Last time that technically happened was Broadwell. Sure we can argue it was only this corner case - high performance desktop - that saw ST regression, we can also argue that i7-5775C was a 65W TDP chip, but the truth is it was unlocked and could not clock high enough to match previous gen ST performance. On top of that it was late and had poor firmware support.
Broadwell did fine! You still have people holding it up as the desktop gaming champion, though as DDR4 speeds go up and as people agree to start using high-speed DDR4 in benchmarks, that claim becomes less true every day.
Board support is really what killed it. Intel never meant for the i7-c and i5-c to go very far.
Uhh, yea. That's why I said Z370 slash CNL-PCH. And if you refer back to the leaked documents(because I sure did), they both offer Next Gen Optane Memory.
The reports that suggested Broadwell didn't clock well because it was based on a mobile chip seems to ring true.
There's also articles like these: http://semiaccurate.com/2014/07/11/intel-castrates-broadwell-gutting-performance/
14nm is the problem. PCWatch had an article where they speculated Skylake had to rearchitect for frequency since 14nm wasn't cutting it.
I do remember that some folks got early 5775c samples to run at around 4.8 GHz. They were outliers, but they did exist. It took high voltage, but it seems that Broadwell-c had very low leakage current.
So, let me get this straight? Intel "re-names" their iGPUs as "UHD", but still lacks the proper ports to actually drive an UHD display (at a proper frame-rate)? LOL. Typical smoke-and-mirrors Intel.
Literally says "4K Output Support" on the slide. What more do you want?
Z370 isn't CNL PCH. It's KBL Refresh PCH. Z390 is the actual CNL PCH.
Then LOL at AMD and nVidia as well, because they have done the same with multiple generations of gpus as well. And AMD has also done it with cpus.So, let me get this straight? Intel "re-names" their iGPUs as "UHD", but still lacks the proper ports to actually drive an UHD display (at a proper frame-rate)? LOL. Typical smoke-and-mirrors Intel.
Starting with, I think Polaris on AMD's side, they support HDMI2.0 natively. NVidia supports it starting with Maxwell (2.0?), although their hardware is apparently flexible enough, and their drivers capable enough, to drive 4K60 at reduced Chroma sub-sampling, over their HDMI1.4 ports, even my GT630 Kepler boards will therefore support 4K60 over HDMI.Then LOL at AMD and nVidia as well, because they have done the same with multiple generations of gpus as well. And AMD has also done it with cpus.
Your entire anti-intel rant has nothing to do with my comment that re-branding is a common practice. In fact your beloved AMD was one of the worst for rebranding gpus.Starting with, I think Polaris on AMD's side, they support HDMI2.0 natively. NVidia supports it starting with Maxwell (2.0?), although their hardware is apparently flexible enough, and their drivers capable enough, to drive 4K60 at reduced Chroma sub-sampling, over their HDMI1.4 ports, even my GT630 Kepler boards will therefore support 4K60 over HDMI.
Intel? Bah, good luck getting 4K60 over any of their iGPU's HDMI outputs, without additional hardware.
And why is this important? Intel apologists will likely (correctly) point out that Intel iGPUs support 4K60 over DisplayPort. That's all well and good, I suppose, but nearly all 4K UHD TVs support HDMI2.0, and NOT DisplayPort. This is an atrocious omission, especially for the HTPC crowd. (*)
(*) I think Apollo Lake may have native HDMI2.0 support, but not totally sure yet.
While I will agree that both AMD and NVidia re-brand GPUs, the fact that Intel is branding this one "UHD", while not support the interface most commonly used for interfacing UHD displays, seems a bit offensive, and somewhat dishonest.Your entire anti-intel rant has nothing to do with my comment that re-branding is a common practice. In fact your beloved AMD was one of the worst for rebranding gpus.
I do remember that some folks got early 5775c samples to run at around 4.8 GHz. They were outliers, but they did exist. It took high voltage, but it seems that Broadwell-c had very low leakage current.
The "5GHz OC" has been bandied around for so long now. 4790K was supposed to do it, then the 6700K, it took Kabylake to reach 5GHz in actual overclocks. Early 5775C rumors had quite high clocks too. That's probably where the 4.8GHz numbers came from.
Most 5775C chips ended up in the 4.2-4.3GHz range. Broadwell-E didn't clock much either. Something is common to both why it refused to clock high enough.
(*) I think Apollo Lake may have native HDMI2.0 support, but not totally sure yet.
Still running mine at 4.2 GHz. At 2133 CL11 MHz
DDR3 I can do 1.365v for the core but at 2400mhz CL10 I have to raise the core voltage to 1.414v! Getting 2400 MHz CL10 RAM wrecked my OC but I refuse to go lower!
I can boot up at 4.3 GHz and do some gaming and Cinebench runs, but it isn't stable. I have gotten it to boot at 4.4 GHz a few times, but I can't imagine the voltage needed.
I really want to kill this chip soon to give myself an excuse to build a new rig
3.4Ghz all core turbo, 4.4Ghz turbo 3.0. Search anandtech. However to maintain all core turbo indefinitely custom water cooling may be necessary. We will see soon.Anyone wants to guess all core turbo boost clock for 7980XE part? My guess is 3.1Ghz for all core sustainable Turbo boost 3.0. If it manages that then it could end up being around ~10% faster than TR 1950x in heavy MT workloads for around 2x the cost (boards and RAM being roughly the same). Given how gaming performance is roughly identical between tweaked SKL-X and tweaked Ryzen, this cannot end up well for XE in price/perf/$/watt comparisons. It will be a monster of a chip that is for sure, fastest one can buy, but not worth the money IMO since the difference between it and TR will be marginal at best in 90% of sue cases. Other("lower" end) SKUs will compete much better due to lower prices.
3.4Ghz all core turbo, 4.4Ghz turbo 3.0. Search anandtech. However to maintain all core turbo indefinitely custom water cooling may be necessary. We will see soon.
On the specification side, the higher-end CPUs get a kick up in TDP to 165W to account for more cores and the frequency that these CPUs are running at. The top Core i9-7980XE SKU will have a base frequency of 2.6 GHz but a turbo of 4.2 GHz, and a Favored Core of 4.4 GHz. The turbo will be limited to 2 cores of load, however Intel has not listed the ‘all-core turbo’ frequencies which are often above the base frequencies, nor the AVX frequencies here. It will be interesting to see how much power the top SKU will draw.
How much of that boost is coming from AVX512 though? They are claiming that 2x memory BW is the key to the speedup. Surely AVX512 helps somewhat but that is not the main culprit as per their explanation. Any platform with massive core count and massive mem. BW will see big gains for all we can tell right now. Needs to be thoroughly tested which will happen as soon as XE launches.For those that want to see AVX-512 optimized program perform: http://x265.org/x265-receives-significant-boost-intel-xeon-scalable-processor-family/
That's per core performance boost. 40% boost in average. Though that's Skylake-SP and x265 guys say memory bandwidth helps a lot there, they say even client SKL-X core sees nice benefits.