Prepare to Skylake OGL Cinebench WRs

China based web TechBang showed (and then quicky removed, but what was once on the net… you know 😄 ) some Skylake comparsions to Haswell. Let's start with CPU-Z and GPU-Z screens:

The Vcore looks too low to me to be true, but whatever. The tests showed that on average, the Skylake is only about 6.1% faster that Haswell. In some tests is even slower. But there are some instances, where it is 29.1% faster, like the Cinebench 15 OpegGL test:

So to put long story short, expect some Cinebench OGL world records on HWbot pretty soon 😄

BTW, as for APU performance, Intel lose big time. AMD Kaveri A10-7870K is getting in the Sky Diver test (witch is very closely tied to the games real world gaming framerates) 6400 points, however Skylake reach only 4650 points. Without eDram's Kaveri stayed unbeated. (removed)

dude your back welcome back!

Still alive  😉

But the news are not that good - yep, 0.4Vcore, but that is probably only w/o load. As you can see, it downclocked to ~800MHz with multiplier x8.

It would be much better if they let run some load on all the CPU cores before taking this screenshot.

What I find odd on the Skylake is, that the TDP is 95W, with allegedly 14nm technology. Now the previous Haswell with allegedly 22nm technology had TDP 88W.
Both CPU's do 4GHz under multicore load, but Haswell do 4.4GHz, while Skylake do only 4.2GHz.

Into the TDP is simply not counted the TDP of the graphic part, so there is no excuse for the higher TDP at all. Unless the "14nm" process have higher leakage, witch lead to the higher TDP that the 22nm process.

That could possibly also explain the lower "turbo" clock…  ::)  All in all, not only the alleged nm is scam ( Comparsion of the manufacturing processes of Intel, Samsung and TSMC - 14nm proces infografika | - to put it short, the "xx nm" does not reflect anything anymore on the reall dimensions on these chips ), but maybe we reached the very end, where the smaller process is not beneficial anymore and best is the Intel "22nm" process used in Haswell…?

  • 3
  • 1134