Charles Wirth
06-12-2007, 09:29 PM
At Computex I had a week of testing under LN2 and would like to share my experiances on what to expect from overclocking of this board.
Sponsors:
K|nppIn LN2 pots
Intel Core 2 Duo and Quad processors
Abit IP35 Pro Retail box #19
Corsair PC10000 Dominator DDR2
Nvidia 8800 Ultra
OCZ 1010Kw PSU
Western Digital 10K RPM hard drive
http://219.84.132.86/blog/uploads/2006/12/computex_07/1597.jpg
Abit ordered a pressurized bottle of LN2 for the show, I was specific for a 22 psi pressurized bottom fed dewer so we could have LN2 on tap, I have been running this way for a few years and hopefully more will adopt variations to work with, simple copper plumbing with two valves made for LN2. The two feeds were to support CPU/GPU and we found the single CPU feed was more than enough to support a quad heavily overclocked.
http://219.84.132.86/blog/uploads/2006/12/computex_07/1594.jpg
http://219.84.132.86/blog/uploads/2006/12/computex_07/1595.jpg
Crotale and I had two boards, we each had one to push for the exhibition, for 3 days I ran LN2 in the worse conditions possible, it rained every day of the show. We have huge frozen areas and even puddles at some points benching at full speed. We used grease to protect the components around the cpu socket but none inside. I used a modified VapoLS base gasket for a tight seal with the pot.
Both boards were unmodded and what you can expect to get from the retail parts on the shelf, both boards exhibited similar traits and limitations.
http://www2.abit.com.tw/upload/products/ip35-pro_top_500.jpg
Complete specifications can be found here on Abits site;
http://www2.abit.com.tw/page/en/motherboard/motherboard_detail.php?pMODEL_NAME=IP35+Pro&fMTYPE=LGA775
The configuration video thanks to madboxpc.com
http://www.youtube.com/watch?v=i-nQEheqY58
I was able to test a large variety of processors starting with the QX6850 quad core G0 stepping supported out of the box. This was questionable if it would even post but I was nearly able to reach 4.8Ghz under LN2. The CPU cold bugged around -150c and I spent a day plugging at going higher with FSB kicked back to 266, the maximum I booted into windows as at 485FSB, uguru up near 500 would cause the system to lock up. We did not measure vdroop with the quad under load.
Thanks image pcnews.ru of Crotales rig
http://www.overclockers.com.ua/news/101806.jpg
A pair of sweet retail QX6700 at the show also checked in at 4.8Ghz and 485ish FSB top end, what made these retails chip nice was they cold bugged at -170c.
I pushed the X6800 to nearly 5.3Ghz and maxed the FSB out at 580Mhz with the vcore set to 1.9v max in bios. There was .02 droop measured, vcore was stable with the four phase power design. The system was very stable and pi was run at 5.2Ghz yielding a nice sub 10 second run.
Thanks images and coverage tweaktown.com
http://images.tweaktown.com/imagebank/news_abitocshowctx0702.jpg
http://images.tweaktown.com/imagebank/news_abitocshowctx0701.jpg
Thanks image and coverage inquirer.net
http://www.theinquirer.net/images/articles/Crotale_at_Work.jpg
We did have two other QX6700 ES chips, one bugged at -120c and the other was a -150c chip and reached 485FSB like the rest of the quads.
On air I pushed the QX6850 to 4Ghz fairly easy, for the demo's sake, all benches were run at 3.6Ghz 460x8. I was running multi-threaded applications for the live demo OC.
I also the the E6320 processor on air and LN2, it reached 585FSB before system lock, we were most impressed with the chip on this board, I believe the vcore was around 1.6v it was stable until it locked at top FSB of the board. On air it passed 3.6Ghz for live demo benches.
In the bios a graphic bug was noticed that would put random charecters on the memory speed line, this bug is graphical only and when highilighted it will dissappear. It caused no stability problems.
The chipset was very cool to the touch while heavily over clocked, little voltage was needed to push it to max FSB, I was setting for 1.4v and noticed no stability problems.
Each day we run for hours plugging away at FSB and voltages, the board is very safe for itself except for memory voltage options over 3v on the bios we had. iirc it went to 3.3v, I will confirm.
We froze over every day, the first day I was operational for 4 hours, the second round I split time with an IN9 for SLI benches and demo. And then on third round I ran all day on LN2, setup at 9:30AM and ran till closing at 6PM non stop. Plugging away with our OC live demos it was a perfect day, heavy rains, large crowds and a stable motherboard frozen over completely. I was asked at noon if I wanted to break down and swap out and I declined as the board was doing so good overclocking the X6800 plugging away over 5.2Ghz with vcore set to 1.9v on my unmodded board.
Both boards survived and performed superb, PCMark2005 was ran on the quad @ 4.6Ghz for a CPU score of over 7100, this was above par and noted.
With LN2 there wasn't much use of bandwidh tests other than I did tighten up the memory to my known good settings with complete stability.
Thanks image inquirer.net
http://www.theinquirer.net/images/articles/GTL_Mod_Abit.jpg
Another unique option in bios is GTL reference mod options that are also adjustable in uguru. This does work and I was able to crash the machine when adjusted too far. This is defaulted to 67% and seems to yeild best results with quad cores. I was not able to breach 600FSB on my baord while testing any of the processors on it. I was not able to breach 500FSB on the quad core at the show. Any attempts would lock the system.
Thanks Abit for having us out to show off at Computex, I am very happy that everything went perfect for the entire show, I could not have expected better results. Often when these events can be plagued with problems but with proper planning and solid hardware we made it look simple.
Thanks to the sponsors listed above, all the supporting hardware performed admirably.
Thanks Crotale, the sweedish overclocking champion. Great to meet you and huge props for the perfect show. Looking forward till we meet again.
Sponsors:
K|nppIn LN2 pots
Intel Core 2 Duo and Quad processors
Abit IP35 Pro Retail box #19
Corsair PC10000 Dominator DDR2
Nvidia 8800 Ultra
OCZ 1010Kw PSU
Western Digital 10K RPM hard drive
http://219.84.132.86/blog/uploads/2006/12/computex_07/1597.jpg
Abit ordered a pressurized bottle of LN2 for the show, I was specific for a 22 psi pressurized bottom fed dewer so we could have LN2 on tap, I have been running this way for a few years and hopefully more will adopt variations to work with, simple copper plumbing with two valves made for LN2. The two feeds were to support CPU/GPU and we found the single CPU feed was more than enough to support a quad heavily overclocked.
http://219.84.132.86/blog/uploads/2006/12/computex_07/1594.jpg
http://219.84.132.86/blog/uploads/2006/12/computex_07/1595.jpg
Crotale and I had two boards, we each had one to push for the exhibition, for 3 days I ran LN2 in the worse conditions possible, it rained every day of the show. We have huge frozen areas and even puddles at some points benching at full speed. We used grease to protect the components around the cpu socket but none inside. I used a modified VapoLS base gasket for a tight seal with the pot.
Both boards were unmodded and what you can expect to get from the retail parts on the shelf, both boards exhibited similar traits and limitations.
http://www2.abit.com.tw/upload/products/ip35-pro_top_500.jpg
Complete specifications can be found here on Abits site;
http://www2.abit.com.tw/page/en/motherboard/motherboard_detail.php?pMODEL_NAME=IP35+Pro&fMTYPE=LGA775
The configuration video thanks to madboxpc.com
http://www.youtube.com/watch?v=i-nQEheqY58
I was able to test a large variety of processors starting with the QX6850 quad core G0 stepping supported out of the box. This was questionable if it would even post but I was nearly able to reach 4.8Ghz under LN2. The CPU cold bugged around -150c and I spent a day plugging at going higher with FSB kicked back to 266, the maximum I booted into windows as at 485FSB, uguru up near 500 would cause the system to lock up. We did not measure vdroop with the quad under load.
Thanks image pcnews.ru of Crotales rig
http://www.overclockers.com.ua/news/101806.jpg
A pair of sweet retail QX6700 at the show also checked in at 4.8Ghz and 485ish FSB top end, what made these retails chip nice was they cold bugged at -170c.
I pushed the X6800 to nearly 5.3Ghz and maxed the FSB out at 580Mhz with the vcore set to 1.9v max in bios. There was .02 droop measured, vcore was stable with the four phase power design. The system was very stable and pi was run at 5.2Ghz yielding a nice sub 10 second run.
Thanks images and coverage tweaktown.com
http://images.tweaktown.com/imagebank/news_abitocshowctx0702.jpg
http://images.tweaktown.com/imagebank/news_abitocshowctx0701.jpg
Thanks image and coverage inquirer.net
http://www.theinquirer.net/images/articles/Crotale_at_Work.jpg
We did have two other QX6700 ES chips, one bugged at -120c and the other was a -150c chip and reached 485FSB like the rest of the quads.
On air I pushed the QX6850 to 4Ghz fairly easy, for the demo's sake, all benches were run at 3.6Ghz 460x8. I was running multi-threaded applications for the live demo OC.
I also the the E6320 processor on air and LN2, it reached 585FSB before system lock, we were most impressed with the chip on this board, I believe the vcore was around 1.6v it was stable until it locked at top FSB of the board. On air it passed 3.6Ghz for live demo benches.
In the bios a graphic bug was noticed that would put random charecters on the memory speed line, this bug is graphical only and when highilighted it will dissappear. It caused no stability problems.
The chipset was very cool to the touch while heavily over clocked, little voltage was needed to push it to max FSB, I was setting for 1.4v and noticed no stability problems.
Each day we run for hours plugging away at FSB and voltages, the board is very safe for itself except for memory voltage options over 3v on the bios we had. iirc it went to 3.3v, I will confirm.
We froze over every day, the first day I was operational for 4 hours, the second round I split time with an IN9 for SLI benches and demo. And then on third round I ran all day on LN2, setup at 9:30AM and ran till closing at 6PM non stop. Plugging away with our OC live demos it was a perfect day, heavy rains, large crowds and a stable motherboard frozen over completely. I was asked at noon if I wanted to break down and swap out and I declined as the board was doing so good overclocking the X6800 plugging away over 5.2Ghz with vcore set to 1.9v on my unmodded board.
Both boards survived and performed superb, PCMark2005 was ran on the quad @ 4.6Ghz for a CPU score of over 7100, this was above par and noted.
With LN2 there wasn't much use of bandwidh tests other than I did tighten up the memory to my known good settings with complete stability.
Thanks image inquirer.net
http://www.theinquirer.net/images/articles/GTL_Mod_Abit.jpg
Another unique option in bios is GTL reference mod options that are also adjustable in uguru. This does work and I was able to crash the machine when adjusted too far. This is defaulted to 67% and seems to yeild best results with quad cores. I was not able to breach 600FSB on my baord while testing any of the processors on it. I was not able to breach 500FSB on the quad core at the show. Any attempts would lock the system.
Thanks Abit for having us out to show off at Computex, I am very happy that everything went perfect for the entire show, I could not have expected better results. Often when these events can be plagued with problems but with proper planning and solid hardware we made it look simple.
Thanks to the sponsors listed above, all the supporting hardware performed admirably.
Thanks Crotale, the sweedish overclocking champion. Great to meet you and huge props for the perfect show. Looking forward till we meet again.