Calling all JIUHB experts. Am I on the right track?

[G|-|oST]

I have been experimenting with some different settings on my system to see what the potential of it might be.

This is the hardware I have:

XP1700+ 'B' 0310WPMW
Volcano 9/SmartFan II <--- to be replaced with SLK900U soon
ASUS A78NX-DLX rev.1.04 bios 1004 uber
512 PC2700 Infenion RAM <--- to be replaced with ???
ElCheapo 300W PSU <--- to be replaced with Antec TP430 soon

Here is what I found so far:

As is, with all the weak links in the system, the best, most stable, and temp/volt friendly setup is:

185 x 12 ~ 2220 @ 1.625V

From here I know that RAM becomes the first limitation, because 190FSB x anything gives Prime95 errors, regardless of voltage, vdimm, or memory timings. Well I already knew this RAM is crap so I decided to try something different.

I decided to test out the FSB while lowering memory speed to around 75% to take it out of equation. I also opened the case, layed it on the floor horizontally, added fans to the top of the NB heatsink (8cm) and also above the chipset (6cm). I also aimed an AC vent pointed right at the computer for cold air. This is to sorta simulate an SLK-900U with Arctic Silver ( i just have some ElCheapo Thermal Goop on right now).

Well, to my surprise I got up to 220 FSB so far. I will keep on Priming for a bit at this speed and then try for more. I also found the system booting up to around 2400MHz, but PSU was crapping out more and more at the higher speeds.

According to my admittedly limited knowledge of computers, I am guessing that with the right RAM and PSU + SLK-900 this thing should be able to run this FSB in sync with memory. (provided I
figure out how to mount all the extra fans of course). I may to be totally wrong on this. Can anyone confirm or contradict my theory? Am I testing anything at all by running the memory at lower % of FSB??? I know its not good to do that performance wise, I am just doing it for testing purposes.

Any helpful response/advice would be much appeciated.

[hitechjb1]

I think you have all the good hardwares and hardware-to-be for good overclocking of the CPU and the mb FSB.

1. Be careful about the SLK-900U, I heard that it may not fit easily with the A7N8X-DLX, without bending some capacitors. I haven't face this problem personally, so I don't know how severe the problem is. SLK-800U would perform almost the same as the SLK-900U, in case it cannot fit.

2. You need to get some good PC3500 CAS2 memory module which officially ensure to run at least to 217 MHz. Some rev2 mb may be able to go to 220 MHz and beyond, so you don't want memory to be a limiting factor. Getting PC3700 CAS2 memory is not easy and also expensive at this time. TWINMOS PC3200 memory with Winbond chips CH-5 are rated to run at 200 MHz, and have a good chance to be overclocked to around 220 MHz, but not guaranteed officially.

You mentioned you tested memory in ASYNC mode w/ memory being slower. This way, you can test at higher FSB (between the NB and the CPU). But this is easier in general than testing the full blown SYNC mode when you have faster memory to test at speed of FSB, which would stress harder on the memory controller in the NB. You may find that the FSB may have to back off a bit under SYNC than that of ASYNC. Will see how it goes.

3. Overclocking CPU and overclocking FSB can be achieved separately, memory speed and FSB speed should not limit the CPU overclock potential. Since the two can be resolved by changing the multiplier for unlcoked CPU such as the Tbred B that you have. I have written a summary about this, link: "Summary for overclocking CPU and FSB" in Page 2 of Tbred B 1700+ DLT3C 2.5+ GHz air results (http://forum.oc-forums.com/vb/showthread.php?s=&threadid=187887&pagenumber=2), plus some other oc stuffs.

[G|-|oST]

Thanks for the reply...

I just checked the mounting dimensions for the 800 and 900 's and they are the same, according to Thermalright website. I guess only the top section is bigger on the 900. It seems that the local retailers stock the 900 and the 800 is the order item, so I figure may as well get the 900. My goal is not the highest numerical speed of the CPU, but instead best balanced system performance. So a stable SYNC'ed setup would be great. The memory prices are insane though, if I want PC3500 CAS2 matched set of 2 x 256. SO I guess untill I get ALL of the upgraded components, I may as well stay with 185 x 12.

I can totally see the law of diminishing returns kick in right here - I would have to invest quite an amount of money to get from 2200Mhz to MAY BE 2600 or so at best. This last part of overclocking is the one that makes the least sense. Overclocking is supoposed to take something inexpensive and speed it up cheaply, but here we are taking all the money we saved on the CPU purchase and dump it all (and then some) into monster heatsinks, watercoolers, exotic RAM, etc.

Hmmmm...

[hitechjb1]

Glad you pointed out price-performance.

The last 100-200 MHz overclcoking of the CPU is very expensive, and may not worth practically the return, except for oc and benchmark satisfaction, ... It may end up with 20-30+% cost increase for much less performance gain. One may have to replace PSU for the last 100 MHz, and even expensive cooling system, ... IMHO, one can hardly feel the difference between 100-200 MHz when the CPU is running at 2400 MHz (< 10%), except from benchmark numbers.

Having said that, on the other hand, one can learn quite a bit of science for getting the last % of overclocking, such as CPU voltage, clock frequency, power, current, temperature relationship, fluid flow, cooling, heat transfer and exchange, semiconductor devices, ...

1. To save money on HSF on air, get the SK7 instead of SK800/900 (40% saving for 2-3 C higher in CPU temp which equates to 1-2% of CPU clock, I think)

2. I don't think the match pair for dual channel memory is necessary, unless there is no premium price to pay. My oc uses noname PC3500 CAS2 memory which does spec (saved 30-50% than brand name). A 512 MB PC3500 C2 module could be had for $90-140, depending on the brand.

3. There are lots of recommendation about Fortron PSU for price-performance in the PSU section. I personally have yet to try them and have no say on this at this time.