Vostro 1400 overclocking, its crazy.

My 8400m GS's "Factory Shipped Clock Frequencies" is 400/500 according to Nvidia's nTune as well as Rivatuner.

Nvidia lists the 8400m GS as 400/600. Hmm, I wonder what the deal is... anyone else have this problem?

I still stand by the "Magic Ratio" theory though. I'm going to delete that post until I can fix my numbers, no need to confuse anyone.

 

Well your magic ratio seemed to have at least some postive effect with me as I found also that it tended to give the best performance and varying a few mhz off resulted in poorer scores and visibly stuttering performance... maybe there is another even better ratio though? I haven't tried 2/3 which is 400/600 ratio..

Anyhow I did some quick tests with mine and preliminarily it seems 656/820 seems to be about as high as I can reliably go with a 3dmark05 score of 4768.

More tweaking and testing might get me more...

 

Anymore on those pictures of shims that cooled everything so well?

 

I score around 2200 3dmark05 and 1100 3dmark06.....

But I can run CS:S 1280x800 everythng on high, no AA/AF @ 95fps average.
and
Oblivion Ultra High 800x600 @ 30-40fps average.
Oblivion Ultra High 1024x768 (Same as 3dmark05) @ 20-30fps average.

So, my card runs oblivion on ultra high with more fps than it runs 3dmark, is that weird or is it just me?

Another note, my Dell XPS m1330 scored 1600ish on 3dmark06, but only 75fps average using the same settings in the source test. I'm now running an SZ680.

 

Someone else opened another thread for this, but I think that you guys will probably be able to answer best -- has anyone figured out whether the 8400 in the Vostro 1400 is DDR2 or DDR3?

 

I can't offer you a definite answer, but I believe it is GDDR3. And I think this mainly because what others have said, and because I don't think its possible for GDDR2 to clock as high as it does. I mean I have mine up to 800Mhz+ on memory. That’s an effective 1.6Ghz.

 

I believe that possibly the 1400 is using GDDR3 and that the 1500 is using GDDR2...this is strictly based upon the posts I have read and the reported memory clocks achieved by some 1400 owners (my tests on a 1500 have yielded stable function only up to 530Mhz)

 

It's scary that I can't just call Dell and ask them. Dell's (or any of the major manufacturers' for that matter) tech support, sales support, etc. is quite poor. I just can't easily get reliable answers from them. They just look at the website like I could do.

 

More research!!!

Now that we know the temps are safe up to the lower 100's in Celcius I searched for my max overclock to test the "diminishing returns in 3dmark with higher overclocks" people have been noticing.

The first things I noticed was a "magic ratio" of sorts, first here is some data...

Score-----Core-----Memory
8596-----619-----765
8537-----618-----767
8540-----617-----770
8728-----616-----770
8556-----615-----770
8533-----614-----770
8541-----613-----771

As you can see 616/770 is the best score by a long shot, but it isn't my best overclock in the chart. This ratio of 4/5, Core/memory consistently resulted in much higher scores then the settings around it, even when it wasn't the highest overclock. I can't explain it but it held true for all the settings I tried to test it. Also Devedander has confirmed these results seperately from me with even higher clocks than I could reach.

To test the diminishing returns I downclocked from my max OC and compared the change in frequencies to the change in score.

Score----Core----Memory--% OC-----% Increase in 3dmark03
8728-----616-----770-----54.0%-----54.0%
7999-----564-----705-----41.0%-----41.2%
7034-----508-----635-----27.0%-----24.1%
6234-----456-----570-----14.0%-----10.0%
5666-----400-----500-----00.0%-----00.0%

As you can see I had nearly equal increases in scores and overclocks across the board. In fact as I increased the frequencies the better the increase became compared to lower overclocks.

The possible loss in gains could be due to going from a frequency in the "magic ratio" to a higher one outside of the ratio which could lead to a less than expected score even with a higher overclock.

My recommendation would be to follow the magic ratio while you look for the best stable overclock and try it for yourself. If you find a better ratio let us know! Hope this information helps people out.

 

I don't believe it is due to any 'magic ratio'. Overclocking the 8xxx series is quite different from previous cards, because there is a shader clock in addition to the core clock.

http://www.xbitlabs.com/articles/video/display/msi8800gts-640_7.html

The shader clock moves in large discrete steps, and cannot be fine-grained controlled like the core clock. There's also no fixed way of controlling at which point the shader clock will ramp up without going into BIOS tweaking.

It's quite possible that up to 615MHz core, the shader clock is lower than at 616MHz core, at which point it increases. However pushing the core beyond 616MHz may be causing thermal issues, which drops the shader clock down, reducing performance. That would be a much more concrete explanation of the sudden jump in performance going from 615MHz->616MHz.

Could you try testing at 616/760MHz for example? If my thoughts are correct, then you shouldn't see much performance loss compared to the 'magic ratio 616/770.