yea, I'll try later on tonight when i have time...Oh and when My PC decides to stop turning on and off
Printable View
yea, I'll try later on tonight when i have time...Oh and when My PC decides to stop turning on and off
^^ hehe your DFi does that too! yeah nice trick... my damn comp shuts off when it's sleeping. Hitting power seems to awaken it as if it was just an advanced mode that makes the board seem off (no fans spinning, just that yellow standby light, but hitting power the desktop arises and fan noise returns), but at reduced clocks and MORE voltage (LDT, vcore, and nb voltage).
If i dont raise my LDT voltage by a tad my OC is not stable so yeah... voltage doesnt increase scores really, jyst might provide more STABLE MHz... even when BIOS is set the same then it's the stable part to harp on!
did anyone experiencing this speedincrease with voltage increase test if the lower voltage was prime/memtest/occt stable for a longer period?
if it is, then this really is strange
it it aint, it actually still is strange, since i thought only intel slowed certain things down when things got a lil hot / ?instable?
got me on this.
wow this is interesting.. voltage normally shouldn't have any influence on performance.. maybe stability but not performance.. very strange.
The only explanation i can come up with is that perhaps Hypertransport works with packets, and if it messes up sending one... maybe they are resent untill they DO get through correctly?
This way you wouldn't experience instability even if it is really there (say 10% of the packets arrive mangled or malformed), just the performance would be lower.
Filter, you're on a 312 HTT right, what's your LDT multiplier?
3x312=936MHz or 4x312=1248MHz ?
If it's 4x that would explain why the HT link has trouble sending data.
Try turning it down so you stay below 1000MHz, and then see if changing the LDT voltage still has any influence on your scores.
my thoughts exactly.Quote:
Originally Posted by Gogar
packet loss packet loss packet loss
When you have a phone modem you need to reduce the packet size for sending and recieving on the internet. Why? Because if it takes too long for the packet to be sent, it will time out and the server will start sending it to you again.
In this case, when a packet is sent it contains a parity bit. So for a 16 bit piece of data it will send an additional parity bit with it. The parity bit makes it so that there is an even number of 1's, that way if a single bit gets flipped (1 to 0 or vice versa) the reciever will know that the data was corrupted (there will be an uneven number of 1's).
Increasing the HTT (LDT) signal voltage will increase the slew rate of the signal. The slew rate is the rate that the voltage changes when you go from a 1 to a 0 or vice versa (think of the slope of the signal when you go from a 0 to a 1). When you increase the HTT freqency, the data lines may not have enough time to switch values, yeilding corrupted data.
I suspect that your HTT is tuned too high, and that you'll get a higher score by turning it down (keep it below 1000 like mentioned before). I'd turn down the voltage too to keep things cooler. As far as stability/preformance is concerned - maybe only the first bit is ever corrupted or some situation like that.
ok im gonna try this now i hope i can get a higher score lol ill try anything
________
YourMary
Only Oskar Wu could answer this theory, but maybe if the LDT multi is at auto and you set maximum LDT voltage it may auto select a higher LDT multi?
I'm at 3x htt multi
i experienced a better "performance" with the onboard nic with incresed ldt-voltage compared to default Before when i was downloading the computor would start to "lag" the cpu usage went up and down and it was near impossiple to move the mouse around and starting and closing windows etc took forever. No it dosen't and i can downlod. Strange since i only took a chance but hey it works. :toast:
What components will heat up with increasing LDT voltage?
Have you guys noticed any difference in the SuperPi benchmarks with a higher LDT voltage?
i did a superpi run once, with a tad lower vcore the score was 0.1sec better,
repeated this few times and got the same thing.
so pumping the procc with voltage ain't good for finetuning your rig:)
so this might be with 3dmark too ;)
LDT is the lightning data transport (http://www.hypertransport.org/docs/spec/ldt_spec1.01a_with_comment.pdf#search='LDT%20light ning%20data')
Using the LDT multiplier and the Bus speed the chips is at is how to calculate it. Using a program like CBID (central Brain Identifier) you can identify how high it is running. You want to keep the LDT to 1000 Mhz or below on the nforce4. Only when running more than 1000 MHZ LDT should you need to increase the voltage to it. I just helped someone last night with a 146 Opteron break 2.2 Ghz overclock - his LDT mult was on auto and he could not go past 230 Mhz, once he dropped to 3 he manged 2.9 Ghz.
Had he upped his LDT voltage he may have gotten stable past 1000 Mhz, the auto detect on the board wasnt working right.
So unless your LDT is over 1000 Mhz you should not have to up that voltage.
Not SuperPi.Quote:
Originally Posted by Daved+
But my Opteron 175 failed dual prime95 at 2.7 GHz after 20 hours.
I then set LDT voltage to 0.1 Volts more, lowered to 2695 MHz and run other RAM and it was now 48 hours stable. Of these factors the LDT voltage is a likely candidate to be responsible for the improvement.
I didn't cross-check the other factors. These runs were both with 4 GB RAM and the DFI board doesn't support ECC, so the prime fail can as well be a bit flipped in one of the RAM sticks.