Terrible Venice experience...

[CMC79]

Did that do anything for performance though? That is, did the extra 70MHz cancel out your increased latencies?

I think I can hit >2600Mhz--is it safe to up the voltage on chipset to 1.8? (and the LDT bus to 1.4?)?

 

[rise]

i run my ldt at 1.4 but not the chip at 1.8, too high for me. i've seen people do it though.

edit- my chip is 1.6=1.58 in ITE and ldt reads 1.36 in ITE

 

[CMC79]

I figured 1.8 was a bit high for the chipset...I'll try upping the LDT voltage...I was fine at 9x280 but 289 died on me. I am a bit perplexed why a Windows file was corrupted if I've got locks on the PCI bus. (I'm using IDE, not SATA, for now.) I do know that I can get better than 2.5 without too much effort though.

 

[Lithan]

CMC you are using value ram with generic chips. You WONT run 1t. Just wont happen. Sorry to tell you. To verify it's the sticks and not controller, use Memtest86. If you fail memtest it's the ram. If memtest can run for hours on end, it's the memory controller. For example.


2x512 Generic 5ns chips with winchester @ 250 1t - Fails memtest, fails windows
2x512 Generic 5ns chips with winchester @ 250 2t -Pass memtest, pass windows
2x512 Generic 5ns chips with Venice @ 250 1t Fails memtest, fails windows
2x512 Generic 5ns chips with Venice @ 250 2t -Pass memtest, pass windows
2x512 TCCD With Winchester @ 266 1t Pass memtest, fails windows
2x512 TCCD With Winchester @ 266 2t Pass memtest, Pass windows
2x512 TCCD with Venice @ 266 1t Pass memtest, Pass Windows
2x512 TCCD with Venice @ 266 2t Pass memtest, Pass Windows

 

[CMC79]

I'm actually glad to hear that--I figured (and hoped!) it was the memory. Time to buy some better stuff. The chip itself seems to have a decent amount of headroom though. Till then, I guess I'll stick with 2T and getting as much out of it as I can. Thanks for all the help everyone--I came from a situation where I was overclocking a T-bred 2100 rather simply on an Nforce 2 board (just raised 133 to 166, with a tiny bump in voltage), and the DFI bios is, well, rather more complicated.

 

[Zebo]

:::YAWN:::

Nothing wrong with using a divider. in fact underclocked ram if you apply tight timings can be just as fast as TCCD.

Try this: set HTT to 180ish, mem voltage to 2.8 or 2.9 and try 2-2-2-8 1T see if that passes memtest 5 AND 6.

Every single one of the 3dmark leader board is using a mem divider. Granted it's not value ram, but it's still using a divider, usually 166.

2-2-2-8 1T is the fastest and you need 80 more Mhz of bandwidth @ 3-4-3 1T to beat it. In other words you're better off underclocking that value ram, using 133 or even 100 memsetting than going for bandwidth of any kind. My value Buffalo runs 2-2-2-8 1T all day long at 180Mhz.

 

[Acanthus]

Originally posted by: ExarKun333
memory timings are KEY on this DFI board...dont overlook them...check dfi-streetcom for more info and check if someone has some good settings for the venice 3000 you are running...u could have a faulty mem controller too...not being able to even use 1T on only 2 sticks is not what it is supposed to be able to do

Show me a real world (NOT SANDRA OR PI) bench where 1t makes more than a 2% difference.

 

[Suspicious-Teach8788]

Do realize a divider doesn't make much of a difference if your CPU speed is not increasing.

ASSUME constant memory and CPU speeds.

1) Use high HTT w/ divider
2) Keep 1:1

1) will be like SLIGHTLY faster, we're talking 20 3D Marks in 2001. In most tests its like 0.1% difference.

 

[jbh129]

Originally posted by: Zebo


Every single one of the 3dmark leader board is using a mem divider. Granted it's not value ram, but it's still using a divider, usually 166.

That is a bit misleading to most people on this forum. They run a divider because they are using BH5/CH5/VX at 260+ with 2-2-2-5 timings.

 

[DaRkPhAnToM]

why does it matter if u are using a SATA drive? i have my main windows drive on SATA port 1 on the dfi nforce4 ultra-d is that gonna limit my oc? how do i lock it? im kinda confused...