XS INTEL CORE i7 STABLE OVERCLOCKING DATABASE

[ReverendMaynard]

were do you see a tjmax of 95c??? idk... the cpus tjmax is 95c... and the reason why the temps arnt as high as they should be on the max side is because I restarted realtemp and it reset itself. Also Realtemp shows the turbo speed idk why?? I have an R2E so I can shut Turbo off.

If you guys wont accept this I can run another 41 hours I will gladly do this again and leave it running this time.

Easy champ, never accused you of not being stable. i7 tjmax is 100C, so add 5C to your temps....that is all lol.

[MarcusFoX]

woh champ what?? just wanna know if this will be accepted so I can restart on making another, the way it was looking before my oC was a fraud because I stopped prime and restarted Realtemp??

idk put urself in my shoes

[rge]

were do you see a tjmax of 95c??? idk... the cpus tjmax is 95c... and the reason why the temps arnt as high as they should be on the max side is because I restarted realtemp and it reset itself. Also Realtemp shows the turbo speed idk why?? I have an R2E so I can shut Turbo off.

If you guys wont accept this I can run another 41 hours I will gladly do this again and leave it running this time.

No one was saying they dont accept it, but if you download a recent version of Realtemp, ie 3.3, it will probably read specs correctly including Tjmax, tjmax 95 is from adding distance to tjmax + core temps.

And to set off prochot, you must be at high vcore/mhz...so the run is believable...it is just nice to have clean data so we can see what different chips can do, different cooling, etc, some dont make sense, yours does it is just messy from restart of Realtemp and using old software But I would not rerun anything for that.

[unclewebb]

MarcusFoX: RealTemp 2.70 was released before Core i7 even existed so it does not properly support Core i7 CPUs.

RealTemp 3.30 works correctly.
http://www.fileden.com/files/2008/3/...alTempBeta.zip

[cstkl1]

MarcusFoX: RealTemp 2.70 was released before Core i7 even existed so it does not properly support Core i7 CPUs.

RealTemp 3.30 works correctly.
http://www.fileden.com/files/2008/3/...alTempBeta.zip

well something not right
all his fft iterations are different for each thread. 8 10 16???
if this was blend its always large fft, two small fft, followed by one large fft.
his shots shows threads faster /slower by an hour/two hours.

[MarcusFoX]

MarcusFoX: RealTemp 2.70 was released before Core i7 even existed so it does not properly support Core i7 CPUs.

RealTemp 3.30 works correctly.
http://www.fileden.com/files/2008/3/...alTempBeta.zip

Thanks ill download that I didnt even now that was out. gunna do the test again just so its clear this time.

well something not right
all his fft iterations are different for each thread. 8 10 16???
if this was blend its always large fft, two small fft, followed by one large fft.
his shots shows threads faster /slower by an hour/two hours.

its not because i was doing a custom test with FFT's in place

[unclewebb]

If you do a long Prime run and you are using your computer at the same time for other things, it is possible for the amount of work that each thread does to get a little out of sync but this is much more than normal.

RealTemp 2.70 does not handle hyper threading correctly on Core i7 so it shows the temperatures for the first 4 threads which belong to core 0 and core 1. It does not show the temperature for all 4 cores. The PROCHOT# flag within the CPU has been set for the first two threads which belong to core 0 so the screen shot shows that this core was throttling or reaching 100C, likely during the test. When this happens, the multiplier cycles down to 12. You can't tell from the screen shot whether this CPU throttled once or twice or was throttling continuously for hours but it definitely did some throttling. With early versions of RealTemp there was no way to reset the thermal throttling flag unless you knew how to edit some registers within the CPU or re-booted. Restarting RealTemp would not clear the flag that is set within the CPU. You should be able to clear this flag with newer versions of RealTemp by clicking on the Reset button. I can't remember what version I added that feature.

RealTemp reports 4195.23 MHz. It didn't know about turbo mode back then so it was using a 20X multiplier. 4195.23 / 20 = 209.76 BCLK which agrees with CPU-Z. If throttling was continuous then it would be only 209.76 x 12 = 2517 MHz. With turbo enabled and functioning, it would be 209.76 x 21 = 4404.96 MHz which agrees with CPU-Z. Hopefully some new screen shots at full load will clear this all up.

[sinister1st]

Nice work everyone, just a reminder:



Prime should be running in your screenshots because otherwise it's rather simple to cheat i.e. just stop Prime then set whatever clocks you want before taking a screenshot

Cheers

UHHHH DUHHHHH for me ok I see lol you want us to take the shot while prime is still going but then you cant see the time its ran? I guess from the start date in the first box you can by the date. I dont think I have the pleasure of software to make mine cheat even if i wanted to mine sucks msi core center CHEATERS NEVER PROSPER! you guys are to much. Well his a nice long run I did before I turned on turbo but its still the same baseclock. 100 passes 20hrs

[Daddyjaxx]

CPU: i7 920
USERNAME: Daddyjaxx
CPU MHZ: 4.1 GHz
HT: ON
CPU BATCH: 3849B202
CPU S-SPECK: SLBEJ
QPI/VTT 1.2875
CPU VOLTS: 1.2875
CPU COOLING: Water
BRAND AND MODEL OF COOLER: Apogee GTZ
MOBO BRAND AND MODEL: ASUS P6T V1
BIOS VERSION: 0603
OS: Vista Ultimate
RAM BRAND: OCZ
RAM MODEL: Gold PC 12800
RAM MHZ: 1640
RAM TIMMINGS: 8-8-8-24

[MarcusFoX]

Yah ok that sounds about right.... because i now when the cpu gets to its tjmax it will throttle itself that could be why.

we'll see on my next pass, im calibrating realtemp right now getting ready to run the test for another 40hours

If you do a long Prime run and you are using your computer at the same time for other things, it is possible for the amount of work that each thread does to get a little out of sync but this is much more than normal.

RealTemp 2.70 does not handle hyper threading correctly on Core i7 so it shows the temperatures for the first 4 threads which belong to core 0 and core 1. It does not show the temperature for all 4 cores. The PROCHOT# flag within the CPU has been set for the first two threads which belong to core 0 so the screen shot shows that this core was throttling or reaching 100C, likely during the test. When this happens, the multiplier cycles down to 12. You can't tell from the screen shot whether this CPU throttled once or twice or was throttling continuously for hours but it definitely did some throttling. With early versions of RealTemp there was no way to reset the thermal throttling flag unless you knew how to edit some registers within the CPU or re-booted. Restarting RealTemp would not clear the flag that is set within the CPU. You should be able to clear this flag with newer versions of RealTemp by clicking on the Reset button. I can't remember what version I added that feature.

RealTemp reports 4195.23 MHz. It didn't know about turbo mode back then so it was using a 20X multiplier. 4195.23 / 20 = 209.76 BCLK which agrees with CPU-Z. If throttling was continuous then it would be only 209.76 x 12 = 2517 MHz. With turbo enabled and functioning, it would be 209.76 x 21 = 4404.96 MHz which agrees with CPU-Z. Hopefully some new screen shots at full load will clear this all up.

[unclewebb]

MarcusFoX: I found this article at Anandtech.

http://www.anandtech.com/cpuchipsets...spx?i=3184&p=4

When pushing a QX9650 hard, they also started seeing some significant differences in the amount of work done for each thread when running Prime95 even during a short run. A 40+ hour run could lead to some large differences in Prime like you were seeing if some of your cores were throttling. This could also be a Prime bug or more likely a Windows bug and how the work load is shared between threads. The picture just above your post above isn't close to any thermal throttling but it also shows some large differences in work accomplished for each thread.

[MarcusFoX]

ill read this article, the cpu is running prime right now im about 5 and a half hours in and I dont see a big difference between the work done on most of the cores except core 6-3 which are just alittle bit behind by 1 test too the other cores.

heres a new screen I was able to bump my voltage down abit and up the core speed alittle, I think I have it were I want it now.

MarcusFoX: I found this article at Anandtech.

http://www.anandtech.com/cpuchipsets...spx?i=3184&p=4

When pushing a QX9650 hard, they also started seeing some significant differences in the amount of work done for each thread when running Prime95 even during a short run. A 40+ hour run could lead to some large differences in Prime like you were seeing if some of your cores were throttling. This could also be a Prime bug or more likely a Windows bug and how the work load is shared between threads. The picture just above your post above isn't close to any thermal throttling but it also shows some large differences in work accomplished for each thread.

[rge]

MarcusFox, looks like you will be stable there, but curious with high temps, are you on just one rad/one loop for your i7 and 4870x2, and what are ambients, rad, and fan speed. And prime longer you run it, more cores will be faster/slower than others, always happens with me.

[unclewebb]

Nice overclock MarcusFox.
No one will doubt your numbers now.

98C is getting very close to the thermal throttling point but RealTemp is showing that hasn't happened yet. It continues to report OK for all 4 cores in the Thermal Status area.

[MarcusFoX]

MarcusFox, looks like you will be stable there, but curious with high temps, are you on just one rad/one loop for your i7 and 4870x2, and what are ambients, rad, and fan speed. And prime longer you run it, more cores will be faster/slower than others, always happens with me.

To answer that question: the ambient is 26c my computer is downstairs but it right by a water heater and its 160F all the time so it raises ambient around the system 4-5c easy. 4870x2 isnt in the loop and maybe has alittle to do with it but I doubt that since i have a TJ07 and the air flow in this puppy is amazing.its hotter at the back of the system than it is at the cpu were you think the heat would rise to but it gets exhausted by the 2 upper 120mm pritty quickly. The temps are like that because of the ambient.


hardware is Thermochill PA120.3 Rad with 3x Scythe S-Flex fans 1/2 loop D-tek V2 Block swift mini res, about 1.34L of water, mixture 10%pc-ice 2%fesner 2% clorine 76% Distilled water.

Nice overclock MarcusFox.
No one will doubt your numbers now.

98C is getting very close to the thermal throttling point but RealTemp is showing that hasn't happened yet. It continues to report OK for all 4 cores in the Thermal Status area.

Yup thank uncle, thanks for the heads up on realtemp I probly wouldnt have found out about for another 2weeks if you wouldnt have told me.

heres the 8hr shot, this should be enough so now its 48hr stable.

[justin.kerr]

Did some HT off testing, did not gain much speed, max bootable B-clock is 221 for this setup. Can get 4.7Ghz stable, but after booting have to use E-leet to bump B-clock to 225, and that doesn't seem like a 24/7 setup. lol

Temps are great when HT is off though.

Is HWmonitor messed up? or is RealTemp off? or is it just the different ways they calc the temps?




USERNAME: Justin.kerr
CPU MHZ: 4.630ghz
HT: OFF
CPU BATCH: 3845B027
CPU S-SPEC: SLBEJ
QPI/VTT 1.35 IN BIOS1.348 with DMM
CPU VOLTS: 1.3875 in bios 1.387 with DMM
CPU COOLING: WATER
BRAND AND MODEL OF COOLER:Heat killer, ddc3.2 + Xspc restop, Feser 360, 2 Swiftech Mcr320's, Xspc rx360,
MOBO BRAND AND MODEL: Evga Classified 760 B3
BIOS VERSION: S610
OS: Vista 64
RAM BRAND: MUSHKIN
RAM MODEL: REDLINE Ascents
RAM MHZ: 1764Mhz
RAM TIMMINGS: 7-7-6-19

[MarcusFoX]

Nice oc man.... If i had to cooling I mite be able to match that. on second thought maybe I can we HT off

[unclewebb]

justin.kerr: It looks like RealTemp is using TJMax=99C for your CPU. Did you make any changes in the RealTemp settings window that would have changed this?

Some of the ES processors I've seen have TJMax set to 92C or 93C by Intel but all of the retail ones I've seen so far have been TJMax=100C. It looks like HW Monitor might be using TJMax=95C.

TJMax is contained within a register within the CPU. I'll show you later how to read this register directly so you can see what program is telling you the truth. Try Core Temp 0.99.5. I'm pretty sure that it is reading this register correctly and it will directly display TJMax on its main window.

[rge]

Realtemp and Coretemp read same and accurate. Everest only reads cores 0 and 1, then duplicates readings for core 2 and 3 (thread/core confusion by Everest). Hardware monitor is out to lunch.

Justin, nice OC and temps, what are your ambients and can you run everest (computer/sensor) and see watts at load.

I am testing mine on different settings/different fans, ie new gentle typhoons 1850 versus scythe 1600, and trying to get an idea of watts vs temps, etc. Your 3 rads will improve your temps 4C over mine by bringing the 6.7C delta down to ~2C...but curious if you have very low ambients or very low watts cpu or both.

Sorry about off topic post...but will give an idea of temps to expect +/- cpu and setup variability. I cut hole back of my desk exact size of computer to expel hot air to back and prevent hot air from becoming intake air, which decreases my 13 hr prime load temps by few C by preventing local ambients from increasing...which then means at about 20 mins you are near max temps in prime...and any rise after 20 mins is from failure to prevent local ambients from rising...like my 1C rise.

This is in deg C with 4.4ghz, 1.34 vcore, 1.34 qpi/vtt, and loads at 170-185W, max core temps 76C. At 4.6 and 1.4v, temps are much higher, will run that later, this is with PA 120.3, 6 fans scythe 1550 rpms (Scythe s flex F 1600 maxed on controller, always little lower on controller). air intake measure directly over all 3 fans, as air out, then avg.

The 41C delta from water temp to Core temp is only affected by block/tim1,2/cpu diff etc, not by rad which just controls air/water delta.

[justin.kerr]

justin.kerr: It looks like RealTemp is using TJMax=99C for your CPU. Did you make any changes in the RealTemp settings window that would have changed this?

Some of the ES processors I've seen have TJMax set to 92C or 93C by Intel but all of the retail ones I've seen so far have been TJMax=100C. It looks like HW Monitor might be using TJMax=95C.

TJMax is contained within a register within the CPU. I'll show you later how to read this register directly so you can see what program is telling you the truth. Try Core Temp 0.99.5. I'm pretty sure that it is reading this register correctly and it will directly display TJMax on its main window.

I have not made any changes, the version is RC 10? I think..
I do swap out different chips a few times a day though. lol Xeon 3540 was in it just before this run?

This is after a 3dmark 05 run,

[zalbard]

Can get 4.7Ghz stable, but after booting have to use E-leet to bump B-clock to 225, and that doesn't seem like a 24/7 setup. lol

Did you try bumping PCI-E frequency? Should let you boot up with higher bclk.

[justin.kerr]

tried up to 110 on Pcie, no go.

[TheCarLessDriven]

Im about to try an OC on my i7 920 I was just wondering. How many of you just simply go for w/e frequency your trying for right away? And how many work up to it? Is it ok either way?

For example, right now im at stock with turbo mode, so FSB is 133. Is it ok to jump straight to 200 FSB for example?

[TheCarLessDriven]

............

[unclewebb]

justin.kerr: RealTemp is using TJMax = 99C for your processor and so is Core Temp. To figure out the core temperature, software needs to know TJMax.

Reported Temperature = TJMax - Digital Thermal Sensor Reading

With Core i7, Intel writes TJMax information into each CPU so as long as software reads that information correctly, which RealTemp and Core Temp are able to do, there shouldn't be any issues. I'm not sure what HW Monitor is doing but it doesn't look like it's following the Intel documentation when it comes to reading TJMax which is screwing up the reported temperatures.

rge brings up another issue about Everest. For the longest time, it has only been reading two cores of a Core i7 and then copying that information and presenting it as the core temperature for all 4 cores. I don't know if that issue has been fixed yet.

When lightly loaded, a CPU can change core temperature instantaneously but when loaded running Prime95, Core Temp and RealTemp should be virtually the same.

RealTemp used to do that too when Core i7 first came out but rge's infinite testing helped me make sure that RealTemp works correctly on Core i7. Core Temp also works correctly though in certain situations it might not order the temperature data in the correct physical order but at least it is able to read all 4 cores and it uses the correct TJMax.

AC3421: Once you know what settings work for you at BCLK 200 MHz then going from 133 to 200 isn't a big deal but if you don't know what settings work then you might have to go up in more gradual steps.