7950x3D die on D15 with Kryosheet included. GLORY OR DEATH!
7950x3D die on D15 with Kryosheet included. GLORY OR DEATH!
I finally did it. It's been on my mind since I noticed MSI's factory included a contact frame for their Z97 MPOWER. The outcome wasn't great with Cinebench R23 across all cores. Just CCD0 managed to clear 19k, which is the best I've seen so far. The top result was 19108 for CCD0. All-core scores stayed around 36k, though I haven't tried more aggressive undervolting yet—getting it to POST proved tough. My main focus is checking if a direct die change enables stable, near-constant 5.25GHz boost, which would surpass what the 7800x3D can achieve without bclk OCing. Even without the thermal cap, the 7950x3D is mostly limited by TDP and voltage. I'll experiment more to see if I can push past this barrier, but it seems unlikely. The solder removal was a challenge—I spent about half an hour on that.
Currently, using just CCD0 offers a clear edge compared to the 7800x3D with the larger multiplier. Running both dies directly doesn’t yield much benefit. CCD0 consistently stays above 5.05GHz, while the 7800x3D maxes out at about 50.5x. It’s not a big gain overall, but it confirmed my idea that CCD0 has thermal constraints. Maybe the Kyrosheet shifted and I’m missing contact on CCD1—something I can check next.
I think using binning for the 7950x3D makes sense. The 7900x3D with the same 52.5x multiplier might be due to having two fewer cores, but it reaches a limit before hitting that in my tests before reaching the die. I'm also attaching fresh screenshots of runs to this thread for my own reference.