View Full Version : Memory Timings and the better OC?

05-09-05, 11:52 AM
I can run my memory at 250/500 with timings of 2.5-3-3-6 with a cpu multi of 10.5 to get a cpu spd. of 2625mhz.

or I can run my memory at 292/584 with timings of 3-4-4-8 with a cpu multi of 9 to get a cpu spd. of 2628mhz.


Thanks guys. By the way, I really appreciate your responsiveness on this forum.

People over on anandtech and extrem oc forums tend to make a lot of useless comments and snide remarks when all someone like me is trying to do is get some help.


05-09-05, 12:07 PM
For sure, this is a great forum :thumbsup:

As for your question....its going to be a hard call, especially using the 10.5 multi(your mem in actually running around 236-237)...I'm not tring to blow you off but you might have run some benchmarks too see which settings are faster.

05-09-05, 12:40 PM
I would agree with Pain, benchmark both and see which one you like better.

I would have to say it's personal choice at this point but I would go for the tighter timings. I probably would keep the one that benchmarked faster though :)

05-09-05, 12:40 PM
PaiN is right IMO. Run some benchmarks. It could really go either way. I'd run 3dmark01 myself. Also might be worth checking Sandra memory bandwidth scores between the two.

05-10-05, 12:08 PM
Its prolly gonna differ depending on the benchmark aswell, some things might be faster with higher FSB, some with lower timings.

05-10-05, 01:45 PM
For sure, this is a great forum :thumbsup:

especially using the 10.5 multi(your mem in actually running around 236-237)

looks like i am missing something here, what does the multiplier has to do with the memory speed, i thought its all about the front side bus when it comes to overclocking the memory, my CPU is locked and i still can get higher speed on the memory by rasing the FSB or using the divider.

05-10-05, 03:39 PM
Using higher CPU multipliers helps boost memory bandwidth with the exception of half multipliers. IOW, using a multi of 10.5 does not produce any more bandwidth vs. a multi of 10.

05-11-05, 11:17 AM
Option 1 should be faster slightly. 3-4-4-8 is bad on A64.
I replace my TCCD for UTT to run on 2-2-2-5 myself now.

05-11-05, 08:20 PM
Timings make very little difference in benchmarks and frame rate in games. I bet I could do a blind test on just about anyone here and they wouldn't be able to tell a difference between slower timings and faster timings. I'm almost willing to bet a couple hundred on it.

I know, I'm a hypocrite. /checks sig


05-12-05, 12:14 AM
Most people would say FSB is King so the higher FSB @ the same clock wins right? Hmmm... well what's the memory IO and what applications are you typically running?

Benching makes sense... but benchmarks are typically synthetic. I have experemented somewhat and found that for some apps higher FSB at the expense of looser timings made some apps run faster while others ran notigeably slower.

I've also found that the slower FSB with tighter timings is noticeably faster in certain apps than the faster FSB with loose timings. The difference in gaming would probably not be noticed at all as most people cannot notice 1 fps drop if your system plays Doom3 typically @ 50fps. 1~2 fps will not be noticed, that's obviously not the case if you're pulling 25fps.

My advice, do some benching, check voltages and temps and play/test the system with both settings and stick with what works best for you. This might seem like obvious advice, but tweeking your system to get 1000 more AquaMark3 points does not make your system surf the net any faster and if that's what you do most, it makes sense to stick with the setting that gives best overall system performance and lowest CPU/system temperatures for longevity.

I run my current rig @ 2.4GHz typically and the temps (water cooled) idle around 41C. I can run my box in "gaming mode" @ 2.6+ GHz but the temps go up substantially to around 46~47C idle and to be quite honest, I don't notice a huge difference in gaming performance unless I crank on the res. and eye candy then it helps somewhat, but really, beating on the system to get that smallest increase in performance sometimes is just not worth the hassle IMHO.