Comments by "Winnetou17" (@Winnetou17) on "" video.

They did already. It's hard to see, because it's been many iterations. A 6700K is much much worse than a flagship 10th or 11th gen CPU, especially with the Spectre+Meltdown mitigations in place. And I'm pretty sure that a 11900K with only 4 cores enabled + hyperthread will be significantly faster than a 6700K, while consuming less power. Or the same power consumed, but much higher clock speeds. I said a 11900K instead of a 11100 because the 6700K was the flagship, so it had the best binning, so it would be fair to compare it to the best binned of this generation.
50
@mukkeshmckenzie7386 Guilty! I'm just a bit salty that people still seem to think that Intel CPUs are the same as Sandy Bridge (last great IPC improvement) or Sky Lake (first 14nm ... ish... 5th gen was mostly missing). It's not like Intel wanted the 10nm fab to go so wrong and be 2-5 years late. They didn't plan for 100-200 MHz improvements for 5 consecutive years. It's surprizing how far they managed to get with 14nm and how they can still somewhat compete with AMD, which are on a much better manufacturing node.
34
@mukkeshmckenzie7386 It's hard to say now, but I don't think they got to more than 4 cores (finally) just to combat AMD. I think they genuinely just managed to make work ok then. 8600K was launched only 6 months after Ryzen 1000, which clearly means it wasn't something that they recently started. And I do think they would've released it without Ryzen too. Between 6th and 7th gen - literally 200 MHz more, a joke of an upgrade Between 7th and 8th gen - 2 more cores. Also a bit better clocks IIRC Between 8th and 9th gen - another 2 more cores. Also a bit better clocks IIRC and turbo boost improvements. IIRC it was here that Specter and Metldown had hardware mitigations, so it should be some IPC improvements over 6-7-8th gen when running in safe mode, to say so. Between 9th and 10th gen another 2 more cores, better thermal interface (made the die thinner or something like that). Really pushed the clocks to the maximum. More core and cache tuning options. They DO have A LOT of anti-consumer practices, to this day. Those should be the focus of criticism on them, not the 14nm part.
5
@SimonVaIe True. That's why I said it's difficult to say. Thing is, for HEDT, they had "Broadwell-E" with 6950X as their top-of-the line CPU, in 2016 (when Kaby Lake was on desktop), with 10 cores. They were just then developing these things, on the new 14nm node, which also had a bumpy start. From this, I'd say that they brought the mainstream 6 core about as early I'd you expect. Most likely they might've delayed them a bit if Ryzen wouldn've appeared.
3
@FullNietzsche Interesting. Theoretically speaking you should have two sources of some IPC improvements: from Broadwell to Skylake and from getting the Spectre and Meltdown mitigations in hardware. Not something huge, indeed. I'd estimate to be something like 20%. But, it's also comparing a flagship to an i5. It's still better in absolutely everything, at half the price and half the power consumption, right ?
2
@davidereverberi5279 You didn't understood what I wrote. It's because Skylake (and before) doesn't have the hardware mitigations that Comet Lake has which makes it kind of an IPC increase. It's basically just much less of a penalty to the IPC on Comet Lake than it is on Sky Lake. Next, you have to add that on top of Sky Lake IPC improvement over Haswell. Here's a review from when 6700K was fresh: https://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation You can see there, near the bottom, that it says 5.7% IPC over Haswell on Sky Lake with DDR 3, and 6% IPC increase when Sky Lake was on DDR 4. That was quite low, buuut, that was with DDR 4 on 2133 MT/s, the lowest most basic DDR 4. Which was normal back then, as it was new. Now, with 3200-4400 MT/s being the norm, with nice timings, that should be easily at least another 5%. All three combined should add up to about 20%. That is, from Haswell to Comet Lake.
2
It's about the same as 10700K. People freaked out because IN A NEW WORKLOAD of AVX512 it did consume much more power. But it was about the same in AVX2 workload
2
@Doyyumhwaji Maybe it's a name thing. If you care about humanity getting better, name your next son Steve!
1