Overclocking Air: Intel i7 920, Asus P6T, G.Skill DDR3-1600 C8, Xigmatek HDT-1284E

JakFrost

JakFrost

Limp Gawd
Joined
Dec 2, 2005
Messages
241
I've been reading a lot of overclocking threads and it seems that folks mostly use water based setups and I'm having a hard time finding accurate and reliable info for air overclocking. I'm asking for any info from folks that did overclocking on air or have read other threads where they saw other people do that also.

I've just started trying to overclock my system and I can get to 3.6 GHz (180 x 20) at CPU: 1.25 V at 85C with RealTemp while doing a Prime95 stress test SmallFFT or LargeFFT. I think that I can push this CPU down to 1.75 V or 1.15 V and maintain stability at 3.6 GHz and lower my load temps to ~77C if I spend a little more time.

However, I'm looking to do 3.8 GHz on air with reasonable temperatures ~80C on load so I need some advice.

Processor Information

Processor: Intel Core i7 920 2.66 GHz
Revision: D0 (SLBEJ)
FPO/Batch#: 3901A238

Date of purchase: 2009-06-29
Place of Purchase: Newegg.com

Articles

X-Bit Labs - Intel Core i7-920 Overclocking Guide

Extreme Tech - Overclocking Core i7 920 on the ASUS P6T Deluxe

The Parts

Below are the parts that I picked out. I used all of them before and they all worked together well.

Processor: Intel Core i7 920 Nehalem 2.66GHz 4 x 256KB L2 Cache 8MB L3 Cache LGA 1366 130W Quad-Core Processor - Retail - $279.99 - D0 (SLBEJ) - Batch# 3901A238


Motherboard: ASUS P6T LGA 1366 Intel X58 ATX Intel Motherboard - Retail - $249.99 - 0603 Firmware


Memory: G.SKILL 6GB (3 x 2GB) 240-Pin DDR3 SDRAM DDR3 1600 (PC3 12800) Triple Channel Kit Desktop Memory Model F3-12800CL8TU-6GBPI - Retail - $99.99


Heatsink: XIGMATEK HDT-S1284EE 120mm Rifle CPU Cooler - Retail - $39.99


Bracket: XIGMATEK ACK-I7363 CPU Cooler - Retail - $9.99


FrostyTech - Xigmatek HDT-S1284EE Heatsink Review



[size=-2]
MOB: ASUS P6T, LGA 1366, Intel X58 ICH10R, 3 SLI, 6 DIMM, 6 SATA - FW: 0603, AiTweaker XMP, Turbo Off, HT On, SpeedStep On, C-Step On, (All) Auto
CPU: Intel Core i7 920, 2.66 GHz, 4C, 8 MB, 4.80 GT/s QPI, 45nm, 130 W, D0 (SLBEJ) 3901A238 - 3.60 GHz, BCLK 180, Multi 20, CPU 1.20 V (Manual), QPI 1.21250 (Auto)
FAN: Xigmatek HDT-1284EE 120mm Hydro Bearing Fan 800-1500 RPM 4 Heatpipes - 41 C Idle, 81 C Load
RAM: G.Skill 6gb (3 x 2 GB) DDR3 1600 (PC3 12800) 8-8-8-21 2T 1.60 V - XMP Profile 1, Ratio 2:8, Frequency 720 MHz (1,440 MHz), DRAM 1.60 V

VID: eVGA nVidia GTX 260 896 MB G92 192c 576/999 MHz PCI-e 2.0 16x 2xDVI 1xSVid 1xHDTV - 600/1100 MHz, DVI to LCD, HDMI 60' to 50" Plasma TV.
CAP: AverMedia AVerTV Combo PCI-e M780R ATSC/NTSC Capture MPEG-2 Hardware Encoder - MCE Remote, IR Receiver

CAS: Lian-Li PC-V1200 Plus Mid-ATX Aluminum 4x5.25 6x3.5 2x120mm - Best case to date in my opinion.
POW: OCZ PowerStream 520W SLI ADJ ATX2.0 EPS12, +3.3V 28A +5V 40A +12V 33A

SSD: Intel X25-M 80GB MLC SSD SATA II 3 GB/s - FW: 8820
HDD: Western Digital Caviar SE16 500 GB WD5000AAKS 16 MB SATA-II - (4) in (2) RAID-1 (Mirror) Arrays

ENC: MacAlly G-S350SU External HDD Enclosure 3.5" eSATA USB 2.0 Aluminum, Universal 12V Adapter - (2) external WD GP 1TB for Movies
ENC: i-Rocks 9410 External HDD Enclosure 3.5" eSATA USB 2.0 Aluminum, Univeral 12V Adapter - (2) external WD GP 1TB for Backups
HDD: Western Digital Caviar GP Green 1 TB WD10EACS 16 MB SATA-II - (2) external Movies, (2) external for Backups

CON: JMicron JMB363 PATA and SATA Controller, PCIe 1x, 1 PATA, 1 SATA, NCQ, Hotplug, RAID 0,1,0+1,JBOD
CON: JMicron JMB322 SATA II 3 GB/s Controller, PCIe 1x, 2 SATA, NCQ, Hotplug, RAID 0,1,JBOD
NIC: Realtek 8111C PCIe GigE
SOU: Realtek ALC1200 8 channel, HD Audio, Coax/Opt S/PDIF - Analog for Speakers, Motherboard Digital S/PDIF to nVidia GTX 260 S/PDIF for DVI->HDMI

DVD: NEC ND-3550A 16x DVD+-RW SL/DL 48X CDR PATA - FW: 1.07
BLU: LG GGW-H20L Blu-ray 6x BD-RE 16x DVD SL/DL 48x CDR SATA - FW: YL05

MON: HP LP2475w 24-inch LCD H-IPS Monitor Wide Gamut 102% NTSC 6-12ms Response - 1920x1200x32 60Hz, DVI, GIG122/GIG052, Bri 15, Con 80, 9300K
KBD: Microsoft Natural Keyboard Pro USB/PS2 Media Keys
MOU: Microsoft IntelliMouse Explorer 3.0 Optical USB/PS2 Scroll-Wheel 5-Button
JOY: CH Products Fighterstick USB 3 axes 24 buttons
THR: CH Products Pro Throttle USB 3 axes 24 buttons
PED: CH Products Pro Pedals USB 3 axes

ROU: Linksys BEFSR81 EtherFast Cable/DSL Router 8-port Switch 100MBps FW: 2.51.3

OSS: Microsoft Windows 7 Release Candidate Build 7100
[/size]
 
Last edited:
As an Amazon Associate, HardForum may earn from qualifying purchases.
Preliminary Results

I have my Ai Tweaker setup as XMP to get the memory timing from the XMP profile on the memory chip to 1600 MHz 8-8-8-21 2T. I'm happy with using this profile and I really do not want to overclock my memory since the effort would be a waste of time because the 2% performance increase would be mostly a wash.

For testing the limits of my CPU I set my memory timings at the lowest possible, 1080 MHz to remove the memory from the equation as a cause of crashes. This allowed me to push the BLCK up past 180 without immediate crashes or freezes.

I was able to get to 3.8 GHz today at CPU 1.25 V but when running Prime95 and RealTemp I spiked at 89C after only a dozen or so minutes of running it.

I think that 3.8 GHz on air is possible for benchmark results but probably not prudent to run on a normally used system with my air cooler. Since I am trying to do the "minimalist" effect for overclocking I want to avoid extra cooling fans or other computer hacks so I'm going to have to concentrate at keeping the thermals in my system more in check and not concentrate so much on the speeds.

I'm now going to concentrate now on falling back to 3.6 GHz (180 x 20) at CPU 1.2 V to see if this will work better in terms of keeping the thermals in check at 80-~85C on load.
 
Half-Way There at 3.7 GHz

I was messing around a bit more and reading some of the articles and posts and I just can't believe that people are pushing 1.35 V and higher on this processor, but I think that folks are had the C0 version and most of them are using water cooling.

I'm trying to do some modest overclocking on air and I've gotten to 3.7 GHz (185 x 20) with CPU 1.225 V but I hit a high of 87C on RealTemp while most of the time it hangs on 85C when doing Prime95 LargeFFT to maximize heat generation and load.

I didn't change any other voltage settings on my system except for setting Ai Tweaker to XMP mode and letting it it set the memory parameters and pumping up the memory to only 1.6 V.

I've been running Prime95 for only 3-hours and I'm going to stop it now to play some games and do some 3D testing. If it passes those I'm going to let it go overnight for the full 8-hour burn.



Intel Core i7 920 2.66 GHz D0, Asus P6T 0603, G.Skill 6GB DDR3-1600 C8 2T, Xigmatek HDT-1284EE - 3.7 GHz (185 x 20), CPU 1.225 V, DRAM 1.60 V, QPI 1.2125 V = Unstable after 7-hours of Prime95 Blend

intelcorei7920266ghzd0a.png
 
Last edited:
I have the same cooler as you, and I had issues with the CPU running too hot at load with 3.6+ GHz overclock. After experimenting a bunch with the amount and pattern of applied thermal paste, I finally got my temps down into the mid 70º's under load.

But to get there, I had to attach the HSF literally 7 different times in order to find the right setup. It was a huge pain. I think what I settled on was to spread a thin layer across the CPU with a credit card, then I added a tiny blob to each of the heatpipes on the cooler itself. Saw my temps drop 10º+ and haven't looked back since.

Good luck with the OC!
 
Bigwillis said:
I have the same cooler as you, and I had issues with the CPU running too hot at load with 3.6+ GHz overclock. After experimenting a bunch with the amount and pattern of applied thermal paste, I finally got my temps down into the mid 70º's under load.

But to get there, I had to attach the HSF literally 7 different times in order to find the right setup. It was a huge pain. I think what I settled on was to spread a thin layer across the CPU with a credit card, then I added a tiny blob to each of the heatpipes on the cooler itself. Saw my temps drop 10º+ and haven't looked back since.

Good luck with the OC!
Click to expand...

I'm in the same boat as you were right now since I can't push past 3.6 GHz without hitting 90C temps at 1.25++ V voltages.

I read that another use had similar problems and it came down to him applying too much thermal paste on this Noctura cooler. I believe that I have also done the same thing and used too much thermal paste since when I saw the gaps in the heat pipes I thought "well let's add extra" to let it get in between the pipes. I think that it was much too much.

Will remove the heatsink and scrape the goop off.
 
I'm pretty sure you can't overclock air, you're welcome to try though ;)

Have to say though, I'm digging the NB clock (~3GHz).
 
visionviper said:
I'm pretty sure you can't overclock air, you're welcome to try though ;)

Have to say though, I'm digging the NB clock (~3GHz).
Click to expand...

That test was a failure and I think that while it might not have been the CPU it might be the memory. I'm doing testing now at 3.6 GHz (180 X 20) at CPU 1.25 V at with the lowest memory ratio 2:6 at 540 MHz (1080 MHz DDR3) and NB 2880 MHz.

Let's see if that passes Prime95 blend overnight after 8-hours and let's see how high the temps climb. It is 80F (27C) ambient here and I already hit 83C in a few minutes. I think that overnight the temps will climb to ~90C.

I wonder if it will clock out stable speeds though at CPU 1.25V at only 3.6 GHz with the lowered memory timings. If it does then I'm going to re-run the same test with memory at the next higher ratio at ~740 MHz (1480 MHZ DDR3).

Intel Core i7 920 2.66 GHz D0, Asus P6T 0603, G.Skill 6GB DDR3-1600 C8 2T, Xigmatek HDT-1284EE - 3.6 GHz (180 x 20), CPU 1.25 V, DRAM 1.64 V, QPI 1.2125 V - Prime95 BlendFFT Stable - 11-Hours+

intelcorei7920266ghzd0a.png
 
Last edited:
Bigwillis said:
I have the same cooler as you, and I had issues with the CPU running too hot at load with 3.6+ GHz overclock. After experimenting a bunch with the amount and pattern of applied thermal paste, I finally got my temps down into the mid 70º's under load.

But to get there, I had to attach the HSF literally 7 different times in order to find the right setup. It was a huge pain. I think what I settled on was to spread a thin layer across the CPU with a credit card, then I added a tiny blob to each of the heatpipes on the cooler itself. Saw my temps drop 10º+ and haven't looked back since.

Good luck with the OC!
Click to expand...

I removed the heatsink and saw that there was too much of the goop there so I cleaned a lot of it off the heatsink and some of it off the CPU and the bracket since some of it got on it. I ended up leaving a very thin film on the CPU itself with the heatsink pretty cleaned up. I don't think that I want to remove any more of the thermal paste since it would probably create some gaps in the coverage. I don't think that messing around with the compound is going to yield any better results than I am getting now since I think that these chips run hot and also the ambient temps in each location being tested matter.

Fan Blow Direction Experiment

I also decided to see if the temps would improve if I changed the direction that the fan was blowing, to blow over the memory instead of towards the back of the case to the exhaust. After a few seconds of Prime95 the CPU hit 90C where before it would slowly climb to 80C and then go up a point slowly. So this little experiment ended in failure as I thought it would. It seems that my original fan mounting idea of putting the fan at the back of the heatsink and pushing air over it towards the back of the case is a better idea than to pull air from the heatsink and blow it over the memory towards the front of the case, creating a vacuum and heat hole.

I think that pulling the air from the memory, pushing it over the heatsink, and then having the case can pull the hot air from the heatsink and pushing it out of the case is the best fan setup since it creates a good heat pipe effect through the hottest part of the case. Here's the picture below.

computerinsideintelcore.jpg


computerinsideintelcorey.jpg
 
Last edited:
Frankly, I get this feeling that I'm only going to be able to do 3.6 GHz at reasonable temperatures with Prime95 Blend stable, and that 3.8 GHz might be possible but at higher temperatures but too close to max temp of 100C. I just don't have the cooling required to break past 3.6/8 GHz. I'll be fine with that but I was really hoping for 4.0 GHz on air but that won't be possible I think.

I've already hit 86C RealTemp on Core 0 after less than an hour of Prime95 Blend. I will be fine if this system tests out 8-hour or more stable on Prime95 since I don't see it ever hitting those kinds of loads and temps during gaming or hardcore encoding jobs or virtual machine usage. The Prime95 is like testing the worst case scenario that is unlikely to happen.

I don't want to invest in a water cooling solution just for overclocking bragging rights since the expense of the whole setup would cost more than paying a little extra for the 940 and getting those speeds on air.
 
I was right so far. I can do 3.6 GHz (180 x 20) at only CPU 1.25 V but with ram set on the slowest 2:6 ratio on 540 MHz instead of 740 MHz.

This was the most minimal overclocking attempt with all settings on auto in the BIOS, Turbo Disabled, Multi at 20, BLCK 180, and RAM at 1.64 (normally at 1.60 from XMP profile). I have SpeedStep and C-Step technologies enabled also to save power when the CPU is not in use since I want this to be a feature-rich overclock and not a forced full-speed one.

Now that I have a baseline overclocking benchmark down at 3.6 GHz that is actually Prime95 BlendFFT stable I'm going to see if I can push the individual components any further. I'm going to try to do the test with the RAM at the next higher memory ratio to get it at 740 MHz but with the little higher DRAM 1.64 V to see if it will keep stable.

After that I'm going to see if I can push it to 185 x 20 for 3.7 GHz and maintain stability at 1.25 V. I really don't want to push the voltage up that much because the temps get out of control but I might go up to 1.275 V to see if the temps stay sane.
 
Did a little more testing today. Was able to push the RAM to 720 MHz (1440 MHz DDR3) and retained stability at 3.6 GHz.

The only problem is that my peak CPU temp hit a high of 94C and was idling around 91-92C in the middle of the day that wasn't so hot. There are definitely thermal issues with my setup but frankly my Lian-Li case is setup well and with all possible intakes filled with 120mm fans. The case is quiet because of this so that is why I like it so much. This chip just runs damn hot and there isn't much that I can do about it without replacing the heatsink and changing the fans to high rotation ones that would generate more noise.

Intel Core i7 920 2.66 GHz D0, Asus P6T 0603, G.Skill 6GB DDR3-1600 C8 2T, Xigmatek HDT-1284EE - CPU 3.6 GHz (180 x 20) 1.25 V, DRAM 720 Mhz 2;8 8-8-8-21 2T 1.64 V, QPI 1.2125 V - Prime95 Blend Stable 9-Hour+ 94C Peak

intelcorei7920266ghzd0a.png
 
Last edited:
I moved some cable around in my case to improve the air flow in it but I'm also going to order a front bezel with a 120mm fan in it to pull some air into the top part of the case and feed it to the CPU. There are two items on the market right now to do this and I've contacted a few online retailers to see if they can order the Lian-Li BZ-502A (Silver) bezel that will match my case. If they can't do it I'll ask around on a few forums. If nobody can do it then I'll settle on the Scythe Kama bay since it looks similar to my case but not exact.


Lian-Li BZ-502A (Silver) Bezel with 120mm Fan - $24.99





Scythe KAMA BAY SCKB-1000SL 120mm 120mm Fan for 5.25" System Cooling - Retail - $19.99


 
Last edited:
Race to the Bottom for Voltage

Now that I've reached the thermal limits of my setup when my peak temp hit 94C at only 3.6 GHz, I'm going back the other way and lowering the voltages for the CPU to determine the stability at 3.6 (180 x 20) at the lowest voltage.

I far I have 1.20 V that is stable at 3.6 GHz but when I tried 1.175 V today my system crashed. I'm going to test the memory, and bump it down from 1.64V to 1.60V to see if it is still stable at 720 MHz.

Intel Core i7 920 2.66 GHz D0, Asus P6T 0603, G.Skill 6GB DDR3-1600 C8 2T, Xigmatek HDT-1284EE - CPU 3.6 GHz (180 x 20) 1.20 V, DRAM 720 Mhz 2;8 8-8-8-21 2T 1.64 V, QPI 1.2125 V - Prime95 Blend Stable 6-Hour+ 81C Peak

intelcorei7920266ghzd0a.png
 
Turning off hyperthreading will lower temps significantly. I really don't think there is much cause for it in a quad-core myself.
 
Except for the fact that it generally increases performance by more than a bump of a couple of hundred MHz does of course.
 
DragonQ said:
Except for the fact that it generally increases performance by more than a bump of a couple of hundred MHz does of course.
Click to expand...

I'm going to leave HyperThreading on since I will be eventually using this system for multi-threaded encoding jobs that I do sometimes when I get a hankering for some H.264/AVC encoding action.

I'm going to also mess around with different multipliers to see if maybe I can utilize this the Intel Core i7 920's x21 multiplier to see if I can work around the BCLK issues without bumping up the voltages to control the thermals.

Overclocking for me is a slow process since I just started with this system and I do one change at a time before I test it out during the day for that quick 30-minute stability test to see if the system will crash and reboot within that time and then to do the longer overnight 8-hour Prime95 test to see if the settings hold. I'm not exactly thrilled with the 3.6 GHz that I got but I'm not upset either. Still, there's enough settings left for me to try to experiment.

I'm a little disappointed to find that most folks overclock and post their settings only to find out that they are using Water or others who overclock on air but don't post their exact settings. This turns out to be a solo project for me when instead I should just be able to plug in someone else settings and test them out since I do have the most popular components with this system, Asus P6T, Intel Core i7 920.
 
I used speed fan.

But here is what I got back in MARCH.

i7920.jpg


since then I have added multiple fans and actually have a very very good amount of air flowing though my case. (Front+Side+Top - to - Rear)
 
JakFrost said:
Overclocking for me is a slow process since I just started with this system and I do one change at a time before I test it out during the day for that quick 30-minute stability test to see if the system will crash and reboot within that time and then to do the longer overnight 8-hour Prime95 test to see if the settings hold. I'm not exactly thrilled with the 3.6 GHz that I got but I'm not upset either. Still, there's enough settings left for me to try to experiment.

I'm a little disappointed to find that most folks overclock and post their settings only to find out that they are using Water or others who overclock on air but don't post their exact settings. This turns out to be a solo project for me when instead I should just be able to plug in someone else settings and test them out since I do have the most popular components with this system, Asus P6T, Intel Core i7 920.
Click to expand...

Well here's what I know from my system: Reaching 3.9GHz was easy. I ran IntelBurnTest overnight just now and it didn't fail once (16 threads, maximum RAM, 62 passes). I am running 185x21 (as long as you leave the mutliplier and Turbo options on "auto", you'll always get 21x under stress) at stock voltages. I'm going to try to drop them a bit at some point but I'm a bit busy this week. Reaching 4GHz is more difficult for my system. I always got blue screens. After bumping up the VCore, at first I thought it was always failing even at 1.3V but it turned out to be the IBT 8 thread bug. When I have found the lowest voltages for 3.9GHz, I will set that to my "safe" OC in the BIOS profile thing and then try doing 4GHz again with the 16 thread IBT test.

With 3.9GHz, I idle at 35-40 and my max load is 75, mainly hovering around the low 70s. When we had a heatwave last week (ambient = 30-33C), my temps were 50/85.
 
Ron FTL said:
But here is what I got back in MARCH.

since then I have added multiple fans and actually have a very very good amount of air flowing though my case. (Front+Side+Top - to - Rear)
Click to expand...

You have C0 stepping processor but looking at your 3.7 GHz (185 x 20) at 1.216 V seems very close to my 3.6 GHz (180 x 20) at now 1.2 V.

I always discounted the 185 BCLK number since I was shooting for 190 x 20 for 3.8 GHz so I never went for 3.7 like you. I might try that though since it seems like 190 BCLK is not reachable for me without hitting against the thermals in the 90+C. I'll do a test run tonight at 185 BCLK to see how it goes and how much voltage it needs. Maybe it'll be stable at my current 1.20 V if not I'll bump it up to your 1.216 V.

DragonQ said:
Well here's what I know from my system: Reaching 3.9GHz was easy. I ran IntelBurnTest overnight just now and it didn't fail once (16 threads, maximum RAM, 62 passes). I am running 185x21 (as long as you leave the mutliplier and Turbo options on "auto", you'll always get 21x under stress) at stock voltages. I'm going to try to drop them a bit at some point but I'm a bit busy this week. Reaching 4GHz is more difficult for my system. I always got blue screens. After bumping up the VCore, at first I thought it was always failing even at 1.3V but it turned out to be the IBT 8 thread bug. When I have found the lowest voltages for 3.9GHz, I will set that to my "safe" OC in the BIOS profile thing and then try doing 4GHz again with the 16 thread IBT test.

With 3.9GHz, I idle at 35-40 and my max load is 75, mainly hovering around the low 70s. When we had a heatwave last week (ambient = 30-33C), my temps were 50/85.
Click to expand...

You have a Noctua Push-Pull fan setup but I only have the single Xigmatek 4-heatpipe Push setup so my thermals are not as good as yours. I just purchased a new Lian-Li front fan bay and a good Scythe S-Flex 1,200 RPM fan to push some cold air into my case since there is a bit of a heat intake deficiency right now since the CPU fan has to pull the air from the middle of the inside of the case with the air coming in from the surrounding holes. While this won't push air directly through the heatsink it should improve the thermals a little bit by providing forced cooler air from the front of the case instead of the middle of the case. We'll see what happens.

Now that I've done enough testing to reach an idea of how high my system goes in voltage and thermals and also how low it goes in voltage I'm going to be messing around with different multipliers to see how it goes. I'm going to try the BCLK at 185 and the x21 multi direct or with Turbo=Auto setting.
 
Memory Voltage Back to 1.60 V same as XMP Profile 1

I went ahead and lowered the DRAM voltage to 1.60 V that is what my memory's XMP Profile 1 is set at for DDR3 1600 at 8-8-8-21 2T timings. My memory runs at 720 MHz (1,440 MHz) with Northbridge (Uncore) at 2,880 MHz. It is below the 1600 rating for this RAM so I am satisfied that at least the memory is performing up to spec so I'll take that.

I really don't care about overclocking the memory or pushing the timings on it since the AnandTech - Memory Scaling on Core i7 - Is DDR3-1066 Really the Best Choice? article proves that there is only an average of ~2% real world performance gain between the slowest and fastest memory timings.

I know that my memory on the XMP profile is working on a 2T (Command Rate) timing that makes it a lot slower than 1T timings. Since I plan on putting 6 GB more of the same memory into this system to max it out at 12 GB I'm fine with 2T timings since from my experience with my older system filling all the memory slots required me to set the memory to 2T command rate just to boot the system.

I could use one of the other memory profiles from the SPD's JEDEC #1-5 profiles and get 1T command rate working but frankly I haven't done my own performance testing yet to determine how much of a performance difference I would get from XMP Profile 1 = 800MHz 8-8-8-21 2T versus let's say JEDEC #5 = 741 MHz 10-11-11-27 1T or one of the other JEDEC profiles.

I had a hell of time with my Mushkin Redline DDR1 memory in my AMD Opteron 175 giving me random corruption, errors, and game checksum failures, PunkBuster kicks for consistency check failures when I tried to push it from 200 MHz to above 225 MHz even on 2T command rate. I am happy with memory working on stock settings.

Intel Core i7 920 2.66 GHz D0, Asus P6T 0603, G.Skill 6GB DDR3-1600 C8 2T, Xigmatek HDT-1284EE - CPU 3.6 GHz (180 x 20) 1.20 V, DRAM 720 Mhz 2;8 8-8-8-21 2T 1.60 V, QPI 1.2125 V - Prime95 Blend Stable 9-Hour+ 80C Peak

intelcorei7920266ghzd0a.png
 
Good Bye Floppy

I was very happy to see that Vista and Windows 7 that I now use to be able to read the USB sticks that I put in there during OS install to get the drivers. I also like the fact that the Asus motherboard can flash itself from reading the ROM files from NTFS volumes on my computer also. Now that I don't have any other hardware that I need to flash from a floppy disk I can do away with the floppy drive and bay.



Ordered Lian-Li Front Fan Bay

I went ahead and ordered this bay and fan to put into the front of my case since I have a lot of spare space in there now that the floppy drive bay is useless with this motherboard that has the floppy connector all the way at the back in the top and I don't want to use a ribbon cable that long to connect it. I have a huge bag of extra long floppy, IDE, and SCSI cables but I do not want to use them anymore.


Performance PCs - Lian Li 3 X 5.25" Fan Module with Mesh Front and Filter - Silver - $29.99 USD





Scythe S-Flex 120mm 1,200 RPM Fan

I did a little research on fans and found this very good and interesting article.

Silent PC Review - SPCR's Fan Round-Up #2: 120mm Fans

Out of all of these fans I chose the Scythe S-Flex 120mm 1,200 RPM fan since I think that for this price this is a good fan for the front intake fan for the bay that I bought. I looked at the Noctua fans with their different fins but frankly these fans are too expensive at $20 USD a pop when the other Scythe fans do almost as good but at 25% off the price.

After reading the article I decided on a traditional fan design but with non-sleeve bearings which lead me to the Scythe S-Flex using the Sony S-FDB (Fluid Dynamic Bearing). I really did not want sleeve bearing fan since I had a heck of a time with them failing after a year or so back in the days of 80mm in my Pentium Pro computer. I learned them to go with ball bearing but now that fluid bearing is available I'll take that instead.


Scythe USA - Sony S-FDB (Fluid Dynamic Bearing)



Below is the fan I chose. I had a choice of 800, 1,200, 1,600, and 1,900 RPM but I went with 1,200 RPM version since my internal fans in the middle or rear of the case run usually at 1,350-1,400 RPM on idle and and 1,500 RPM under load. Since this fan will be positioned at the front I want it a little slower and quieter so I chose 1,200 RPM. Also there are reports of the slowest 800 RPM fan making a ticking sound because of the slow rotation or low voltage being used.

Scythe USA S-FLEX S-FDB 120mm Quiet FAN - [URL="http://www.performance-pcs.com/catalog/index.php?main_page=product_info&cPath=49&products_id=4072"]$16.95 USD

sff21e.jpg
[/URL]



Noctua Fad

Silent PC Review - Recommended Fans

I could have chosen one of the Noctua fan that are so popular but frankly even with their slightly better air push and lower noise they strike me a little bit like a fad recommendation especially when you start reading some of the claims on their web site and also look at their strange diagrams and obviously made-up graphs showing curves but no numbers on the axes. I learned from my mistakes of following fads that they usually aren't long lasting and aren't worth the effort or money invested in them so I stay away.

I looked at the numbers just now from the article and it seems like the similar S-Flex and Noctua both do about the same but the S-Flex using half the power and pushing almost the same amount of air with just slightly lower RPMs. Frankly, the power usage puts the Scythe S-Flex above the Noctua fan and it is also a little bit quieter and pushes almost the same amount of air. The newest Noctua fan release uses the same 9-fin traditional fan blade design for all of their current CPU coolers and abandons the aircraft propeller blade design also showing that the company isn't standing behind their radical design.

As much as I don't like some of things that Sony has done with their music division and their overpriced TV and AV equipment I trust them more with their S-Flex S-FDB (Fluid Dynamic Bearing) design engineering than a company like Noctua who make dubious claims on their site.

Code: 
[B]
Make   Model            Voltage    Sound        Speed       Volume    Power
[/B]
Scythe S-Flex SFF21E    12V        23 dBA@1m    1230 RPM    34 CFM    0.62W
Noctua NF-S12-1200      12V        25 dBA@1m    1250 RPM    38 CFM    1.32W

[I]Source: [URL="http://www.silentpcreview.com/article695-page5.html"]Silent PC Review - SPCR's Fan Round-Up #2: 120mm Fans[/URL][/I]
 
Last edited:
JakFrost said:
You have a Noctua Push-Pull fan setup but I only have the single Xigmatek 4-heatpipe Push setup so my thermals are not as good as yours. I just purchased a new Lian-Li front fan bay and a good Scythe S-Flex 1,200 RPM fan to push some cold air into my case since there is a bit of a heat intake deficiency right now since the CPU fan has to pull the air from the middle of the inside of the case with the air coming in from the surrounding holes. While this won't push air directly through the heatsink it should improve the thermals a little bit by providing forced cooler air from the front of the case instead of the middle of the case. We'll see what happens.

Now that I've done enough testing to reach an idea of how high my system goes in voltage and thermals and also how low it goes in voltage I'm going to be messing around with different multipliers to see how it goes. I'm going to try the BCLK at 185 and the x21 multi direct or with Turbo=Auto setting.
Click to expand...

Push-Pull only offers a 1-2C temperature improvement according to the reviews I have read. I am considering using my current pull fan as a case intake fan instead cos my case's only intake fan doesn't feel like it's doing much. I recommend setting both the CPU multiplier and Turbo to auto - not only does this lower your power consumption (because it will not go to 21x unless it's needed) but it also is the only way (AFAIK) to allow one core to go to 22x when it is the only loaded core. There's no disadvantage to leaving all the power saving features like SpeedStep on either (unless you maybe have a rubbish PSU that can't handle the voltage changes).

I lowered my VCore to 1 or 2 steps lower than stock yesterday and my system is still stable. Load temps only dropped by 1-2C but it's always nice to use the lowest stable voltage. 4GHz is still gonna be tough I think without getting into 85+C range. I am currently running my QPI voltage at 1 or 2 steps above the normal 1.2V to be on the safe side (running 185x21 automatically puts your QPI speed above stock even at the lowest setting) but it probably isn't necessary.

JakFrost said:
I could have chosen one of the Noctua fan that are so popular but frankly even with their slightly better air push and lower noise they strike me a little bit like a fad recommendation. I learned from my mistakes of following fads that they usually aren't long lasting and aren't worth the effort or money invested in them so I stay away.
Click to expand...

I love my Noctua fans but then I've never been able to hear how quiet or loud they are by themselves - they are drowned out by my case fans (which aren't exactly loud either). I'm not sure if I'd buy them separately though, £15 is a bit much...
 
DragonQ said:
Push-Pull only offers a 1-2C temperature improvement according to the reviews I have read. I am considering using my current pull fan as a case intake fan instead cos my case's only intake fan doesn't feel like it's doing much. I recommend setting both the CPU multiplier and Turbo to auto - not only does this lower your power consumption (because it will not go to 21x unless it's needed) but it also is the only way (AFAIK) to allow one core to go to 22x when it is the only loaded core. There's no disadvantage to leaving all the power saving features like SpeedStep on either (unless you maybe have a rubbish PSU that can't handle the voltage changes).
Click to expand...

The first thing I did before I started overclocking is to enable SpeedStep and also C-Step sleep states in my Asus P6T BIOS to enable the power saving features of the processor. Even with a manual BCLK set to 180 and Multi at 20x I see the processor running at 2.16 GHz with 12x Multi because of SpeedStep and I really like that since I'm not wasting power running at maximum multi when I'm browsing the web or e-mail.

I also left HyperThreading on because as I mentioned earlier I'm not really interested in the highest clock speed for my overclock when I have to disable features to get there. It just doesn't make sense when a feature like HT gives just as much or more performance than a slightly higher OC like you mentioned yourself earlier. All features On is my overclocking motto.

I'm going to enable Multi=Auto and Turbo=On when I do some more testing, but I just plain forgot to run a burn-in last night after doing some scripting work.

DragonQ said:
I lowered my VCore to 1 or 2 steps lower than stock yesterday and my system is still stable. Load temps only dropped by 1-2C but it's always nice to use the lowest stable voltage. 4GHz is still gonna be tough I think without getting into 85+C range. I am currently running my QPI voltage at 1 or 2 steps above the normal 1.2V to be on the safe side (running 185x21 automatically puts your QPI speed above stock even at the lowest setting) but it probably isn't necessary.
Click to expand...

I noticed that the Asus BIOS leaves the QPI at standard 1.2125 V until you start pushing the BCLK up and then it automatically pushes the QPI higher. I guess in this instance Auto really does something. I still haven't seen the need to push up the QPI past the Auto setting at BCLK 180 but as other people mentioned and overclock articles too I might have to bump it up a bit to get stability. We'll see what happens.
 
I'm testing right now with CPU Ratio (Multi) = Auto and Turbo = On for 3.78 GHz (3.96 GHz Turbo) at BLCK 180 and trying to find the stable CPU voltage and QPI voltage.

I just had a 6-hour Prime95 Blend stable system until I tried to make this post and it blue screened on me and that was at CPU 1.23750 V and QPI 1.21250 V stock.

I'm trying now with same CPU and QPI bumped up two knotches at 1.2250 V. Let's see what happens.

At CPU 1.23750 V my peak RealTemp hit 87C and I can live with that knowing that the front intake fan is coming soon.
 
I've been testing overclock with 180 BCLK with Turbo = On giving me 3.6 GHz with standard 20x multi along with 3.78 GHz with 21x Turbo for all cores and occasionally 3.96 GHz with 22x Turbo on a single core load.

I've been following the minimalist overclocking method and also changing only one parameter at a time while I do Prime95 Blend testing for a few hours to determine stability. I've learned some funny things along the way and they are what I expected. I know full well that 2D stability in Prime95 does not mean that the screen won't blue screen when you try to do something like opening a web browser, since that actually happened to me while running a 6-hour+ stable system on CPU 1.23750 V.

I had to bump up voltage from my previous stable low of 3.6 GHz at 180 x 20 manual multi set of CPU 1.20 V, to 1.23750 V which passed stress test but blue screen with web browser opening, to come back to 1.25 V to retain stability, at least for now until I throw some 3D tests at this machine.

When I was at CPU 1.23750 V and experienced the crash I decided to keep the same CPU voltage but to bump the QPI voltage up two knotches from the default 1.21250 V to 1.2250 V. The higher QPI did not stabilize the core at all and I dumped it back down to the default 1.21250 V and upped the CPU voltage to 1.250 V to get stability.

I've read a bunch of overclocking articles and they all claimed that they had to dump the QPI voltage up a knotch or two to get stability but frankly I did not find that this helped when I had stability issues. The only thing that did help was CPU voltage increase. Maybe it is because I'm working with slow clocks at 180 BCLK that I don't stress the QPI yet.

Anyway, below is my temporarily stable system with all features enabled including Turbo and SpeedStep & C-Step with only a change in the BCLK to 180 and CPU voltage to 1.250 V.

I did hit a high peak of 93C at a ~29C ambient but this normal at CPU 1.250 V and before I get the parts to install the front bay air intake.

I'm going to see if I can push the BCLK to 185 but I doubt it from my previous testing. Still, a 3.6/3.78/3.96 GHz overclock is quite good for me.

Intel Core i7 920 2.66 GHz D0, Asus P6T 0603, G.Skill 6GB DDR3-1600 C8 2T, Xigmatek HDT-1284EE - CPU 3.6 GHz (180 x 20+T) 1.25 V, DRAM 720 Mhz 2;8 8-8-8-21 2T 1.60 V, QPI 1.2125 V - Prime95 Blend Stable 5-Hour+ 93C Peak - Unstable: Blue Screen Crash on USBPORT.sys

intelcorei7920266ghzd0a.png
 
Last edited:
My last attempt just did a blue screen crash on USBPORT.SYS. I guess that CPU 1.250 V is too low when it hits 3.96 GHz for 22x turbo so I set the CPU to 1.2750 V. My thermals are now definitely out of whack but I should get the intake fan tomorrow or next day to test it with better air flow.

I'm also getting a lot of Intel RAID failures on this computer with random drives losing connection or after a blue screen crash getting unrecognized device on ports. Makes me wonder if my older 550 Watt power supply might need to be replaced with a newer model.
 
Try 1.35 Vcore initially @ around 4ghz and lower it as much as you can.
 
med1kl said:
Try 1.35 Vcore initially @ around 4ghz and lower it as much as you can.
Click to expand...

That is about the most unhelpful comment I've read in a while.

I've already been reporting that at 1.250 V my peak CPU temps hit 94C and are just 6C under the thermal limit of the processor. You go ahead blindly recommending 1.35 V just like the overclocking articles who all used water cooling.

Man, I hope that nobody listens to your suggestion about overclocking.
 
Yeah, bumping the voltage up to 1.35V will definitely net you a temperature-related shutdown very quickly.

I appreciate the methodical approach you're taking. I have a very similar setup using an ASUS P6T Deluxe v2.

What I find interesting is that your QPI Voltage can stay so low. My QPI has to run at 1.35 or so to maintain stability. I guess that's due to the fact that I'm running my Corsair ram at 1600 Mhz (spec), but still I'd love to put less voltage on it.

Carry onward...and get some cooling for the case. =)
 
Paradoxex said:
Yeah, bumping the voltage up to 1.35V will definitely net you a temperature-related shutdown very quickly.

What I find interesting is that your QPI Voltage can stay so low. My QPI has to run at 1.35 or so to maintain stability. I guess that's due to the fact that I'm running my Corsair ram at 1600 Mhz (spec), but still I'd love to put less voltage on it.
Click to expand...

I've seen people recommend high QPI voltage but when I did my testing at 3.60 GHz at CPU 1.20 V and DRAM at 1.60 V I found that I didn't have to touch the QPI 1.21250 V at all. One of the AnandTech articles even recommended the CPU PLL voltage to be bumped up by .02-08 V. I see a lot of people saying to set the CPU at 1.35 V right off the bat also and that is just crazy for me when I consider the thermals.

I found that for me the sweet spot seems to be 3.60 GHz at BLCK 180 x Multi 20x at the very low 1.20 V and stock QPI with manageable ~80C peak thermals on air with 29C ambient.

To go up from there like I tried recently with Turbo On for 21/22x multi really seems to require major voltage boosts like 1.30+ V and thermals in the 100+C peak. The clock difference will only be 3.78 GHz at 21x and 3.96 GHz at 22x for only 10% frequency increase but 20% temperature increase 80C-100+C and even lower real-world performance result difference my guess at ~4%.

It doesn't seem worth it to push the CPU over the 1.20 V and ~80C peak for such low performance and chance at instability in higher heat ambient temperatures.

I'll still try to push the system though to see if I can manage a bit more out of it.



Turbo On and CPU 1.2750 V = Blue Screen

Well, it seems that Turbo is a bane of my overclocking existence. I just had a blue screen crash even after I bumped up CPU to 1.2750 V thinking that it should be enough juice for 3.96 GHz with 22x turbo multi. It isn't enough juice.

At this point I have to make myself stop because I think that my thermal limit is 1.250 V with a peak of 94C giving a little breathing room before the 100C CPU thermal limit maximum. I already went off on a prayer thinking that 1.275 V would stabilize the system but I didn't run any stress tests against this voltage because I knew that it would bump up against 100C peak.

I'm still running at QPI 1.21250 V minimal voltage and I'm stable on the memory front end portion, at least I think that I am from my Prime95 stress testing. I haven't really thrown a lot of other 3D testing at this machine because I've had problems getting 3DMark Vantage to work on Windows 7 RC 7100 because it complains about losing screen focus. It might just be some application that I'm running that's causing a pop-up or taking the focus away will have to deal with it later. I do my 3D testing by actually playing STALKER Clear Sky for a few hours at a time. I did do the whole PCMark Vantage yesterday with 3x each category with Turbo On and it passed but it seems that this benchmark doesn't really stress the system enough to determine stability.



Turbo Off for now, Back to 3.60 GHz 180x20

I had to make the tough choice just now after the blue screen and went back to Turbo = Off and 3.60 GHz at 180 x 20 manual at CPU 1.20 V and QPI 1.21250 V. Last night I played for 4-hours with the same settings and everything was fine but now it blue screened on me. It is possible that the 22x multi doesn't kick in most of the time since I've seen the system be stuck at 21x when doing Prime95 testing but I have a feeling that when the CPU actually kicks in to 22x for any period of time and hits that 3.96 GHz speed I blue screen pretty quickly. My guess is that at this speed the CPU really needs 1.30+ V to maintain stability and unfortunately that voltage is beyond my thermal limit on this air cooler.

I'm going to wait for the intake fan to come to see see if it drops my peak temperatures. I'm also going to do some research on forums and testing about real need to increase QPI/DRAM Voltage because my memory is stable at XMP Profile 1 2:8 1440 MHz (720 MHz) at 1.60 V with 8-8-8-21 2T timings. The spec on this DDR3 set is 1600 so I'm staying within it. Like I said before, I'm not interested in overclocking the memory because the net gain is 2% at best.

I'm trying to get Turbo = On to work but it is difficult getting the system to comply with testing the 22x multiplier because I can't control when it's enabled or not. Even when I run a 1-thread Prime95 it sticks to 21x.

This overclocking thing is difficult and the difficulty really comes at the very end. I am a newbie to overclocking when I judge myself by other people's standards like on XtremeSystems forum. I also have a hard time finding real good step-by-step and exact information that people have for their Intel Core i7 920 overclock settings on air. Plenty of folks report their speeds on water but few on air.

More research, more testing...
 
Last edited:
Interesting.

As I said, I run a very similar system to you.

I have a:
Core i7 920
6gb of Corsair Dominator 1600Mhz Ram
2x Sapphire 4890 Toxic GPU
Asus P6T Deluxe V2
Corsair 850TX
1x 1TB WD Caviar Black
Antec 1200

I've been running 3.6 Ghz at 1.2V until recently, when I pushed the computer up to 4.0 Ghz at 1.25V. I've tested it with Prime95, etc...and it all runs stable. QPI Voltage is 1.3125V or 1.35V (I forget atm), and Ram voltage is 1.6V. I have better thermals than you do at the moment, almost certainly due to the fan heavy case I'm using. I'm running every feature on, minus Turbo Mode.

That said, I tried decreasing the QPI Voltage, and it caused instability at the settings above. I've also tried decreasing the CPU Voltage independently, and it also causes a crash. I haven't touched PLL Voltage or any other setting. Thus I have two suggestions:

1. Perhaps the QPI Voltage is the bottleneck? I know you've theorized that it isn't, but for me, it was a significant factor. Take it for what you will.
2. In general, I have heard it is better to increase BCLK and disable Turbo Mode for OC'ing. Why? Because Turbo mode throttles more quickly due to thermal factors than a normal OC would. It also makes the overclock more predictable, as you're finding out. With turbo turned off, it is a lot easier to figure out if your overclock is actually stable, as the processor won't throttle itself until it starts to brake 100 degrees C. With Turbo mode on, it will throttle itself a little before this.
 
Paradoxex said:
1. Perhaps the QPI Voltage is the bottleneck? I know you've theorized that it isn't, but for me, it was a significant factor. Take it for what you will.
Click to expand...

You and other are very probably right. I've been reading of folks hitting 3.60 GHz like I did with just simple tiny CPU voltage bumps but when they head out for the 4 GHz+ zone they need to start juicing the CPU a lot more and many of these people also juice up the QPI at the same time.

I just merely said that at my low CPU 1.20 V testing the QPI 1.21250 V is pretty much in-parity so it makes sense that I'm able to keep it at stock and not experience problems. I really haven't done very thorough testing of QPI voltage and its effects on my stability or overclocks. I've been concentrating on the CPU voltage-to-thermal relationship. I'm going to bump up the QPI to 1.35 and CPU to 1.25 to see if I can get higher. Turbo will be Off of course. I fear though that my system might be BCLK ~180 limited from my experience with previous ad-hock testing of setting CPU and QPI to 1.35. We'll see how it goes.

No Manual 21x Multi

One thing I noticed is when I set my Asus P6T with BIOS 0603 to CPU Ratio (Multi) of 21 and I started CPU-Z right after boot-up to Windows 7 I noticed the ratio jumping to 22x even though I assumed that Turbo was Off since the Turbo option disappears from the BIOS when you set the Multi manually. However, when I just checked the BIOS, when I got to Multi Auto it shows Turbo as Disabled, when I set Multi to 21x and I check CPU-Z just now it jumps to 22x. I'm going to have to limit my testing to 20x Multi because setting it to 21x automatically enables Turbo and the system hits 22x without telling it to.

It sucks because I think I could stabilize the system at BCLK 180 x 21 Multi for 3.78 GHz but I can't since Turbo auto-enabled itself. Maybe Asus will release a new BIOS that fixes this or maybe this is a CPU level feature that gets enabled if the multi is set to 21x.



Odd Multi's? Why? - Facts, Options, Rumors!

One thing that I saw on XtremeSystems forum is one person mentioning to use the odd-multi's 17,19,21 for overclocking on this Core i7 processors but they didn't specify a reason. That makes things even more strange for me to test.

I really hate doing research in a medium like forum posts because the information is so fragmented that it is hard to piece it all together. Then was is even harder people write their opinions along with their facts in the same post and I cannot figure out which information is actual and useful and which is just conjecture. What is worse is that some of the gems that are posted are actual facts and they are corraborated by other people's testing and findings but the references to this information is sometimes spread between different pages on the same thread of different threads or different sites and articles. Other times opinions and rumors also spread like facts and others repeat them without knowing the "why".

It really pisses me off because it is so time consuming to find good information in forum posts unless a senior member steps forward and writes up an explanation and reference thread with links to sources. Those posts are absolutely golden but are few and far in-between.

This thread itself is turning into a hodgepodge of random information and testing mixed with my wrong assumptions and flawed test results. This is not good also. I think I'll need to delete it and condense it later to a summary post with what voltages give me what temps and what is stable and what is not.
 
Last edited:
4.0 GHz - No Go!

I've tried it again like I did before but I cannot get this system to 4.0 GHz while maintaining my CPU 1.250 V temperature to keep my thermals in check.

I set CPU to 1.250 V and QPI 1.350 V and when I set the Multi to 19x or 20x and move the BCLK past 195 I get hit with instant freeze. From what others have said on XtremeSystems forum a freeze means lack of voltage and I can't push my system past 1.250 V because the thermals would hit past 94C peak and definitely come up against 100C CPU thermal limit.

I did the same type of testing when I first got the chip by trying to use other people's voltage numbers to see if I could get the system to work at their reported 4.0 GHz and higher speeds. I hit my system with CPU 1.350 V and QPI 1.350 V but anytime I go above BCLK 195 it freezes or crashes. I even went higher with the voltage than that just to see if I could get stability for even a minute but it was a no go situation also. It feels almost like this system has a low BCLK limit.

BCLK Limited to ~180

It seems like this system is BCLK limited to ~180. I tried 190 and it is unstable and I didn't bother with 185 because it is such as small jump that it isn't worth the instability at 3.70 GHz.

I thought that it might be the memory so I set it to the slowest 2:6 ratio and overvolting it to DRAM 1.64 V but not dice. I tried CPU PLL bump from 1.80 V to 1.90 V like the AnandTech article and no dice either.

At this point I think that CPU voltage is the key and I can't push past CPU 1.250 V without better thermals.

I'm hoping that the intake fan might provide a few degrees of drop but I think that if there is any difference with it it will be minimal since my case is already pretty well ventilated since it has the Lian-Li PC-V1200 Plus aluminum design with the whole front being full of intake holes.

Could it be the memory?

It could very well be the memory that I'm using. I assumed that the G.Skill DDR3 1,600 MHz at 8-8-8-21 would be good memory only to find out that it requires 2T rate while other memories are 1T at 8-8-8-24 or so. This wasn't listed in the product specs. From my experience with my AMD Opteron and DDR1 memory I could run my Mushkin Redlines DDR1 200MHz at ~225MHz with 1T for 2GB and 2T for 4GB. Here I have to start off with 2T right off the bat. It makes me think that maybe the BCLK limit is not motherboard or CPU related but memory related since even at the lowest 2:6 ratio it still might be causing the instability. I can't test any other memory since I don't have DDR3 anywhere so I can't tell. I might try only using one stick to see if I can get to BCLK 195-200 and not freeze.

XBitLabs Article - Close Results

X-Bit Labs - Intel Core i7-920 Overclocking Guide - CPU Overclocking (Page 11)

The article above that I linked to in the first thread got very similar results to what I'm seeing with my system except that they have the C0 revision and I have D0 processor, Deluxe version of the motherboard, better memory, and a different air cooler. They pushed their CPU and QPI to 1.35 V and hit 3.8 GHz at 190 x 20 with stability. I can't do that with my system though since I retain stability at ~180 BCLK. So this article isn't so far from the mark for the experience that I'm having.
 
Last edited:
Memory Benchmarking - 1T vs 2T Command Rate = Same Bandwidth Results, Latency Lower for 2T by 3 ns

I started thinking of what could be blocking my BCLK from going above 180 and I thought that it might be the memory timings. I originally set the Asus BIOS AiTweaker to X.M.P. (Xtreme Memory Profile) to get the memory timings from XMP Profile 1 from the memory SPD chips to set it to 8-8-8-21 2T. I did all my stress testing like that and everything worked well.

Today I decided to set the AiTweaker mode to Manual. This caused the BIOS to pick up the memory timings from the SPD on the memory DIMMs for 9-9-9-24 1T.

I the Multi to 20x and BCLK to 180. I set the DRAM Frequency to DDR3-1443MHz to keep it below the 1,600 MHz rating for this memory. I couldn't get closer to the rating since the next frequency ratio was 1,886 MHz.

I did SiSoft Sandra 15.99 SP3c tests for Memory Bandwidth and Memory Latency tests and got the same results for the Bandwidth, 24.48 but for latency the 1T came back with 71 ns and 2T with 68 ns.

I'm going to leave the AiTweaker in X.M.P mode since it seems to work just fine without having me enter all the memory timings manually and gives a slightly lower latency.

I really like this XMP profile stuff since on my DDR1 memory I had to set all ~20 memory parameters manually by hand to get a stable system. What a pain the butt that was.
 
JakFrost said:
That is about the most unhelpful comment I've read in a while.

I've already been reporting that at 1.250 V my peak CPU temps hit 94C and are just 6C under the thermal limit of the processor. You go ahead blindly recommending 1.35 V just like the overclocking articles who all used water cooling.

Man, I hope that nobody listens to your suggestion about overclocking.
Click to expand...

Wasn't trying to be unhelpful, just didn't realize your processor was binned so low man.:(
 
Hiya Jak!

Your temps seem kind of high. Have you considered reseating your heatsink? The reason I ask is because I was also having extremely high temperatures, and it turned out the culprit was that I'd put too much TIM on. After I cleaned it off and applied properly, while making sure to evenly bolt my V8, my temps dropped more than 10c. As a reference, here's what I'm running:

capturea.png


As you can see, I'm @ 3.8 @ 1.211v, and 75c under load. Understand, I'm not bragging, or trying to be snide, I'm just telling you what I did wrong. Before, to get 3.8, I tried 1.325, and would hit 99 or sometimes even 00.

I do think my chip is strange, or maybe my mobo, because I have to set CPU core voltage VERY low in bios -- 1.625v!! Under load, it shoots up to the 1.211 you see there, but normal useage rarely sees it go over 1.2. I sort of wonder if the voltage setting of my board is defective. It works, though, so I haven't RMA'd it, nor will I until it doesn't. As for hitting 4.0, I've tried, with no luck, but I'm happy where I am. Good luck with your OC!

capture1o.png
 
You must log in or register to reply here.
Back
Top