It's a G0 1.3125v VID one. I left all the GTL settings at AUTO, because I didn't know what else to do with them.
Printable View
It's a G0 1.3125v VID one. I left all the GTL settings at AUTO, because I didn't know what else to do with them.
Well look at my settings for all of them and the NB one too as those are the most common for the Q6600's.
Your Vid is not that much higher then mine .. so i would say maybe 1.45-1.46 3.6ghz for you with other settings adjusted .... but all you have to do is try.
Try em out like i did .. trial and error and thks to you making linpack easier to use .... i now have v3 linpack stable, beating my head about what i had to change for settings.... lol
All i can say is update your bios to 1201 and try my settings .... just be sure to back up your bios incase you have to flash back as some ppl claim RAID/SATA was not working for them after flashing to 1201 ... but that can happen to anyone no matter what Bios they flash too.
I not have a raid setup, but i do want raid 0 again ... i had 2 x 320mb seagates about 4 months ago, but i sold them to pay some bills off and i settled for this 160gb for now.
It's a dead end... most chips 1.3000v VID and over can't do 3.6GHz without water.
There's a HUGE difference between 1.2625 and 1.2875, and a bigger one between 1.2875 and 1.3000.
Now if you got a 45nm chip, VID has almost no correlation with overclocking ability, it all comes down to how well your chip scales... there isn't a definite measurement.
I would give a shot as with good air cooling which he has a decent cooler, he could get 3.6ghz maybe around 1.45-1.46-1.48, but it will be trial and error on other settings in BIOS too.
There is no 100% such a thing that 3.6ghz from 1.30 VID and over can't do so with air cooling as it can be done and i am sure has been done ... need good air flow is all if the chip needs high volts.
Oh just to add my friend has Q9450 45nm and his temps under load folding is very similar to mine and he runs a Scythe Ninja Rev B .. i had one sold it for my Thermalright Ultra 120 Extreme.
He is at 3.6ghz ... point is .. even a Q9450 45nm at 3.6ghz similar HEAT and he is at 1.35-1.38 volts for it to be Orthos stable 24hrs and linpack v3 stable 10 loops.
Thanks for the prog AngenGod, looks like my system is still stable after 10 loops. My 3.4 q6600 only hit 66c using real temp. Not bad!
And for your reference regarding your q6600, mine has a vid of 1.300, and takes 1.45 volts for 3.6, so id say yours seems right needing a little more. I dont like how hot it gets at those volts using air so I keep it at 3.4(1.375 volts.)
There is no "try"... The chips I've gotten have all been hit or miss. On the bad chips I tried countless sets of voltages and settings and whatnot, and in the end it just comes down to what chip you got. 3.6GHz was a easy feat in the early days of G0, but not anymore.Quote:
Originally Posted by BeastNotro
There is no 100% thing, but there is a 99% thing. You're a lucky one, just don't expect everyone else to get the same results.
The Q9450 I got couldn't do 3.6GHz, period.
There are a lot of people who take good chips for granted, and there are a lot more who get smacked in the face with a brick wall again and again.
There was one guy here who got around SIX 1.3000v+ VID Q6600s IN A ROW... none of them could do 3.4GHz, much less 3.6GHz. There are a lot of people here who would KILL for an AVERAGE chip, and it all comes down to luck.
@BeastNotro
The 1201 BIOS helps with memory compatibility. With 1103 i was hitting the limit hard at 400MHz FSB because memory was already at 800MHz. Now i can boot with 410MHz FSB where it seems something just can't go any further. With better components i'm sure it could go higher than just 10MHz more. But still, thats 20MHz more on memory and whole 80MHz on RAM. Not much but always better than nothing.
@AgentGOD
I don't like the latest verion too much. It's not flexible enough.
You cannot manually select memory size, only 3 predefined modes. There should be forth mode called "Custom" where you can manually enter memory size.
And whats up with the "Run IntelBurnTest" confirmation? It's a completelly unecessary step in between. If you run IntelBurnTest by mistake you can always close it with Close button top right corner like any other Win app...
Now running Xp-64, everything works perfectly.
If it's possible to make it more 32-bit compatible, that would be cool, but doesn't really affect me anymore =P.
-Zigosity
Both my program and Linpack is fully 32-bit and 64-bit compatible. However, the problem arises with 32-bit when the user has more than 2 GB of system RAM, because 32-bit windows doesn't allow programs to allocate more than 2 GB of memory for each program. And since you are trying to use maximum free memory and have more than 2 GB of RAM, Linpack fails to use it.Quote:
Originally Posted by Zigosity
Yeah I'll take that into consideration.Quote:
Originally Posted by RejZoR
Would it be possible to just set the maximum RAM used to a fixed 2GB in 32-bit mode, then?Quote:
Originally Posted by AgentGOD
AgentGOD, on the 32bit exe make sure you add /LARGEADDRESSAWARE to the link.exe switches or cmdline when linking dynamic libraries. This might allow you to use more than 2GB private memory space. I don't know if Linpack is compiled with this option, may require you to compile it from source.
Also which Linpack are you using? Compiled from which version of MKL?
I just installed the MKL 10.1b2 and gave the new Linpack a whirl to see if it improved from the previous 10.0 I was using and all I can say is damn. Intel have definitely done some optimizing. 64-bit exe pushes 100% cpu during the whole problem which the older ones never did.
Difference overall is about 8GFlop between 10.0 and 10.1b2
Current date/time: Fri Aug 15 18:24:33 2008
CPU frequency: 3.870 GHz
Number of CPUs: 4
Number of threads: 4
Parameters are set to:
Number of tests : 2
Number of equations to solve (problem size) : 2000 14000
Leading dimension of array : 2008 14008
Number of trials to run : 6 4
Data alignment value (in Kbytes) : 4 4
Maximum memory requested that can be used = 1569180256, at the size = 14000
============= Timing linear equation system solver =================
Size LDA Align. Time(s) GFlops Residual Residual(norm)
2000 2008 4 0.155 34.5401 4.657913e-012 4.051814e-002
2000 2008 4 0.157 34.1272 4.657913e-012 4.051814e-002
2000 2008 4 0.158 33.8337 4.657913e-012 4.051814e-002
2000 2008 4 0.150 35.5305 4.657913e-012 4.051814e-002
2000 2008 4 0.219 24.4316 4.657913e-012 4.051814e-002
2000 2008 4 0.171 31.2108 4.657913e-012 4.051814e-002
14000 14008 4 40.694 44.9630 1.832709e-010 3.309677e-002
14000 14008 4 40.512 45.1650 1.832709e-010 3.309677e-002
14000 14008 4 40.790 44.8575 1.832709e-010 3.309677e-002
14000 14008 4 40.478 45.2025 1.832709e-010 3.309677e-002
Performance Summary (GFlops)
Size LDA Align. Average Maximal
2000 2008 4 32.2790 35.5305
14000 14008 4 45.0470 45.2025
End of tests
Actual frequency is 3440MHz. Temps hit highest of 80c on one core. The earlier one struggles to hit 70! If you don't have the beta build, shoot me a message and I'll send your way. It isn't static linked anymore though, requires 32bit and 64bit libiomp5md.dll to be on the system which for portability sake isn't all that good.
I'm using the Linpack binaries that came with 10.0.3.021.
Great program.
Have been testing my system with v1.3.
Vista 64bit Ultimate
ASUS P5E WS-Pro
Q6600 G0 vid 1.325v
Ram DDR2 OCZ 2 x 2GB 6400 5-4-4-15 @1.9V
Corsair HX1000 PSU
ASUS 3870 GPU
Normally I have 4 x 2GB but one stick is faulty and am returning it.
I noticed that on a 20 pass run with FSB of 355, the max temp was 55, I then ran Prime for 1 hour and its max temp was 45.
It also uses more CPU Volts, that is, CPU-Z shows the Core Voltage about 1 step (0.00625v) higher when running this than when running Prime.
After much testing, I am running a stable ( 30 passes ) with FSB 360, Vcore in Bios of 1.3215, Voltage Damper is Enabled, all other voltages set to AUTO.
To run a higher FSB the Vcore needs to be pushed a lot more (1.34375) at FSB 370 max temp of 55 but not stable.
Room temp is 24C
First picture is whilst test was running (5 passes ) and the second picture was just after it completed.
Yep, the recent version of Linpack that AgentGod is using, is much better at finding unstable OC's than the previous version which I tried in last thread.
Testing this version, I tried some known settings.
4Ghz at my normal vcore bios setting is 15+ hours 10ffts orthos stable, 15+ hrs stable prime small ffts, and ran 30 runs Linpack on AgentGod's max setting without difficulty.
But for testing, I dropped 2 notches in bios, where I fail orthos 10ffts/prime small ffts reliably in 2-5hrs, shown is a fail in 3 hours.
Then ran IntelBurnTest using the obviously improved Linpack. First off, runs of 5 were too few to reliably find a 2-4 hours orthos stable OC, on my computer anyways. I had to run 4 runs of 5's on max setting to find an error. Also settings 2 and 3, I found to be useless as well, unless looking for a very unstable OC. I passed multiple runs of 15 passes on settings 2 and 3. However, setting 1, using max memory, was very useful using 15 passes, I failed it 5 times in a row. So using max memory setting 1, I would discover 100% of the time in 20 minutes, what takes orthos or prime to tell me in roughly 2-4 hours.
Just my opinion, I would get rid of setting 2 and 3, I cant find any use for them, maybe a custom and max. And using setting 2 and 3 versus 1 is clearly NOT equivalent to primarily testing memory versus cpu. IMO, I dont see a use for any setting but 1, but that one works very well.
Also, it is kind of wordy in the interface, but I like the idea of having an interface. But as that improves, I think it will be useful, using runs of 5 on max to quickly find settings, then runs of 15? 20? 30 to show stable OC. But to be sure, need to test 1 notch below stability which fails orthos in 6-10 hours, to see how many passes are needed to reliably create an error and fail and show the OC as unstable.
Agentgod, check your inbox. Sent you a present.
Lol, looks like the new beta binaries are more bloated.
Can't wait for the next release. I had one thought though, in the menu options, can you set it to Key input instead of prompt so that I only have to hit 'y', and not enter? It's not that big of a deal, but any time you can get it to a running point faster is better in my opinion. I also agree that there should be a custom amount of memory setting.
There shall be a custom stress level option in v1.4. However, it is cumbersome to do the "Key input" in C++, as I would have to sacrifice efficiency to do so.Quote:
Originally Posted by WoZZeR999
No problem, just thought if it wasn't a big deal then it would be nice.
Updated to v1.4
- Fixed a bug with percentage display that resulted in displaying 0% instead of the actual
pass rate
- Removed redundant prompt at the beginning asking if user wants to run the program
- Added customize option for selecting the stress level
I have a great question(s).
Since many PC's are stable on v1 and v2, but many have failed v3 ...... this brings me to the conclusion ... was all the millions of PC's running v1 or v2 actually stable?...... lol
Also all those who passed v1 and v2 were Orthos stable as well .... now were they really stable ... if they failed v3 ... ?
What i am getting at is v3 fails many ppl and adjustments in Bios can get you stable again, but previous versions were stable for many ppl at the lower bios settings .. so were the PC's actually stable when using v1 and v2?
If orthos Blend was stable before on v1 and v2 linpack ....... is it stable now in prime/orthos?
Glad to see you were able to get stable BeastNotro :up:
I agree 1.3 is certainly a ball breaker...I now wonder what V 1.4 will have in store for us....
John
If it's stable in daily situations like gaming, folding or co would you not call it stable as it fails a program like this... I like what Agentgod did here and I will use the program as a quick reference but you cannot call it stable if it passes this... it still can cap out on any benchmark, game etc... but it's a good step in the right direction...for some...
Some people are way too obsessed with running stress programs, most users don't even stress their rigs as hard as a normal prime let alone a linpack test ... and there again there's no bulletproof testprogram... a mix is okay but don't rely only on this, as eg ya cooling too could be a factor why your rig caps out...
But anyway thx again Agentgod
Version 1.4 i do believe has the 1.3 binaries as there is no updated official binaries above 1.3 yet.Quote:
Originally Posted by JohnZS
Agentgod has released 1.4 with some requested updates.
That is correct, 1.4 still uses the binaries that came with the latest official (non-beta) Intel MKL.