Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

After the sensor test:

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

and sensor test results after changing Tj max to 85:

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

and for good measure this was my latest attempt at TIM application:



I'll probably re-apply after the exams and show you how that one worked out!

I'm off to bed, last big push of studying tomorrow.

My continued thanks for all your help and support!

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

The TIM on your cpu doen't look like a grain of rice, it looks like....uhh....umm....err....oh never mind.
BTW, the blob of TIM is somewhat off-center and the resulting TIM pattern will be slightly off-center.

RealTemp displayed data:
1st row - cpu model -- current cpu speed -- cpu FSB x cpu multiplier <--- small text
2nd row -- CPUID -- current cpu load % -- elapsed time since RealTemp started or was reset <--- small text

4th row -- reported current core temps for core0, core1, core2 and core3 <-- large text
6th row -- reported distance (how many degrees) each core is from TjMax <-- large text

8th row -- the minimum temperature for each core <-- small text
9th row -- the individual wall clock time that each of the minimum core temps occurred <-- small text

11th row -- the highest temperature for each core <-- small text
12th row -- the individual wall clock time that each of the highest core temps occurred <-- small text

Thermal status = OK when none of your cpu cores barfed.


Homework assignment:

Step #1:
Add one core's current temperature to it's corressponding Distance to TjMax.
Do this four separate times, once for core0, core1, core2 and core3.

Step #2:
Repeat step #1 for your RealTemp window display after you set TjMax = 85.
Do you feel bold enough to make a general observation?

Step #3
Do you notice any similarities between your Sensor Movement Tests with TjMax = 100 and TjMax = 85? Explain!!


You're free to defer the homework assignment until after you've completed your exams.


You have repeatedly said "thank you" in many of your posts and I appreciate it.
Thank you's are the exception in the forum, not the rule.


TjMax = 100



TjMax = 85

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

1 exam down, one to go! Gonna' have a go at that homework and then do a bit more studying for tomorrow :)

Step #1:
Add one core's current temperature to it's corressponding Distance to TjMax.
Do this four separate times, once for core0, core1, core2 and core3.

So... They all add up to TjMax (100)

Step #2:
Repeat step #1 for your RealTemp window display after you set TjMax = 85.
Do you feel bold enough to make a general observation?

These also add up to TjMax (85), general observation: because they're measured from TjMax, the temperatures aren't actually any cooler

Step #3
Do you notice any similarities between your Sensor Movement Tests with TjMax = 100 and TjMax = 85? Explain!!

The distances to TjMax for both 85 and 100 are the same. I would (try to) explain this by acknowledging that I understand that the temperature is measured as a distance from TjMax and that during the test my CPU ran between 56 and 62.5 degrees away from its TjMax regardless of whether it's 85 or 100.

I've read that you should aim to stay 30-40 degrees from TjMax to ensure stability and longevity and therefore even though I don't know what my correct temperatures are, maybe I should change to 'distance from TjMax' setting when overclocking?

I don't know if this is the answer you were looking for and I'm really hoping that the X5460 TjMax is 85 and that my cooler isn't so much worse than those that you have found data on but I will work with what I have and be content with getting the best result I can.

Where do we take it from here profJim? :)

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

You've done goodly and you understand the relationships in RealTemp.

Next step? Set your cpu to it's stock speed: 9.5 x 333 = 3.16GHz and enable the bios power saving options EIST and C1E.
Your windows power plan should use the Balanced option.

Boot your system and let it idle for 30 minutes.
Note the single current CPU temperature reading in HWiNFO which should be close to the ambient temperature.
Your four individual cpu core temperatures should be in the low 20^o(C) range in RealTemp and in the 35^o(C) range in Core Temp and/or HWiNFO.

Next you should run a 30 minute Prime95 test and note the maximum cpu core temperatures.
I recommend running the default Blend test for now and for each test as you overclock higher.
We want to be able watch your temperatures as you overclock higher.
If the RealTemp core temperatures are reasonable, you can run LinX or IBT later.
Core Temp, HWiNFO and RealTemp all can log their own sensor readings, but only have one of these programs logging at a time.
My preference would be for HWiNFO to do the logging because it logs almost everything in it's list and we might see some voltage settings or other settings that could be part of the problem.

This test is to help verify as best we can what your real DTS value is.
Assuming that TjMax for your cpu should be 85, you could merely subtract 15^o(C) from the HWiNFO or Core Temp displayed temperatures.
Your single CPU temperature reading should have no relationship to setting TjMax to 85 or 100.

After everything is OK at stock speeds, you can overclock to 3.30GHz or 3.60GHz and repeat these steps.
You can try leaving EIST and C1E disabled as you overclock but this might cause some instability issues.

When I run my tests I leave all my other programs running in the background because at idle my cpu utilization is only about 1% - 2%.

I have no problem with 80^o(C) core temperatures when running LinX or IBT stress tests (with TjMax=100) but I wouldn't be happy with my cpu core temperatures that high when running intensive programs such as video encoding for hours on end. My earlier 8 hour Prime95 Blend test had maximum core temperatures in the 60^o - 65^o(C) range with my earlier, higher cpu core voltage settings. My most recent LinX and IBT temperatures were still below 70^o(C), but these tests only lasted for 10 - 12 minutes each.

It will be interesting to see if your Xeon cpu has excess heat problems if the cpu core temperatures (with TjMax = 85) go higher than about 70^o(C), which would be like my Q9650 running at 85^o(C), where 85 minus 70 is "equal to" 100 minus 85.

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

So I changed the BIOS settings like you said and rebooted only to find it didn't POST. So I reduced the vcore right down to 1.23750V and it loaded Windows fine. I then changed to balanced power settings and am now waiting til 9:30pm to do the blend test. Just thought I'd post my BIOS settings since I haven't done so in a while:

Motherboard: Gigabyte GA-P35-DS3L Rev 2.0: BIOS Version: F9C
CPU: Intel Xeon X5460
RAM: Crucial Model: Ballistix Sport DDR2 800 Part#: BLS2CP2G2D80EBS1S00CEU (http://uk.crucial.com/ProductDisplay?urlRequestType=Base&catalogId=10153&categoryId=17510&productId=10743&urlLangId=-1&langId=-1&top_category=&parent_category_rn=17510&storeId=10153)
Power Supply Unit: Corsair HX520W
Robust Graphics Booster___________ [Auto]
CPU Clock Ratio __________________ [9]
Fine Clock Ratio __________________[0.5]
CPU Host Clock Control____________[Enabled]
CPU Host Frequency (MHz)_________[333]
PCI Express Frequency (Mhz)________[101]
C.I.A. 2__________________________[Disabled]
Performance Enhance = ____________[Standard]
System Memory Multiplier (SPD)_____[2.00]
DRAM Timing Selectable_____ SPD ___[Manual]
 
CAS Latency Time_______________5_[5]
Dram RAS# to CAS# Delay________5_[5]
Dram RAS# Precharge Delay_______5_[5]
Precharge Delay (tRAS)__________12_[12]
 
ACT to ACT Delay (tRRD)_________3_[Auto]
Rank Write to READ Delay________3_[Auto]
Write to Precharge Delay_________6_[Auto]
Refresh to ACT Delay____________52_[Auto]
Read to Precharge Delay__________3_[Auto]
Static tRead Value_______________8_[8]
Static tRead Phase Adjust_________0_[Auto]
 
System Voltage Control____ [Manual]
DDR2 OverVoltage Control__ [1.8V]
PCI-E OverVoltage Control_ [+0.1V]
FSB OverVoltage Control___ [+0.1V]
(G)MCH OverVoltage Control [+0.1V]
CPU Voltage Control_______ [1.23750V]
Normal CPU Vcore________1.23750V

Hard Disk Boot Priority.....: [Press Enter]
1.	Kingston SV300S37A240G
2.	ST3500320AS
3.	ST3500320AS
First Boot Device...........: [Hard Disk]
Second Boot Device......: [CDROM]
Third Boot Device.........: [Disabled]
Password Check...........: [Setup]
HDD S.M.A.R.T. Capability......: [Disabled]
Limit CPUID Max. to 3.........: [Disabled]
No-Execute Memory Protect.........: [Enabled]
CPU Enhanced Halt (C1E)...........: [Enabled]
CPU Thermal Monitor 2(TM2) ......: [Enabled]
CPU EIST Function...............: [Enabled]
Virtualization Technology..........: [Disabled]
Full Screen LOGO Show.........: [Disabled]
Init Display First............: [PEG]

SATA AHCI Mode.........: [AHCI]
SATA Port0-1 Native Mode.........: [Enabled]
USB Controller...........: [Enabled]
USB 2.0 Controller............: [Enabled]
USB Keyboard Support............: [Disabled]
USB Mouse Support..........: [Disabled]
Legacy USB Storage Detect��.[Disabled]
Azalia Codec............: [Auto]
Onboard H/W LAN............: [Enabled]
SMART LAN.................: [Press Enter]
Onboard LAN Boot ROM.......: [Disabled]
Onboard IDE Controller........: [Enabled]
Onboard Serial Port 1..........: [3F8/IRQ4]
Onboard Parallel Serial Port...: [378/IRQ7]
Parallel Port Mode................: [SPP]


ACP1 Suspend Type........: [S1(POS)]
Soft-Off by PWR-BTTN....: [Delay 4 Sec.]
PME Event Wake Up.......: [Disabled]
Power On by Ring.......: [Disabled]
Resume by Alarm......: [Disabled]
HPET Support..........: [Enabled]
HPET Mode.............: [64-bit mode]
Power On by Mouse.............: [Double Click]
Power On by Keyboard.........: [Disabled]
AC Back Function..................: [Soft-Off]


PCI1 IRQ Assignment.............: [Auto]
PCI2 IRQ Assignment.............: [Auto]
PCI3  IRQ Assignment............: [Auto]

Reset Case Open Status...........: [Disabled]
CPU Warning Temperature.......: [80�C/176�F]
CPU FAN Fail Warning...............: [Disabled]
SYSTEM FAN2 Fail Warning.......: [Disabled]
POWER FAN Fail Warning..........: [Disabled]
SYSTEM FAN1 Fail Warning.......: [Disabled]
CPU Smart FAN Control.............: [Disabled]

Do you think the vcore being high before might have caused some overheating?

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

So here is the idle screenshot after 30 mins (using the above BIOS settings):

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

and here is a screenshot at 30 minutes of the blend test:

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

... and the sensor log for good measure:

:)

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

OK, missed communication, I was giving you credit for (mostly) knowing what you're doing as you started the thread at 3.60GHz, even if it wasn't fully stable.

Any voltages that are too high can cause boot problems, instability, and excess heat. Most of the the time there is a range for most voltages that will be tolerated and possibly be reasonably stable. It's very time consuming to optimize all settings at each step as you are overclocking. It's up to you if you want optimize all voltages at each step so that each setting is at or near the lowest stable setting. I save multiple bios profiles along the way as I'm climbing the o/c ladder. In some cases I'll save an unstable profile that I think is close to being stable.

Your cpu and cpu vcore temperatures look good at 3.16GHz during the Prime95 blend test and are similar to my 3.00GHz Prime95 Blend maximum temperatures. When you adjust for the fact that Core Temp and HWiNFO temperatures are ~ 15^o(C) higher than RealTemp's values, everything is in order.

Suggestion, posted screen-shots with temperature displays should include comments that describe the TjMax values used by each of the temperature monitoring programs. Otherwise people who find the screen-shots while doing image searches will not understand the apparent mismatched temperatures.

Next:
Run LinX (or IBT) using as much available free memory as possible, probably 1536MB or 1792MB.
Running 3 or 5 loops is OK, and save a screen-shot of the completed test which will be your baseline benchmark score. Be sure to include the temperatures at the end of the test.
Remember my best LinX score is 55+GFlops at 4.0GHz.
As I recall my 3.0GHz LinX score was ~ 39 GFlops, and IBT usually gives a slightly lower score.
Be sure to save the HWiNFO log for this test.

The next step: a 9% overclock

• cpu = 8.5 x 401MHz = 3.408Ghz
• ram = 2.0 x 401MHz = 802MHz

Cpu core voltage will probably require a modest increase, try 1.26875
The remaining four voltages should be fine with their current +0.10 volt settings.

There is an alternative to running several long Prime95 stress tests.
The program is called HCI Memtest and it runs in windows.

• download link: HCI MemTest: RAM reliability tester for Windows
• online manual: HCI MemTest Manual
• HCI Memtest FAQ: MemTest Support @ HCI Design

I run it with four simultaneous windows.
If I have 1800MB of free memory I would allocate 400MB of memory to each test window and let them run until the overall coverage for each window was at least 300% - 500%, longer is better. If there are any reported errors, the test has failed and should be stopped, just like you would if one of the Prime95 worker threads got an error.

I recently ran each of the three Prime95 tests, one at a time, with each test running for at least 8 hours.
I'm now going to run HCI Memtest overnight.
I've used HCI Memtest several times before and it completed with no errors on systems that had previously completed full Prime95 testing.

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

One last question before I run HCI Memtest at 3.408GHz - should I try without changing the voltage and increase it by a single increment if it fails or just run it straight at 1.26875V?

I will try to adjust all the thermal monitoring programs to Tjmax=85 and post the results.

P.S. Sorry if I mislead you with my 3.6 overclock - still feel like an absolute newbie but I'm learning! :)

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

Cpu voltage at 3.408GHz
My 1.26875 recommendation is 5 clicks up from your cpu's "normal" voltage.
Your current 9% overclock might run with the normal cpu voltage and it might not.
My preference is to go a bit higher than I think is necessary.
Even if we had identical hardware, each system will require somewhat different settings.

Adjusting TjMax in monitoring programs
Core Temp is very easy to adjust TjMax, but it's window still displayed TjMax = 100, even though the core temps were adjusted.
I didn't find a TjMax setting in HWiNFO, but I didn't look in the main system information panel.

HCI Memtest
I'm running 4 Memtest windows where each task is working with 1200MB out of 5400MB that was originally available.
My cpu temp is 36^o(C) and my average core temps are in the mid 40^o(C) range.
Based on the displayed memory coverage, it will take about one hour to cover 100% of the memory and letting it run for 10 - 12 hours should be an excellent test.

Check out Minimum Guesses To Find A Number Between 1 and 1000 and understand how the binary search approach works in guessing a number between 1 and 1000 in 10 steps or less. Once you understand how and why this works you can then use a similar approach find the optimal cpu voltage setting.

Each click up or down is 1/160th of a volt or 0.00625v or 625 one hundred thousandths of a volt..
For example, to go from 1.25000 to 1.30000 is a change of 0,05 volts or 8 clicks of memory increases.
I'll let you determine how many clicks it is for memory voltage increases if your cpu eventually needs up to 1.3750 or 1.40000 volts.

I'm not suggesting that you use the middle of this very large voltage range at any time when you try to determine the voltage, but the splitting the difference approach when using a much smaller voltage range will let you find a better voltage setting much quicker than going up or down one click at a time.

_ Have I thoroughly confused you yet?

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

I think I understand!

I guess I'm just worried about the heat production of the extra voltage. I'll give it a go without any increase and if it fails I will try at 1.26875V as you you suggest. I know that it's not completely necessary to want to find your lowest viable voltage (and memory timings) until you've reached your maximum stable overclock, I'm just unsure about when to reduce the vcore if you think it is the cause of the instability (vcore [and therefore temperature] being too high rather than vcore being too low).

I'm sure you'll be able to shed some light on this for me!

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

This is the main reason to keep logs while stress testing, as we will have a record of temperatures, voltages, cpu utilization that might help pinpoint what might have caused the problem. Windows Event Viewer might also have some useful information when we know the time and date when a problem occurs.

If another BSOD occurs, we can use WhoCrashed or Blue Screen Viewer to quickly and easily get the major information from saved mini-dumps in your system.

Both programs are freeware.