Early Core i7-7700K benchmarks suggest minor speed gains over Skylake

[Scorpus]

Intel aren't expected to formally announce their Kaby Lake desktop processors until early next year, but somehow Tom's Hardware have received and benchmarked a Core i7-7700K more than a month early. Their findings indicate that Kaby Lake won't be a significant upgrade over Skylake, at least on desktop platforms.

The Core i7-7700K is clocked at 4.2 GHz with a boost clock of 4.5 GHz, which is a small 200-300 MHz clock speed gain on the Skylake Core i7-6700K. The processor also comes with improved Intel HD 630 graphics, and the same four cores and eight threads as its predecessor. The chip's TDP has been increased from 91W to 95W to account for the clock speed gains, noting that Kaby Lake is built on largely the same 14nm process.

Tom's Hardware's testing revealed that the Core i7-7700K is only 3.6% faster than the Core i7-6700K across the benchmarks they performed. This is slightly lower than the 5-7% improvement in clock speeds, and due to the increased power draw that Tom's Hardware recorded, Intel's flagship Kaby Lake chip is actually less efficient than its Skylake predecessor.

Tom's also spent some time overclocking the i7-7700K, and discovered that while the Kaby Lake sample they received did overclock slightly higher than their Skylake CPU (4.78 GHz versus 4.59 GHz), it produced a lot more heat in the process. The i7-7700K ran at 82°C above ambient while overclocked, which equated to a recorded temperature of a whopping 97°C. The i7-6700K, in contrast, ran at just 60°C above ambient (75°C).

As it was unclear whether Tom's Hardware received an engineering sample or a retail Core i7-7700K, these performance characteristics may not carry over to units that will actually go on sale next year. You'll just have to wait until we get our hands on Kaby Lake in the coming months.

Permalink to story.

https://www.techspot.com/news/67224-early-core-i7-7700k-benchmarks-suggest-minor-speed.html

 

[Kunming]

In other news, water is wet.

 

[Uncle Al]

Have we finally reached the upper limits of speed on these processors? It wasn't all that long ago that the claims were that all silicone based processors had reached their limits for speed and heat......or was it all just a bad dream?

 

[Nobina]

Have we finally reached the upper limits of speed on these processors? It wasn't all that long ago that the claims were that all silicone based processors had reached their limits for speed and heat......or was it all just a bad dream?

It could be because AMD isn't giving them competition so they have no reason to make faster processors. If you look at Nvidia, they are doing the same thing.

 

[MoeJoe]

Too many iterations from Intel.

Time to get more efficient ...

 

[theBest11778]

Have we finally reached the upper limits of speed on these processors? It wasn't all that long ago that the claims were that all silicone based processors had reached their limits for speed and heat......or was it all just a bad dream?

It could be because AMD isn't giving them competition so they have no reason to make faster processors. If you look at Nvidia, they are doing the same thing.

Doubt that. Intel is genuinely worried about Zen, but it seems their engineers and MFG process progression have stalled. This happens to companies that have gone unchecked for awhile. Considering the Kaby lake Architecture isn't much different than Nahlem and pretty much the same as Sandy Bridge Intel hasn't produced a new CPU architecture in over 5 years. Remember Intel did well with the core 2 before the Core series, but 1 architecture before that... Netburst... So 2 successful architectures in a row doesn't mean they're really as far ahead as you think.

On the GPU side AMD needs to step away for a few years and let Nvidia discourage their own customers with massive price hikes (similar to what we saw this year.) Reemerge a few years later when people are begging for another GPU option.

AMD is in a good spot right now if they play their cards right.

 

[alabama man]

Take all the useless graphics **** out and double the cores. If AMD screws up with the zen I guess I'll be using my current intel processor for 10+ years, just overclock the 100 or so megahertz they add yearly.

 

[Amet Monegro]

Zen is the future

 

[Theinsanegamer]

Take all the useless graphics **** out and double the cores. If AMD screws up with the zen I guess I'll be using my current intel processor for 10+ years, just overclock the 100 or so megahertz they add yearly.

This style of comment always pops up on CPU articles concerning intel's new arch of the day. They already have that. It's called HDET. The 6900k is a 8 core, the 6800 and 6850 are six cores. The 6950x is a 10 core. They have had 6 cores since nehalem. So why are you complaining? They have the chips you want, with a superior chipset to boot.

 

[meric]

Perhaps Intel is right to keep 6 or more core CPUs in prosumer space right now. Perhaps we must first check if the software companies provide proper multi core support in their software. In year 2016 how many new game titles use the full potential of your 4 core i5?

 

[theBest11778]

Perhaps Intel is right to keep 6 or more core CPUs in prosumer space right now. Perhaps we must first check if the software companies provide proper multi core support in their software. In year 2016 how many new game titles use the full potential of your 4 core i5?

Pretty much all 2016 games are too much for an i5. At this point an i7 4c/8t cpu is needed to push any GPU at the 1070 level and beyond. Gamers Nexus has shown that 6c/12t CPUS are ideal for battlefield 1. Even more important than IPC and core speed as their Haswell i7 5930k 3.5ghz was faster than the i7 6700k Skylake cpu at 4ghz.

 

[BadThad]

Intel as usual, issuing new processors that are just slightly faster.

 

[MilwaukeeMike]

Have we finally reached the upper limits of speed on these processors? It wasn't all that long ago that the claims were that all silicone based processors had reached their limits for speed and heat......or was it all just a bad dream?

It could be because AMD isn't giving them competition so they have no reason to make faster processors. If you look at Nvidia, they are doing the same thing.

They'd still love to sell upgrades to people who already own an Intel. Look at Apple - they keep coming out with more iPhones with small improvements even though there is a lot of competition. The problem isn't a willingness, it's just really hard to improve.

Making a processor on 14nm isn't easy - they're running out of improvements they can make. I'm sure they'd make them at 6nm if they could instead of 14. But you're already bumping up against what's even physically possible. You can't run a processor at 6.0 Ghz on air cooling or it'll melt. I think they're running out of ways to improve.

I think next we'll need to move past silicon.

 

[ScubaRhys]

I don't see much to complain about, most of us have decent Intel chips that perform better than we need them to. For me this is just another reason why my 4790k is still more than just fine.

 

[Vulcanproject]

Have we finally reached the upper limits of speed on these processors? It wasn't all that long ago that the claims were that all silicone based processors had reached their limits for speed and heat......or was it all just a bad dream?

It could be because AMD isn't giving them competition so they have no reason to make faster processors. If you look at Nvidia, they are doing the same thing.

Doubt that. Intel is genuinely worried about Zen, but it seems their engineers and MFG process progression have stalled. This happens to companies that have gone unchecked for awhile. Considering the Kaby lake Architecture isn't much different than Nahlem and pretty much the same as Sandy Bridge Intel hasn't produced a new CPU architecture in over 5 years. Remember Intel did well with the core 2 before the Core series, but 1 architecture before that... Netburst... So 2 successful architectures in a row doesn't mean they're really as far ahead as you think.

On the GPU side AMD needs to step away for a few years and let Nvidia discourage their own customers with massive price hikes (similar to what we saw this year.) Reemerge a few years later when people are begging for another GPU option.

AMD is in a good spot right now if they play their cards right.

Intel aren't that worried about Zen. Reports put it's IPC efficiency (instructions per clock) at no better than Haswell which is 3 years old. AMD are also debuting Zen on a 14nm process in the same year Intel transition to their 10nm process after more than 2 years of 14nm parts. They are still quite clearly one step ahead.

Zen is going to be a big jump for AMD and make them much more competitive, but it's not a leap ahead of what Intel have delivered for several years now. AMD should have been this competitive the past ten years, but have failed at nearly every turn since Core 2 arrived in 2006. All they are doing is FINALLY not screwing up constantly. This is better for competition, but it's not blowing the doors off the market.

Intel have delivered incremental upgrades and not significant architectural changes because they just haven't been forced into it by any worthy competitor. But also because yes, we are fast reaching the limits of conventional silicon lithography. Intel project 6nm on silicon, which is potentially just a couple steps away now.

Back in 2008 when they launched Nehalem to a huge IPC jump, they were at pains to point out much of it was from integrating the memory controller on die. That we aren't likely to see jumps like this in future because you can't make significant gains like that often.

I never forgot that. Most companies now are simply chipping away at their cutting edge silicon, grinding their way to the theoretical limit of the conventional processes we have. Scraping the barrel is the precise situation the market is in after nearly 50 years of commercial integrated circuits, waiting perhaps for a revolution elsewhere.....

 

[cliffordcooley]

noting that Kaby Lake is built on largely the same 14nm process

I'm not ashamed to admit, my 32nm Sandy Bridge is purring right along, waiting for CPUs built on a 10nm process.

 

[theBest11778]

Intel aren't that worried about Zen. Reports put it's IPC efficiency (instructions per clock) at no better than Haswell which is 3 years old. AMD are also debuting Zen on a 14nm process in the same year Intel transition to their 10nm process after more than 2 years of 14nm parts. They are still quite clearly one step ahead.

Zen is going to be a big jump for AMD and make them much more competitive, but it's not a leap ahead of what Intel have delivered for several years now. AMD should have been this competitive the past ten years, but have failed at nearly every turn since Core 2 arrived in 2006. All they are doing is FINALLY not screwing up constantly. This is better for competition, but it's not blowing the doors off the market.

Intel have delivered incremental upgrades and not significant architectural changes because they just haven't been forced into it by any worthy competitor. But also because yes, we are fast reaching the limits of conventional silicon lithography. Intel project 6nm on silicon, which is potentially just a couple steps away now.

Back in 2008 when they launched Nehalem to a huge IPC jump, they were at pains to point out much of it was from integrating the memory controller on die. That we aren't likely to see jumps like this in future because you can't make significant gains like that often.

I never forgot that. Most companies now are simply chipping away at their cutting edge silicon, grinding their way to the theoretical limit of the conventional processes we have. Scraping the barrel is the precise situation the market is in after nearly 50 years of commercial integrated circuits, waiting perhaps for a revolution elsewhere.....

It's impossible to know exact IPC of zen right now but looks more like broadwell level at the moment. Intel won't be jumping to 10nm until 2018 or the very end of 2017 at the earliest. Considering the IPC gains from broadwell to Skylake and Skylake to Kaby lake are insignificant this means it's irrelevant. The main thing will be price and amd are currently selling 8 core CPUS at $150. 8c/16t zen should be between $399-599. This will compete with Intel's hedt lineup which is... Broadwell. Not Skylake, not Kaby lake, but zen will be competing with broadwell. Intel will either 1.) cut prices in 1/2 to compete (and lose untold millions in the process,) or keep prices inflated and lose a ton of market share. Either way 2017 will be a bad year for Intel.

 

[Vulcanproject]

It's impossible to know exact IPC of zen right now but looks more like broadwell level at the moment. Intel won't be jumping to 10nm until 2018 or the very end of 2017 at the earliest. Considering the IPC gains from broadwell to Skylake and Skylake to Kaby lake are insignificant this means it's irrelevant. The main thing will be price and amd are currently selling 8 core CPUS at $150. 8c/16t zen should be between $399-599. This will compete with Intel's hedt lineup which is... Broadwell. Not Skylake, not Kaby lake, but zen will be competing with broadwell. Intel will either 1.) cut prices in 1/2 to compete (and lose untold millions in the process,) or keep prices inflated and lose a ton of market share. Either way 2017 will be a bad year for Intel.

It's impossible to know for certain just yet but most reasonable reports put it Broadwell, at best. Haswell/Broadwell, either way it's several years behind Intel's consumer parts at an IPC level. Intel's prices have been as such because they do not have competition to lower them. AMD bring competition, Intel can easily lower them. That is how this works.

Broadwell itself is still 2 years old and it's obsolete- LGA1150 is a legacy (I.e dead) platform, if you didn't know.

As for 10nm you're maybe not quite up to date on the news. Intel have had 10nm in production trials for several months now already and will go into full production in 2017. The line is running in the Israel plant and has been since the summer. Volume production is maybe 6 months away. 2017 is the year of 10nm Cannonlake, it is likely to be the 2nd half but it still shows how Intel are indeed a step or two ahead of AMD.

It'll not be a bad year for Intel at all, they will still completely dominate the server market, they still dominate the high end consumer, they still are a step ahead on process technology and they still have a more efficient architecture.

Zen closes the gap but it's nothing that shouldn't have been done years and years ago. AMD has always managed to compete at the lower end with price cuts, hence their regular ongoing losses. I am sure Zen will at least mean the company become profitable again, which is long overdue as far as investors are concerned.

 

[Pewzor]

Have we finally reached the upper limits of speed on these processors? It wasn't all that long ago that the claims were that all silicone based processors had reached their limits for speed and heat......or was it all just a bad dream?

It could be because AMD isn't giving them competition so they have no reason to make faster processors. If you look at Nvidia, they are doing the same thing.

Doubt that. Intel is genuinely worried about Zen, but it seems their engineers and MFG process progression have stalled. This happens to companies that have gone unchecked for awhile. Considering the Kaby lake Architecture isn't much different than Nahlem and pretty much the same as Sandy Bridge Intel hasn't produced a new CPU architecture in over 5 years. Remember Intel did well with the core 2 before the Core series, but 1 architecture before that... Netburst... So 2 successful architectures in a row doesn't mean they're really as far ahead as you think.

On the GPU side AMD needs to step away for a few years and let Nvidia discourage their own customers with massive price hikes (similar to what we saw this year.) Reemerge a few years later when people are begging for another GPU option.

AMD is in a good spot right now if they play their cards right.

Intel aren't that worried about Zen. Reports put it's IPC efficiency (instructions per clock) at no better than Haswell which is 3 years old. AMD are also debuting Zen on a 14nm process in the same year Intel transition to their 10nm process after more than 2 years of 14nm parts. They are still quite clearly one step ahead.

Zen is going to be a big jump for AMD and make them much more competitive, but it's not a leap ahead of what Intel have delivered for several years now. AMD should have been this competitive the past ten years, but have failed at nearly every turn since Core 2 arrived in 2006. All they are doing is FINALLY not screwing up constantly. This is better for competition, but it's not blowing the doors off the market.

Intel have delivered incremental upgrades and not significant architectural changes because they just haven't been forced into it by any worthy competitor. But also because yes, we are fast reaching the limits of conventional silicon lithography. Intel project 6nm on silicon, which is potentially just a couple steps away now.

Back in 2008 when they launched Nehalem to a huge IPC jump, they were at pains to point out much of it was from integrating the memory controller on die. That we aren't likely to see jumps like this in future because you can't make significant gains like that often.

I never forgot that. Most companies now are simply chipping away at their cutting edge silicon, grinding their way to the theoretical limit of the conventional processes we have. Scraping the barrel is the precise situation the market is in after nearly 50 years of commercial integrated circuits, waiting perhaps for a revolution elsewhere.....

Kaby Lake's IPC isn't much better than Haswell neither.

 

[Vulcanproject]

Kaby Lake's IPC isn't much better than Haswell neither.

True enough. But Kaby Lake is just an optimisation. Kaby Lake is the weird one and an anomaly because of the delays. I have said though this is the situation the industry is in now, there are no more big gains to be had without construction material advances.

As was pointed out Zen is competition to Broadwell-E, which is replaced by a brand new socket 2066 platform about 6 months after Zen launches anyway. LGA1150 is dead, the Broadwell LGA2011 platform is at the end of it's lifecycle as are the Broadwell based chips. The Skylake X and Kaby Lake X processors (on that new 2066, X299 series platform) replace them before the end of 2017.

Long story short you're only getting incremental gains no matter what Intel platform you're on. It's just news that AMD have probably sorted their house out after a decade of mediocrity.

 

[cliffordcooley]

Broadwell itself is still 2 years old and it's obsolete- LGA1150 is a legacy (I.e dead) platform, if you didn't know.

LGA1156 and older are the only platforms I am calling dead. There is not a great enough difference in the others to label any of them dead.

 

[Vulcanproject]

LGA1156 and older are the only platforms I am calling dead. There is not a great enough difference in the others to label any of them dead.

Difference isn't really the definition to label a platform dead though. If it has no more current or future CPU models and is being phased out of the channels it is on its way out.

LGA1150 boards for all the suppliers I have seen are labelled as a legacy platform, their selection is dwindling. I tried to find a good new board to replace a customer's failed H87 last week and found there wasn't a great selection around, particularly of the H97/Z97 refresh chipsets. It quite surprised me.

It's still a fine platform but it's obviously mostly swept out of the channel for LGA1151.

 

[scorpian007]

Looks like I'll be hanging onto my i7 4770K for another year or two

 

[BadThad]

Looks like I'll be hanging onto my i7 4770K for another year or two

I'll be hanging on to mine for at least 3-4 more years. I reluctantly upgraded from a Q6600 mainly due to chipset/mobo improvements. The old Q6600 system still runs anything I throw at it just fine!

 

[Jack Attack]

I'll stick with my dual Xeon X5650 Hex cores as the cost for both processors and board is cheaper than buying a single current processor but with better more core counts for what I need is for (Video Editing)