Vostro 1400 overclocking, its crazy.

Chuck232 - Thanks for that bit of info, I didn't know that. I'm away from my Vostro today, but I'll post some results on that tomorrow for you.

 

While I am certainly no expert I do recall many times when core and mem worked best locked at the same speed or some such...

And while the shader clock theory sounds good it doesn't really explain why varying from the ratio has detrimental effects even at lower overall clocks... where the shader should be still in it's previous large discrete step...

 

I did a quick glance through the past 100 posts or so of this thread and wasn't able to find any definitive tests of this ratio idea at lower clocks, only what pattertj posted a few posts back of the 616/770 ratio. Which benches are you referring to?

 

I don't know if I mentioned it or not but it was consistent at lower clock frequencies. I'll post some results about that tomorrow though too for everyone.

I'll use the same numbers as I did for the diminishing returns test, 616/770, 564/705, 508/635, 456/570, and 400/500. Then I'll adjust the memory up and down as well as the core to test the shader clock idea and to make sure that it holds true for both small and large overclocks. That sound alright?

 

Yeah, that would be a great, definitive test. A test suite like this should sort things out:

616/775
616/770
616/765

564/710
564/705
564/700

508/640
508/635
508/630

etc...

I'd refrain from changing the core clock since we don't know where the shader clock may jump. Changing the VRAM clock will have the intended effect if the ratio theory is correct.

 

really good thread thus far although I don't agree with some of your clocks (*cough* devilsknight *cough*). I can see 600/750 to be just fine though, its when you start pushing 700/800 or whatever you will see your laptop die within a year + its just not worth the small marginal gain.

patter: those were some impressive gaming results when you increased the clock speeds. 20-30% higher is always worth it!

 

I don't have as many results as patterj but I do recall that I had much lower clocks at one point when I was messing around and I noticed that if they weren't on the magic ratio I often got stuttering or some odd performance isues that looked like some part of the system was waiting for another part of the system and not synced up properly...

It seems to me entirely possible that the memory and core work best at a certain ration...

 

This is really great stuff, and I've been following this thread since it was first started. Coming from a guy who has never tinkered around with overclocking, it's great to see all these tests being run to optimize performance.

But after all this, what are the clocks you guys have been settling with? What are you running right now? I'm trying to gauge the difference between what is good for the computer and what is good in general. Thanks!

 

Upping my stock clocked GS (400/600) to 575/725 took my avg fps in CS:S from 75 to 100. (Avg of about 5 video stress tests)

Also, I noticed that lowering my core to 525/725 only took my fps avg down to 95. (lowest of 3 tests)

 

Well some interesting results...

I switched from nTune to ATItool due to some errors from ntune during 3DMark even at stock speeds. When I did this and tested per Chuck232's idea, everything changed.

The magic ratio was gone and nearly every result I had from using nTune in earlier tests was different from my ATItool results.

I went back to nTune and noticed when using the built in frequency monitor that the clock speeds reported were not always what I had them set at. I wonder if nTune skewed my frequencies and gave me my odd results. Here are my ATItool results...

Core-----Mem-----3DMark03 Score
616-----770-----8639 < supposed Magic Ratio
616-----765-----8650
564-----710-----8052
564-----705-----7968 < Supposed Magic Ratio
564-----700-----7958

I didn't have time to run more tests on the other frequencies I said I would. Sorry. But the results seem fairly cut and dry to me. The Magic Ratio is not real. The new question now, Is why did nTune give such different results, and (more importantly) Can I push my clocks further?