Shall i try and go for the gold?

[bluestreakLB]

I'm thinking about getting another 512mb of ram making htat 1gig but i've also been thinking if maybe i should sell my current two sticks of PC3200 and get a gig of some faster ram...pc3700...pc4000.

I am just not sure if the NF7-s v2.0's can even go that high, i know it's a guessing game but without any modding would it even be worth it, on my xms pc3200 i get about 215, is that most likely the limit of my mobo or my ram?

 

[9mmCensor]

nope, it probably be best for lower speeds (pc3500) and tighter timings.

 

[squads]

well w/o any mods my nf7-s is topping out at 227 fsb, so some BH-5 will run tight timings at least that fsb, so i'd get some BH-5 stuff

 

[maxima88]

If you're going to run in dual channel mode, then the highest you can look to get to is 222FSB or a few mhz higher. I have gotten my PC3500 C2PT to 246FSB stable in single mode but only stable at 222FSB in Dual channel mode.

So if it's worth 5-10mhz more in the FSB range, then go for the faster RAM but if not, then stick with what you have right now.

 

[itshondo]

a great benefit of having a gig is that you can set windows to for no swap/paging file.

 

[mongoled]

Maxima,

A number of us are running 1gig of ram at considerably higher mhz then u are presuming. Granted that it isnt many of us but it is very possible.

The settings in my sig are rock stable for 24/7 use. I can run the mem in this config at 235mhz rock stable.

bluestreakLB, high mhz are hit and miss, if u dont try then u will never know what may be achieved. U just have to do what u think is right for you.

Hope this helps

L8rz

 

[maxima88]

mogoled, what bios are you using?
Glad to see that you can run at 231fsb stable. You're one of the lucky ones.

 

[mongoled]

I think that i have should have stated that my board is modded. I am not sure if you guys are talking abt non modded boards.

If you are then I would have to say that u need to be very lucky to achieve high fsb in dual channel.

Maxima im currently running 9.5x231=2200mhz with d20 bios. I cannot run with three sticks populating all dimm with d10 bios over 200mhz. I mentioned in a thread on some forum that I believed that the reason they cant sought out newer bioses to run with high fsb on multis of 10-10.5 is because of this problem with the 3 dimms being populated at fsb over 200mhz.

Ive been trying to find peeps to test with 3 dimms and d10 bios over 200mhz to c if they experience the same problem as me. Its quite strange, when u try to boot up u get no beep, no siren, nothing! The comp just sits there with fans turning and does nowt.

 

[maxima88]

Mongoled. Thanks for knocking some sense into my empty head. I just realized that I tried to go past 222fsb @ 10.5x multi using bios 20 which won't work unless I'm using d10 bios.

I just dropped my multi to 9.5 and now I'm 242fsb Dual Channel.

 

[Hoot]

Okay, Humor ole Hoot. Running 9x242 (2299) 2.5-3-3-9 run 3D Mark 2001se. Then switch to 11.5x200 (2300) 2-2-2-10 or 11 and tell me the difference in the score as an indicator of the benefits from such a high FSB with retarded timings. BTW, I'm impressed that you got your TI-4200 to overclock so far. Even with big sinks on everything, I cant get that high.

Hoot

 

[maxima88]

BTW, I'm impressed that you got your TI-4200 to overclock so far. Even with big sinks on everything, I cant get that high.

Hoot

I was pleasantly surprised that it o/c this high. Finally found my chance to get close to 17000 in 3dmark01. Maybe if I added giant heatsink I might finally make it to 17000.
http://service.futuremark.com/compare?2k1=7294546

I have a 120mm Sunon fan on the side cover blowing a lot of cool air onto the vid card that helps it o/c a lot higher.

 

[maxima88]

Ok Hoot, I ran3dmark2001 11.5 x 200 = 2300mhz 2-2-2-10
w/ 333/733 vid card o/c. scored 15,983.


@ 9.5 x 242, 3dmark kicks me out during Dragothic high detail test. Don't know why it does, I'm already using 2.0vcore.

But here's a run I did a few days ago @ 9 x 245fsb = 2214mhz with RAM settings of 2.5-3-3-11 w/ a 15,918 score.

So even running 86mhz slower than the 11.5 x 200 and even with relaxed RAM timing, higher fsb will produce higher overall scores in 3dmark2001. Except for the Nature test, I noticed that
this test is more depedent on CPU speed than FSB. All the other test benefit from the higher FSB speeds.

Hope this helps you.

Gotta go to bed now.

Edit 1:
I just realized that the 9x245 run was in single channel mode only. Imagine what the score would have been had it been in dual channel mode. And based on the FPS rate in the first 3 tests in 3dmark2001 when I ran it @ 9.5x242 dual channel, it was noticeably faster than the fps running @ 11.5x200.

 

[ninety6]

What Hoot is trying to say is that you should really shoot for the "balance" of "high fsb" with "tight timings".

Try to shoot for 11-2-2-2 timings at around 220 or 215 and retest.

96

 

[maxima88]

I believe Hoot just wanted to compare the performance of really high FSB w/ relaxed timing to fairly high FSB with tight timing. That's all he was interested in.

 

[mongoled]

Congratulations Maxima,

Glad to be of help. We all fall foul to these types of user 'error'. I say 'error' becuase its more like u just forgot rather than made a mistake with the multis!

By the way, you seem to suffer from the same problem many of us suffer with the NF7 with high clockspeeds and high fsb. I too get kicked out of 3dmark01 in the same way as yourself, its not dependent on the type of tests which is running at the time but random.

Also from my testing it seems the problem which is occuring does not seem to be linked with Vcore or cooling. Example i can run 3dmark stabily at 208x12 which is around 2500mhz @ 2.06v Vcore but cannot run 3dmark stabily at 223x11 which is around 2450mhz @ 2v Vcore. Also i can run 3dmark at 240x8.5 which is around 2050mhz. All these tests are run at 11-2-2-2 with 2xBH5 sticks in dual/Sync at 3.5volts Vdimm and 3.6volts for the 3.3v rail.

Increasing/decreasing volts does not help, easing the latencies for the ram does not help, providing extra cooling in form of fans blowing across several components does not help. From what i have read in numerous forums nobody has been able to provide a universal solution to this prob. It seems to stem from the combination of mobo with CPU, some peeps get lucky and have a 'good' combo others of us have less luck.

Over the Christamas period I will be doing some more research into this phenomena. I am setting up my first WC rig and will be cooling nthbridge, gfx and CPU. I havent been able to fully test to c if by increasing the vdd to very high levels, would help benefit stabilty for higher fsb speeds in 3dmark01. I have crystal orb on my nthbridge and the 'best' vdd from my experiments seems to be between 1.8-1.9 volts any higher cause more instabilty.

Oh be4 i forget one thing which did help stabilty a little while not definitive is increasing AGP frequency to abt 80mhz. Yup i dont get it either but i found that at this high frequency tests which would kick back to desktop instantly now had a chance of completing a full run. i have to point out that i still am getting kicked to desktop but not instantly with AGP @ 66 mhz.

If u want to try this out the best way to test this is to set your fsb to a mhz which u know u will be kicked out of 3dmark instabtly. Keep your multis low and voltage low. Then simply raise your AGP to the highest level u know is stable and try to run the test. I hope your results mirror mine.

Hope this info may help you

mongoled

 

[Hoot]

On my nF7-S, I found my AGP and PCI dependent devices did not mind setting the AGP to 80 also. Maybe they would have tolerated more, but 80 was as high as I tried. It gave me about 100 more points on 3D Mark 2001se, but I didn't see any impact upon stability. The reason I chose 80 initially was because it resulted in an AGP divider of exactly 5 when I run 200 FSB at tight timings. I don't know if picking an AGP/PCI speed that results in a whole number divider has any effect upon stability, I was just experimenting during an idle moment...

Hoot

 

[maxima88]

Hoot, I'll try running 3dmark2001 tonight using single channel at the 2 combinations of FSBXmultis and RAM timings you gave me. At least I know I won't be kicked out of 3dmark2001 using single channel RAM even up to 250fsb.

I'll post back w/ the scores.

BTW, what kinda voltage rails do you have with that Antec 550 of yours. Did you adjust any of it's pots?

 

[bluestreakLB]

wow...nothing like leaving for a few days only to find that the thread exploads with info

Finally no work tonight, time to play a little.

 

[maxima88]

ok HOOT, I've got the scoop you've been looking for.

I ran all these 3DMark2001se tests 1.85v 2.8v 1.7v cchipset and 1.8 agp all in single channel mode.

200 x 11.5 = 2300mhz 10-2-2-2 = 15,720

219 x 10.5 = 2300mhz 10-2-2-2 = 16,164

230 x 10 = 2300mhz 10-2-2-2 = 16,317

So as you can see, the higher the FSB, the higher the score.
The only exception was the nature test where the score was almost the same on each test which means that it's mainly dependent on the cpu speed and not the fsb.

 

[Zyl]

I believe the nature test is based mostly on the video card and not cpu-dependant isnt it?

and I think Hoot meant the tests as not just FSB and Multiplier but rather whether the score will be faster with highest FSB you can get with tight timings VS. loose timings with highest FSB for that. IE: (we can see you do 230x10 10-2-2-2) try that vs. 240x9.5 11-3-3-2.5 and see which score is higher