About ten days ago we noted how Greek Extreme overclocker $@39@ (pronounced Saega we think) had taken down a bunch of Global First Place scores including HWBOT Prime, GPUPI for CPU 1B, wPrime 1024M and wPrime 32M. The scores elevated $@39@ to the heady heights of seventh place in the HWBOT world rankings which is pretty incredible for a non-Elite level overclocker. Today however, we bring your attention to a few more impressive submissions from the prolific Greek that now sees him rise into fourth place in the rankings.
Over the weekend $@39@ submitted a score of 355,858 marks in the classic 3DMark03 benchmark, a new World Record score. The submission was made using an Intel Core i7 7700K 'Kaby Lake' processor clocked at 7,038MHz, a massive +67.57% beyond the chip’s stock settings. The riog was based on an ASUS ROG Maximus IX Apex motherboard with system memory configured at 1,836MHz using 12-12-12-28 timings. In terms of GPU $@39@ used an Nvidia GTX 1080 Ti card with the Pascal GPU clocked at 2,544MHz (+71.89) and graphics memory clocked at 1,526MHz (+10.90). The next fastest 3DMark03 score is from Dancop with 353,600 marks using a Titan XP card.
Also adding to his current tally of 2,641.3 points we have Global First Place ranked scores in 3DMark Time Spy and GPUPI 1B. $@39@ used a Broadwell-E Intel Core i7 6950X processor clocked at 5,175MHz (+72.50%) plus a GTX 1080 Ti clocked at 2,580MHz (+74.32%) / 1,526MHz (+10.90%) to score of 13,448 points, the highest ever single GPU score. The rig was based on a GIGABYTE X99-SOC Champion motherboard with DDR4 configured at 1,515MHz (12-13-15-14). In GPUPI 1B the same rig was used to complete a run in just 8sec 857ms, also a Global First Place score for a single GPU.
It’s interesting to note that $@39@ has been experimenting with a direct mounted delided CPU without IHS in recent sessions. It seems enhanced performance is possible, but benching without an IHS also brings a few unique challenges:
"Indeed no IHS benching helps with 8 threaded benches if you are very careful with it. But is not as easy as it looks. I had to modify my pot base a bit and also try to make the chip bend as little as possible. Sometimes while testing I could bench perfect high core frequencies but anything above 6g for cache was no go. Also some of the times with no IHS my CPU needed high DMI although i was benching 2D with GPUs in the last slot... As for single threaded 2D I did not saw any significant change in frequencies but maybe that depends on each CPU."