this is a contuation of the thread: "Any ideas for me" where the thread was hijacked by myself and someother members wanting to discuss timmings in relation to dividers and FSB.

so in response to the last post there from Peteyboy23:
when you do the dvider you need to tighten up the timmings a little tighter than 7-3-3-2.5 or whatever it was, "tight" here i will refer to 6-2-2-2 and that, my friend is tight. on my NF7-S i have that XMS3500 running @ 460 @6-2-2-2 and i have not had a single issue. in the bosses machine it has ran 464 (290FSB, 5:4 divider) @ 6-3-2-2. today i shoud have the results of that test. the challenging RAM will be OCZ Gold 4000 dual channel, 2 x 256. My ram is Corsair TWINX 3500 dual channel 2x256.
i will test these pairs of ram in a Asus p4c800 dlx 875p (i think that is the name, correct me if i am wrong), with a 2.4c SL6WF Processor, 9800 AIW and a Zalman Cu7000 something or other (it is on the front page). the testing media will be:

PCmark2002
SiSoft Sandra
3dmark2k1

If you can think of anything else i should test with Post it here. I should have the resusts by 5PM EST.

Maybe run a couple of gaming tests too...like UT2K3 and/or MOHAA with FRAPS counts or something like that...maybe HALO/FRAPS?

yea, add in some games;

Quake 3
UT2k3
Halo if you can
3dMark2003 games 2 and 4 FPS and overall score

unfortunatly all my game stuff is at home, is there a DL anywhere that i can get that i could use? The Boss's machine is only for Overclocking, not playing games, or so he says...lol
but i will find what i can, any linkage would be great!!

You can use the demo version of UT2k3 to benchmark...

Is there a demo for HALO? You can install that and FRAPS to make a makeshift benchie...same with something like Max Payne 2 and/or Splinter Cell...

Try Comanche too...that's a pretty intensive benchmark...

but where???

Here a link to download the HALO demo:

http://www.gamepro.com/computer/pc/g...les/4852.shtml

This is the link to download FRAPS:

http://www.fraps.com/

Go here for the Comanche 4 download:

http://www.fileplanet.com/files/80000/83003.shtml

You can download all the Unreal 2003 stuff from here:

http://www.unrealtournament2003.com/downloads.php

Why overclock if not for games? To see how well you can run Word and AOL instant messenger?

lmao @ Vada! He wants to make MS paint run faster.

can't wait to see the results of the test, I'll PM you with another program you might wanna test with.

the boss is 53. Benchmarking is his "games". so go figure

I was starting to get caught up in that too. Just wanting to get the biggest number. Then thinking my Ram was holding me back and was about to spend some big cash for some ram and then I thought "how does this help my computing experience?" So, now im saving up for a bigger monitor.

Quote: Originally posted by Chrome_Rust this is a contuation of the thread: "Any ideas for me" where the thread was hijacked by myself and someother members wanting to discuss timmings in relation to dividers and FSB. so in response to the last post there from Peteyboy23: when you do the dvider you need to tighten up the timmings a little tighter than 7-3-3-2.5 or whatever it was, "tight" here i will refer to 6-2-2-2 and that, my friend is tight. on my NF7-S i have that XMS3500 running @ 460 @6-2-2-2 and i have not had a single issue. in the bosses machine it has ran 464 (290FSB, 5:4 divider) @ 6-3-2-2. today i shoud have the results of that test. the challenging RAM will be OCZ Gold 4000 dual channel, 2 x 256. My ram is Corsair TWINX 3500 dual channel 2x256. i will test these pairs of ram in a Asus p4c800 dlx 875p (i think that is the name, correct me if i am wrong), with a 2.4c SL6WF Processor, 9800 AIW and a Zalman Cu7000 something or other (it is on the front page). the testing media will be: PCmark2002 SiSoft Sandra 3dmark2k1 If you can think of anything else i should test with Post it here. I should have the resusts by 5PM EST.

I realize my timings weren't as low as possible, that's why I put the word tight in quotes. I only used those timings, because those were the ones you told me to test with, and compare them to my current 3-8-4-4. I am not entirely sure that the ram I have will run @ 2-6-2-2, even at 200mhz. However, I'll sure try them, and also present some testing data, once I have the time.

Quote: Originally posted by PeteyBoy23 I realize my timings weren't as low as possible, that's why I put the word tight in quotes. I only used those timings, because those were the ones you told me to test with, and compare them to my current 3-8-4-4. I am not entirely sure that the ram I have will run @ 2-6-2-2, even at 200mhz. However, I'll sure try them, and also present some testing data, once I have the time.

Good deal, i feel that this will a very constructive test. but i am going to have to wait until Boss Man gets his new OCZ. it was failing MEMtest-86. and i want this to be a fair test. it was failing on one address on only one chip. So OCZ is replacing it. i will run the testing AS soon as i can. so please keep checking back!

Oh hell yeah, I will. I was personally considering getting more ram. If I find ram that can give me much tighter timings, and can still run at 250mhz+, I would sell my current ram, and just get a gig of that. Thanks for going the extra mile, I for one appreciate it.

Quote: Originally posted by PeteyBoy23 Oh hell yeah, I will. I was personally considering getting more ram. If I find ram that can give me much tighter timings, and can still run at 250mhz+, I would sell my current ram, and just get a gig of that. Thanks for going the extra mile, I for one appreciate it.

the XMS 3500 is the hot ticket right now. 168 bucks at newegg per chip 512mb. i was going to mess with it today but we found out that the board is bad, not the memory. Tried my RAM and tried a PC3200 512mb stick of Geil Value crap ram. all failed at the same place. the new board should be here next week. hey, you got MSN messenger or yahoo??

Update on where I am:

I'm going to sell the OCZ (I'll test it first @ DDR500) and then I'm going to get some Mushkin Level 2 PC3500 instead and run it at 2-2-2-5 DDR400 @ 5:4 for testing. I should have this thing nailed down by the end of next week.

I've already benched the OCZ at 3-4-4-8 in Sisandra, 3Dmark2001, Super PI, PC Mark, and Aquamark3, but I'll save the results so I can post it all at once.

THanks Anon. Let me guess......about 6k in sandra, and if not tighten the timmings to about 2.5-4-3-7 that is what the Boss's box ran @ 260 1:1

Well I'm not going to test anything but the loosest@ 1:1, DDR500 vs. the tightest @ 5:4 DDR400 because it seems like that's what most of the discussion is about. Testing overclocked sticks is pointless because you can't really nail down the overclockability of one rating of RAM. It varies from brand to brand, stick to stick. Also, varying the timings wouldn't show much either, beacuse it would also be very brand, module intensive. The RAM with loose timings would just sacrifice bandwidth for lower latency, whereas the modules with tight timings would do the opposite, and like I said, the modules' tolerance to these changes vary too much for it to yield any form of conclusive evidence.

True. Now that i think about it more. you are right.

The Results

The results below are the average of the results of 5 seperate runs in each test. Nothing was altered but the sticks themselves and the respective BIOS settings included below.

DDR500 @ 1:1 @ 3-4-4-8

PCMark2002 Memory Score---9098

3DMark2001 Score (default settings)---13205

Aquamark3 Score (default settings)---17030

Super Pi Time (In seconds calculated to 2 million places [lower is better])---110

Sandra Buffered (Avg. of Int and Float scores)---5573

Sandra Unbuffered (Avg. of Int and Float scores)---2683

Unreal Tournament (dm-asbestos flyby @ 1024x768 and high quality [average fps])---188


DDR400 @ 5:4 @ 2-2-2-5

PCMark2002 Memory Score---9391

3DMark2001 Score (default settings)---13424

Aquamark3 (default settings)---17032

Super Pi Time (In seconds calculated to 2 million places [lower is better])---106

Sandra Buffered (Avg. of Int and Float scores)---5404

Sandra Unbuffered (Avg. of Int and Float scores)---2895

Unreal Tournament (dm-asbestos flyby @ 1024x768 and high quality [average fps])---188

A few notes:
---The modules used to run the 2-2-2-5 tests were double-sided whereas the DDR500 ones were only single-sided. According to the findings in this article this may account for much of the difference between the 2 tests and actually cause the results to be much closer.
http://anandtech.com/memory/showdoc.html?i=1849&p=5

---Although the Average FPS were the same for both configurations in UT2K3 the low and high FPS were both higher in the 2-2-2-5 tests....go figure.

---Aquamark3 seems to be much more GPU intensive, like 3Dmark2003 so the difference in results is very small.

---The high bandwidth sticks are usually more overclockable (although not nearly as much as most reviews would have you believe) which may push them over the low latency sticks when this is taken into consideration, however, my conclusions are still the same when the extra price is taken into consideration.

---Sandra Unbuffered better measures the real-world performance of modules than the Buffered tests.

---Finding modules with such a low latency can be difficult. Most now only run at 2-2-3-6 or so (without getting higher bandwidth sticks with similar timings and underclocking them), which would probably fall short of the higher bandwidth sticks.

The conclusions are simple. Are higher bandwidth modules with very loose timings worth the extra money in order to run them 1:1 with the processor over cheaper, lower latency sticks run with a 5:4 divider?........HELL NO!!!

There you have it. Do with it what you will.