Leaked specs for upcoming Intel Skylake CPUs reveal high clock rates, 95W TDP and more

Shawn Knight

Posts: 15,240   +192
Staff member

intel skylake core processors specifications leaked tdp 2ghz boo intel skylake

Leaked specifications from a pair of upcoming Intel Skylake processors have hit the web. If accurate, we could see Intel pumping out CPUs with high clock rates on par with the company’s existing Devil’s Canyon offering.

The chips in question are the Intel Core i7-6700K and the Core i5-6600K, both being unlocked with overclocking in mind as evident by the “K” designation.

The faster i7-6700K is said to consist of four cores with Hyper-Threading clocked at 4.0GHz (Turbo Boost up to 4.2GHz), 8MB of last-level cache, a dual-channel DDR3/DDR4 memory controller, 95W TDP and Intel HD Graphics 5000-series, all on an LGA1151 packaging.

The slightly slower (albeit less expensive) i5-6600K will ship with a standard operating frequency of 3.5GHz (Turbo Boost up to 3.9GHz). It’ll also include 6MB of last-level cache as well as the same dual-channel DDR3/DDR4 memory controller, 95W TDP and Intel HD Graphics 5000-series on the LGA1151 package.

intel skylake core processors specifications leaked tdp 2ghz boo intel skylake

Skylake will represent a “tock” in Intel’s Tick Tock design model, meaning it’ll be built on the same 14-nanometer manufacturing process as Haswell but will feature a new microarchitecture. With microarchitecture advancements, Intel generally seeks to improve energy efficiency and performance as well as functionality and density of features.

It’s worth reiterating that these are simply leaks and haven’t yet been confirmed. As such, one should take this information with a healthy dose of salt.

Permalink to story.

 
@shawn
I read that the performance jump between Broadwell and Skylake will be like what we saw with Prescott to Conroe. Do you believe this is possible with the unlocked Skylake-S parts? Have you heard this rumour?
 
@shawn
I read that the performance jump between Broadwell and Skylake will be like what we saw with Prescott to Conroe.
Definitely not. The move from Netburst to Core was a change in the entire philosophy of how Intel designed processors, with the two architectures having entirely different goals. That's not the case today, the way they develop CPUs has changed only slightly from Core to Nehalem, and has remained essentially the same since Sandy Bridge. And Broadwell has already shown that a move to 14 nm doesn't yield performance improvements that large.
 
Funny, I had a feeling we would see the return of the 6600 moniker.

I'd like to believe the rumours about a large jump in performance but frankly I can't see how it could possibly be true.

The jump from Conroe to Prescott was akin to 486 to Pentium. A 75% increase in MIPS from the data I found. 50% improvement on benchmarks was common.

Think about it, we went from single to multi core!? What on earth could Intel have up their sleeve that would match that?

It's also worth noting Intel HAD to make that jump, because AMD's x2 was vastly superior to Conroe in pretty much every concievable way. The competition is now non-existent, sadly.

I wouldn't be suprised if Intel's desktop business is largely fueled by businesses replacing units after x number of years because the whole system becomes grotty and worn down, rather than a need to improve performance. That, and the desirability of motherboard features like the latest SATA variant.

Finally, Prescott to Conroe saw a whole new product nomenclature, and here we are just seeing the continuation of the I series' sequential move up the decimals (the 'jump' from 4 to 6 seems to be explained by the apparent floundering of Broadwell for desktop)

I think 10-20% more IPC, better cache performance, DDR4, and six cores for desktop would be about right. I'd also expect the S, T and E variants to offer performance slightly closer to the base models. And although integrated graphics are of zero interest to enthusiasts, improvements in 4k and h265 performance may tick a few boxes for some organisations.
 
Last edited:
The jump from Conroe to Prescott was akin to 486 to Pentium. A 75% increase in MIPS from the data I found. 50% improvement on benchmarks was common.
Think about it, we went from single to multi core!? What on earth could Intel have up their sleeve that would match that?
The change was more drastic than you think, because they didn't go from single core to multi-core. The Pentium D processors are from the Netburst era and are basically two Prescott CPUs in the same die. Even then, we still saw a 2.4 GHz Core 2 Duo outperform a 3.6 GHz Pentium D.
On the other hand, it's unlikely that DDR4 will bring any benefits to Skylake CPU performance. The only thing it brings to the table is higher bandwidth, which dual-channel DDR3 is already more than enough for 99% of tasks. Timings on DDR4 needed to be loosened to allow for the higher frequency (and bandwidth), which means that in the end there will be no improvements in latency. It will only be a benefit for laptops with integrated graphics, where more bandwidth is very useful, but desktop will probably see zero benefits.
 
The jump from Conroe to Prescott was akin to 486 to Pentium. A 75% increase in MIPS from the data I found. 50% improvement on benchmarks was common.
Think about it, we went from single to multi core!? What on earth could Intel have up their sleeve that would match that?
The change was more drastic than you think, because they didn't go from single core to multi-core. The Pentium D processors are from the Netburst era and are basically two Prescott CPUs in the same die. Even then, we still saw a 2.4 GHz Core 2 Duo outperform a 3.6 GHz Pentium D.
Oops, whatta mistake-a-to-make-a. In my defence, my recollection of that period is hazy, due to.... reasons....
On the other hand, it's unlikely that DDR4 will bring any benefits to Skylake CPU performance. The only thing it brings to the table is higher bandwidth, which dual-channel DDR3 is already more than enough for 99% of tasks. Timings on DDR4 needed to be loosened to allow for the higher frequency (and bandwidth), which means that in the end there will be no improvements in latency. It will only be a benefit for laptops with integrated graphics, where more bandwidth is very useful, but desktop will probably see zero benefits.
Ahh, yeah, I knew the APU's loved DDR4, and I see you're right here. New iterations of DDR don't really scale under other conditions.

Do you think HMC and HBM will be significant for desktop users?

Even disregaring DDR4, I'm downgrading my relative performance estimate. I had based that on Ivy Bridge (3770k) to Haswell (4790k), where we saw something like a >15% gain. But that's in part because of the higher clock speed of the 4790k (4ghz vs 3.5).
Comparing chips running at the same speed, the effective performance gains appear some way under 10%.

I'm in the market for an i7, coincidentally. As we already have z-97's with Sata-express and m-2 support, the six core chips we've seen don't match the 4790k's single thread performance, and it seems we'll be on 95w TDP for a while, I'm finding fewer and fewer reasons to hold out for Skylake.
 
So, i5 keeps the predecessors clocks, i7 have them cut. Add it to a previous leaks of problems with TDP on 14nm and one may think Skylake won't be a great overclocker. Plus, previous leak of internal slide where Intel boasts new features and energy efficiency rather than massive impromements in IPC, and one may really think Intel hit the performance wall and owners of Sandy/Ivy/Haswell can sleep tight. Plus new platform at the beginning will cost You some 20-30% more, poor choice of motherboards at launch and possible teething problems. Just ordered Haswell today. Not because of that news, more to do with a recently released game that 'destroyed' my ancient Q9550 and Witcher 3 in a pipe, too.
 
I would like to see more cores in a consumer chip, like 8, 10, or 12, or more, instead of a speed increase.
 
I think im just gonna stick with my i5 4690k devils canyon. I can hit 4.7 ghz on air no problem and have plenty of headroom when it comes to gaming.
 
Looks like the same old CPU's with a new name, was expecting 8 to 16 core with 5 ghz at least. I have 3970 3.5 That already what they are advertising in this article. And its 3 years old!
 
I have 3970 3.5
Then what the hell are you complaining about? Why would Intel release a CPU that is 4 times faster than anything AMD currently has available for mainstream market?

Perhaps you should look into a Dual Xeon platform with 2x E7-8890 v2, if you are not happy with a single i7 3970. Then you would have 30 cores with 60 threads to be disappointed with. Or go with Intel Xeon E7-8890 v2 that has a max of 8 CPU's configured together, for a grand total of 120 cores with 240 threads to be disappointed with. All because the Processor Base Frequency is 2.8GHz! Instead of 5GHz where AMD is forced to be with their low single-core efficiency.
 
"Comparing chips running at the same speed, the effective performance gains appear some way under 10%."
That is because in consumer applications we won't see much avx2 code (most probably), and that's (sadly for gamers) the haswells real strength.
 
Looks like the same old CPU's with a new name, was expecting 8 to 16 core with 5 ghz at least. I have 3970 3.5 That already what they are advertising in this article. And its 3 years old!
yeah, then buy a i7 5960x (the same price as 3970 was) with water cooling and overclock it to 4,5ghz...
 
I would like to see more cores in a consumer chip, like 8, 10, or 12, or more, instead of a speed increase.
You've had 8 core CPUs from AMD from 2011 and 8 thread Intel CPUs from 2008. You can buy 8 core/16 threads consumer CPU from Intel since last year.
 
Looks like the same old CPU's with a new name, was expecting 8 to 16 core with 5 ghz at least. I have 3970 3.5 That already what they are advertising in this article. And its 3 years old!

lol 8-16 cores at 5Ghz on what planet??

talk about unrealistic expectations.

High clock rate my a**, clock rates have been parked at 3.5 GHZ for the last 15 years,

The 4790k comes at 4Ghz stock clock.

15 years? we didn't have 3.5 ghz cpu's in 2000.

10 years is alittle more accurate
 
Good... It'll make buying the current set of chips that much cheaper as their prices will come down form this. I don't see a real reason for me to be buying this new chip yet.
 
Nice article!
I wonder if this will be a worthwhile upgrade over my i5-4460. I use my desktop for gaming only, not even heavily threaded games (e.g. SC2). What do you think Shawn?

Cheers
 
I wonder if this will be a worthwhile upgrade over my i5-4460.
My name not Shawn, but I think your i5 CPU will be find for a few more years yet.
Does this all make me an old fart because I am still running on an LGA 1156 Lynnfield i7-860???
If it still works for you, there is no shame in running an old system. There is still life in LGA775 for those that can use the system. I just wouldn't buy into those systems, if I had the option to get newer.
 
Do you think HMC and HBM will be significant for desktop users?
HBM will be significant due to being used on desktop GPUs. If you meant as system RAM, my guess is not so much. It's the same boat as DDR4, it's not very useful to provide more bandwidth when the software doesn't require it. It would become useful if software demands came to change, but to my knowledge that hasn't happened much since the DDR2 days.
That is unless HBM brings improvements in latency that I'm unaware of, which might be useful it it's a big enough improvement for a reasonable price. But I know using GDDR as system RAM is counter-productive exactly due to the higher latency compared to DDR, is that the case with HBM as well?
I'd think AMD's HSA has the potential to make a bigger impact than the RAM technology itself, because it would make it easier for GPGPU to finally take off everywhere. Suddenly having many applications tapping into some TFLOPS of processing power previously unused is a much bigger impact that new RAM types could ever provide.
 
My name not Shawn, but I think your i5 CPU will be find for a few more years yet.
If it still works for you, there is no shame in running an old system. There is still life in LGA775 for those that can use the system. I just wouldn't buy into those systems, if I had the option to get newer.
I handles games just fine for me, and my gtx660ti seems to run games at max settings on my 1080p monitor, so I don't see a reason to upgrade. :)
 
Back