Overclocking at Intel 2007

Overclocking at Intel has been one of our most popular articles so far and we hope that you will like this followup article on this event that happened this year on October 5-7. Enjoy!

After trying to write the first page of this article six, seven or perhaps even eight times, I came to the conclusion that maybe it’s just best to say it like it is. Right now I’m traveling home after having experienced a weekend that will forever be remembered as one of the most exciting events of my life. I got a chance to actually meet with some of the legends of the overclocking community; Sampsa Kurri of muropaketti and his fellow companion Pekka Rousu from the Finnish computer magazine MikroBitti, our friends from Denmark; Tom (xTom) and Peter (Nosfer@tu) of OC Team Denmark and last but certainly not least; Sturla, also known as Stoolman, from our neighboring country Norway, and our Swedish colleague from Sweclockers, Öjvind (KnightCharger).

Last year Intel held a gathering much like this one, but back then it was only myself [Andreas/Delph1], Robert (crotale) and Marcus (Kinc) at the event. It was a much smaller event, but thanks to the success of the first event we were invited back, and this time we wouldn’t be alone. Like I said above, Intel had invited overclockers from Norway, Denmark and Finland, beside us from Sweden. The idea wasn’t exactly to battle us against each other, but more to combine our different skills into one huge overclocking think tank.

Very unfortunately, Marcus caught the flu right before the event. He phoned me the day before and told me the bad news. He had been flying back and forth across the globe to display extreme overclocking with liquid nitrogen all through August and September, but if there was any event he was looking forward to more than any other, it was this one. And I don’t blame him, last time we had a blast. The build-up before this year’s event left me tingling with excitement and I felt really bad for Marcus having to stay home.

On the following pages we will tell our version of what happened up at Intel. I’m pretty sure that the other articles covering this event will be very different as we all had different experiences up in Stockholm.

I know we started the last article with some funny anecdotes about Marcus’ navigational skills, but since he was home with fever I actually thought that things would be different this year. I know for a fact that he was planning to go by train this year, but to be frank that didn’t really calm me down that much. It wouldn’t surprise me if he all a sudden called me saying he had accidentally ended up in Malmö instead of Stockholm, but that’s also one of the reasons we love him.

My journey didn’t exactly go as planned either. I was just suppose to take the train from Kalmar to Alvesta and then switch to another train that would bring me to Linköping where I would meet Robert and from there on we would travel together to Intel and Kista, Stockholm. We had been chatting with Sampsa on the night before and we were planning to pick up him and Pekka at Arlanda right before we went to Intel.

I got on the right train and everything was going smoothly. I sat down and kept hearing a voice in my head singing: ”I’m going to Intel, I’m going to Intel, I’m going to Intel…” I put on my headset and started listening to Toy Dolls’ album ’Covered in Toys’ when all of a sudden the left channel dies. I started investigating the cord and I noticed that there is small cut and that the copper inside had been exposed and damaged. I sighed loudly … not a good start.

SMS 1 to Robert: I’m sitting on the train, but my headset is broken. 🙁 Only got music in one ear, considering just going home. … Bad omen you know (;-)) //Andreas

There was quite a lot of adrenaline moving through my body at this time. My legs were shaking and the gentleman sitting face to face with me on the train was starting to wonder why I was looking so nervous, and I told him that I was just hoping that nothing would go wrong.

I reached the first station and things were going just fine, however we got stranded there for 7-8 minutes because a train traveling in the opposite direction was late. [”Crap, crap, CRAP!” echoes in my head] I only had 10 minutes to switch trains in Alvesta and now I’m down to 2-3 minutes. That’s going be tough if the trains are a long way apart (Alvesta is a rather big train station where a lot of trains meet, switch lanes and continue to almost every other part of the country). I calm myself down by thinking that I’m quite fit and fast, so it shouldn’t be a problem even if there are several tracks and platforms between the trains.

The train left the station in a hurry and was moving a bit faster than usual to try and catch up with the schedule, but right before we’re about to enter the next station in Växjö the train stops a few kilometers outside of town. [”No way … this isn’t happening?!”] I got up and try to walk off my anxiety and stop my heart from beating so damn fast. I eventually returned to my seat and I saw that my co-passenger was looking a bit worried. I decided to explain to him that I was traveling to Stockholm and that I was going to visit Intel, in a ”once in a lifetime” event. Although in my case it’s actually twice in a lifetime (hoping to make it thrice. Please, pretty please?), but realizing I’ve already been there once didn’t exactly reduce my heart rate.

SMS 2 to Robert: First I have to wait for 7-8 minutes in Nybro, which leaves me with 2 minutes margin when switching trains in Alvesta, but then it turns out that a train has an engine failure at Växjö station! Argh!

My co-passenger tries to clam me down with some encouraging words that the train will most likely wait in Alvesta. I already knew that, but somehow it helped hearing it from someone else. Who ever you are, thank you for those calming words. I also want to apologize to the conductor for nagging him about how we were doing. Sorry. You were kind enough to tell me that the train would be waiting and that it would be right across from where we would be stopping so there’s no need to worry or get stressed.

I still sprinted across the platform to the X2000 heading for Linköping, [”There’s just no way I’m taking any chances here”] The train switching went fine. When we left Alvesta the conductor announced through the speakers that unfortunately we have to slow down north of here because the track is being repaired. [”Why SJ? Why today? On the single most important day of the year you decided to repair the track and on one of the only two tracks that I will be traveling on this very day? You knew I was coming right?”] *evil stare at the SJ logo on the ticket*.

SMS 3 to Robert: Had to sprint across to the next train … 🙂 but damn it now I’m coming to you in about an hour

Eventually we got past the track repairs and the train could speed up and try to make up for lost time. Everything was moving on nicely, when all of a sudden the train hits the emergency brake! I opened my mouth and I was just about to shout something like: ”ARE YOU F*CKING KIDDING ME?”. But then I noticed that there were children sitting on the seat right next to me so I managed to control my feelings, a bit. Sorry about that, but I was starting to get a bit frustrated. SJ is usually flawless, but for some reason pretty much everything was going wrong today.

SMS 4 to Robert: Should be arriving in 10 min

SMS from Robert to Andreas: Goodie, I’ve filled up the dewars with nitrogen, so we’re good to go

Eventually I arrive in Linköping and can meet up with Robert at the station. He had already been to AGA and picked up 50 liters of liquid nitrogen. We were suspecting that we would be the only ones with nitrogen so we were worried we would have to share ours, but we were wrong (we actually ended up borrowing some nitrogen from the Danes).

From there we travel to Stockholm and Arlanda to pick up Sampsa and Pekka. I’ve been wanting to meet him for a long time so I didn’t really mind the detour to Arlanda. Sampsa called us on the way up and said he had found someone to share a cab with from Arlanda to Kista. This saved both us and our Finnish friends an hour of wait and travel.

Finally we arrive at Knarrarnäsgatan 15, Kista.

The great thing about sharing a ride with a guy like Robert, or Marcus for that matter, is that you don’t have to hold back on the extreme details when discussing hardware. There’s just no such thing as too extreme hardware and you can really just let the thoughts fly wild. It’s quite a relief in some ways as usually you have to soften and keep the number of details to a minimum so you don’t confuse the person your talking to. And as always, we had some quite interesting discussions, unfortunately they would fall under more than a couple of NDAs, so enough about that.

Robert and I were the first to arrive in Kista. First we got a short tour of office, there were some changes since last year. For example, they had set up two gaming rigs where you could go for some head-to-head racing, shooting or whatever, if you needed a break from the benching (although I actually think that Thomas from Intel was the one using them the most, but then again he was playing against the rest of us then). I never got around to try the gaming rigs, but I know that others had a great time gaming with him.

Thomas decided to show us the Room of Secrets where no photography was allowed (hence no photos) and he opened a package containing five tiny, but fierce beasts; 5x Intel Core 2 Extreme QX9650 week 34, B3 stepping. As close as you can get to mass production samples at the time, which is also why we tried to share as much as we could back at the beginning of the month, when we were actually at Intel. Thank you for being so understanding, and here today we now have the results you have been asking for.

Before we could start to bench we had to pick an area in the lab to set up the gear. Thomas had prepared two laboratories for us. One larger and one smaller, but they had the same basic equipment, low humidity and stable temperatures. Since drinks were free (not just for us, but the Coke machine at Intel doesn’t charge you anything), it was good idea to drink as much as possible. The low humidity combined with the liquid nitrogen can really damage your vocal cords and mucus membranes in the nose and upper throat.

I’m not saying we were trying drink nitrogen, although I did hear that some nitrogen were being tossed around (I think our Danish and our Norwegian friends know more about this). It’s just that the cold air coming from the pots is not very healthy and you should try to drink a lot when benching with liquid nitrogen. On the other hand, both Robert and Sampsa had some issues keeping the thermos and Coca Cola Light cans apart (same silver color), so avoid Coca Cola Light when benching, it’s not good for you!

Since we were the first we got to pick area first and we decided to go for the smaller lab with a single work area, mainly because you had move through the other area to get to this one. Less running behind our backs that way, although it wasn’t much running at all. We were all quite focused on the benching for the 49 hours we had at Intel.

Eventually both of our Danish friends from OC Team Denmark arrives and right after them, Sampsa and Pekka came through the door. Thomas gave them the tour while Robert and I sat back and took it easy.
The Norwegian representative Stoolman was running a bit late, but he was making good speed so he would arrive shortly. He had one huge compressor cascade with him and of course plenty of liquid nitrogen.

Thomas decided that it was perhaps for the best that we checked in at the Hotel before Öjvind called, but just as we got the keys for our rooms, he called and Thomas had to get and pick him up. We put the bags in our respective rooms and went back to Intel to set up the gear.

When we had all gathered, Thomas fetched the goodies. Unfortunately I don’t have a photo of all five because it was something a ravenous pit of benchers throwing themselves at Thomas when he entered the room with the five Intel Core 2 Extreme QX9650.

There was no cherry-picking or even major consideration of which batch number that could possibly be better/lower/higher than the others. They were all week 34 and that was enough to make us all more than happy. The benchmarks by Marcus we had posted news about was with week 30, and even though they looked good, week 34 is even better.

Anders and Mikael from ASUS arrived at 16:00, one hour late… with some more goodies. A bunch of Maximus boards, both Formula and Extreme versions, a Striker Extreme, two GeForce 8800 GTX, one GeForce 8800 Ultra and one ASUS Radeon HD 2900 XT.

I had also brought a Be Quiet Dark Power Pro 1000W to see if could handle the load of true overclocking, and it did. Thomas had some power supplies from OCZ and PC Power & Cooling, but I was really curious to find out what the Be Quiet PSU could do. Marcus is using it now. Along with the power supplies, OCZ had also sent some brand new DDR3 memory for us to use.

We decided to team up with the Fins to see if we together could perhaps break Marcus’ 3DMark 06 world record. If we did that we could move on to SuperPi 1M. Sampsa brought two Radeon HD 2900 XT and pots for them, along with two quite nice Kingston HyperX PC6400 modules that did 400 MHz with 3-3-3-3 timings. Robert brought his good ol’ DragonPot (CPU pot) and we also had the nitrogen for the pots.

We chose the ASUS Maximus Formula, which uses DDR2 memory, just to play it safe. The DragonPot was mounted onto the CPU, but the graphics cards had to settle for the stock coolers until further. We wanted to get to know the processor a bit before mounting pots on the graphics cards too. We kept the temperatures at somewhere around 0 °C and bumped up the frequency. The processor was able to reach around 4.5 GHz with a slight increase in voltage at about -45 °C. The temperature was lowered to near the -100 °C mark and the voltage was raised a bit more. It ran past 5 GHz and when it reached 5.45 GHz at -150 °C and around 2.0 V we were quite pleased with what we had accomplished the first day. There was still more to get from the processor, but now we had been able to reach stable frequencies that would be enough to break records.

We felt that it would be better stop here, get up early and start overclocking the graphics cards tomorrow, than continuing now. We were all rather tired from the trip and we had gotten a feel for the processor by now. A fresh start would give us more room for testing and if anything would go wrong we would have a lot of time to recover.

We four and Öjvind, who was running everything on air cooling, were actually the only ones that were having any real progress. Our Danish and Norwegian friends were using ASUS R.O.G. Maximus Extreme and DDR3, and they were having a lot of problems with their processors. It was probably memory or motherboard-related because later on day two they started to catch up with us.

Sampsa had flashed the Radeon HD 2900 XTs to 850/950 MHz before the event, which is slightly higher than the stock frequencies at 740/825 MHz. We had some discussions on how to overclock them further, but since Sampsa was used to doing it through the software method we decided that this was probably for the best. The alternative would’ve been to flash the cards with new BIOSes.

We had a slight problem with the graphics cards as they for some reason used different voltages. We thought it was because one of the cards was a press sample, which are worse overclockers thanks to a different manufacturing process than the retail cards, while the other one was a retail card. As long as we used air cooling this didn’t present a problem, however it didn’t take that long before we swapped the air coolers for two pots so we could start pouring liquid nitrogen and reach sub-zero temperatures on the graphics cards too. We had the processor running stable at 5.2 GHz and were happy with that for the moment.

With the pots mounted to the cards we kept the cards at the flashed 850/950 MHz frequencies during boot and then overclocked the cards using the software/desktop switching method. We reached 1 GHz GPU frequency without much fuzz, but when we started to climb above that we started getting crashes and odd lockups. Those of you who haven’t benched Radeon HD 2900 XT at sub-zero temperatures have probably not experienced these issues, but we can tell you that condensation is a major issue with these cards. They are very picky. We also discovered that the contact between the probes and tubes was quite poor.

So we simply disassembled the whole system and tried to get rid of as much condensation as possible. Unfortunately, none of us had a hairdryer with us (that’s usually Marcus’ department), but luckily our Norwegian friend Stoolman did. We managed to get rid of the worst using toilet paper, which I stole from one of the toilets, and then finished the drying with Stoolman’s hairdryer.

This was also good place for us tp pause because Xover TV was about to arrive. They would do interviews and take some pictures of us relaxing in the dining area, and then continue by filming us while we were benching. The footage will be shown on Swedish TV7 November 15th.

Both OC Team Denmark and Stoolman of Norway was starting to get some results now. OC Team Denmark had been running dual setups, Peter with QX9650 and Tom with QX6850. Both of them had now reached some nice SuperPi results and Stoolman was starting climb up the frequency ladder and prepare for the world’s first sub-8s SuperPi 1M, which he also managed to achieve later that night.

We were starting to get a bit annoyed by now as the system kept locking up now and then. The processor had a coldbug at somewhere around 155-160°C and we were constantly benching at temperatures within a few degrees of the coldbug. Robert was the one balancing the temperature of processor, and certainly did a great job, but there was something missing.

Sampsa decided to go and fetch some Heineken for us and all of a sudden things started to work. It’s hard to explain, but when we started to drink more, the hardware started to cooperate. Even though Heineken is far from my favorite it seemed to get the job done so why stop? So to summarize, beer good, Coca Cola Light bad. I’m not sure what the others were drinking but I have a hunch that it wasn’t enough Heineken.

We rebuilt the system and worked our way up again. The GPU frequencies was raised again and the processor was pushed up to the same frequency again. When the graphics cards were stable at 1050/950 MHz we decided to work up the sweat on the processor. It had been running at 5.2 GHz, but we needed more if we wanted to break Marcus’ record. We worked up the frequency to 5.4 GHz on the CPU, 1135 MHz on the GPUs and 970 MHz on the graphics memory, and this resulted in a 3DMark 06 score of 28,054 points. The cards had been flashed with a new BIOS now because we wanted a higher GPU voltage, 1.4 V, which is why the screenshot shows a default speed of 1000/980 MHz.

The picture below (taken by Pekka) describes our feelings pretty well.

Throughout the day I kept contact with Marcus to check how he was feeling and perhaps get some advice from him. He was of great help and should of course be credited for in the results. In between questions he managed to squeeze through the entire eighth season of X-Files and when we were getting tired and thinking of going to bed, he was about to start on season nine.

We kept pushing, but after some lockups we decided that this is probably as far as we’re going to get today. At the same time, Öjvind had been pushing his QX9650 to more than 4.6 GHz on air cooling and as much as 1.6 V! He was using two Radeon HD 2900 XT in CrossFire and actually had the second best 3DMark 06 score at the end of the day. The processor was was 3DMark 06 stable at 4551 MHz and SuperPi 1M stable at 4622 MHz, which resulted in a SuperPi 1M time of 9.812 seconds. The best score in 3Dmark 06 was achieved at 4.45 GHz with two Radeon HD 2900 XT in CrossFire (850/900 MHz); 22,578 points.

And then we have this just hilarious conversation that took place in the middle of the day when Stoolman came and asked for help overclocking Radeon HD 2900 XT CrossFire;

Stoolman: Did you have the bios on a USB-drive?
Sampsa: I have them on a Floppy
Stoolman: Ah, an USB?
Sampsa: No, a floppy.
Stoolman: A USB-floppy?
Sampsa: No, a floppy!
Stoolman: Like a floppy floppy?
Sampsa: Yes, a floppy floppy floppy.

We slept until late on Sunday, with the exception of Sampsa who wanted to see the Formula 1 race. Unfortunately the hotel didn’t carry TV6 so he set his alarm two hours earlier for no good. I take partial blame to this because I was pretty sure that the hotel carried TV6, I was wrong.

We started day three by trying out a second Intel Core 2 Extreme QX9650. The reason we chose to switch processor was that eventually we would have to pick one to take home, and we were two groups benching together. However we didn’t have much luck with the second one, but we think it was because we switched to Sampsa’s much sturdier pot and it seems it didn’t get any real contact with CPU due to the capacitors around the socket.

Stoolman had started to do some Aquamark 3 benching with a GeForce 8800 Ultra and was able to break 300,000 points and reach 301,034 aquamarks. Last night after we had went to bed he broke the 8 second mark, 7.984 seconds to be more precise. He was able to improve this to 7.879 seconds when he got home. He did some single-card GeForce 8800 Ultra benches later on and was able to achieve 98,447 points in 3DMark 2001SE with the processor running at 5,377 MHz, and 30,662 points in 3DMark 05 with processor now clocked at 5,460 MHz and GeForce 8800 Ultra at 970/1270 MHz.

Alas, Robert and I had to leave a bit early. Robert had practice at 18:00 back in Linköping and I wanted to catch an early train since I had to be back in the lab early Monday morning. We had to leave by 14:00 since it’s a two hour trip at best back to Linköping from Kista, worst case scenario it takes a whole hour longer depending on the traffic. But before we could leave we had to choose between the two processors. We knew that one of them was a gem, the other was still untested. I did my best to cheat, but failed (I’m lousy at it). Luckily the one we got home was a good one, but to be honest I sincerely hope that the one Sampsa got is just as good.

We were quite happy with what we had achieved during the weekend. I can imagine that Adam and Eve felt pretty much the same way when they were kicked out of Paradise as we did when we realized that we were actually about to leave this wonderful place and return home.

The trip home went flawless. I decided to save some time by writing parts of this article on the train and beside me I had an elderly gentleman who peeked at what I was writing. Every time he made a glance he also made a surprised and almost startled face as if he was wondering what the heck have these guys been up to? It was like if he wanted to turn to me and say: ”That’s not normal…” And he would have been absolutely right about that.

I can’t blame him for feeling that way. I’ve been trying to explain to my closest friends and family what I did this weekend but I just can’t find any good layman’s term for “overclocking a not yet released Intel Core 2 Extreme processor with liquid nitrogen in Intel’s own laboratory and smashed the current 3Dmark 06 world record.” I just summarize it to “a once in a lifetime experience.”

It’s really hard for me to explain how I feel about having experienced this. Last year I was one out of three blessed souls to visit Intel’s lab in Kista. This year we were a few more with Sampsa, Pekka, Sturla, Peter and Tom, and Öjvind, but to be honest that only made it better. Thomas at Intel Sweden is looking more and more like a wizard.

Marcus was really down when he phoned me the night before the event and told me he had the flu. There wasn’t much I could do, but if we’re lucky there will be another one in a year or so, and I do believe that such an event will be even more special than this one.

Until then I hope you had a good time reading about our visit to Intel, because we certainly had a great time being there. We can’t thank Thomas Melzer at Intel Sweden enough for his interest in overclocking, for arranging this event and of course being such a good sport when we kept breaking hardware and asking for more.