Overclocking at Intel

We visited Intel and its lab in Kista, Stockholm, to take a closer look at its first quad-core processor, Kentsfield, and 48h non-stop overclocking session. 40 liters of liquid nitrogen can’t go wrong.

At the launch of Intel’s first Quad-core processor Core 2 Extreme QX6700 we decided to take a closer look at it. That was far from everything we wanted to do with it though, and in the true spirit of NordicHardware we decided that to do things a bit more unorthodox. Robert ‘crotale’ Kihlberg and Marcus ‘Kinc’ Hultin are both famous for their overclocking skills and numerous worlds. Accompanied by ex-overclocker and assisting editor in chief Andreas ‘Delph1’ Galistel they would test Intel’s new processor under completely different circumstances, but not at the homes of any of the X-legends but Intel’s very own labs.

Before the launch of Kentsfield, the codename of the core, we at NH got an offer we simply couldn’t refuse; come up to Intel’s office and lab in Kista (pronounced sheesta), Stockholm. Also known as Sweden’s answer to Silicon Valley. Before visiting Kista it felt a bit sad that Sweden would even attempt something like this, but as soon as we arrived we realized that it was a lot bigger than we could’ve ever imagined. Around Kista, or perhaps Chipsta, you can find representation for practically all major IT and electronics companies known around the globe: Intel, IBM, Philips, Symantec, Citrix, Novell, MathWorks, Ericsson, AT&T, and many many more. The list is simply too long to be presented here.

Before we start we thought we might offer you an insight into what can happen when one gets so excited he forgets how to take directions.

The goal of the journey was to travel from each home of its own and meet in Linköping and then travel to Stockholm. Sweden isn’t all that big so it shouldn’t take more than five hours even for him with the longest distance, but that is only if you travel in the right direction (more about this soon). When we arrived at Kista and Intel we would unpack our gear and install it in Intel’s temperature-controlled laboratory for some really serious overclocking. We weren’t sure what kind of hardware Thomas and Intel had waiting for us, but we can assure you that we were pleasantly surprised.

The map for how Marcus ‘Kinc’ drove (purple) and how he should’ve have driven (brown) to meet up with Robert ‘crotale’ and Andreas ‘Delph1’, coming from Kalmar (Blue) and continue to Stockholm (Red)

Marcus’ trip stated in Varberg and continued to Gothenburg where he picked some metal dealies we needed. His trip would then continue to Linköping. “So far, so good” as the saying goes, but somewhere around here things went horribly wrong. Instead of traveling via Jönköping, Marcus somehow managed to travel west of the big lake Vättern (not shown, but located north of Jönköping). At about 11:00 (AM) he decided to call Robert to check where he was going. He wasn’t certain, but at least he was traveling east… They agree that it’s perhaps best if he continues the same way and one hour later Marcus calls again and tells us that he is closing in on Örebro and on the way to Västerås. Below is the SMS conversation that occurred at this very moment:

Andreas to Robert: What the heck is this?! Did I leave my dry and sunny home for this rainy crap, almost wanna turn back 🙂
Robert to Andreas: … to sit in a cold lab at Intel the entire weekend! 😀 By the way, Marcus has missed Linköping by 100-150 km, so we will meet up with him in Stockholm, hehe
Andreas to Robert: 100-150 KM?! HAHAHAHAHA
Andreas to Anton a.k.a. Woxxman: Marcus accidentally missed Linköping a little with about 100-150 km 🙂
Anton to Andreas:
Anton to Andreas: oops, something went wrong there. 🙂 my real reply; LÖL!!

Anton decides to call Marcus and say: ‘KINCY!! The human GPS, haha'”. Marcus replies: “Yea, I guess something went wrong there, accidentally turned left instead of right at the great lake.”

When we finally arrived to Stockholm, after some shortcuts, Robert and Andreas headed for Intel’s building, while Marcus was somewhere “in the area.” After some consulting of an Information Board and a local description of the area by Marcus, the two decided to fetch him from the outskirts of Kista where he was located. Finally, we could now meet up with Thomas at Intel’s office.

The Golden Gate with Robert on the right and Thomas Melzer from Intel to the left. A silent admiring of our home the coming days. Marcus isn’t lost again, he’s just unpacking.

Above you can see some of the goodies we had to play with. We had more than 40 liters of liquid nitrogen in two containers, two compressor cascades, one single-stage, two toolboxes, a whole roll of paper, one liter of isopropyl, an already mounted test system with Radeon X1950XTX and an Intel Core 2 Extreme X6800.

We still didn’t know what Thomas had managed to gather for us, but it would soon be revealed. Up in the temperature-controlled, but most of all very dry room (it was so dry that Andreas’ vocal cords immediately made him sound like Barry White) we were pleasantly surprised.

You will have to excuse the poor picture quality, the lighting was pretty poor and the photographer was a bit ecstatic at the moment. Just as the picture shows there were five Intel Core 2 Extreme X6800 for us to play with… and five more Intel Core 2 Extreme QX6700.

Left to right, the ten beauties, a close up of one the X6800 and then a QX6700. All five of the QX6700 would have a turn in the test bench.

We arrived relatively late on Friday, which unfortunately forced us to delay any serious benching to Saturday, but it was still enough to unpack our gear and prepare some air overclocking to get a feel for the processors. Andreas was a bit too eager and ruined one of Robert’s voltage mods on one of the motherboards. We realized that we had to drag our personal stuff (hygiene, clothes and other things) off to the hotel before they got cranky. Not everyone can understand the joy of benching up until 4:30 and then walk into the lobby expecting them to have a room ready.

We took our stuff and walked over to Mr. Chip Hotel, 150 meters away, and checked in.

We went back to Intel pretty much immediately and setup our gear and prepared for testing the new Kenty, Marcus nickname for it, with these extremely good air conditions. To begin with we used an ASUS P5W64 Pro. The board was ready to meet a new challenger.

Previously mentioned soldering someone couldn’t keep his hands off…

More precision work by Robert with the same board

Some X1950 coolers we had lying around. The cards had been equipped with better cooling now.

Soon, soon …

On the next page the benching begins.

Since we weren’t sure what would be waiting for us we couldn’t plan that far ahead. Well now we had Kentsfield(s) and simply threw ourselves at them and tried to get to know Kenty a bit better. We had a fully modified ASUS P5W64 WS, but we were unable to push it any higher than 375MHz FSB for some reason and we have to say that we were a bit disappointed. We decided to continue and do some overclocking with air cooling.

P5W64 was used to create a fresh installation of Windows XP

ATI Radeon X1950 XTX without the mousepot

We built a system consisting of a Kenty, an ASUS P5W64, two X1950 XTX and some faithful Corsair modules. We chose to use only a single fan with the tower cooler, with less than optimal mounting and airflow. We decided to remount the cooler.

Air cooling mounted

Marcus admires Andreas’ bottle of isopropyl

We reached the limits of X1950 CrossFire with air cooling on the processor and the results can be seen below. Marcus was very eager to move on and swap the slave card for a Radeon X1900 XTX and equip it with a mousepot. As you might remember this was a combination that Marcus scored some great results with in the past, especially with 3DMark 06. The reason we chose to use a mousepot was that we were too lazy to install the stock cooler, and we couldn’t find it. We think it was left at home, but the nitrogen worked just as good.

Robert refused to drink anything but Mer during the entire weekend

Liquid nitrogen with the X1900 slave card and a remount of the CPU cooler

3DMark 06 is the first benchmark from FutureMark where the processor performance has a test of its own and where this has a major impact on the final result. So we decided to use this to test the processor during the air benching. 5682 points was the best we could achieve with air cooling. The best result on the ORB at this time was about 5950, but it didn’t say what kind of cooling that was used. At 375MHz FSB and a multiplier at 10x we managed to push Kenty to 3750MHz, pretty decent with air cooling.

The results of the first day on the upper half. The lower half is tomorrow’s results.

Since we arrived a bit late, it’s not easy finding your way through Sweden. The roads turn and no signs (or so they say). Tomorrow we will switch to some heavier stuff, but first some more pictures.

Below we’ve gathered some mixed pictures from day 1.

Day 2 we stripped the air coolers and started planning how to position the cascades and remaining equipment.

After moving around some stuff and testing a little we decided to try a new motherboard. Said and done, we fed our Kenty to our AW9D-MAX, board we thought has support for the processor, but no. The Abit board reported completely inaccurate values. This could’ve been solved through a BIOS update but it wasn’t available until Monday.

It took some time planning the positions of the cascades, but we managed to solve it eventually. One of the graphics cards would be cooled by one of the cascades, and the processor to begin with.

With the cascades in place it was time to prepare a new motherboard, another ASUS P5W64.

You need some well prepared graphics cards when you’re about to bench 3DMark.

More on the next page.

Some mixed pictures from Day 2. As we mentioned before we weren’t able to reach the results we were hoping for, but then Thomas stepped up and performed some magic. A short while later we had a BadAxe 2 to play with.

BadAxe 2 unpacked and waiting to be used. It just needed some modifications before we could use it the way we wanted.

Down in the container you can see the nitrogen waiting to come out, hit the "hot" bottom of the copper container and then rise up again.

A so called temp solution and a sacrific…. uhhhh sample. We decided to test all five samples of Kentsfield now. It turned out that two of these were slightly better than the rest and were thus branded "NH approved."

The two best reached a 395MHz FSB completely stable. With air cooling. Since these two turned out to be more or less identical and the remaining three slightly worse we decided to move on from the air cooling. There was simply no time for it.

Below are some mixed pictures of a beautiful decor, voltage mod, processors and other sweets.

We decided to close down fairly early and hit the sack at a pretty decent hour and instead get started quite early on Saturday. Thomas started to get some weird colors in his facial area as well, so I guess it was with some respect for our host as well.

When we started to mount our CrossFire system on the new motherboard it turned out that the space between the boards wasn’t as generous as we had hoped for; the evaporator on the master card rubbed against the slave card.

So we decided to modify the tubing to create some more room between the cards. You have to be careful when doing this though, there are some delicate things hiding beneath all that isolation.

The result was a slightly less strained system. Below is a picture of the mousepot in use. It’s a lot bigger than you think. Corsair’s PC8500 modules could be sighted beneath the fog and the horde of cables from time to time. They handled the stress- as always- without the slightest complaint.

Coldbugs are far from fun and it happened more than once that we had to sit and wait for the temperatures to rise before we could start over.

It’s both very concerning and adrenalin-pumping to see how much ice that is created during an overclocking session. Despite minimal moist in the lab it was still impossible to avoid condensation.

Next we took a tour of Intel’s office.

As you can see below there are worse places to work. Clean and organized and we found out that Intel extremely careful with keeping the working environment homogenous, clean and pleasant. Let’s see what kept us alive during the weekend.

We went around and noticed some products based on Intel’s platforms.

You probably recognize a few of the gadgets above. On the next page there are some picture of the cafeteria and the sleeping room, where they also have their meetings.

We almost felt sad having to leave. Later on we enjoyed some Thai food and Ice Age 2 in the cafeteria.

Since there was no way we could set any new world records with a X1950 CrossFire system we chose to focus on the processor performance and as we said before, the CPU test in 3DMark 06 and completed with Cinebench. We’d found the sweetspot of the processor, or the system, to be at about -125°C. With loads of nitrogen left we just had to load it and run some tests.

The goal was to reach 7,000 points in 3DMark 06 processor test, which we unfortunately didn’t manage to do, but had to settle for 6,995. After testing more than plenty of configurations we had to give up and start packing.

The journey to Stockholm didn’t go as planned exactly and the trip home was pretty much the same. Since Robert had to take a detour and switch cars, Andreas traveled with Marcus to Linköping. After discussing both Kentsfield and overclocking in general they accidentally missed Linköping and ended up in Ödeshög instead, a detour of an hour or so. They turned around and met up with Robert so he could give Andreas a ride home to Wiktor, a.k.a. Coore, since the train home to Kalmar had already departed. Marcus continued home to Varberg, but had to ask for a few hours off the day after since he wouldn’t be able to get some sleep otherwise.

We want to thank Intel for giving us the chance of looking around their office and not the least give us access to their labs.