Less voltage is good voltage?

S

SoCalDJ

Limp Gawd
Joined
Jan 7, 2006
Messages
501
So at first i had some issues getting this Gskill TCCD pc-3200 stable at higher freq's. I couldn't get it for the life of me i tried everything, i had it running at 2.7Vdimm in the bios, 2.73 in windows damned DFI undervolt ><, anyways so i had this crazy idea to humor myself and try less voltage. Dropped it to 2.6 +.03V in bios, 2.67 in windows, and what do ya know. The mem stables out at TIGHTER timings, go figure right?
 
tccd can do that. my LE's most optimal voltage is 2.73v, 2.7 isn't quite enough, 2.8 is too much
 
Just the memory undervolts on the DFI board? I used a multimeter to check core voltage.. sets to exactly what I set in the BIOS.
 
Borgschulze said:
Just the memory undervolts on the DFI board? I used a multimeter to check core voltage.. sets to exactly what I set in the BIOS.
Click to expand...

Nice board. My DFI undervolts about ~0.02V on vDIMM and vCore.
 
Same, so it ran prime for a few hours at lower voltage. now for some reason it's even less stable than before i just don't get it. I haven't done much other than mess around and tweak it faster. for now i just turned it to a 140 divider and 2-2-2-5 for stabilities sake. I mean earlier today those settings worked just fine, now they woln't go for 1 min.....weird
 
SoCalDJ said:
Same, so it ran prime for a few hours at lower voltage. now for some reason it's even less stable than before i just don't get it. I haven't done much other than mess around and tweak it faster. for now i just turned it to a 140 divider and 2-2-2-5 for stabilities sake. I mean earlier today those settings worked just fine, now they woln't go for 1 min.....weird
Click to expand...
and mind telling us everything else that's set in the bios? ;)
 
Weel the only things i've actually personally set is the main timings and the 1T timing. Everything else is set to auto. i dunno give me a few mins and i'll post a screenshot of A64 tweaker. It's not loading atm so i'll have to give a reboot.

Also too couple more weird things. When i change the mem divider the voltage also seems to go up and down for instance 2.5 @ a 140/200 divider = 2.73Vdimm, BUT 2.6 @ 180/200 divider = 2.73 Vdimm.....*shrug*

Running more tests now seems to be stable today atm. I'd like to get it back to 306 tho so i'm still gonna post that screen of the A64 Tweaker timings. Eclipse if ya got and ideas let me know :)
 
LOL @ Bona, nah not stable yet. The stability of these sticks has yet to be determinded as they seem to be abit moody depending on the day!
:confused: Trying to get a tad more info in A64 tweaker to eclipse so he can tell me what sucks :)
 
SoCalDJ said:
Also too couple more weird things. When i change the mem divider the voltage also seems to go up and down for instance 2.5 @ a 140/200 divider = 2.73Vdimm, BUT 2.6 @ 180/200 divider = 2.73 Vdimm.....*shrug*
Click to expand...
that's strange. i know the dfi's give2.5 and 2.6v and extra .1v when running at trcd 2, but never seen that happen with dividers. i'll give you some suggestions on settings as soon as i see what you got now ;)

bona fide, my tccd can do 300mhz in memtest at 2.5-3-3 no problems, but my memory controller hates the stuff, and wants 2.5-4-3 to hit (kinda amusingly in this case) 306mhz, which is the limit of my memory controller :p
 
Ok stupid A64 tweaker isn't working so heres everything set in bios written ut as it is in Bios
Edit - http://server1.pictiger.com/img/73260/computer-games-and-screenshots/timings-2.php

LDT/FSB - x3.0
LDT Transfer Width - 16x16
CPU/FSB - x9.0
PCi Express Freq. - 100mhz
K8 cool and quiet - disabled
CPU Vid Startup -1.475
CPUvid control - 1.475
CPU Vid special control - auto
LDT Volt - 1.4
Chipset Volt - 1.7
Dram Volt - 2.7

Mem Page settings
Dram Freq. Setting - 180 = 9/10
CPC - Enable
Cas Latency = 2.5
Ras# to Cas# delay (Trcd) - 4
Min Ras# active time (Tras) - 7
Row Precharge (Trp) - 3
Row Cycle (Trc) - Auto
Row Refresh (Trfc) - auto
Write recovery time (Twr) - auto
Write to read delay ( Twtr) - auto
Read to write delay (Trwt) - auto
Refresh period (Tref) - auto
Dram bank interleave - enable
DQS Sqew control - auto
DQS Skew value - 0
Dram drive strength - auto
Dram data drive strength - auto
Max Async latency - Auto
Dram response time - Normal
Read Preamble time - auto
Idle Cycle limit - 256 cycles
Dynamic Counter - disable
R/W Queue Bypass - 16x
Bypass Max - 7x
32 byte granularity - Disable (4 bursts)

if u have any idea why A64 tweaker woln't refresh to the right settings let me know and i'll get you a screenie if it ever works ><
 
open cpu-z, then open a64tweaker ;)

i hope you didn't take the time to actually write all that out. i have the timings in the dfi's bios basically memorized by now, optimal for your tccd will proably be something like this
enabled
2.5
4
7
3
8
15
2
2
2
2-3

and play with drive strengths, it varies between rigs and sticks, so it's hard to say what yours will like ;)
 
it says weak and normal in a64tweaker, but you have 8 levels of drive to choose from in your bios, along with 3 data drive levels. mess with em all and find the one that gets the most mhz. make sure you actually max out the ram, not test for stability at x mhz, otherwise, no obvious trend will arise ;)
 
what do u suggest is the best way to do that? memtest isn't the best to tell. Prime seems to be the best but i dunno u tell me :)

And should i just stay with 2.73Vdimm?
 
i like superPI or windows memtest, just find the max at 8/3, then 8/2, then 8/1.. then whichever data drive is the best, keep it. then mess with the drive strength, lowering until you find the optimal drive. i think 2m or 4m should be good to pose somewhat of a stability test (hardly though) and not take up too much time.
 
ok

so what about prime95, thus far i've been doing some testing and at all of them and it seems to have a wall at 255ish in prime95, i upped the voltage to do some testing with that, but u suggest superpi over prime? cuz prime seems to fail more than superpi. Wouldn't it be a better test of stability?
 
ok so if i do a single instance of superpi i can run it at a 1:1@306 (which is where i'm trying to get) 4m passed but if i setup two instances of pi and run one on each core i instanlly get an error....

should i be running 2 instances or just one?
 
you can only run one instance of superPI PER FOLDER, copy the folder and run two out of each one

also, your p95 shot looks strange. it should be doing ~4000 itinerations of 1024k, for a total of like 4-6 tests per fft size
 
Prime95 is just default settings i haven't changed anything at all......so i dunno

Also super pi passed 2M and 4M @ 306 1:1
 
PWMIC is the temperature of those 3 heatsinks by the cpu socket. i actually find that they don't really play a very large role in how well stuff clocks, but the actual temperature of what you're oc'ing instead, the fan is keeping the ram cool, and it probably getting you an extra 5mhz or so :D
 
lol ok. Anyways so it'll pass super pi 2M and 4M with dual instances and 32M in one. But if i try to run 32M dual instances then i get an error on one of them. Whats next eclipse? :p

PS - It'll run prime for 1-2 hours too before it errors...usually due to the CPU not mem tho.
 
hmm, bump the vcore! :D

also, you can probably tweak a bit more stability out of the ram by messing with the timings. what i gave you were the best for my rig and tccd, might be different for you
 
Sorry it took me so long to get back to ya. Been testin and stuff. Anyways i've tried timings as loose as 2.5-4-4-8, hesitant to loosen them to 3-4-4-8 cuz wouldn't i get better preformance if i just used the 2.5-4-3-7 timings to get 275 on a 2180/200 divider? I dunno if it would make preformance better or worse. *shrug* So i've gotta be on the verge of getting this right tho @306 cuz i can pass a dual instance of 16M super pi without errors but for some reason soon as i try to get a 32M one going it'll error out after 3-20 loops :(

Had to up the voltage too. 2.78 is what it's at atm. Upped the CPU volts to to 1.47 hoping maybe it would stabilize the mem controller a bit (it did too :p)
 
i could but i wanna get the full potential out of my proc, seems it's max is like 2.8ish and stable at 2.75, so heres the question which would be faster, 2.75@306 3-4-4-8, or 2.70 @ 300 2.5-4-3-7?
 
probably gonna end up being really really similar, i'd test both to see what it shows


also, just a random thought, try switching memory stick order, see if it helps any ;)
 
Ok heres the results, they're all pretty similar cept for a few diffences,
SiSoft Sandra:
2.75@306 3-4-4-8
-CPU Multi: 52092 / 560701 <----both are higher!
-CPU Arith: 24758 / 11071 <----First is higher, but second is lower (the difference is like 36pts)
-Memory Bandwidth: 7277 / 7204 <-----This is the biggest difference by far beating the 2-2-2-5 setting at 140 divider by almost 1000!

PCPitstop: CPU:8569 Memory:10631 (Both scores are faster by 30+ pts)

now this is compared to 2.75 @180 divider and 2-2-2-5 @ 140 divider. I've still gotta test the 2.7 @ 2.5-4-3-7 but i think it'll be slower. Brb gonna go test it.
 
can you make a nice looking chart with the
Code: 
 tag when you're done to make the results easily readable? i need to have answers broken down for me :( :p
 
Ok Eclipse heres the breakdown. The
Code: 
 didn't really make it much better, mainly because i didn't know how to use it lol ><

2.75 @ 306x9 (3-4-4-8) 1:1 divider, Mem=306
Sisoft Sandra:
-CPU Multi- 52092 / 56701
-CPU Arith- 24758 / 11071
-Memory Bandwith- 7277 / 7204
PC Pitstop Bench:
-CPU- 8569
-Memory- 10631

2.75 @ 306x9 (2-2-2-5) 7:10 Divider, Mem=214
SiSoft Sandra:
-CPU Multi- 52049 / 56681
-CPU Arith- 24760 / 11249
-Memory Bandwidth- 6242 / 6237
PC Pitstop Bench:
-CPU- 8264
-Memory- 10580

2.75 @ 306x9 (2.5-4-3-7) 9:10 Divider, Mem=275
Sisoft Sandra:
-CPU Multi- 51946 / 56777
-CPU Arith- 24794 / 11165
-Memory Bandwidth- 7151 / 7073
PC Pitstop Bench (this one is abit off i think)
-CPU- 8287
-Memory- 10605

2.70 @ 300x9 (2.5-4-3-7) 1:1 Divider, Mem=300
Sisoft Sandra:
-CPU Multi- 51059 / 55143
-CPU Arith- 24310 / 10869
-Memory Bandwidth- 7246 / 7168
PC Pitstop Bench:
-CPU- 8398
-Memory- 10552

There ya go  :D
 
hmm, kinda all really synthetic tests that don't give you much info one way or the other. something like hl2's VST or an equivilent game bench might have been wiser, though we'd be picking over like .5fps differences :p

also, in setting #3, shouldn't you be able to run 2.5-3-3 at those speeds?
 
Nah it wouldn't boot, dunno maybe i got some tard stick *shrug*, and as for in game tests meh it's like u said how are ya gonna tell difference? But i'll run some Aquamark, 3DMark03 later. just got off work and i've over at a buddies house testing his new 175 :p
 
Ok some more benches here...
Aquamark03 - 105,392
-CPU = 13,237
-GPU = 17,502

3DMark06 - 4436
-CPU = 1889
 
You must log in or register to reply here.
Back
Top