Printable View
some companies won't honor a subsequent transfer of a warranty to a second person...better make sure CellShock will allow that...i would go to their website to read up on their warranty guarantee.
EDIT...here is their warranty page...ugh, you have to send the ram to Germany for RMA: http://cellshock.flagbit.com/en/service/guarantee.html
Yup. That's why I'm not going to sell the Gskill. That and I may not like 2GB with vista and switch back to 4. We'll see.
Thx Ace.
I'm currently on vista with 2 gb and I don't like it. I think I'm going to put xp on a raid0 tomorrow and can this install.
nothing like XP...it is a comfort zone..huh!:up:
I only installed the vista to run some vantage and I was foolish to think that I could actually enjoy using it daily with only 2gb ram :down:
Maybe next month I'll get a new 4gb kit and give vista64 a try.
Looks like for now I'll stick with f5d. Really no reason to mess with success.
http://lakesidepc.com/336x48t.jpg
and MIT settings
http://images6.theimagehosting.com/3433.372.th.jpg http://images6.theimagehosting.com/3435.th.jpg
Timings are kind of loose but I'll take it for this kind of stability with these temps, really doesn't seem noticeable to me when I game or work on vids. Much more noticeable is the lag from only running 2gb. It was failing agentgod's linpack util until I changed the MCH skew from auto to 100 and bumped MCH overvoltage to +.575v. I use LLC on and while there's still a pretty good drop, it keeps droop to about .03v or less which is good enough for me.
Thx for sharing loony. I did some testing last night, with LLC on and off. Personally I don't see much difference, but then I only did one test on each.
LLC Enabled 3.2GHz - 1.325v BIOS -> 1.292 idle - .029v droop
1.264 load - .061v droop
LLC Enabled 3.2GHz - 1.300v BIOS -> 1.280 idle - .020v droop
1.248 load - .052v droop
Now, the first thing I notice here is the disgusting amount of voltage I need to keep 3.2GHz stable on this board. If you'll look at this cap
3.0GHz previously only took 1.2000v BIOS / 1.176v reported:
http://img142.imageshack.us/img142/1...000v25hja6.png
3.2GHz ran at that voltage, but prime stability only took a pinch more at 1.2125v.
What I did last night was screw around with different combinations of values for GTLREF1/2. After trying the first combination -3/-3 I found I could come down from 1.35 to 1.3Vcore and remain stable. Whereas b4 prime would blow up after a few seconds at 1.3000v. I really don't know enough about these settings to know if they would effect how much Vcore could be used stably. But after reading a few articles on the subject I think it may.
For now, this has to be my main focus. I need to know how to get this board to undervolt or lowvolt b4 I go for the big clocks. I've had this chip up to 4.0GHz b4 and lets just say it didn't scale as well as I'd like. So far with this board I really don't think it will do it.
So, Ace, looney, shanks et al. In your experience, what might be the best chance for me to get this board to give me better clocks per voltage on a 65nm Quad? Is it even possible? I think I may already be on the way, but I haven't really done enough testing yet.
Perhaps I need to try a 45nm chip?
First of all I know it's nice to try and keep voltages low but there's really no need to be overly conservative with the kent quads. I have some and they seem to love abuse. I see the same question all the time "how can I get better clocks" and 99% of the time the answer is voltage. :D Try playing with those skews too. One setting at a time and also one increment at a time. I don't know if it'll help but it never hurts to try.
As a side note I finally got the ds3l back that I sent for rma at the same time as my x48t. Not sure why this took so long since it was a cut and dried replacement. I'll definitely be putting some small sinks and a fan on the mosfets on this one. ;) Great budget board but not really built to take the strain of crunching an oc quad 24/7.
Oh I'm no stranger to punishing my poor little quad with abusive amounts of voltage, believe me.
I'm just a tad bit disappointed that my $375 x48 board won't clock as high with the same voltage out of the box as my $150 680i. It's silly tbh.
I'll do some research on the skews, they are the only thing I haven't messed with.
Also, when you say crunching, what does that mean exactly? I've been folding for quite some time, but not to familiar with crunching.
Cheers,
I run boinc WCG. We have a subforum in the teams section. Check it out some time if you're curious. We have a blast and my opinion is there's no better cause. Lot's of people bash the nvidia chipset boards but I really like 'em and I think sli is lots of fun. For a high end board, I think gigabyte's bios is sadly lacking. This isn't to say that it's not sufficient because it is but it's not really more than sufficient. They should have gone further with fine tuning and voltage options for their top models.
Biostar's TPower I45 for $150 USD is the best bang for the buck right now...they supplanted Abit's IP35 Pro, IMHO.Quote:
Originally Posted by Ol'Baditude
the biostar is out performing or even with giga's X48T and it is a DDR2 board.
Ok guys, a bit of news. I spent the night playing around with the GLTREF values from the hidden settings menu. I tried every combination with an overclock of 3.2GHz and 1.2500 Vcore. It turns out that -3, -3 worked out the best and allowed a prime stable overclock. At least for an hour, but that's miles better than the instant blue screen I was getting with the default GTLREF values. So, as it stands I'm sittin pretty with a nice low voltage overclock, and I'll continue the Prime test over the next 24hrs. Hopefully there are no adverse long term effects of changing these values.
Btw. did I mention I got my new 1.2000 VID Q66 today???? $160 over at OCF. Sweet 65nm goodness. It's a peach so far.
If my temps seem a bit high, it's because the ambient in here is still 30degC. It was a muggy sh*tty day here.
IDLE:
http://img125.imageshack.us/img125/2...w32idlelh5.png
LOAD:
http://img125.imageshack.us/img125/4...w32loadpd7.png
what happen to your images???
Dunno, they were there last night. Hmm. Anyhow.
good overclock for a low VcoreQuote:
Originally Posted by Ol'Baditude
Yea, Thx. I've since had the opportunity to do a decent idle calibration. Hehe.
Also, I've since bumped it up to 4.14GHz, but she BSD'd about 5mins after I started Prime95. It also BSD'd about halfway through 3Dmark06. Too bad.
Any Tips guys??? FSB at that point is 460. Vcore is 1.65, then tried 1.7v. With LLC enabled the vcore was shooting up to 1.806-1.9v in CPUID. Yikes, but that being said, I would give it more.
Not sure why some people suggest a 100 PCIe Freq. and some 104 or 105. That's kinda been confusing me. I've locked it in at 100 even though it doesn't seem to make any difference from not doing it.
Those Skews have got to be good for something, whataya think? What the hell do they really do I wonder. I've been doing a bit of research, and can't seem to get a solid answer.
BIOS Settings are as follows.
GIGABYTE GA X48T DQ6 Clock Settings
MB Intelligent Tweaker (MIT)
Robust Graphics Booster...........: Turbo
CPU Clock Ratio...................: 9
CPU Frequency.....................: ?
CPU Host Clock Control............: 3,750
CPU Host Frequency(Mhz)...........: 460
PCI Express Frequency (Mhz).......: 100
C.I.A.2...........................: Disabled
Performance Enhance...............: Standard
System Memory Multiplier (SPD)....: 3.2B
Memory Frequency(Mhz).............: 14something
DRAM Timing Selectable ...........: Manual
Standard Timing Control
CAS Latency Time (tCL)............: 7
DRAM RAS# to CAS# (tRCD).........: 7
DRAM RAS# Precharge (tRP).........: 7
Precharge Delay (tRAS)............: 18
Advanced Timing Control
ACT to ACT Delay(tRRD)............: 5
Rank Write to READ Delay (tWTR)...: 5
Write To Precharge Delay (tWR)....: 10
Refresh to ACT Delay (tRFC).......: 60
Read To Precharge Delay (tRTP)....: 5
Static tREAD Value (tRD)..........: 9
Static tREAD Phase Adj............: Auto
Command Rate (CMD)................: 2T
Clock Driving and Skew Control
CPU/PCIEX Clock Driving Control...: 800
CPU Clock Skew Control............: Normal
(G)MCH Clock Skew Control.........: Normal
System Voltage Control () = Base Voltages
DDR3 OverVoltage Control...(1.50v): +0.40v
PCI-E OverVoltage Control.,(1.50v): +0.15v
FSB OverVoltage Control....(1.10v): +0.30v
(G)MCH OverVoltage Control,(1.45v): +0.35v
CPU GTLREF1 Voltage*.......(67%)..: +3%
CPU GTLREF2 Voltage*.......(61%)..: Auto
Loadline Calibration..............: Enabled
CPU Voltage Control...............: 1.7000v
Normal CPU Vcore..................: 1.2000v
Note: * Hidden Features Press [CTRL] + [F1] from main Menu to Access
Advanced Bios Features
Limit CPUID Max to 3..............: Disabled
No-Execute Memory Protect.........: Disabled
CPU Enhanced Halt(CIE)............: Disabled
CPU Thermal Monitor 2.............: Enabled
CPU EIST Function.................: Disabled
Virtuaalization Technology........: Disabled
Idle, Load and quick Benchmark at 23Deg. Ambient
http://img176.imageshack.us/img176/2...ghzidleps4.png
http://img379.imageshack.us/img379/6...ghzloadef2.png
http://img365.imageshack.us/img365/6...hz3dm06up5.png
that one radeon 4800 really pumps out the 3D2006 score...that is impressive for a single card!:up:
suggestions for 4.1 ghz in red next to your above settings....BUT...4.1 for a Q6600 is going to be real tough to pull off unless you are using water.Quote:
Originally Posted by Ol'Baditude
Yup, she's all under water. Loop Specifications are in my sig, but with spanky new homemade coolant. Hehehe.
http://img353.imageshack.us/img353/2558/img0702pd6.jpg
Thx for the suggestions. I'll tweak some more and post back later.
also, forgot to ask why you do not use "Loadline Calibration?"
Sorry about that Ace. In the text above the settings listing, it says that LLC was enabled. I just neglected to add it in the settings list.
I'd say 460fsb is maxing that q6600 of yours unless it could get more fsb voltage.
Yea, probably. This is the highest FSB I tried with my low multi testing, I honestly don't know where the wall is on this CPU. We shall see.
generally, on quads it works better at 445 fsb...usually the max is 475-480 fsb...i can get higher on my biostar but the chipset latencies have to be really relaxed.
Oh yea, 445x9 is no problem as in the screenshots above. Giver 1.65 and it'll prime until I'm ready to shut it down. Although, honestly I've never felt the need to test 4.0GHz for 24+ hours, since I know I'm not going to be running that 24/7.
Now, when you say chipset latencies, what exactly does that mean. It's a new term. What BIOS parameters are we talking about there.
And what the heck is with that Biostar??? What made you decide to pick that up???
Thx again for all your help Ace and Loony.