-1 at 240 at 2-2-2-5, against say mrspecs bt-d43 running 292 fsb 1-1 with memory running 292 at 3-4-4-8 is going to get killed by mrspec's setup..

both your examples have the higher fsb outperforming.

my point is that whatever your machine can run at highest fsb, is what gives the most performance

always go with the highest fsb you ca attain stable vs lower fsb and tighter timings

yep thats the way to go, but it is important to understand that with all things equal tight timings give the best memory bw. but not at sacrificing fsb and cpu speed. i forget where but i read in some review that cpu speed much more important than pure memory bandwidth. and your testing has shown that, also it shows that u are not relying on what people say and are thinking for yourself which i respect.... yours truly flapperhead...

About a year ago and before most memory reviews and articles focused on the timings as if the memory was worthless if it couldn't run at fast timings but over the last six months or so, it seems that everyone (except the vendors which keep listing CAS as if it was the only thing that matters) has realized that, at least on the Intel systems, high fsb at 1:1 at 3-4-4-8 provides just as much performance as 5:4 at 2-2-2-5. You just proved it again.

I am curious about your benchmarks at each of those different settings you used. Could you provide those? It would be interesting to see how they would plot out (I'm into math and stuff).

http://service.futuremark.com/compare?2k3=2241406

thats 5/4 2-2-2-5 just did this last night

http://service.futuremark.com/compare?2k3=2227296

thats 1:1 looser timings


http://service.futuremark.com/compare?2k3=2227331

1:1 2-2-3-5

bah, did so many tests i cant find them all and deleted many, sorry

must have run over 100 diferent benchmarks on different programs

most differences were very small, so stability is key in my opinion

If you want to read up about mem timings/dividers and bandwidth check this out

http://www.pcstats.com/articleview.cfm?articleID=873

Colin found that in the end

Thanks to the i865PE/i875P's dual channel memory controller things are much brighter. On average, the system with the memory running at 400 MHz (5:4 memory divider enabled) with aggressive memory timings performed 2-3% faster than the system using high speed memory with loose timings.

Click to expand...

So as voodoothenoob said stability should be the end goal