I wonder if this is even good for overclockers. We would presume, on a "better" process, they got higher MHz at lower voltages (to stay within that 95W TDP envelope).
But...honestly.....the i5-6600K (rated at 91W) maxes out around 70W. Technically, Intel may have just pre-overclocked Skylake a smidgen.
If true, they'll run hotter than Skylake...because they just raised the voltage a smidgen, too.
Well, this is a win for non-overclockers, though. Intel finally gives them that 'free performance that overclockers enjoy.
My 6600k needs 1.44 Vcore at 4.6GHz to not bluescreen on the x264 stress test. It will also begin underclocking itself before the first pass finishes, as it hits 95C or so.
I have it in a small form factor build, with a Thermalright AXP-100R with the fan replaced by a very large 140mm*26mm fan. The heatsink is rated for 180W TDP. So the thermals on extremely CPU intensive workloads like x264 aren't good.
So yeah, I don't feel the TIM used under the integrated heat spreader is good at all. I've tried different thermal compounds over the integrated heat spreader, but none are effective at preventing the inevitable throttling that happens over a lengthy 100% utilized CPU stress test.
That being said, it'll only be used for gaming, and no games stress the CPU 100% of the time, at worst it'll be ~75C in game benchmarks, which is perfectly acceptable. So I'm OK with the CPU melting (but stable) in unrealistic workloads, since it's just fine and never throttles on actual realistic workloads.
Maybe you should delid and put liquid ultra. It decreased temps over 10C. Its pretty easy, i just did 6 chips in a row. Didn't really help my overclocks that much though.
You're talking about a laptop, you don't have to delid those (seeing as those processors come without a lid). You just changed the thermal paste as you would when installing a heatsink on a desktop. The problem with desktop processors is that you have a double layer of thermal paste, first between the actual chip and the lid, then between the lid and the heatsink. When Intel decided to cheap out on the paste between the chip and the lid, thermal performance has gone way down.
But anyway, it's true that Dell isn't known for using good thermal paste (and for applying it properly).
They need something easy to put on for thousands upon thousands of systems, fast and hard to screw up. Toothpaste fits the bill. Or you wouldn't see dozen of companies selling thermal paste, I mean such a niche product!
Overclocking headroom has steadily decreased since Sandy Bridge, in part due to higher default frequencies, and in part (until Devil's Canyon on Haswell) due to the heat spreader mating.
Even with tweaked fins, 14nm+ isn't going to yield additional overclocking headroom over Skylake (which was already a pretty mature 14nm process) but it will obviously give better default speed.
My Haswell i5-4690k maxed out at the same 4.5GHz my Sandy i5-2500k did. In comparison, I can't get my friends Skylake i5-6600k stable at even 4.4GHz on a really good cooler and motherboard, so just luck of the draw I guess, and kind of unfortunate when you consider the i5-6600k has an identical clock speed to the i5-4690k...
The IPC improvement, clock for clock, since Sandy Bridge hasn't been all that good up through Skylake. Sandy Bridge offered a good jump from Nehalem but since then the generational performance has been around 2% average, making a 4.5GHz i5-6600k about 8% faster than a 4.5GHz i5-2500k, albeit much more energy efficient on a newer platform.
And let's not ignore the elephant in the room here. The benefit of new Core chips are ALWAYS the platform. CPU performance simply hasn't been improving and isn't the selling point of a new PC. It's the I/O.
If my old X58 system had native USB 3.0, m.2, PCIe 3.0 and better energy efficiency, I'd still have a Core i7-950.
They don't call it Silcone Lottery for no reason. You can see the rarity of 4.7+ by the prices. I played with 5 delidded Skylakes 4700k and 4.6 was the max of one of them and thats with hefty bump of 1.375v. Temps don't reach much above 80C at full load on a Hyper 212.
I know they have improved the packaging. This is overall a fair upgrade to Skylake all and all. The GPU side of things isn't getting much love, seems few care about that.
Intel doesn't have particularly great GPU tech and it sharing 90-odd Watts of heat removal with the CPU doesn't give great performance. I don't understand the fuss about Iris -- it's useless.
On the bright side, as Kaby Lake doesn't add a lot of performance, lots of people that are currently running "good enough" processors won't be as compelled to upgrade. Those C2D and 1st/2nd generation Core iX CPUs remain relevant.
C2D is most certainly not relevant. It's missing many instruction sets and accelerated encryption. I recently put my last C2D system out of it's misery. A dual core Skylake at 3Ghz is the equivalent of a C2D at probably 5Ghz (which is obviously impossible). The 5-10% clock for clock increase per generation really adds up after 7+ generations. It's almost multiplicative at this point when compared to those old architectures. Not to mention the perf. per watt is off the charts these days by comparison.
Everyone complains, "oh it's just another 300mhz increase" but that +300mhz for Kaby Lake is the same as +350mhz for Skylake, +400mhz for Broadwell, +475mhz for Haswell, and so on, by the time you get back to Nehalem it's just laughable. (These of course aren't real numbers, it's just to make a point). 4ghz Kaby lake does not equal 4ghz Haswell. You simply can't compare clock speeds between generations.
No arguments about the incremental increases in performance adding up over time or about the added features that are useful to a lot of people. However, it does really depend on workloads. My daily use computer at home is a Gateway MA8 laptop with a C2D T2310 at 1.46GHz and it's perfectly adequate for the time being. Yup, I'm not going to push out any high end games on the old Intel x3100 graphics card and it takes about 25 seconds to boot up thanks to the 5400 RPM 500GB hard drive, but I don't see myself buying a replacement for at least another year. I could see wanting an upgrade if I was still using my Atom n270 Eee PC as a primary box. Web browsing is a pretty miserable experience on it, but the T2310 is more than enough for me.
You certainly can compare clock speeds between generations, that's how you determine IPC gains. Anandtech did a great roundup and found the averages in IPC improvement from Nehalem to Haswell around 12% (that's really 5 generations since Westmere and Clarkdale fit in between Nehalem and Sandy Bridge, not four, but who's getting technical) making an average gain of ~2.5%.
But as you said, that doesn't tell the whole story, because clock speeds have definitely scaled up (good luck getting Nehalem beyond 4GHz when 4.5GHz is a breeze for Devil's Canyon at nearly half the platform power consumption)
Also, there are tremendous spikes in performance improvements in a variety of key tests, specifically rendering and compression/decompression where Haswell leads Nehalem by 25%+
Gaming metrics are also important as prior to Ivy Bridge, PCIe 2.0 caused slightly less GPU throughput especially when calling system DRAM (which was also slower on older chipsets, especially X58)
Overall, IPC gains have been pretty disappointing, and actually on-par with AMD's IPC gains over the last few years. The difference is Intel started out significantly stronger with Core than AMD did with K10 back in 2008 (Deneb, etc) as clock for clock Intel was somewhere around 40% faster than the Phenom X4, and AMD was stuck on 45nm for years.
The relevance of chipset and platform features to the end user ultimately are workload dependent. In my case, an ancient C2D is does just fine because newer hardware doesn't offer any useful benefits.
"Usable" is definitely not the same as "relevant." I am a curator of old systems, and while I do have an Athlon 64 X2 system and a fairly tricked out Core 2 Quad system set up as League of Legends guest PCs (a role they perform admirably; an 8800 GTX can run LoL at 1080p maxed at over 100FPS, fyi), to expect either of them to keep up with even my wife's Haswell i3 as a daily driver is eyebrow raising.
My strapped-for-cash friend's 2500K, however... maybe Cannonlake for that one.
If that were true it'd be a 12% performance increase from a 5% clock speed increase. A non-linear performance increase would be pretty newsworthy. It also says 12% 'process performance supporting modern processors'. This is referring to the actual manufacturing process, not the computational performance of the processor.
Kaby Lake just sounds like a more fleshed out Skull Canyon re-release; higher base clock/turbo frequencies, but essentially IPC is the same as the previous "gen".
I'm at least hoping for better thermal interface than what we got for Skylake... Going back to real flux solder like from the 2500k/2600k days would be nice, Intel.
Most propably there is still bad toothpaste I there and Intel will release skull Canton 2 if Zen is really good. If it is not, Intel can save 0.01 penny for each prosessor by using substandard paste in there... and qurantee that the paste is no good after the warrant is over...
I'm not so sure about Zen. Yes, they hit their performance targets internally at AMD, but the rumoured clock speeds and the fact that Zen+ was announced pretty much at the same time as Zen hints at Zen+ being a refined Zen, the Zen we're hoping to see but might not to begin with due to process immaturity plus a few performance tweaks that won't make the original design. Something akin to getting Phenom instead of Phenom II, though not anywhere near as bad. It'll be a significant leap for AMD and the low 3GHz clock range probably makes sense for the 8C/16T variants but I can imagine that the 4C/8T variants may be a tad underwhelming.
Look at me, I'm beating up on Zen already, and it's nowhere near release.
its not about the money for intel, it's about the environmental impact. Millions upon millions of grams of solder going into landfills someday is not good.
Thats basically what they are supposed to be in Intels new 3-stage cycle.
Process -> Architecture -> Optimization, and we're in the last step with Kaby Lake, so its basically just an optimized Skylake, and Cannonlake will be the 10nm shrink of that, restarting the cycle.
Rumors have it there is an intermediate step even in there, Coffee Lake, because 10nm is not quite ready yet, but we'll have to wait and see.
Only after Cannonlake with Icelake do we actually get a new architecture again.
I have to wonder how much scope for improvement through architecture there is. It's not just the rate of node shrinkage which has slowed. Apple is clearly extracting more out of ARM, but Intel has been evolving Core for ten years now.
The issue for intel, as Apple will find soon enough, is the easy gains are well gone, and it keeps getting harder and more expensive to improve architecture time after time.
Pentium G4620 is 2C/4T, isn't it? Most publications, including the chinese one linked here, says it is. It should be a very interesting chip, similar to i3 6100 and able to do well enough in gaming.
2cores are not enough these days for gaming, specially console ports. You maybe see high fps. but also see the game freezing from time to time when it can't access more cpu resources regardless of OC.
Athlon X4 860K > any skylake i3 in the new tomb raider, less max fps but fluid experience, when the game is out of cpu resources the i3 will make the fps go to the low 20's + stuttering while the 860K maitain a minimum of 45-50fps.
I have an 860K and I must respectfully disagree with you that it's a better alternative based on my experiences with it in the past few months. Mine is running on a Gigabyte A68H micro ATX board with 16GB of 2133 RAM and it replaced a Xeon 3065 (dual core 2.33GHz LGA775) with 4GB of DDR2 800. To say I'm underwhelmed is a serious understatement and it created so much heat over the summer that I shoved it out to my porch and now use it as a headless box to stream games to my laptop (which is almost never since the gaming experience on it is pretty lackluster).
I have an i3 6100 with a R9 380 and 8gb 2133 ddr4. I have noticed a few occasional dips in Rise of the Tomb Raider under DX11 but since they have gotten DX12 smoothed out frame rates have increased greatly. DX 12 on RotTR really gives a major boost to fluidity and frame rates when running on an i3 with AMD graphics.
You can't overclock an 860K enough to bring it to parity with a Skylake i3 when it comes to gaming. A quick check over at cpuboss shows a crushing 58% advantage for the i3 in single core performance so the small SMT threads on the i3 aren't that much slower than a stock 860K thread.
The sleeper in that batch seems to me to be the i3-7300. Dual core with HT, and 4.0ghz base. If they ever released an unlocked multiplier version or at least allowed base clock adjustments (damn them for locking down the i3 Skylakes after the fact), they'd have a huge hit on their hands.
I guess that's what happens with no competition. Let's hope for some AMD success in a few months. I can almost guarantee that they'll unlock their lower binned parts.
These days all the dual cores have a very high base clock.
Lots of idiot readers like to knock mobile-class dual-core HT i7's as not being "i7" worthy, but given that typical users do nothing more intensive than watch YouTube videos and make awfully egotistical posts on Tumblr and Facebook, then yeah, a dual-core with HT and that much higher base clock and lower TDP than a lower clocked quad core is actually a good match for the workload on that system.
Consider that a relatively meager percentage of PC sales are actually used for somewhat intensive applications (such as games), and that most are used for mundane tasks, so yes, even in 2016 a dual-core is actually fine for most users.
The issue with the 15W i7, dual-core/HT is, there's no appreciable difference between the i5 and i7. The difference between i5-6300U and i7-6500U is 100MHz and 1 MB of cache. They are both dual-core, they both have HT, yet there's a 40% (tray price) premium on the i7 ($393 vs $281).
I would think as a matter of branding, Intel would like to separate the i7 and i5 more than that. All this is, is marketing. Kinda like tuning a car to run 0-60 in 7.8 seconds instead of 8.0 seconds, calling is "sport edition" or "SS", mark it up a couple grand and then justify yourself by saying "well the way most people drive cars, it's a good match for the majority of drivers out there".
No word on whether or not any of these will have the new Iris Pro+ HD620 IGP with Crystal Well GT3e eDRAM/L4. Ok. so obviously I have no idea what the latest and greates Intel IGP is really called, nor it's exact feature set, that's because nowhere have I seen mention of the terms HD 620 and GT3e together.
So far all the specs I've seen call for an HD 620 GT2. Why not just put the best available as thermal is not an issue in a desktop? (and NO, I don't want to purchase a discrete video card).
You're the idiot, minijedimaster. Not everyone uses and treats PCs as upgradable game consoles. People like you make me sick, and you should be shot in the head. If PC gaming ever dies, we'll still be here and you'll be jerking off in some corner like an idiot to some game.
He just wants the best freebie IGP he can get with a fast CPU. I'm looking at the 7700K for the same reason.
another lame update for the imbeciles who think they need the latest.
intels baby steps when it comes to performance is really annoying.
lets hope AMD´s ZEN can deliver and will not be another bulldozer crap. not that i will buy AMD but when it forces intel to really push it.... than i wish AMD all the best.
Gonna have to agree here. The baby steps forward and charging full price for essentially the same as last gen is pretty appalling, but that's what's allowed when there's little competition.
Here's to AMD for bringing that competition back and putting Intel's feet over the fire.
With any luck AMD will be able to bring an unlocked 4C/8T CPU with an average IPC between Haswell and Broadwell at the $199 price point. This would force Intel to drop Their Skylake range of i5's and i7's down a lot to compete with Zen. A Skylake i5 6600K at $169 or an i7 6700K at $239 would be great for customers as Intel will surely keep Kaby Lake prices high.
It doesn't matter to me though as I'm grabbing a Zen CPU and Vega GPU so I can build the all AMD system I've been wanting for years. It really irritated me that my last PC build had to have an Intel CPU but I wasn't throwing my money away on a 4 year old AMD CPU that couldn't come close to even competing with an i3 6100 on a performance per dollar basis.
My oldest CPU is a C2D T7200 inside a notebook. My newest a Skylake i5-6600K. Im fine with both of those and i wont by a new kaby lake soon. Actually my next money dump is a 1 TB samsung SSD and a new digital camera.
I'm very eager to see what clockspeed improvements Kaby Lake gives with 35w and 45w mobile CPUs. So far not leaked yet. Could settle for a MSI GS73VR with Kaby if a bit quicker than a 6700HQ. Assuming not more than 200MHZ more. .. But may want to wait for Coffee Lake, the 3rd iteration of 14nm that is focused more on performance than Cannon Lake (Coffee and Cannon releases around the same time a year from now -ish).
I doubt you'll see a 12% gain in a .200mhz increase in clock speed. Sorry, not buying it. The 4790k was and is still the best cpu to own atm with a base clock of 4.0 turbo'ing to 4.4ghz stock, with the great TIM used, the DC is still far superior than Skylake and Kaby lake CPUS.
It will not be any good at all My i7 6700k is also not able to overclock more than 100 mhz when i try it crashes horrible So i am stuck at the boost speed 4.2 Ghz thats it if i add the 100 mhz the temps fly skyhigh. Thats why i do not even dare pump up the voltage after 1 try it became clear that it never can oc. It needed 1.46 volt to stay steady at 4.3 ghz I bet the 7700 will be exactly the same, and i am confident i will not buy any i7 for my systems from now on it will be either i5 or another AMD. The old i5 6600k is en was quicker in almost everything, and that was an easy overclocker to 4.7 Ghz it runs circles around the i7 ..... Really expect nothing special from this release Ofcourse the hardware sites will get payed to give a good review as usual, but for me no upgrade for coming year unless amd is releasing something good, but i do not expect much of the new zen. More threats is useless for most application and will not make a huge difference for gaming either.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
72 Comments
Back to Article
Lolimaster - Monday, October 31, 2016 - link
12% performance increas is just from the increased clocks, this things don't even deserve news.ikjadoon - Monday, October 31, 2016 - link
I wonder if this is even good for overclockers. We would presume, on a "better" process, they got higher MHz at lower voltages (to stay within that 95W TDP envelope).But...honestly.....the i5-6600K (rated at 91W) maxes out around 70W. Technically, Intel may have just pre-overclocked Skylake a smidgen.
If true, they'll run hotter than Skylake...because they just raised the voltage a smidgen, too.
Well, this is a win for non-overclockers, though. Intel finally gives them that 'free performance that overclockers enjoy.
saratoga4 - Monday, October 31, 2016 - link
>Technically, Intel may have just pre-overclocked Skylake a smidgen.FWIW, my Skylake system hit 4.5 GHz ... at stock voltage.
I wonder if they've improved the TIM like on Devil's Canyon.
etamin - Monday, October 31, 2016 - link
Luck of the draw. My 6700K needs 1.344V to hit 4.5GHz and refuses to go any higher no matter the voltage.JoeyJoJo123 - Monday, October 31, 2016 - link
My 6600k needs 1.44 Vcore at 4.6GHz to not bluescreen on the x264 stress test. It will also begin underclocking itself before the first pass finishes, as it hits 95C or so.I have it in a small form factor build, with a Thermalright AXP-100R with the fan replaced by a very large 140mm*26mm fan. The heatsink is rated for 180W TDP. So the thermals on extremely CPU intensive workloads like x264 aren't good.
So yeah, I don't feel the TIM used under the integrated heat spreader is good at all. I've tried different thermal compounds over the integrated heat spreader, but none are effective at preventing the inevitable throttling that happens over a lengthy 100% utilized CPU stress test.
That being said, it'll only be used for gaming, and no games stress the CPU 100% of the time, at worst it'll be ~75C in game benchmarks, which is perfectly acceptable. So I'm OK with the CPU melting (but stable) in unrealistic workloads, since it's just fine and never throttles on actual realistic workloads.
Byte - Monday, October 31, 2016 - link
Maybe you should delid and put liquid ultra. It decreased temps over 10C. Its pretty easy, i just did 6 chips in a row. Didn't really help my overclocks that much though.Notmyusualid - Tuesday, November 1, 2016 - link
1 for that.My Alienware 18 had been throttling constantly out in the Asian heat - if I even mentioned the word overclock. (Tmax for 4940XM is a mere 100C)
Added Collabrotory Liquid Ultra, max temp is 85 - 87C now - and my basclock is upped to 4.1GHz too!
Honestly that much of a difference. I wonder if the Dell tech was using toothpaste, (or more likely just had a bad batch of thermal insulaton paste?).
Macpoedel - Tuesday, November 1, 2016 - link
You're talking about a laptop, you don't have to delid those (seeing as those processors come without a lid). You just changed the thermal paste as you would when installing a heatsink on a desktop. The problem with desktop processors is that you have a double layer of thermal paste, first between the actual chip and the lid, then between the lid and the heatsink. When Intel decided to cheap out on the paste between the chip and the lid, thermal performance has gone way down.But anyway, it's true that Dell isn't known for using good thermal paste (and for applying it properly).
Byte - Tuesday, November 8, 2016 - link
They need something easy to put on for thousands upon thousands of systems, fast and hard to screw up. Toothpaste fits the bill. Or you wouldn't see dozen of companies selling thermal paste, I mean such a niche product!colinstalter - Monday, November 7, 2016 - link
I am able to hit 4.7GHz on my 6600k at stock voltages. I use a corsair single-fan liquid cooler and have yet to get over 55-60 C during stress tests.StevoLincolnite - Monday, October 31, 2016 - link
My 3930K does 5ghz happily... And would give these Kaby Lake chips a good run for it's money, despite being 5 years old.Samus - Tuesday, November 1, 2016 - link
Overclocking headroom has steadily decreased since Sandy Bridge, in part due to higher default frequencies, and in part (until Devil's Canyon on Haswell) due to the heat spreader mating.Even with tweaked fins, 14nm+ isn't going to yield additional overclocking headroom over Skylake (which was already a pretty mature 14nm process) but it will obviously give better default speed.
My Haswell i5-4690k maxed out at the same 4.5GHz my Sandy i5-2500k did. In comparison, I can't get my friends Skylake i5-6600k stable at even 4.4GHz on a really good cooler and motherboard, so just luck of the draw I guess, and kind of unfortunate when you consider the i5-6600k has an identical clock speed to the i5-4690k...
The IPC improvement, clock for clock, since Sandy Bridge hasn't been all that good up through Skylake. Sandy Bridge offered a good jump from Nehalem but since then the generational performance has been around 2% average, making a 4.5GHz i5-6600k about 8% faster than a 4.5GHz i5-2500k, albeit much more energy efficient on a newer platform.
And let's not ignore the elephant in the room here. The benefit of new Core chips are ALWAYS the platform. CPU performance simply hasn't been improving and isn't the selling point of a new PC. It's the I/O.
If my old X58 system had native USB 3.0, m.2, PCIe 3.0 and better energy efficiency, I'd still have a Core i7-950.
Meteor2 - Tuesday, November 1, 2016 - link
That's being a bit disingenuous; generational IPC improvements have been more than 5%, compounding to 25% from Sandy Bridge to Skylake.Byte - Tuesday, November 8, 2016 - link
They don't call it Silcone Lottery for no reason. You can see the rarity of 4.7+ by the prices. I played with 5 delidded Skylakes 4700k and 4.6 was the max of one of them and thats with hefty bump of 1.375v. Temps don't reach much above 80C at full load on a Hyper 212.ryrynz - Tuesday, November 1, 2016 - link
I know they have improved the packaging. This is overall a fair upgrade to Skylake all and all.The GPU side of things isn't getting much love, seems few care about that.
Meteor2 - Tuesday, November 1, 2016 - link
Intel doesn't have particularly great GPU tech and it sharing 90-odd Watts of heat removal with the CPU doesn't give great performance. I don't understand the fuss about Iris -- it's useless.Morawka - Monday, October 31, 2016 - link
not necessarily, they could increase performance without increasing power if the improved fin pitch is to be belived.BrokenCrayons - Monday, October 31, 2016 - link
On the bright side, as Kaby Lake doesn't add a lot of performance, lots of people that are currently running "good enough" processors won't be as compelled to upgrade. Those C2D and 1st/2nd generation Core iX CPUs remain relevant.Ej24 - Monday, October 31, 2016 - link
C2D is most certainly not relevant. It's missing many instruction sets and accelerated encryption. I recently put my last C2D system out of it's misery. A dual core Skylake at 3Ghz is the equivalent of a C2D at probably 5Ghz (which is obviously impossible). The 5-10% clock for clock increase per generation really adds up after 7+ generations. It's almost multiplicative at this point when compared to those old architectures. Not to mention the perf. per watt is off the charts these days by comparison.Everyone complains, "oh it's just another 300mhz increase" but that +300mhz for Kaby Lake is the same as +350mhz for Skylake, +400mhz for Broadwell, +475mhz for Haswell, and so on, by the time you get back to Nehalem it's just laughable. (These of course aren't real numbers, it's just to make a point). 4ghz Kaby lake does not equal 4ghz Haswell. You simply can't compare clock speeds between generations.
BrokenCrayons - Monday, October 31, 2016 - link
No arguments about the incremental increases in performance adding up over time or about the added features that are useful to a lot of people. However, it does really depend on workloads. My daily use computer at home is a Gateway MA8 laptop with a C2D T2310 at 1.46GHz and it's perfectly adequate for the time being. Yup, I'm not going to push out any high end games on the old Intel x3100 graphics card and it takes about 25 seconds to boot up thanks to the 5400 RPM 500GB hard drive, but I don't see myself buying a replacement for at least another year. I could see wanting an upgrade if I was still using my Atom n270 Eee PC as a primary box. Web browsing is a pretty miserable experience on it, but the T2310 is more than enough for me.kitty4427 - Tuesday, November 1, 2016 - link
From what I've seen, there's been minimal improvement in performance-per-clock for the last few generations.Samus - Tuesday, November 1, 2016 - link
You certainly can compare clock speeds between generations, that's how you determine IPC gains. Anandtech did a great roundup and found the averages in IPC improvement from Nehalem to Haswell around 12% (that's really 5 generations since Westmere and Clarkdale fit in between Nehalem and Sandy Bridge, not four, but who's getting technical) making an average gain of ~2.5%.But as you said, that doesn't tell the whole story, because clock speeds have definitely scaled up (good luck getting Nehalem beyond 4GHz when 4.5GHz is a breeze for Devil's Canyon at nearly half the platform power consumption)
Also, there are tremendous spikes in performance improvements in a variety of key tests, specifically rendering and compression/decompression where Haswell leads Nehalem by 25%+
Gaming metrics are also important as prior to Ivy Bridge, PCIe 2.0 caused slightly less GPU throughput especially when calling system DRAM (which was also slower on older chipsets, especially X58)
Overall, IPC gains have been pretty disappointing, and actually on-par with AMD's IPC gains over the last few years. The difference is Intel started out significantly stronger with Core than AMD did with K10 back in 2008 (Deneb, etc) as clock for clock Intel was somewhere around 40% faster than the Phenom X4, and AMD was stuck on 45nm for years.
etamin - Monday, October 31, 2016 - link
The chipset changes alone since C2D make an upgrade very relevant.BrokenCrayons - Tuesday, November 1, 2016 - link
The relevance of chipset and platform features to the end user ultimately are workload dependent. In my case, an ancient C2D is does just fine because newer hardware doesn't offer any useful benefits.VoraciousGorak - Monday, October 31, 2016 - link
"Usable" is definitely not the same as "relevant." I am a curator of old systems, and while I do have an Athlon 64 X2 system and a fairly tricked out Core 2 Quad system set up as League of Legends guest PCs (a role they perform admirably; an 8800 GTX can run LoL at 1080p maxed at over 100FPS, fyi), to expect either of them to keep up with even my wife's Haswell i3 as a daily driver is eyebrow raising.My strapped-for-cash friend's 2500K, however... maybe Cannonlake for that one.
VoraciousGorak - Monday, October 31, 2016 - link
Or even later, if Cannonlake is going to be ULV processors only as rumored....Dritman - Monday, November 7, 2016 - link
If that were true it'd be a 12% performance increase from a 5% clock speed increase. A non-linear performance increase would be pretty newsworthy. It also says 12% 'process performance supporting modern processors'. This is referring to the actual manufacturing process, not the computational performance of the processor.JoeyJoJo123 - Monday, October 31, 2016 - link
Kaby Lake just sounds like a more fleshed out Skull Canyon re-release; higher base clock/turbo frequencies, but essentially IPC is the same as the previous "gen".I'm at least hoping for better thermal interface than what we got for Skylake... Going back to real flux solder like from the 2500k/2600k days would be nice, Intel.
haukionkannel - Monday, October 31, 2016 - link
Most propably there is still bad toothpaste I there and Intel will release skull Canton 2 if Zen is really good. If it is not, Intel can save 0.01 penny for each prosessor by using substandard paste in there... and qurantee that the paste is no good after the warrant is over...silverblue - Monday, October 31, 2016 - link
I'm not so sure about Zen. Yes, they hit their performance targets internally at AMD, but the rumoured clock speeds and the fact that Zen+ was announced pretty much at the same time as Zen hints at Zen+ being a refined Zen, the Zen we're hoping to see but might not to begin with due to process immaturity plus a few performance tweaks that won't make the original design. Something akin to getting Phenom instead of Phenom II, though not anywhere near as bad. It'll be a significant leap for AMD and the low 3GHz clock range probably makes sense for the 8C/16T variants but I can imagine that the 4C/8T variants may be a tad underwhelming.Look at me, I'm beating up on Zen already, and it's nowhere near release.
Lolimaster - Monday, October 31, 2016 - link
Zen+ will go with the 7nm process. Just because of it it will be a huge evolutionary path, specially for APU's2048SP Zen+ APU in 2019?
Michael Bay - Wednesday, November 2, 2016 - link
Both must first exist outside of marketing.Morawka - Monday, October 31, 2016 - link
its not about the money for intel, it's about the environmental impact. Millions upon millions of grams of solder going into landfills someday is not good.JoeyJoJo123 - Monday, October 31, 2016 - link
And millions upon millions of grams of thermal interface (grease, usually) is any better?The solder flux was leadless, so it's not like they were sending lead into landfills with their products.
nevcairiel - Monday, October 31, 2016 - link
Thats basically what they are supposed to be in Intels new 3-stage cycle.Process -> Architecture -> Optimization, and we're in the last step with Kaby Lake, so its basically just an optimized Skylake, and Cannonlake will be the 10nm shrink of that, restarting the cycle.
Rumors have it there is an intermediate step even in there, Coffee Lake, because 10nm is not quite ready yet, but we'll have to wait and see.
Only after Cannonlake with Icelake do we actually get a new architecture again.
Meteor2 - Monday, October 31, 2016 - link
I have to wonder how much scope for improvement through architecture there is. It's not just the rate of node shrinkage which has slowed. Apple is clearly extracting more out of ARM, but Intel has been evolving Core for ten years now.TheinsanegamerN - Monday, October 31, 2016 - link
Intel also has not had solid competition for half a decade. that may slow things down a bit.doggface - Monday, October 31, 2016 - link
The issue for intel, as Apple will find soon enough, is the easy gains are well gone, and it keeps getting harder and more expensive to improve architecture time after time.Kalelovil - Monday, October 31, 2016 - link
It is in many ways the old Core 2 Quad G0 stepping or Core i7 9xx D0 stepping revisions, but this time with new model numbers.ZeDestructor - Tuesday, November 1, 2016 - link
> Going back to real flux solder like from the 2500k/2600k days would be nice, Intel.They physically can't. If they do, too many dies crack from repeated thermal shock because of how physically small (mm²) they are.
ZeDestructor - Tuesday, November 1, 2016 - link
If you really want your fluxless solder die-attach tech, then pony up for the big socket platform.BOMBOVA - Monday, October 31, 2016 - link
Not real news, the fibs on this were coming out 3 months ago. unfortunate for power computer users.neblogai - Monday, October 31, 2016 - link
Pentium G4620 is 2C/4T, isn't it? Most publications, including the chinese one linked here, says it is. It should be a very interesting chip, similar to i3 6100 and able to do well enough in gaming.Lolimaster - Monday, October 31, 2016 - link
2cores are not enough these days for gaming, specially console ports. You maybe see high fps. but also see the game freezing from time to time when it can't access more cpu resources regardless of OC.Athlon X4 860K > any skylake i3 in the new tomb raider, less max fps but fluid experience, when the game is out of cpu resources the i3 will make the fps go to the low 20's + stuttering while the 860K maitain a minimum of 45-50fps.
BrokenCrayons - Monday, October 31, 2016 - link
I have an 860K and I must respectfully disagree with you that it's a better alternative based on my experiences with it in the past few months. Mine is running on a Gigabyte A68H micro ATX board with 16GB of 2133 RAM and it replaced a Xeon 3065 (dual core 2.33GHz LGA775) with 4GB of DDR2 800. To say I'm underwhelmed is a serious understatement and it created so much heat over the summer that I shoved it out to my porch and now use it as a headless box to stream games to my laptop (which is almost never since the gaming experience on it is pretty lackluster).SquarePeg - Monday, October 31, 2016 - link
I have an i3 6100 with a R9 380 and 8gb 2133 ddr4. I have noticed a few occasional dips in Rise of the Tomb Raider under DX11 but since they have gotten DX12 smoothed out frame rates have increased greatly. DX 12 on RotTR really gives a major boost to fluidity and frame rates when running on an i3 with AMD graphics.You can't overclock an 860K enough to bring it to parity with a Skylake i3 when it comes to gaming. A quick check over at cpuboss shows a crushing 58% advantage for the i3 in single core performance so the small SMT threads on the i3 aren't that much slower than a stock 860K thread.
vFunct - Monday, October 31, 2016 - link
Which ones are likely to go into the next iMac?TheinsanegamerN - Monday, October 31, 2016 - link
didn't you hear? Apple is going to make a "courageous" decision and use the A10X in the next iMAC.Lolimaster - Monday, October 31, 2016 - link
AMD Raven Ridge APU is going to the iMACs next year.bill.rookard - Monday, October 31, 2016 - link
The sleeper in that batch seems to me to be the i3-7300. Dual core with HT, and 4.0ghz base. If they ever released an unlocked multiplier version or at least allowed base clock adjustments (damn them for locking down the i3 Skylakes after the fact), they'd have a huge hit on their hands.I guess that's what happens with no competition. Let's hope for some AMD success in a few months. I can almost guarantee that they'll unlock their lower binned parts.
damianrobertjones - Monday, October 31, 2016 - link
"3-7300. Dual core with HT, and 4.0ghz base"I also read that and thought, "Does anyone need anything else?" It's simply fantastic that an 'i3' reaches 4.0Ghz base.
JoeyJoJo123 - Monday, October 31, 2016 - link
These days all the dual cores have a very high base clock.Lots of idiot readers like to knock mobile-class dual-core HT i7's as not being "i7" worthy, but given that typical users do nothing more intensive than watch YouTube videos and make awfully egotistical posts on Tumblr and Facebook, then yeah, a dual-core with HT and that much higher base clock and lower TDP than a lower clocked quad core is actually a good match for the workload on that system.
Consider that a relatively meager percentage of PC sales are actually used for somewhat intensive applications (such as games), and that most are used for mundane tasks, so yes, even in 2016 a dual-core is actually fine for most users.
jaydee - Tuesday, November 1, 2016 - link
The issue with the 15W i7, dual-core/HT is, there's no appreciable difference between the i5 and i7. The difference between i5-6300U and i7-6500U is 100MHz and 1 MB of cache. They are both dual-core, they both have HT, yet there's a 40% (tray price) premium on the i7 ($393 vs $281).I would think as a matter of branding, Intel would like to separate the i7 and i5 more than that. All this is, is marketing. Kinda like tuning a car to run 0-60 in 7.8 seconds instead of 8.0 seconds, calling is "sport edition" or "SS", mark it up a couple grand and then justify yourself by saying "well the way most people drive cars, it's a good match for the majority of drivers out there".
HardwareDufus - Monday, October 31, 2016 - link
No word on whether or not any of these will have the new Iris Pro+ HD620 IGP with Crystal Well GT3e eDRAM/L4. Ok. so obviously I have no idea what the latest and greates Intel IGP is really called, nor it's exact feature set, that's because nowhere have I seen mention of the terms HD 620 and GT3e together.So far all the specs I've seen call for an HD 620 GT2. Why not just put the best available as thermal is not an issue in a desktop? (and NO, I don't want to purchase a discrete video card).
minijedimaster - Monday, October 31, 2016 - link
You're an idiotHardwareDufus - Tuesday, November 1, 2016 - link
Why thank you. And all along I was just striving for dufus.Flying Aardvark - Saturday, December 3, 2016 - link
You're the idiot, minijedimaster. Not everyone uses and treats PCs as upgradable game consoles. People like you make me sick, and you should be shot in the head. If PC gaming ever dies, we'll still be here and you'll be jerking off in some corner like an idiot to some game.He just wants the best freebie IGP he can get with a fast CPU. I'm looking at the 7700K for the same reason.
zodiacsoulmate - Monday, October 31, 2016 - link
my 6700k can hit 4.4ghz at -55mv... is it even good my 3770k can get to 5.2ghz...Gothmoth - Monday, October 31, 2016 - link
another lame update for the imbeciles who think they need the latest.intels baby steps when it comes to performance is really annoying.
lets hope AMD´s ZEN can deliver and will not be another bulldozer crap.
not that i will buy AMD but when it forces intel to really push it.... than i wish AMD all the best.
JoeyJoJo123 - Monday, October 31, 2016 - link
Gonna have to agree here. The baby steps forward and charging full price for essentially the same as last gen is pretty appalling, but that's what's allowed when there's little competition.Here's to AMD for bringing that competition back and putting Intel's feet over the fire.
SquarePeg - Tuesday, November 1, 2016 - link
With any luck AMD will be able to bring an unlocked 4C/8T CPU with an average IPC between Haswell and Broadwell at the $199 price point. This would force Intel to drop Their Skylake range of i5's and i7's down a lot to compete with Zen. A Skylake i5 6600K at $169 or an i7 6700K at $239 would be great for customers as Intel will surely keep Kaby Lake prices high.It doesn't matter to me though as I'm grabbing a Zen CPU and Vega GPU so I can build the all AMD system I've been wanting for years. It really irritated me that my last PC build had to have an Intel CPU but I wasn't throwing my money away on a 4 year old AMD CPU that couldn't come close to even competing with an i3 6100 on a performance per dollar basis.
ryrynz - Tuesday, November 1, 2016 - link
Multicore performance will likely be quite strong but single core? Also games? Don't expect much.powerarmour - Monday, October 31, 2016 - link
Intel are still peddling the same old 4C/4T or 4C/8T desktop CPU's, why not give us an affordable 8C/8T CPU that doesn't need a server chipset?Seriously, it's all just getting stale these days. If only AMD can be competitive again to end this monotonous BS.
eyk - Monday, October 31, 2016 - link
Better a 8C/16T like Zen will bring.ryrynz - Tuesday, November 1, 2016 - link
You're forgetting how little extra cores sometimes help performance wise..Also single threaded performance will likely be considerably behind Intel.
ryrynz - Tuesday, November 1, 2016 - link
Coffee Lake (6 Cores) is coming to mainstream, It's Intel's answer to Zen's eight cores.peterwhitehouse - Tuesday, November 1, 2016 - link
X299 is the new HEDT chipset for Skylake-E and current roadmaps put it around Q3 2017.jeffry - Saturday, November 5, 2016 - link
My oldest CPU is a C2D T7200 inside a notebook. My newest a Skylake i5-6600K. Im fine with both of those and i wont by a new kaby lake soon. Actually my next money dump is a 1 TB samsung SSD and a new digital camera.danwat1234 - Sunday, November 6, 2016 - link
I'm very eager to see what clockspeed improvements Kaby Lake gives with 35w and 45w mobile CPUs. So far not leaked yet. Could settle for a MSI GS73VR with Kaby if a bit quicker than a 6700HQ. Assuming not more than 200MHZ more. .. But may want to wait for Coffee Lake, the 3rd iteration of 14nm that is focused more on performance than Cannon Lake (Coffee and Cannon releases around the same time a year from now -ish).Byte - Tuesday, November 8, 2016 - link
Any word if the 200 series has native usb 3.1? Waiting to upgrade my Ivy Bridge and waiting for that. Really like Skylake.SeanJ76 - Tuesday, November 8, 2016 - link
I doubt you'll see a 12% gain in a .200mhz increase in clock speed. Sorry, not buying it. The 4790k was and is still the best cpu to own atm with a base clock of 4.0 turbo'ing to 4.4ghz stock, with the great TIM used, the DC is still far superior than Skylake and Kaby lake CPUS.bronan - Thursday, November 10, 2016 - link
It will not be any good at allMy i7 6700k is also not able to overclock more than 100 mhz when i try it crashes horrible
So i am stuck at the boost speed 4.2 Ghz thats it if i add the 100 mhz the temps fly skyhigh.
Thats why i do not even dare pump up the voltage after 1 try it became clear that it never can oc.
It needed 1.46 volt to stay steady at 4.3 ghz
I bet the 7700 will be exactly the same, and i am confident i will not buy any i7 for my systems from now on it will be either i5 or another AMD.
The old i5 6600k is en was quicker in almost everything, and that was an easy overclocker to 4.7 Ghz it runs circles around the i7 .....
Really expect nothing special from this release
Ofcourse the hardware sites will get payed to give a good review as usual, but for me no upgrade for coming year unless amd is releasing something good, but i do not expect much of the new zen. More threats is useless for most application and will not make a huge difference for gaming either.