For the first time, an HWBOT OC Challenge featured three different stages. Although all three stages were very similar in style, each required a different approach and different accents to be won. In the end, USA overclocker Chew*, known for his in-depth knowledge on AMD overclocking, took the win. Hampti, Chew* and yours truly Massman took one of the stage wins.
It was most certainly an interesting challenge. Not only because of the impressive memory frequencies that were reached on an AMD platform, but also because of the choice of mainboard and CPU proved to be vital for winning the competition. The less cache on board, the more performance. The higher the CPU frequency, the higher the score … switching from mainboard A to mainboard B sometimes led to a significant drop in performance and so on. In the end, a lot of testing was needed.
Link to the competition: http://hwbot.org/competition/hoc_jun10
Everyone expected Norwegian AMD nutcase Knopflerbruce to take the win with ease, but a late submission of Austrian overclocker Hampti moved Knopflerbruce to place #2, followed by Chew*, Massman and Moonman. Initially, everyone expected the highly overclockable Samsung TCCD to be the right choice, but the ultra-low latencies of BH-5 memory turned out to be more effective in this competition. Several overclockers reached frequencies up to 280MHz CL2-2-2-5 and very tight sub timings … even years after the reign of BH-5 this still is incredibly impressive!
Overclocking the CPU was quite important for the simple reason that the IMC (integrated memory controller) of the K8 series was running 1:1 with the CPU frequency. This means that if you wanted to increase the memory bandwidth, you needed to increase the CPU frequency as well! A small issue was found with the .5x multipliers which made the IMC run slower than the CPU causing worse performance.
As for the mainboards, I don’t think I have to draw a picture: DFI! DFI! DFI! The now out-of-business LANparty series of DFI reached the top of its popularity with the NF4 series on Socket 939. No wonder virtually everyone was running one of the NF4 DFI mainboards in this stage.
Congratulations to Hampti for winning the stage!
As said before, Chew* was the winner of this stage with an impressive 1124.2 points, which roughly 60 points more than, for the second time, runner-up Knopflerbruce. Arman.n, Massman and Moonman complete the top-5 of a stage that proved to be a battle between old and new-skool hardware.
For the winning submission, Chew* had to pull out a 6.5GHz Phenom II Thuban combined with a NB frequency of 4.5GHz whereas runner up Knopflerbruce used a 3.6GHz Windsor. Further down the ranking, we see a Phenom II at 4.5GHz being almost 150 points behind the K8-Windsor of Knopflerbruce … clearly K8 is still a good choice. The memory clocks were very decent, albeit not as impressive as what we saw in the DDR1 stage. DDR2-1200CL4 was needed to take the win.
The choice of mainboard proved to be vital for this competition. The only way to really get up in the rankings was by running HTT:MEM at 1:2. Especially for mainboards this proved to be not that simple since booting at 300MHz HTT frequency combined with 600MHz memory … was difficult to say the least. Combine that with a BIOS that is hardly tuned for the new Thuban and you know you’re in trouble. To quote winner Chew*: “Extremely difficult to achieve on Thuban with current bios. , , Easier to make a list of what’s not wrong with latest beta”
Congratulations to Chew* for winning the stage!
And then comes my moment of glory for winning the DDR3 stage!
In contrary to the two previous stages, there’s no sign of Norwegian Knopflerbruce at the top of the rankings. This means losing a lot of vital points to beat Chew* in the overall ranking. Chew*, on the other hand, managed to grab second place and secure the overall win, followed by Arman.n, PcCI2iminal & Yoko and very, very old-skool MrIcee. Most of the overclockers used Elpida Hyper-based memory at CL6-6-6 or CL7-7-6 to run tight timings combined with decent frequency. No one besides MrIcee managed to run higher than 1GHz, while this US clocker secured DDR3-2100CL7-7-6. Impressive!
In the end, this also proved to be about mainboards. Since the MSI 790GX-GD70 was the only board ready to be put under LN2, I had to settle for reasonably low memory frequencies with the board ‘walling’ at around DDR3-1880. Against all expectations, 920MHz CL6-6-6-18 proved to be more performant than the 1GHz CL6-6-6-18 of Chew* using the Asus Crosshair IV Formula.
As well as in the two other stages, CPU and IMC clock frequencies boosted the score by quite a bit, given the high memory frequencies, so liquid nitrogen was needed to take the win in this one.
Congratulations to Massman for winning the stage!
I have to say that this OC Challenge was one of the best we’ve had so far. Not only the challenge itself, which again proved memory overclocking can be extremely fun, but also the competition. More than ever, the competition was active with submissions coming through nearly every day. Pushing each other to the limit, that’s what overclocking is all about!
For your information, in this competition we had no less than 181 submissions! Nice!