Originally Posted by McBacker
@ SHAFT Brazil
Indeed a sweet CPU you havecould you share the batch number on that?
@ SHAFT Brazil
Indeed a sweet CPU you have
[SIGPIC][/SIGPIC]
That's why he asked my i7 batch before buying his one... he was after a chip with the same batch like mine, 3845A687
@ SHAFT Brazil
Indeed a sweet CPU you have
Like u Backer?hahaha its just fine man...be cool
Still need help with the P6T little noob?
Thanks NBFIndeed a sweet CPU you have
My batch is 3845A687
...Did i know u?KKKK
Does any one here gets more than 229 bblock on this board?
Thats right!That's why he asked my i7 batch before buying his one... he was after a chip with the same batch like mine, 3845A687
I was impressed when i see your overclock test with this batch
Luck of mine get on of those
Can someone please tell me exactly how you did to flash the 0006 and I'll try to start from the beginning.
How come no one else but me has had the "Warning - Out of memory" issue?
I really want the 0006 bios but I'm about to give up, been spending way to many hours on this now.
Edit: Fixed!
I now have 0006 up and running, I have not noticed any overclocking improvements becides the turbo issue resolved (which is good).
I have a rather annoying issue, I can't use speedstep, C1E or C-State when I'm using the 21x multi and 4+ ghz.
With the v2 bios 0504 (and all previous) it gave me a BSOD when the Windows logo came up, now with the 0006 bios it freezes when the logo comes up.
If I turn of the powersaving feautures it is rock stable.
Anyone had simular problems? Can it be fixed?
Edit 2: Okay, I'm ditching the 0006 bios after all. I get BSOD "system_service_exception" on settings that have been stable before.
The turbo throttling issue did not go away either, it drops to 18-19x once in a while.
Last edited by WeeMaan; 07-25-2009 at 12:38 PM.
Sorry to hear that you ran into so much problems WeeMaan. I just flashed my board to 0006 and it went really smoth. I'll post later on about throtteling results.
| Asus P6T Deluxe V2 | Xeon W3520 | 6GB OCZ Reaper CL8 | Gigabyte GTX460 OC 1GB | Corsair TX750 | 74GB Raptor + 1TB Samsung | WC loop with Supreme LT |
Yeah, it's a shame when I spent several hours trying to make it work.
But atleast I'm one experience richer, now i know.
The multi even drops down to 15 sometimes now, don't now if it's because of the 0006 bios or not.
You think itäs worth trying the 0007 bios or is it about the same?
Might be worth a try. What CPU are you using?
Since I flashed to 0006 i got mem and qpu-ratio settings added to my BIOS. According to funkypunk I will lose these settings again if i flash to 0007 but maybe it will help with the throtteling issue. Anyhow will do some OC first to try 0006 out.
| Asus P6T Deluxe V2 | Xeon W3520 | 6GB OCZ Reaper CL8 | Gigabyte GTX460 OC 1GB | Corsair TX750 | 74GB Raptor + 1TB Samsung | WC loop with Supreme LT |
I'm using a i7 920 D0, I have not noticed any other change in the bios except for the SAS option.
I'll give it a try, if it doesn't work then I go back to 0504.
That has nothing to do with turbo throttling. If your multiplier is dropping at idle and you don't like that then make sure your Control Panel -> Power Options -> Minimum processor state is set to 100%. That might help. On some boards, turning off EIST in the bios doesn't actually turn EIST off. i7 Turbo will report the true state of EIST.WeeMaan: The turbo throttling issue did not go away either, it drops to 18-19x once in a while.
Without a stable system and without some proper testing, I think it's a little early to jump to the conclusion that 0006 doesn't work. It has been proven to work for many happy users here. When you change to this bios version, it's not unusual to need some slightly different settings including more vcore.
Last edited by unclewebb; 07-26-2009 at 09:45 AM.
It does not give me that option in Win7 build 7600 when I have speedstep (eist), C1E and C-State off.
And it is 10h Prime stable now, alltough the multiplier jumps between 15 and 22 once in a while.
And just to clearify, I am not saying 0006 doesn't work! Obviously it does work for many users here, just not for me.
Edit: And it is ofcourse in 100% load that it drops.
The reported multiplier can get a little jumpy on some motherboards when running monitoring software that samples a lot of voltages. HW Monitor and Everest sometimes cause problems.
I know ben805 had to raise his core voltage with bios 0006. When the turbo was throttling and dropping down to 20X at full load, it obviously didn't need as much core voltage to remain stable. With bios 0006, his multiplier was now steady at 21X at full load so extra core voltage was necessary to remain stable.
I think EIST has to be enabled to get the Minimum processor state option available in the Power Options. If you can't boot up with that set then I understand that's not an option for you.
Yes I know som other software can cause misreadings, I read alot in this and the i7 turbo throttling page where you explained this.
So I have only prime, realtemp, cpu-z and turbo v8 or the other turbo software you made on when I am monitoring.
But it looks like it din't need a voltage bump after all, if I reduce QPI / Dram voltage with .025v then it works out OK.
Infact, I am testing now with slightly lower vcore (with the reduce in qpi voltage aswell) and Prime has managed 1½ hour now.
So perhaps the bios wasn't that bad afterall.
But are you saying that the turbo throttling issue can be because I have speedstep off, and because of that can't change the powersaving settings?
I have always assumed that the setting is not there because you can't turn it on when speedstep is disabled.
Lower vcore is a good thing. What MHz are you testing at?
Post a piece of the i7 Turbo log file if there are any multiplier drop outs observed.
Post a screen shot of i7 Turbo V8 at idle. With EIST turned off in the bios and no way to adjust the Minimum processor state, I'm not sure what this is actually set to by default. It might be different from one board to the next. Some boards need EIST enabled for turbo mode to work correctly. Now that your combo is running more stable you might be able to enable EIST in the bios so that you can have some control over your Minimum processor state.
I'm going to take a wild guess and say the turbo v8 isn't really working correctly.
Or is this common?
Edit: This picture is taken only seconds after I started the v8 program, so no throttling has yet occured.
Will take pic of that later together with the log.
Last edited by WeeMaan; 07-26-2009 at 12:35 PM.
I hope it's not common.I'm going to take a wild guess and say the turbo v8 isn't really working correctly.
Or is this common?
Looks like a big bug to me. At least 3 of your 4 cores look correct.
Maybe I can blame this on Windows 7.
On most motherboards at full load, the results are very consistent. PM me a big chunk of your i7 Turbo log file so I can have a closer look. If you copy and paste it into a message just use the code html tags before and after the data enclosed in square brackets [] like this:
[codeX]
paste data in here
[/codeX]
Remove the X in the html code tags. This will make it easier for me to have a look.
You can also upload the log file to somewhere like SendSpace and send me the link.
Edit: I made a quick update to i7 Turbo V8. It should show up as version 6.3.
http://www.fileden.com/files/2008/3/3/1794507/Turbo.zip
Hopefully this tool can show you if your multiplier is steady at 21 or not.
One random value can be ignored but continuous multiplier values below 21 is a good sign of turbo throttling. If there is a problem, I like looking at the log file to get a more accurate picture of what's going on. If you continue to get random values then try turning off CPU-Z and see if that makes a difference. I haven't had any feedback yet on the newest version (1.52) of CPU-Z and how it effects i7 Turbo.
Last edited by unclewebb; 07-26-2009 at 09:11 PM.
I got some data for you.I hope it's not common.
Looks like a big bug to me. At least 3 of your 4 cores look correct.
Maybe I can blame this on Windows 7.
On most motherboards at full load, the results are very consistent. PM me a big chunk of your i7 Turbo log file so I can have a closer look. If you copy and paste it into a message just use the code html tags before and after the data enclosed in square brackets [] like this:
[codeX]
paste data in here
[/codeX]
Remove the X in the html code tags. This will make it easier for me to have a look.
You can also upload the log file to somewhere like SendSpace and send me the link.
Edit: I made a quick update to i7 Turbo V8. It should show up as version 6.3.
http://www.fileden.com/files/2008/3/3/1794507/Turbo.zip
Hopefully this tool can show you if your multiplier is steady at 21 or not.
One random value can be ignored but continuous multiplier values below 21 is a good sign of turbo throttling. If there is a problem, I like looking at the log file to get a more accurate picture of what's going on. If you continue to get random values then try turning off CPU-Z and see if that makes a difference. I haven't had any feedback yet on the newest version (1.52) of CPU-Z and how it effects i7 Turbo.
This log is done with 4.2ghz @ 1.3v cpu, 1.3v qpi. Speedstep ON in bios (it worked some how), minumim state set at 100%.
I'm using Win7 build 7600 (RTM), the only programs that are on is realtemp, prime95 and turbo v8 6.3.
http://www.sendspace.com/file/rvm1nj
Here is a old one with about the same except speedstep OFF in bios, + cpu-z running and Turbo v8 6.2.
http://www.sendspace.com/file/w2gt2r
If you need any more then let me know what settings etc you want me to use.
And by the way, perhaps it would be good to att an option for showing all 4 cores in the log just like in the program itself?
Thanks WeeMaan. I've had a close look at your data from i7 Turbo 6.3 and I think I understand where those occasional glitches are coming from.
First of all, there are absolutely no signs of turbo throttling. Your multiplier is rock steady at 21.000 at full load like it should be.
As much as I hate to admit it, I believe the glitches are being caused by RealTemp 3.00. If you use RealTemp 3.30 instead then I think your i7 Turbo V8 multiplier results will be 100% steady.
http://www.fileden.com/files/2008/3/...alTempBeta.zip
What i7 Turbo is doing is it is comparing two high performance timers within your CPU. When Turbo is engaged on your system, one timer should be running at 4200 MHz and the second reference timer should be running at 4000 MHz. If you divide those two you get a ratio of 1.05. The default 20X multi multiplied by that ratio (1.05) equals your turbo ratio of 21.0.
The problem is that Intel cheaped out a little bit and these timers are only 40 bits wide compared to most timers within a CPU that are 64 bits wide. A 64 bit timer can count continuously for over 100 years but a 40 bit timer running at 4200 MHz will overflow and wrap around about every 4 and a half minutes. I had a look at the frequency of your glitches and they correspond very closely to when these timers overflow.
Since version 3.00 came out 6 months ago, I've changed how RealTemp handles timer overflow so I hope with version 3.30 you can monitor your temperatures and your multiplier and get excellent results with each program.
Thanks for your help with this and your suggestion to add all multiplier data to the log file. I usually look at the standard deviation column. During actual turbo throttling, this number tends to be very low, usually less than 0.005, which shows that all multipliers are reporting pretty much the same thing. All of your glitches showed a very high standard deviation which was a sign that one or more multis were out to lunch, and they were.
Your overclock is looking great now.
Edit: Here's a i7 Turbo V8 log file from rge's i7-950.
http://www.sendspace.com/file/zup4gn
No glitches during 4 hours of testing at full load. As long as no crappy programs like RealTemp 3.00 are screwing up your readings, the Calculated Multi should be very consistent.
Last edited by unclewebb; 07-27-2009 at 06:39 PM.
Hi folks,
I was wondering a few days ago if it would be possible to use something I saw on my previous P45/Q9550 E0 setup to our present i7 systems. Back then people had, and still have, to mess with the GTL references to avoid errors on an analog level of the aging AGTL+ bus of the CPU.
The thread on the workings of the GTL bus ehre on xtremesystems came to the conclusion at some point that it was actually possible to see indirectly how well the system was doing, beyond just the foggy testing when the system got unstable and then put the GTL reference setting in the middle between both unstables.
One indirect indicator of how "well" your setup felt was to see if Intel Burntest/LinX was showing increasing or decreasing fluctuations of the GFLOPs performance. IMHO this can be an indicator for excessive self-correction algorithms at work in the chipset. Also for overshoot and undershoot problems on GTL; but that's not an issue for us i7 users. Hail QPI bus!
To make the story a bit shorter.
I ran a series of tests on Windows Server x64 here on my i7 that showed that the IOH voltage is having a visible, but small, influence on the performance. Ideal seems to be 1.26 or 1.28 volts as it was to be expected. 1.22V led to a significant drop from 45.8 GFlops average to 40 GFlops. The QPI voltage was fine at 1.30 V and higher.
What was most interesting was seeing that my C0 920 was losinga quite big amount of performance if the voltage was just a bit off. 1.25V were dropping to 40 GFlops, 1.2625V and 1.275 were fine at 46.6 GFlops, and 1.30V gave the same weird dropping effect to 40 GFlops.
I was running LinX 1.5.8 using 2048 MB RAM. P6T Del. V2, xflashed to BIOS 0007. Win2003 Server x64. i7 920 C0 Batch 3839A499, BCLK200x19.
It might be possible that those here who try to run their CPUs at as low as possible VCore are losing performance as I do (assuming it's not just a kink of IBT/LinX) while being at the verge of instability without knowing.
Any comments on this?
Last edited by Amurtigress; 07-27-2009 at 06:06 PM.
Amurtigress
imho use newer versions of LinX
0.6.1 current now and it use linpack ver.10.2
there was a problems with i7 (HT also) with earlier versions.
i don't see any perfomance fluctuations - if system is stable to run and pass linpack and the temperatures below 100C.
Which voltage I should add is USB is flaky (just a little bit) and I want the USB to be a bit more stable.
Hi,
thanks for the hint at the update. 0.6.1 gave me some insights, however, the performance dropping effect still persists. Here's a table:
Linpack 0.5.8
vcore
1.250 perf drop
1.275 stable
1.300 perf drop
Linpack 0.6.1
vcore
1.250 cold reboot/LinX error
1.275 perf drop
1.300 stable
(perf drop: GFLops result drops from 45.7 to about 40 for several or all passes)
(stable: GFlops stay around 45.7-46.3 GFlops at 19x200)
This looks to me as if the older linpack code did not work well on i7s, per chance not being able to put the i7 chips to their limits as needed. Apparently i have had a latent instability that never caused me a problem, since my setup only gets under seriously high load when benchmarking.
I still think tho this 15% drop in the floating point performance might be a good thing to keep an eye on when testing stability/looking for sufficient voltages for certain clock rates.
Last edited by Amurtigress; 07-28-2009 at 10:53 AM.
Interesting read there.
Asus Rampage IV X79 Extreme 1305
Intel Core i7-3930K Processor @ 4.8GHZ 1.45V
G.Skill Ripjaw 1866MHZ 4X4GB 9-10-9-28
EVGA SC GTX 580 SLI 880/2200
Corsair H100
Coolermaster HAF932
Corsair 1000W HX Power Supply
im trying to get 920 stable at 4.0ghz (200 x 20)
i can get into windows at 1.18v, but even at 1.234v
its not stable (reboots after 3 hours of prime 95)
i left everything auto except for qpi (1.35), c1e and intel speed something.
should i try to mess around with other voltages?
i thought id get it stable upping only cpu voltage, then mess with
others to get the lowest cpu voltage possible.
or should i try 191 x 21?
if i go with this, i can only select 1533?mhz for my memory.
is this ok? i already ran memtest for several hours at 1600mhz
any suggestion would be helpful.
thx.
I would try 191X21. In some cases the 20X multiplier can be alot more fickle then 19 or 21. (Someone correct me if I am wrong on this.) Perhaps set the DRAM voltage to 1.66 and CPU/PLL to 1.88 and keep the QPI @ 1.35. I am still learning as well, so if someone has better advice please chime in.
Last edited by hetsaq; 08-01-2009 at 08:55 AM.
Asus Rampage IV X79 Extreme 1305
Intel Core i7-3930K Processor @ 4.8GHZ 1.45V
G.Skill Ripjaw 1866MHZ 4X4GB 9-10-9-28
EVGA SC GTX 580 SLI 880/2200
Corsair H100
Coolermaster HAF932
Corsair 1000W HX Power Supply
Reply With Quote
Bookmarks