Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
Kaby Lake IPC: same as Sky Lake
http://www.hardocp.com/article/2016/12/09/intel_kaby_lake_core_i77700k_ipc_review/1#.WE4hpor5510
For clarity, what the review did was to set a Sky Lake CPU and a Kaby Lake CPU to exactly the same fixed clock speed of 4.5 GHz, with turbo boost and various power-saving features that alter clock speed disabled. Then they ran a bunch of benchmarks and found that the two CPUs performed the same.
Basically, the advantage of Kaby Lake over Sky Lake is the ability to clock a little higher at a given voltage or power envelope, or alternatively, to use a little less power at the same clock speed and performance. It apparently tends to offer slightly higher max overclock speeds in addition to the slightly higher stock clock speeds. There are a few fringe things like integrated Thunderbolt 3 support, but those basically merit a "who cares?"
Thus, moving from Sky Lake to Kaby Lake is akin to moving from one bin of a CPU to a different bin of exactly the same CPU that is clocked 200 MHz or so higher. That's not to say that Kaby Lake is terrible in isolation. Given the choice between 200 MHz faster or slower for the same price, of course you want faster. But it does mean that Kaby Lake is a huge step forward, either; to the contrary, it's the smallest gains from a new generation of CPUs that Intel has offered in many years and possibly ever.
For clarity, what the review did was to set a Sky Lake CPU and a Kaby Lake CPU to exactly the same fixed clock speed of 4.5 GHz, with turbo boost and various power-saving features that alter clock speed disabled. Then they ran a bunch of benchmarks and found that the two CPUs performed the same.
Basically, the advantage of Kaby Lake over Sky Lake is the ability to clock a little higher at a given voltage or power envelope, or alternatively, to use a little less power at the same clock speed and performance. It apparently tends to offer slightly higher max overclock speeds in addition to the slightly higher stock clock speeds. There are a few fringe things like integrated Thunderbolt 3 support, but those basically merit a "who cares?"
Thus, moving from Sky Lake to Kaby Lake is akin to moving from one bin of a CPU to a different bin of exactly the same CPU that is clocked 200 MHz or so higher. That's not to say that Kaby Lake is terrible in isolation. Given the choice between 200 MHz faster or slower for the same price, of course you want faster. But it does mean that Kaby Lake is a huge step forward, either; to the contrary, it's the smallest gains from a new generation of CPUs that Intel has offered in many years and possibly ever.
Comments
By making it a new generation - exactly what Quiz has just mentioned happened: Intel has now orchestrated their least meaningful generational x86 update in recent memory, and possibly in their entire history.
It's not that Kaby is bad. It's just that Kaby doesn't have any competition.
http://www.anandtech.com/show/10907/amd-gives-more-zen-details-ryzen-34-ghz-nvme-neural-net-prediction-25-mhz-boost-steps
AMD just showed off Zen being competitive with a Core i7-6900K in both performance and power at the same time and at the same clock speed. Two big caveats are:
1) We don't know how cherry-picked that benchmark was.
2) We don't know how high Zen will clock with turbo.
AMD did say 3.4+ GHz, but there's a huge difference between a max turbo of 3.4 GHz that you only reach when you're only using a few cores and a 3.4 GHz base clock with turbo up to 4 GHz. If Zen can do the latter and the benchmark was a typical result for well-threaded programs, then Zen is going to be a hit. But that's two huge "if"s.
거북이는 목을 내밀 때 안 움직입니다
and thats at 95W TDP. No doubt they can release even higher clocked chip with 140W TDP to match Intels 140W TDP 6900k.
so, back to past again ?
It would explain all the leaked performance benchmarks for the past 2 months, where 7700k would be slower, hotter, and using more power at same clocks as 6700k.
Lets see how it looks like for kabylake so far :
0.1% or less IPC gain thats within a margin of error.
Official TDP 95w compared to 6700k 91w.
Mature 14nm slightly improving perf/watt ratio
Offset due to stronger GPU producing more heat using more power.
Offset due to higher clocks.
Offset due to worse TIM between chip and heat spreader.
Offset due to bad voltage regulation on z170 motherboards without multiple additional bios updates.
Yup it's looking real nice, kabylake compared to skylake is shaping up to be what ivy bridge was to sandy bridge.
Kaby is .... new IGP on the same CPU. So no one should be expecting anything different from the CPU.
Clock for clock, Kaby = Skylake. Same power, same IPC. Kaby gets you an extra 5% clock speed stock (so it tends to draw more power stock). Or for that crowd that absolutely refuses to believe that power draw is proportional to clock speed and only varies based on voltage - well you guys can all believe that power increase is from whatever you want to believe.
But in the GPU market they are again being trashed by the (severely overpriced) Nvidia chipsets.
Also: does AMD finally have the motherboards to support Ryzen fully?
Because if you still only have access to crappy motherboards compared to the Intel chipset ones Zen isn't even gonna pretend to be a contender.
"going into arguments with idiots is a lost cause, it requires you to stoop down to their level and you can't win"
I had a Radeon 7950 for years and it finally died (fan died and the GPU burned up). I ended up with an Nvidia 1060 - 6 GB model. The 2 are the same speed on games that I play and the NVidia drivers are the thing that is hampering their cards. AMD drivers are so much better.
One side getting their high end out before the other is nothing new. In April 2012, Nvidia wasn't competitive at all at the high end either because their new 28 nm cards weren't out, but that generation went decently well for them. At the start of 2010, the highest price point at which Nvidia was competitive was around $100.
Motherboard support hasn't been a problem for AMD in desktops for many years. There are a number of flagship motherboards available for Socket AM3+, and quite a few quality though not really flagship motherboards for Socket FM2+. It would be extremely shocking if there aren't plenty of good Socket AM4 motherboards available in time for Ryzen.
Brenics ~ Just to point out I do believe Chris Roberts is going down as the man who cheated backers and took down crowdfunding for gaming.
Intel doubled back, put an emphasis on efficiency, not just clock speed. And the Core brand came out of that. We ended up with faster CPUs even though they were running on lower clocks.
Better efficiency means you can make the fast chips faster, the low power chips lower power, and everything running cooler for better performance.
Right now, I think we are seeing a stalled CPU performance because of lack of need and competition. Most consumers and businesses don't need a faster CPU right now - the most CPU-taxing thing most people run will be Facebook in a browser, or Excel for work. Sure, a few people actually do intensive stuff regularly - programmers compiling large code bases, video encoders, etc - but those markets are definitely not big or market-driving. Most people (not necessarily those on this forum) are buying their computer at WalMart or Costco and, apart from remembering that catchy Intel sound byte from a few years ago (and maybe some commercials with Blue Man Group), the CPU in that computer is more or less inconsequential and they won't notice one from another.
All of the really taxing work is being shifted to GPU cores. And that's where you are seeing a lot of continued development.
Secondly, those leaked reports of 7700k over 5GHz are all at ridiculous voltages, and more often than not with "golden samples". Like the promised 5Ghz for skylake and before, it's highly likely that it's not going to be exactly like that (not 24/7, not stable, not 3D).
Thirdly using more power at same clocks is indicative of what kind of compared performance one can expect, and i fail to see how your vague power consumption comparison factors into this.
I distinctly remember remember a 5Ghz discussion about 6700k skylake before it came out, and also vaguely remember you yourself looking forward to have a 5Ghz cpu. And like 99% of everyone out there even with custom water cooling setups, neither you nor those 99% have stable 5Ghz 6700k, that were shown on some major tech websites a few weeks before 6700k release with golden sample engineering samples ( and prophesied on the forums of internet en masse)
https://www.pugetsystems.com/labs/articles/Delidding-the-4790K-for-a-quick-look-at-the-TIM-573/
At the level of entire CPU chips, how would you make a purely desktop focused chip different from what they build now? Well, no ECC support in the memory controller, for starters, and strip out any other server-centric features. Perhaps leave out the integrated GPU in some cases. But things like that only get you a smaller die and don't really gain you anything else. The reason Intel puts all that extra stuff in even though desktops don't need it is that it's cheaper on their end to build a "too big" die that can serve as a server, desktop, or laptop chip and selectively disable parts in various bins than it is to design separate dies for every market. So that smaller die really doesn't gain you anything at all.
http://wccftech.com/intel-core-i7-6700k-skylake-overclocked-52-ghz-air-cooling-135v/
And thats worrying, 7700k is barely hutting 5 GHz in these hype articles lol