The Pook

Wondered why 460FSB was only a 3 seconds faster than 400FSB, but 380FSB was 5 seconds slower than 400.

__________________

Ihatethedukes

It starts to make those "mid-400 FSB prime-stable" boards look less attractive. It's either ~500 or nothing.

money11465

I second that! I wonder if Gigabyte will release strap control for us, soon, too. Here's a list of where new straps may take effect, for quick memorization:
200FSB
266FSB
333FSB
400FSB
466FSB
533FSB
600FSB
That's as far as we need, I think. The straps occur every 66FSB, or every 266MHz of Intel's quad-pumped FSB. Basically, push as close to one of these numbers as possible without going over, or jump to the next strap. Then adjust your FSB with Clockgen for screenies/benchies/SuperPi.

__________________

Money11465's Conroe Build [56k warning] - Updated 1-18-07!

How to: Code in Visual Basic (vB) - OOP (Objects) added | Calculator for Conroe straps and NBCC | How To: Set a memory divider on Gigabyte DS3 | Secure your Thermalright Ultra Extreme (TRUE)

pauldovi

I hope you also realize what this means for overclocking. ASUS, Gigabyte, DFI, or any other motherboard company can just put you on the 2400Mhz strap. The northbridge latencies will be extremely high and thus you will have great overclocking. The reviews will look great... But the performance won't be.

This is very important when considering a new motherboard. Tony is pushing for BIOS Engineers to give the end user control over the strap limit and many other tweaks for the north bridge. Who knows how close we are to seeing that ability.

It is clearly important for people to know that a higher overclock does not result in better performance.

__________________

Vasco

Oooo thx man. Thats a really good explanation. Is it true though that the DFI RD600 has a strapless NB?

Jimmy2Shoe

I actually did some memory bandwith tests and SuperPI Mod tests from 395 to 465FSB in 5FSB increments this afternoon.

Super PI seems to be pretty linear, though I'm only eye-balling it and haven't graphed it.

As for memory bandwith, there are some really strange ones:
395 --> 6502MB/s
400 --> 6619MB/s
405 --> 6644MB/s
410 --> 6646MB/s
415 --> 6836MB/s big jump
420 --> 6806MB/s ***?
425 --> 6993MB/s
430 --> 6961MB/s ***?
435 --> 7210MB/s big jump
440 --> 7257MB/s
445 --> 7236MB/s ***?
450 --> 7383MB/s
455 --> 7486MB/s
460 --> 7541MB/s
465 --> 7629MB/s
470 --> Doesn't boot to windows...

I did the tests with SiSoft Sandra, 4-4-4-12 on the G.Skill Ram, always 1:1. I might graph the SuperPiMod times, but it seems pretty linear (1M). Bios is F6 (Gigabyte's Latest) The ***? numbers proove that from a memory bandwith standpoint, you can actually LOOSE speed. There weren't any instances of SuperPi being slower with additional FSB though.

MADMAX22

hey jimmy if you want try that but use one lower multi on your cpu, its pretty interesting

__________________
E6400@3856E4300@4158E6600@4510,Opti148@3102,opti170@2918,Xeon3220(B3)@3430(490fsb)
UD3P@8400600FSB

Jimmy2Shoe

I think my POS 512MB sticks of RAM have capped off at DDR2-930. I tried dropping the multi to get +FSB but it just won't go.

In any case, it took me nearly 1h30min to do that, I don't really feel like it again.... pauldovi is trying it out with his new board I think sometime soon.

MADMAX22

ahh thats cool, i would but my mobo doesnt support multi changes

__________________
E6400@3856E4300@4158E6600@4510,Opti148@3102,opti170@2918,Xeon3220(B3)@3430(490fsb)
UD3P@8400600FSB

pauldovi

I just got back from my trip to South Florida. The new board is here. I will be installing it tonight or tomorrow, and I will do a bunch of tests!

__________________